CN101083637B - System and method for performing decoupling to a plurality of control loop - Google Patents

System and method for performing decoupling to a plurality of control loop Download PDF

Info

Publication number
CN101083637B
CN101083637B CN 200710109830 CN200710109830A CN101083637B CN 101083637 B CN101083637 B CN 101083637B CN 200710109830 CN200710109830 CN 200710109830 CN 200710109830 A CN200710109830 A CN 200710109830A CN 101083637 B CN101083637 B CN 101083637B
Authority
CN
China
Prior art keywords
value
data
data pattern
signal
group
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN 200710109830
Other languages
Chinese (zh)
Other versions
CN101083637A (en
Inventor
日高康雄
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fujitsu Ltd
Original Assignee
Fujitsu Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US11/753,164 external-priority patent/US7804921B2/en
Application filed by Fujitsu Ltd filed Critical Fujitsu Ltd
Publication of CN101083637A publication Critical patent/CN101083637A/en
Application granted granted Critical
Publication of CN101083637B publication Critical patent/CN101083637B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • Y02B60/31—
    • Y02B60/33—

Landscapes

  • Dc Digital Transmission (AREA)
  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)

Abstract

The invention provides a system and method for performing decoupling to a plurality of control loop. The method includes steps of: applying at least one of a loss compensation for frequency-dependent distortion and an offset compensation for DC-offset distortion to a signal before or after the distortion occurs to generate an output signal; using a clock signal, sampling the output signal to generate a plurality of data values and boundary values; monitoring the sampled values for a first set of data patterns; detecting in the sampled values a data pattern in the first set of data patterns; adjusting at least one of the loss compensation and the offset compensation applied to the signal based on one or more of the sampled data values and boundary values associated with the detected data pattern in the first set of data patterns; after adjusting at least one of the loss compensation and the offset compensation, monitoring the sampled values for a second set of data patterns; detecting in the sampled values a data pattern in the second set of data patterns; adjusting at least one of the loss compensation and the offset compensation applied to the signal based on one or more of the sampled data values and boundary values associated with the detected data pattern in the second set of data patterns.

Description

A plurality of control loops are carried out the system and method for decoupling zero
Technical field
Present invention relates in general to signal communication, more particularly, relate to a kind of system and method that a plurality of control loops is carried out decoupling zero.
Background technology
When transmitting signal by communication media, these signals can be subjected to decay owing to the phenomenon such as skin effect and DIELECTRIC ABSORPTION.Signal receiver can comprise this decay is compensated with the precision that improves signal communication and the equalizer of efficient.The compensation rate that expectation is applied by equalizer is as far as possible closely mated the degree of the decay that produces owing to medium, with irrespectively the output characteristic of inhibit signal is consistent in order to the particular communication path that transmits signal.
Summary of the invention
In one embodiment of the invention, a kind of method that is used for conditioning signal comprises: to take place signal before or after the distortion apply at the compensating for loss and damage of frequency dependent distortion and at least one of the migration of DC bias distortion to generate output signal.Described method also comprises uses clock signal that described output signal is sampled to generate a plurality of data values and boundary value.Described method also comprises: at the first data pattern group, and the value that monitoring is sampled.Described method also is included in the data pattern that detects in the value of being sampled in the described first data pattern group.Described method also comprises: based on one or more data value of sampling and boundary value of being associated with data pattern in the detected described first data pattern group, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated.Described method also comprises: after in described compensating for loss and damage and the described migration at least one having been carried out regulate, and at the second data pattern group, the value that monitoring is sampled.Described method also is included in the data pattern that detects in the value of being sampled in the described second data pattern group.Described method also comprises: based on one or more data value of sampling and boundary value of being associated with data pattern in the detected described second data pattern group, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated.
A technological merit of specific embodiment is to make the output signal equilibrium.Specific embodiment is at compensating by being used to send the signal attenuation that communication media caused of signal.This makes the output characteristic of signal be independent of to be used to transmit the communication path of signal and is consistent.Because signal level can be selected as falling in the dynamic range of system component, can comprise the improvement of component responds with the advantage that consistent output characteristic is associated.In addition, signal can be remained on enough level places, to prevent drop-out.
The other technologies advantage of specific embodiment comprises the adaptability to the different communication medium.Specific embodiment use variable gain amplifier is regulated the degree of compensation that signal applied to arriving.Such embodiment can be so that regulate compensation rate at different medium, thereby increase the versatility of the equalizer of these technology of realization.In addition, these embodiment can also be adapted to the change of the dielectric property that is associated with processing, voltage and variations in temperature.
The another technological merit of specific embodiment is to have improved the maximum service speed of equalizer and/or reduced the power that equalizer consumes.Specific embodiment uses existing clock and data recovery (CDR) function to generate output signal value, and this output signal value is used for the residual gain error and/or the residual direct current offset of monitoring output signal.Generate output signal value by use existing C DR function, and no longer need to use special-purpose observation circuit.By not using special-purpose observation circuit, can reduce load to equalizer output, improve the maximum service speed of equalizer and/or reduce the power that equalizer consumes.In addition, chip area be can reduce, and existing equalizer element and/or function in specific embodiment, reused.In addition, owing to do not use special-purpose observation circuit, can also reduce design effort at equalizer.
The another technological merit of specific embodiment is to have increased and the relevant flexibility of data pattern that can be used by equalizer.Specific embodiment can be used the data pattern with single transition, and is not limited to high frequency data pattern and/or has the data pattern of at least twice transformation.This flexibility has many advantages, and these advantages for example comprise can be so that other filter mode (that is, having the pattern of single transition) be applied to output signal, to offset the Duty Cycle Distortion that will discuss hereinafter.
Another technological merit of specific embodiment be can only use output signal value to regulate to be associated with input signal more than one independent Control Parameter.As mentioned above, compare with using observation circuit, the use of output signal value provides a plurality of advantages.In addition, can increase the effect of equalizer when compensating signal is decayed to adjusting more than one independent Control Parameter.
The another technological merit of specific embodiment is as one man the adaptive equalizer that is used for periodicity, quasi periodicity and abundant randomized sequence to be controlled.Another technological merit of specific embodiment is by filter mode being applied to the negative effect that gain-adjusted signal has before reduced Duty Cycle Distortion and quasi periodicity and cyclical signal.The pattern of the value that equally produces basically in the sequence that filter mode begins corresponding to the even data in signal or odd data place.Filter mode is applied to signal, and this can be by being avoided unacceptable result in the gain controlling in that balance is carried out in the action biasing of the self adaptation in the sequence that starts from even data or odd data place (Duty Cycle Distortion).
The another technological merit of specific embodiment is to use useful filter mode tabulation in conjunction with (standard) cyclical signal according to the mode of balance, to reduce the negative effect of Duty Cycle Distortion and (standard) cyclical signal.In these embodiments, the useful filter mode tabulation that is associated with (standard) cyclical signal can be that be scheduled to and fixing.In alternative embodiment, (standard) cyclical signal that this useful filter mode tabulation can be adapted to arrive.Filter mode in this tabulation can be used according to balance mode, to strengthen its applicability to (standard) cyclical signal.By from tabulation, sequentially, randomly or side by side selecting filter mode in this tabulation, can use filter mode in this tabulation according to balance mode.In specific embodiment, can use timer to skip nd filter mode, thereby increase the frequency of self adaptation action.
The another technological merit of specific embodiment is to regulate at observed residual direct current offset in the output signal.Regulate DC-offset compensation and can improve component responds (as, the sensitiveness of increase).In specific embodiment, can use output signal value to regulate DC-offset compensation, and can in signal, not produce data error.As mentioned above, compare, use output signal value (use existing C DR function and generate) that a plurality of advantages are provided with using observation circuit.In addition, because in signal, do not produce data error, not only in the receiving course of the signal that only carries test service, can also in the receiving course of the signal that comprises true data traffic, carry out adjusting to DC-offset compensation.By in the receiving course of the signal that comprises true data traffic (not being test service), carrying out adjusting, can improve the assembly sensitiveness in the receiving course of true data traffic to DC-offset compensation.
The another technological merit of specific embodiment is that the possible error lock problem that the adjusting of carrying out with observed residual direct current offset in output signal is accompanied is proofreaied and correct.When clock recovers and offset cancellation device generation error lock problem when influencing each other improperly, this cause sampling border and data value are exchanged.Regulate DC-offset compensation by high value or low value based on each boundary value, and no matter whether this boundary value is comprising between the continuous data value of conversion, specific embodiment is proofreaied and correct at the error lock problem.For the error lock problem in having used (standard) cyclical signal of double sampling is proofreaied and correct, specific embodiment at first uses output signal value to come Monitoring Data direct current imbalance (at the synonym of error lock problem).If the imbalance of detecting is then regulated DC-offset compensation based on detected imbalance.If the imbalance of not detecting is then based on only regulating DC-offset compensation at the high value or the low value that comprise those boundary values between the continuous data value of conversion.Like this, data direct current imbalance can change within the acceptable range, even use double sampling too at (standard) periodic data sequence.
The another technological merit of specific embodiment is to use output signal value to offset skew in each path in the multipath equalizer.In specific embodiment, skew in each path can be cancelled, and need not use any other circuit to monitor inside residual bias in the equalizer, in addition, in specific embodiment, in the operating process of equalizer, (that is, do not close any part of equalizer) and comprising that (being not only in the receiving course of the signal that only carries test service) can use offset cancellation control in the receiving course of signal of true data traffic.By in the receiving course of the signal that comprises true data traffic (not being test service), carrying out adjusting, can improve the assembly sensitiveness in the receiving course of true data traffic to DC-offset compensation.
Another technological merit that has used the specific embodiment of double sampling is to avoid double sampling cycle and (standard) cyclical signal are carried out any locking.When locking occurred, observed data pattern was different from the data pattern in whole (standard) cyclical signal in the data of double sampling, has postponed potentially by the performed control action of equalizer controller.By changing (standard) cyclical signal at each point of being sampled again in sampling period again, specific embodiment can be avoided locking.
The another technological merit of specific embodiment is that a plurality of control loops (for example being adaptive equalizer controller and offset cancellation device) are carried out decoupling zero.Can avoid sluggish in convergence time and the potential unsteadiness in control loop to the decoupling zero of a plurality of control loops.Specific embodiment can be mutually insensitive and to a plurality of control loop decoupling zeros by making the loop.For example, can make adaptive equalizer control insensitive, and make the offset cancellation device insensitive residual intersymbol interference (ISI) to residual bias.In order to make adaptive equalizer control and offset cancellation device insensitive mutually, in specific embodiment, use two groups of complementary data patterns according to the mode of balance by adaptive equalizer control and offset cancellation device.
The another technological merit of specific embodiment be unnecessary converge to the concrete mean value that under equilibrium state, generates the binary target variable in zero start stop mode (bang-bang) control system (as, ISI degree, equalization of level or other target variables arbitrarily), as the situation in typical start stop mode control system.The binary target variable for example can be the inverse correlation function that is applied to the data value of 1.5 positions (or symbol) before the boundary value between the opposite data value and this boundary value.Under concrete situation, the optimum mean value of the binary target variable in the balance (as, ISI degree, equalization of level or other target variables arbitrarily) can be greater than or less than zero according to various conditions (for example, the signal of channel loss and arrival self).Thereby, generated and restrained to such an extent that the embodiment of mean value of binary target variable of more approaching optimum mean value (promptly zero) is favourable.
The another technological merit of specific embodiment is the controlled target that dynamically generates under the equilibrium state mean value at the binary target variable (as, ISI degree, equalization of level or other target variables arbitrarily).In specific embodiment, optimum average ISI degree may be for height loss channel for high, and be low for low loss channel.Thereby, comprised that the embodiment at the controlling value of the mean value of binary target variable (this variable along with the value of control variables dynamically change) can be favourable.
With reference to the accompanying drawings, specification and claim, those skilled in the art will easily understand other technological merit.And although above enumerated concrete advantage, specific embodiment may comprise some or all in the cited advantage, does not perhaps comprise cited any advantage.
Description of drawings
Fig. 1 has been the illustration block diagram of example digital signal transfer system;
Fig. 2 be in more detail illustration the block diagram of example digital signal transfer system of Fig. 1;
The block diagram according to the example receiver of specific embodiment of the example digital signal transfer system of Fig. 3 is illustration Fig. 2;
A plurality of examples of clock signal that Fig. 4 A, 4B and 4C and the equalizer output signal that has showed polytype effect of inter-symbol have contrasted the ground illustration;
Fig. 5 is an illustration, and according to a particular embodiment of the present invention being used for makes an explanation with the flow chart of method that residual intersymbol interference is compensated to output signal value;
The table of the exemplary gain controlling schemes that Fig. 6 is an illustration is associated with the method for Fig. 5;
Fig. 7 is the illustration flow chart that is used for exemplary method that simulation second dervative equalizer is made an explanation at the output signal value of a plurality of independent Control Parameter according to a particular embodiment of the present invention;
The table of the exemplary gain controlling schemes that Fig. 8 is an illustration is associated with the method for Fig. 7;
The flow chart of Fig. 9 is the illustration exemplary method that the output signal value that is used for a plurality of parametric equalizers of 3 tap FIR filters according to a particular embodiment of the present invention makes an explanation;
The table of the exemplary gain controlling schemes that Figure 10 is an illustration is associated with the method for Fig. 9;
Figure 11 illustration the example boundary information that influenced by Duty Cycle Distortion;
Figure 12 is the illustration flow chart that is used for the selective filter pattern with the exemplary method of the negative effect that reduces Duty Cycle Distortion according to a particular embodiment of the present invention;
The table of the example distribution of six bit data patterns in even 8B10B idle data sequence that Figure 13 has been an illustration and the strange 8B10B idle data sequence;
The table of the exemplary gain controlling schemes that the gain that does not change component, first derivative component and second dervative component that Figure 14 is an illustration with use the example filter pattern of deriving from the table of Figure 13 to regulate imposes on input signal is associated;
The table of the example distribution of six bit data patterns that Figure 15 has been an illustration in even 8B10B CJPAT data sequence and strange 8B10B CJPAT data sequence;
The table of the exemplary gain controlling schemes that the gain that does not change component, first derivative component and second dervative component that Figure 16 is an illustration with use the example filter pattern of deriving according to the table of Figure 15 to regulate imposes on input signal is associated;
Figure 17 is the illustration flow chart that is used for dynamically generating the exemplary method of useful filter mode tabulation according to a particular embodiment of the present invention;
Figure 18 is the illustration flow chart that is used for dynamically generating another exemplary method of useful filter mode tabulation according to a particular embodiment of the present invention;
Figure 19 has been the illustration flow chart of an exemplary method again that is used for dynamically generating useful filter mode tabulation according to a particular embodiment of the present invention;
Figure 20 has been the illustration flow chart that is used for using the exemplary method of filter mode according to a particular embodiment of the present invention by balance mode;
Figure 21 has been the illustration flow chart that is used for using another exemplary method of filter mode according to a particular embodiment of the present invention by balance mode;
Figure 22 has been the illustration flow chart that is used for after certain time period, skipping the exemplary method of undetected filter mode according to a particular embodiment of the present invention;
A plurality of examples of clock signal that Figure 23 A, 23B and 23C and the equalizer output signal that has showed polytype residual direct current offset have contrasted the ground illustration;
Figure 24 is an illustration, and according to a particular embodiment of the present invention being used for makes an explanation with the flow chart of method that residual direct current offset is offset to output signal value;
The table of the example skew controlling schemes that Figure 25 is an illustration is associated with the method for Figure 24;
Figure 26 has been the illustration flow chart that is used for method that the mistake locking that takes place in the process of offsetting residual direct current offset is corrected according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 27 is an illustration is associated with the method for Figure 26;
Figure 28 has been the illustration flow chart that is used for other method that the mistake locking that takes place in the process of offsetting residual direct current offset is corrected according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 29 is an illustration is associated with the method for Figure 28;
Figure 30 and clock signal have contrasted the ground illustration DC path output that shows negative residual direct current offset in the example first derivative equalizer, the first derivative path output that shows just residual direct current offset and the example that mainly shows the equalizer output signal of zero residual direct current offset;
Figure 31 is the illustration flow chart that is used for exemplary method that the residual direct current offset of first derivative analog equalizer is offset according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 32 is an illustration is associated with the method for Figure 31;
Figure 33 is the illustration flow chart that is used for another exemplary method that the residual direct current offset of first derivative analog equalizer is offset according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 34 is an illustration is associated with the method for Figure 33;
Figure 35 is an illustration the flow chart of an exemplary method again that the residual direct current offset of first derivative analog equalizer is offset of being used for according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 36 is an illustration is associated with the method for Figure 35;
Figure 37 is an illustration the flow chart of an exemplary method again that the residual direct current offset of first derivative analog equalizer is offset of being used for according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 38 is an illustration is associated with the method for Figure 37;
Figure 39 is the illustration flow chart that is used for exemplary method that the residual direct current offset of second dervative analog equalizer is offset according to a particular embodiment of the present invention;
The table of the example skew controlling schemes that Figure 40 is an illustration is associated with the method for Figure 39;
The flow chart of the exemplary method of the influence that is used to reduce Duty Cycle Distortion according to a particular embodiment of the present invention that Figure 41 is an illustration;
The flow chart of another exemplary method of the influence that is used to reduce Duty Cycle Distortion according to a particular embodiment of the present invention that Figure 42 is an illustration;
Figure 43 has been the illustration flow chart that is used for changing the exemplary method of the point that double sampling takes place according to a particular embodiment of the present invention in each double sampling circulation;
Figure 44 has been the illustration flow chart that is used for changing another exemplary method of the point that double sampling takes place according to a particular embodiment of the present invention in each double sampling circulation;
Figure 45 is an illustration according to a particular embodiment of the present invention be used for changing in each double sampling circulation the flow chart of an exemplary method again of the point of sampling again takes place;
Figure 46 has been the illustration flow chart that is used for the exemplary method of a plurality of control loop decoupling zeros according to a particular embodiment of the present invention;
Figure 47 has been the illustration flow chart that is used for another exemplary method of a plurality of control loop decoupling zeros according to a particular embodiment of the present invention;
The flow chart of the exemplary method of the certain average value that is used to generate binary target variable under the poised state (for example, ISI degree, EQ degree or residual bias) according to a particular embodiment of the present invention that Figure 48 is an illustration;
Figure 49 is the illustration flow chart that is used for dynamically generating at the exemplary method of the controlled target of the mean value of binary target variable under the poised state (for example, ISI degree) according to a particular embodiment of the present invention;
Figure 50 is an illustration uses example controlled target formula dynamically to generate the curve chart at the result of the example controlled target of the mean value of binary target variable under poised state in equalizer gain controlling according to a particular embodiment of the present invention;
Figure 51 has been the illustration table that is used for the high-frequency gain code is converted to the exemplary scenario of DC path gain code and single order path gain code according to a particular embodiment of the present invention; And
The result's of the exemplary scenario of using Figure 51 in order to convert the high-frequency gain code to DC path gain code and single order path gain code according to a particular embodiment of the present invention that Figure 52 A and 52B are illustrations curve chart.
Embodiment
Fig. 1 has been the illustration block diagram of example digital signal transfer system 10.Digital signal transfer system 10 comprises transmitter 20, communication channel 30 and receiver 40.Transmitter 20 can comprise can be operated to send any suitable transmitter of the signal that carries digital information to receiver 40 by channel 30.In a particular embodiment, transmitter 20 can transmit information by fast relatively speed.Channel 30 can comprise any suitable channel or other communication medias.Channel 30 can comprise cable, the insulator with this cable insulation, this cable encapsulation and/or the connector on every side that for example transports signal.Channel 30 can be operated to give receiver 40 from transmitter 20 received signals and with these signal forwarding.Receiver 40 can comprise can operate with by channel 30 from transmitter 20 received signals and any suitable receiver that the digital information the received signal is suitably handled.
In typical digital signal transfer system (as high-speed communication system), as illustrated in Figure 32, receiver 40 received signals usually can the distortion owing to the frequency dependence decay.In general, for the signal attenuation in the conductive electrical communication medium, there are two important reasons.First major reason is the skin effect that produces along the conduction of communication media owing to signal.Second major reason is the DIELECTRIC ABSORPTION of communication media to signal.Usually, the loss of signal amount of representing by decibel that causes owing to skin effect is product a sX √ f, wherein a sBe the skin effect coefficient of material, x is the length of propagating along material, and f is the frequency of signal.The waste that causes owing to DIELECTRIC ABSORPTION is product a dXf, wherein a dIt is the DIELECTRIC ABSORPTION coefficient of material.
According to material and signal frequency, the relative importance of these effects can alter a great deal.Therefore, for example, cable may have the DIELECTRIC ABSORPTION coefficient more much smaller than skin effect coefficient, makes that the loss that causes owing to skin effect plays a leading role except under high-frequency.On the other hand, base plate trace (backplane trace) may have higher DIELECTRIC ABSORPTION coefficient, make since loss that DIELECTRIC ABSORPTION causes and the waste that causes owing to skin effect quite or bigger.In addition, the variation of operating condition (as variations in temperature) also may influence characteristics of signals.
May also present residual direct current offset distortion by the signal after receiver 40 processing.For example, may cause residual direct current offset owing to manufacturing technology (as device geometries mismatch or threshold voltage mismatch) and/or receiver parts oneself.As further describing, can use equalizer to come the decay of compensating frequency dependence, and can use the offset cancellation device to offset residual direct current offset below in conjunction with Fig. 2 and 3.
Fig. 2 be in more detail illustration the block diagram of example digital signal transfer system of Fig. 1.As can be observed, transmitter 20 comprises transmitter logic 22 and transmitter equalizer 24.Transmitter logic 22 can comprise any suitable logic that can operate information is encoded and send.Transmitter equalizer 24 can comprise any suitable equalizer, and this equalizer can be operated the distortion that may stand because frequency dependence is decayed sending signal with (for example by regulating the gain of signal to be sent) and compensate on channel 30.Only, in a particular embodiment, can compensate gain like that as illustrated in Figure 26 as example.By this mode, equalizer 24 can be before distortion occurring (for example utilize transmitter strengthens in advance (pre-emphasis) equilibrium) carries out precompensation (or balanced) to signal.In a particular embodiment, transmitter equalizer 24 can be operated with the receiver balance device of describing below in conjunction with Fig. 3 42 similarly based on the feedback from receiver logic 47.Yet, should be pointed out that equalizer 24 can compensate signal before distortion occurring, and equalizer 42 can compensate to signal after distortion occurring.It should also be noted that the some or all of logics in (following) logic 47 can be arranged in transmitter 20 or any other suitable position, and not necessarily all in receiver 40.
Receiver 40 comprises receiver balance device 42, equalizer output 46, receiver logic 47, gain control signal 48 and offset control signal 49.Receiver balance device 42 can comprise can be operated to receive input signal that comprises input data signal and any suitable equalizer that received input data signal is applied gain and/or skew at the input port place.Receiver logic 47 can comprise any suitable components or the parts group that can operate with the receive clock signal, as sampler.This clock signal can comprise any suitable clock signal that can be recovered by clock and data recovery (CDR) circuit from input signal, as recovered clock signal.Utilize received clock signal, as further describing below in conjunction with Fig. 3, receiver logic 47 can be operated so that equalizer is exported 46 and sample, and based on this sampling the gain control signal 48 and/or the offset control signal 49 that impose on input data signal is regulated with the compensating signal distortion.Only, in a particular embodiment, can as Figure 52 is illustrated, compensate gain as example.In a particular embodiment, can generate as Figure 54 is illustrated the compensation of gain and finished the equalizer output 46 that has compensated the frequency dependence distortion.In alternative embodiment, equalizer output 46 can be not by full remuneration.In a particular embodiment, receiver 40 can be operated in any suitable way the information in the equalizer output 46 is transmitted and sends to the parts of one or more any suitable quantity downstream.
Such as described in more detail below, receiver 40 compensates distorted signals not using the dedicated monitor circuit to detect under the situation of distortion.Owing to do not use the dedicated monitor circuit, so receiver 40 can be realized one or more technological merit.These advantages for example can comprise improve equalizer 42 the highest operation rate and/or reduce the power that receiver 40 consumes.In addition, can dwindle chip region, can reuse existing equalizer component and/or functional, and/or (because needn't design specialized monitoring circuit and) can reduce the required design effort of design receiver 40.
Should be pointed out that in some specific embodiments, can not carry out precompensation and transmitter 20 can not comprise equalizer 24.In these embodiments, receiver 40 can use 42 pairs of distortions of equalizer to compensate.In some alternative embodiment, transmitter 20 can comprise equalizer 24, and can carry out precompensation (that is, transmitter strengthens in advance).Among some embodiment in these embodiments, receiver 40 can also use 42 pairs of distortions of equalizer to compensate.Among other embodiment in these embodiments, receiver 40 can not use 42 pairs of distortions of equalizer to compensate, and receiver 40 can not comprise equalizer 42.
Should be pointed out that in a particular embodiment, can be called the part of adaptive equalizer signal being carried out parts (for example, logic 47 and equalizer 42, logic 47 and equalizer 24 and/or logic 47 and a plurality of equalizer) at the compensation of distortion.It should also be noted that adaptive equalizer can compensate by mode described herein in the environment except that (as described) signal transfer system.For example, adaptive equalizer can compensate by mode described herein (or in a similar manner) in record channel (for example, magnetic recording channel or optical recording channel).And, can use the equalizer (for example comprising linear equalizer and DFF) of any suitable type to compensate as described herein.
Block diagram in Fig. 3 has been the illustration example digital signal transfer system 10 of Fig. 2 according to the example receiver 40 of specific embodiment.Equalizer 42 can be operated so that the decay of using communication media 30 to send the signal of equalizer 42 to is compensated.In the embodiment shown, receiver logic 47 comprises the adaptive controller of the amount of gain that imposes on each signal path among 3 signal path 101A, 101B and the 101C being regulated based on the output signal of being sampled out by sampler 104 102.The performance of equalizer 42 may suffer residual direct current offset.Therefore receiver logic 47 can also comprise offset controller 106, and offset controller 106 is regulated the DC-offset compensation amount of the signal that imposes on arrival based on the direct current offset of the output signal of being sampled out by sampler 104.The miscellaneous part of equalizer 42 comprises variable gain limiting amplifier 110, mathematical operator (S) 112, postpones maker 114, variable gain amplifier 116, combiner 118 and driving amplifier 120.The miscellaneous part of receiver logic 47 comprises sampler 104 and clock 105.To be illustrated as output 50 from the output signal of sampler 104.
For the distortion of compensating frequency dependence, equalizer 42 can (use any suitable separator) received input signal 108 is divided on 3 signal path 101A, 101B and 101C, and uses variable gain amplifier 116 optionally the signal section on each path to be amplified.The first path 101A does not partly perform mathematical calculations to received input signal.The second path 101B carries out the single order mathematical operation to signal, operates as derivative.This operation can and be illustrated as mathematical operator (S) 112 based on the frequency of signal.As following also as described in, Third Road footpath 101C carries out the second order mathematical operation to signal, operates as second dervative.This operation also can come illustration based on the frequency of signal and by using two mathematical operator (S) 112.Optionally amplify the compensation that the frequency dependence loss effect in the channel 30 of 42 couples of Fig. 2 of equalizer is similar to by single order and two order components to signal.In alternative embodiment, equalizer 42 can have the path of any suitable quantity, for example only has a path.Equalizer 42 can be the equalizer example that compensates at distortion concurrently.Should be noted that, (for example can adopt any suitable balancing technique (for example transmitter strengthens equilibrium and/or receiver balance in advance) and any suitable equalizer, simulation first derivative filter continuous time, simulation continuous time second dervative filter, many taps finite impulse response filter and/or many taps DFF) in any suitable way (for example, before distortion occurs and/or afterwards) carry out compensation concurrently at distortion.What should also be pointed out that is, in alternative embodiment, (for example can adopt any suitable balancing technique (for example transmitter strengthens equilibrium and/or receiver balance in advance) and any suitable equalizer, linear equalizer and/or DFF) in any suitable way (for example, before distortion occurs and/or afterwards) carry out compensation serially at distortion.
Adaptive controller 102 can comprise and is used for any suitable components or unit construction that the information relevant with the output signal of equalizer 42 is analyzed and is used for the corresponding gain of each variable gain amplifier 116 is regulated.Adaptive controller 102 can comprise simulation and/or digital electronic component, as transistor, resistor, amplifier, constant-current source or other similar parts.Adaptive controller 102 can also comprise the suitable components that is used for that the signal from analog conversion of signals become digital signal or converts analog signal from digital signal to.According to specific embodiment, adaptive controller 102 comprises digital processing unit, as microprocessor, microcontroller, embedded logic or other information processing apparatus.
In a particular embodiment, adaptive controller 102 receives data and the boundary value information that is associated with output signal from sampler 104.This value information can comprise high value or the low value (as " 1 " or " 0 ") that for example is associated with each data from the sample survey and/or boundary value.Such as described further below, based on this value information, adaptive controller 102 can be operated so that suitable adjusting is carried out in the gain that imposes on input data signal.For gain is regulated, in a particular embodiment, adaptive controller 102 can be regulated the gain that is applied to regulate to the bias current that imposes on each variable gain amplifier 116.Use bias current to come an advantage of control amplifier 116 to be: the amount of gain that it can pair amplifier applies under the situation of the bandwidth that does not change amplifier is regulated, even make that amplifier also can keep its dynamic range when gain increases.
Sampler 104 can comprise any suitable components, and these parts are designed to from for example driving amplifier 120 reception equalizer outputs 46 and from for example clock 105 receive clock signals and by at interval equalizer output 46 being sampled by the setting that this clock signal limited.This sampling can be to the sampling of equalizer output 46 data values that are associated and/or boundary value and can represent the high value or the low value of each value in these values.Sampler 104 also can be operated with data value of will be sampled out and boundary value and be transmitted to adaptive controller 102 and/or offset controller 106.In a particular embodiment, sampler 104 can comprise the judgement latch of carrying out sampling and 1 (1-bit) analog to digital conversion.In alternative embodiment, sampler 104 can comprise that simulated sampling and maintenance (S/H) circuit are to sample to analog information and to transmit to carry out analog.In other alternative embodiment, sampler 104 can comprise multidigit AD converter (ADC) and digital information is transmitted to carry out Digital Signal Processing.
Offset controller 106 can comprise and being used for exporting any suitable components or the unit construction that 46 relevant information are analyzed and are used for the DC-offset compensation amount that one-level or more multistage place at variable gain amplifier 116 apply is regulated with the equalizer of equalizer 42.In a particular embodiment, offset controller 106 can comprise microprocessor, microcontroller, embedded logic and/or any other suitable components or unit construction.
In a particular embodiment, offset controller 106 receives and equalizer output 46 data that are associated and boundary value information from sampler 104.This value information can comprise the high value or the low value of each data from the sample survey for example and/or boundary value.Such as described further below, based on this value information, offset controller 106 can be operated the bucking voltage that imposes on input data signal is carried out suitable adjusting (that is, proofreading and correct) to proofread and correct or any residual direct current offset of compensation (that is, offsetting).
Can especially, give DC-offset compensation by the various parts of equalizer 42 to signal by variable gain amplifier 116.In multistage variable gain amplifier, between level and level, direct current offset can be accumulated.For skew is proofreaied and correct, offset controller 106 can be to being applied direct voltage by variable gain amplifier 116 amplifying signals.According to specific embodiment, 106 fens multisteps of offset controller apply compensation (that is, proofreading and correct) voltage, wherein carry out each step at the place not at the same level of variable gain amplifier 116.In such an embodiment, can determine the voltage that applies in each step in any suitable way.For example, can between these steps, divide cumulative correction voltage equably, perhaps the cumulative correction voltage distribution can be become the proportional amount of gain with corresponding stage.Should be pointed out that can be alternatively by carry out some or all task of carrying out by offset controller 106 such as any other suitable components of sampler 104.
Variable gain limiting amplifier (VGLA) 110 expression is used for the parts that the input signal 108 that is received by equalizer 42 is adjusted or the set of parts.This adjustment is handled the overall level of input signal 108 is regulated this signal remained on mathematical operator (S) 112 and to postpone in the dynamic range of maker 114.In a particular embodiment, control the amplification quantity that applies by VGLA 110 by the bias current that imposes on VGLA 110.
(incoming) signal of mathematical operator (S) 112 expression generations and arrival is with respect to any parts of the output of the linear ratio of derivative (being called as " single order operation ") of time or the set of parts.Mathematical operator S112 can comprise any suitable electronic unit or Circuits System, as is used for the high pass filter of the mathematical operation of carry out desired.According to specific embodiment, this operation is derivative operation, and it gets signal of coming with respect to the derivative of time, as the change in voltage of per 100 psecs of signal that arrive.Mathematical operator S112 can apply one or many to signal, the number of times that applies based on S112 and obtain and the signal that arrives with respect to the single order of time, second order, three rank or the proportional output signal of higher derivative more.
Postponing maker 114 is illustrated in the transport process of signal and introduces any parts of time delay or the set of parts.Postpone maker 114 and can comprise any suitable electronic unit or Circuits System.According to specific embodiment, be approximately equal to the required time quantum of mathematical operator S112 that will be applied to signal by the delay that postpones 114 pairs of signals introducings of maker.Therefore, can use and postpone maker 114 each several part of the output signal required time quantum of advancing downwards in the path of correspondence 101A, 101B or 101C is equated.By this mode, when arriving merging (mixing) device 118 places, the each several part of signal can make them synchronous.
Variable gain amplifier 116 expression is used for any parts or a plurality of parts that signal is amplified.Variable gain amplifier 116 can comprise any suitable electronic unit, and in a particular embodiment, controls each variable gain amplifier 116 by the bias current that imposes on particular variable gain amplifier 116.In some cases, the response time of carrying out the specific features of processing and amplifying may be oversize, makes amplifier not amplify effectively fast-changing high-frequency signal between high value and low value.Therefore, variable gain amplifier 116 can comprise a series of levels, and each grade all carried out the part in the whole processing and amplifying.Owing to do not have which grade to have the burden of carrying out all processing and amplifying, so the corresponding required time of gaining of each progressive Xingqi is also shorter.This makes this multistage variable gain amplifier 116 to respond the signal of higher frequency.
Variable gain amplifier 116 can also carry out DC-offset compensation to signal.In casacade multi-amplifier, each level can be carried out DC-offset compensation.The method that direct current offset is proofreaied and correct is to apply correction voltage with the direct current offset in the correction signal.Can before initialize signal is exaggerated, this correction voltage all be imposed on initialize signal.Yet, all apply this voltage at a some place and may cause this signal to surpass the dynamic range of one or more grade of amplifier 116.In addition, when increasing new level at every turn the voltage that is applied is calculated and regulates, if thereby at different levels gain be variable, then direct current offset may be distributed in unevenly in these grades.In order to solve this difficulty, a plurality of grades of places that specific embodiment can be included in amplifier 116 apply correction voltage.This possibility that makes the direct current offset that can proofread and correct this grade at places at different levels, reduction trimming process can make signal exceed the dynamic range of amplifier is also eliminated necessity of at whole array direct current offset having been calculated again when at every turn having increased level.In addition,, carry out correction voltage at places at different levels and make and be convenient to direct current offset is proofreaied and correct, make different levels can have different gains and can give different direct current offsets when gain at different levels when being variable independently.
Combiner 118 expressions are used for the signal on communication path 101A, 101B and the 101C is combined into the set of the parts or the parts of individual signals.Combiner 118 can comprise any suitable electronic unit.Combiner 118 provides composite signal to driving amplifier 120.Driving amplifier 120 expression is used for any parts that composite signal is amplified or the set of parts.120 pairs of composite signals of driving amplifier are carried out any suitable amplification to generate equalizer output 46 from equalizer 42, and equalizer output 46 has sufficiently high signal level so that can effectively send output signal to sampler 104.
In operation, equalizer 42 receiving inputted signals 108, input signal 108 comprise because the input data signal of having decayed through the communication process of communication media.110 pairs of these signals of VGLA are adjusted, and make signal level in the dynamic range of mathematical operator (S) 112 and delay maker 114.Equalizer 42 is divided this input signal between 3 path 101A, 101B and 101C.By postpone maker 114 with twice of the signal delay on the 101A of path so that the signal on the signal on the 101A of path and the path 101B (its stand mathematical operator 112 handle once and be delayed maker 114 postpone once) synchronously, and synchronous with the signal (it stands mathematical operator 112 and handles twice) on the 101C of path.Like this, input signal component on 3 path 101A, 101B and the 101C corresponds respectively to the input signal that does not stand mathematical operation, stood the single order operation and stood the second order operation, and (use postpone maker 114 makes) these 3 components arrive combiner 118 places synchronously at approximate synchronization.
Equalizer 42 then utilizes the signal on 116 pairs of each paths of corresponding variable gain amplifier to amplify.Control the gain of each amplifier 116 by adaptive controller 102, and this gain can be different for each path 101A, 101B and 101C.This make equalizer 42 can at and the frequency of signal between have a plurality of loss effects of different proportion relation and in various degree compensation be provided.In general, proportional to the compensation rate and a ratio of specific effect with respect to basic signal (base signal), this ratio is promptly to the amplification of respective path and ratio to the amplification that does not change signal on the 101A of path.Therefore, path 101A can not apply gain or apply slight negative gain (by dB) imposes on the compensation in other paths with increase relative effect.Offset controller 106 is proofreaied and correct any direct current offset of giving the corresponding signal on each path 101A, 101B and the 101C by corresponding amplifier 116 and/or any other suitable components.
To be combined into individual signals from the amplifying signal in each path by combiner 118.120 pairs of these output signals of driving amplifier are amplified so that can send this output signal to another destination effectively.Sampler 104 receives from the equalizer output signal 46 of driving amplifier 120 and the clock signal of coming self-clock 105.Sampler 104 is by being sampled to equalizer output signal 46 at interval by the setting that this clock signal limited, to generate data value and the boundary value that is associated with equalizer output signal 46.Select as another kind, sampler 104 can be sampled only to generate data value and data value and other suitable phase informations of being sampled out are transmitted to adaptive controller 102 and offset controller 106 to equalizer output signal 46.Adaptive controller 102 and offset controller 106 then can utilize the data value and the phase information of coming transmitted to derive one or more boundary value.Usually, if phase place early, then the high value of boundary value or low value are identical with the high value or the low value of the last data value that is right after.If phase place is later, then the high value of boundary value or low value are identical with the high value or the low value of a back data value that is right after.
Such as described in more detail below, 102 pairs of data from the sample survey that are associated with equalizer output signal 46 of adaptive controller and boundary value are analyzed so that the amount of gain that imposes on one or more path among path 101A, 101B and the 101C is regulated, thereby suitable compensation is carried out in residual frequency dependence decay.106 pairs of data from the sample survey that are associated with equalizer output signal 46 of offset controller and boundary value are analyzed so that the correction voltage amount that imposes on one or more path among path 101A, 101B and the 101C is regulated, thereby suitably offset residual direct current offset.
Do not use (using in many canonical systems) dedicated monitor circuit in above-mentioned adaptive equalizer, an one advantage is to reduce the load of equalizer output 46 in a particular embodiment.Especially in high speed circuit, the load that reduces equalizer output 46 can improve the highest service speed of equalizer 42 and/or the power that reduction equalizer 42 consumes.Do not use the dedicated monitor circuit can also dwindle chip area, can effectively reuse existing receiver parts (for example clock 105), and can reduce the required design effort of design specialized monitoring circuit.
Although the specific embodiment of equalizer 42 is described in detail, there are many other possible embodiment.Possible modified example is as comprising: path 101A, 101B and 101C are carried out different or additional mathematical operation, to compensate at different loss character; Increase or reduce the quantity in path; Adopt Artificial Control at controller 102 and 106, but not automatic feedback control; Use one-stage amplifier 116; Receive the signal that (and suitably regulating) comprises difference sequence (as low voltage difference signaling (LVDS)); And other variations of advising of above description.Generally, can reset, revise or omit assembly in any suitable way, and can be in any suitable way be distributed in the function of carrying out by these parts in difference or the additional parts or be incorporated in the single parts.Therefore, should be understood that the realization to receiver 40, equalizer 42 and receiver logic 47 can comprise any this variation, and can in any suitable equalizer environment, use specific embodiment of the present invention.Obtain the more details relevant with operable specific example equalizer component, seeing also the title of submitting on February 20th, 2004 is the non-interim U. S. application (sequence number 10/783,170) of " Adaptive Equalizer with DC Offset Compensation ".
As discussed above, can stand the frequency dependence decay by channel 30 transmissions and at the signal that receiver 40 receives.At receiver 40 places, equalizer 42 can apply gain so that the decay that signal was shown is compensated to received input signal.Receiver logic 47 can be analyzed the equalizer output signal 46 after regulating at residual decay, and based on this feedback the gain that is imposed on input signal by equalizer 42 is regulated.Specifically, sampler 104 can receive equalizer output signal 46 (input signal after the adjusting) and clock signal, and at the specified point place that is determined by this clock signal this output signal is sampled to generate data value and boundary value.Sampler 104 can then be transmitted to adaptive controller 102 to carry out (as follows) suitable analysis with these data and boundary value.Analyze based on this, adaptive controller 102 can be regulated the gain of the input signal that imposes on arrival.
A plurality of examples of clock signal that Fig. 4 A, 4B and 4C and the equalizer output signal that has showed polytype effect of inter-symbol have contrasted the ground illustration.In a particular embodiment, sampler 104 can receive the signal such as illustrative those signals in these figure, and comes output signal is sampled according to 2x oversampling clock and data recovery (CDR) scheme.By this scheme, sampler 104 can be sampled twice every data bit period (it can be limited by clock signal) to received signal.For a data bit period, sampler 104 can be in output signal should be corresponding to the some place of data value to the output signal sampling once, and in output signal should be to the output signal sampling once corresponding to the some place of boundary value.Based on the analysis to particular data and boundary value, such as described further below, adaptive controller 102 can be regulated the gain that imposes on the signal that is received by equalizer 42.
Clock signal example 200 that Fig. 4 A and equalizer output signal 46 have contrasted the ground illustration, this equalizer output signal 46 is with the clock signal homophase and do not show effect of inter-symbol.This clock signal defines data point (being illustrated as the arrow to D5 corresponding to D0) and boundary point (being illustrated as the arrow to E4 corresponding to E0).Sampler 104 can be sampled to generate data value (that is, D0 is to D5) and at the boundary point place equalizer output signal 46 to be sampled to generate boundary value (that is, E0 is to E4) to equalizer output signal 46 at the data point place.Each data from the sample survey value and boundary value can comprise low value (being illustrated as " L "), high value (being illustrated as " H ") or get the high value randomly or get the random value (being illustrated as " X ") of low value.In a particular embodiment, low value can comprise " 0 ", and high value can comprise " 1 ", and random value can comprise " 0 " randomly or comprise " 1 ", and the mean value of random value can comprise " 0.5 ".In alternative embodiment, low value can comprise " 1 ", and high value can comprise " 1 ", and random value can comprise " 1 " randomly or comprise " 1 ", and the mean value of random value can comprise " 0 ".Sampler 104 can be transmitted to adaptive controller 102 so that suitable processing and adjusting are carried out in gain with the data value and the boundary value of being sampled out.
Will the variation to the high value be called transformation from the high value to low value or from low value between two continuous data values.In illustrated example 200, between low data value D2 and the high data value D3, between high data value D3 and the low data value D4 and between low data value D4 and the high data value D5 transformation has taken place.In the signal that does not show residual effect of inter-symbol (as in example 200), comprising that each boundary value (for example, boundary value E2, E3 and E4) between two continuous data values of inverse value all comprises random value (being illustrated as " X ").For sort signal, because effect of inter-symbol is by full remuneration or do not exist, so the gain that adaptive controller 102 can impose on input signal is regulated randomly up or down.If equate basically with the quantity of regulating downwards to adjusted, the gain that then imposes on input signal is said the level that keeps identical on average.If is not to equate basically to adjusted with the quantity of regulating downwards, the gain meeting that then imposes on input signal is drifted about slightly from initial level.This drift of gain level can produce slight residual intersymbol interference.As following illustrative, the equalizer receiver can detect this interference and average initial level is returned in this gain calibration.
Clock signal example 300 that Fig. 4 B and equalizer output signal 46 have contrasted the ground illustration, this equalizer output signal 46 still shows the residual effect of inter-symbol of undercompensation with the clock signal homophase.In the case, equalizer does not fully compensate signal, so there is the low frequency tendency in this signal.In low frequency tendency signal, if occur data pulse (for example, the data pulse at D3 place) afterwards at the several successive data value (for example, D0 is to D2) that has passed through same high value or low value, then this data pulse highly can reduce owing to lacking high fdrequency component.And, boundary value (for example E2) before this data pulse or boundary value afterwards (for example E3) may with the high value or the low value identical (that is, they will not comprise random value) of data value (for example D2) before this pulse.Therefore, such as described further below, when particular data value and boundary value are analyzed, adaptive controller 102 can increase impose on input signal gain so that low-frequency distortion is compensated.Yet, should be pointed out that in a particular embodiment and as described below that adaptive controller 102 may not compensate the low-frequency distortion that output signal showed before (for example between D2 and the D3) transformation occurring.Boundary value (for example at the E4 place) after a few changes continuously can comprise random value " X ", because this continuous transformation may make the high fdrequency component in the signal increase and low frequency component is reduced, therefore, may reduce the susceptibility of this boundary value to residual intersymbol interference.
Clock signal example 400 that Fig. 4 C and equalizer output signal 46 have contrasted the ground illustration, this equalizer output signal 46 still shows the residual effect of inter-symbol of overcompensation with the clock signal homophase.In the case, equalizer has carried out too many compensation to input signal, so there is the high frequency tendency in this signal.In high frequency tendency signal, if occur data pulse (for example, the data pulse at D3 place) afterwards at the several successive data value (for example, D0 is to D2) that has passed through same high value or low value, the high fdrequency component after then pulse height can be enhanced is raised.And, boundary value (for example E2) before this data pulse or boundary value afterwards (for example E3) may with the high value or the low value opposite (that is, they will not comprise random value) of data value (for example D2) before this data pulse.Therefore, such as described further below, when particular data value and boundary value were analyzed, adaptive controller 102 can reduce to impose on the gain of input signal so that high frequency distortion is compensated.Yet, should be pointed out that in a particular embodiment and as described below that adaptive controller 102 may not compensate the high frequency distortion that output signal showed before (for example between D2 and the D3) transformation occurring.Boundary value (for example at the E4 place) after several successive changes can comprise random value " X ", because this continuous transformation may make the high fdrequency component in the signal increase and low frequency component is reduced, therefore, may reduce the susceptibility of this boundary value to residual intersymbol interference.
Fig. 5 is an illustration, and according to a particular embodiment of the present invention being used for makes an explanation with the flow chart of method 500 that residual intersymbol interference is compensated to output signal value.This method begins at step 510 place, at step 510 place, uses clock signal that output signal is sampled.Such as above described in conjunction with Figure 3, this output signal can be the output of equalizer, and can sample to this output signal according to clock signal.
In a particular embodiment, can sample to output signal at the base value strong point and the boundary point place that determine by clock signal.Select as another kind, can output signal not sampled, and can derive and the corresponding boundary value of these non-sample points at the boundary point place.In a particular embodiment, adaptive controller 102 can be derived boundary value according to the data value of sampling out and other phase informations (that is, the phase place of output signal is morning or late).For example, if the phase place of output signal early, then adaptive controller 102 can determine that the high value of boundary value or low value are identical with the high value or the low value that are right after the data value before this boundary value.If the phase place evening of output signal, then adaptive controller 102 can determine that the high value of boundary value or low value are identical with the high value or the low value that are right after the data value after this boundary value.
At step 520 place, after output signal is sampled, can analyze, to determine in these values, whether having occurred transformation to the data value of being sampled out.For example, can carry out this analysis by adaptive controller 102.At step 530 place, if do not detect transformation, then this method turns back to step 520.If detect transformation between the continuous data value, then this method proceeds to step 540.Should be pointed out that in a particular embodiment, can received data value relatively be detected transformation mutually by direct.In alternative embodiment, can detect transformation by received data value and boundary value and the predefined binarization mode that comprises transformation (and corresponding to specific adaptive control action) are compared.It should also be noted that in a particular embodiment, can after only detecting a transformation, carry out the self adaptation action.
If detect transformation, then the data value of 1.5 positions (or symbol) compares before step 540 place will comprise boundary value between the continuous data value of this transformation and this boundary value.In a particular embodiment, the relation between the data value of 1.5 positions (or symbol) can be determined the adaptive equalizer action response before boundary value and the boundary value.For example, in a particular embodiment, can carry out XOR (XOR) operation (or operating together or (XNOR)) to these two values.In such an embodiment, the possibility of result of xor operation (or XNOR operation) is corresponding to the intersymbol interference of the particular type that output signal showed, therefore, and can be in order to determine the adaptive equalizer action response.In alternative embodiment, can use inverse correlation function (or correlation function) to these two values.In such an embodiment, the possibility of result of inverse correlation function (or correlation function) is corresponding to the intersymbol interference of the particular type that output signal showed, therefore, and can be in order to determine the adaptive equalizer action response.In an alternative embodiment again, can be by comparing data value to compare received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) to 1.5 positions (or symbol) before boundary value and the boundary value.And, in alternative embodiment, can use, and not necessarily use the data value of 1.5 positions (or symbol) before the boundary value apart from the nearer or farther data value of boundary value.Although should be pointed out that with the position to be that some discussion is here explained by unit, if suitablely this discussion alternatively can be construed to the expression symbol.
Should be pointed out that the analysis to above-mentioned (and following) boundary value is an example.Be more typically, based on the sampling to output signal, error amount (certain edges thereof dividing value as already described) can be represented the residual quantity of distortion (the following frequency dependence distortion that will further discuss and/or direct current offset distortion).Based on the error amount that is generated, can regulate the loss balancing (and/or migration) that imposes on data-signal.In a particular embodiment, for example, error amount can comprise pulse width values (wide, narrow or typical), and 2 continuum boundary values that can be from 3 continuous data values that have two transformations and intermediate data value are derived pulse width values.Can use this pulse width values to regulate the loss balancing that is carried out.
At step 550 place, whether the data value of 1.5 positions (or symbol) before boundary value and the boundary value is had identical high value or low value determine.In a particular embodiment, as mentioned above, can these two values be compared with XOR (or XNOR) operation.In alternative embodiment, can use inverse correlation function (or correlation function) to these two values.In an alternative embodiment again, can be by they and predefined pattern be compared these two values of comparison.If these two values have identical high value or low value (for example, if XOR result equals " 0 ", if the inverse correlation function result equals " 1 ", if perhaps these values are corresponding to specific predefined binarization mode), then this method proceeds to step 560.At step 560 place, the gain that equalizer will impose on signal increases.In a particular embodiment, the residual intersymbol interference of undercompensation that shown of the increase meeting compensating signal of gain.As illustrated in Fig. 4 B, this interference is disclosed by " 0 " XOR result.
Should be pointed out that in alternative embodiment adaptive control action can be any suitable adaptive control action that adopts various conventional adaptive control algorithms.For example, adaptive control action can be based on the conventional adaptive control algorithm such as lowest mean square (LMS) algorithm, symbol one symbol lowest mean square (SS-LMS) algorithm, ZF (ZF) algorithm or the like.
If the data value of 1.5 positions (or symbol) (for example has opposite high value or low value before step 550 place determines boundary value and boundary value, if XOR result equals " 1 ", if the inverse correlation function result equals "+1 ", if perhaps these values are corresponding to specific predefined binarization mode), then this method proceeds to step 570.At step 570 place, the gain that equalizer will impose on signal reduces.In a particular embodiment, gain reduce can compensating signal shows the residual intersymbol interference of overcompensation.As illustrated in Fig. 4 C, this interference is disclosed by " 1 " XOR result.
Should be pointed out that in a particular embodiment, can be by adaptive controller 102 execution in step 550,560 and 570, and can use 116 pairs of gains that applied of variable gain amplifier to regulate.And, in a particular embodiment,, then can in a path, regulate and fix the gain in other paths to the gain that is applied if an above signal path (for example to the path in the example equalizer 42 101) is applied gain.In alternative embodiment, can utilize specific function that independent control variable is mapped to a plurality of paths, and can gain to these paths according to this mapping process.Select as another kind, as further discussing, can regulate gain independently at each path below in conjunction with Fig. 7 to 10.
The table 600 of the exemplary gain controlling schemes that Fig. 6 is an illustration is associated with the method 500 of Fig. 5.Each row 602 is carried out specific adaptive equalizer control action all corresponding to the particular value pattern at this particular value pattern.Row 610 comprise the series data of being sampled out and each high value or the low value (although it can be " 1 " or " 0 ", being any other suitable value perhaps, is "+1 " or " 1 ") in the boundary value in this specific example in other examples.Row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " D3 " comprise the 3rd data from the sample survey value of output signal, and row " E2 " comprise the boundary value between second data value and the 3rd data value.These values are similar to Fig. 4 A illustrative those values in the 4C.As can be observed, between the data value of row " D2 " and " D3 " transformation have appearred in each pattern.
Should be pointed out that can by sampler 104 sample out the value in each row 602 pattern and send it to adaptive controller 102.In a particular embodiment, adaptive controller 102 can receive the value than illustrated more quantity, for example, comprises the boundary value between the data value among row " D1 " and " D2 ".Select as another kind, as discussed above such, adaptive controller 102 can receive only data from the sample survey and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E2) (and can can't help sampler 104 sample out) according to these data values and phase information.
Row 612 comprise the ISI degree.The ISI degree is to derive out from the particular value that is associated with output signal.For example, the ISI degree can be the data value of 1.5 positions before the boundary value between two data values that comprise transformation and this boundary value to be used the result of inverse correlation function.In a particular embodiment, can utilize the long-pending anti-value that the ISI degree is calculated as boundary value and data value with " high/low " value corresponding "+1/-1 " value.In table 600, the ISI degree in the row 612 is to use the result of inverse correlation function with the data value (1.5 positions before the boundary value) among boundary value among the row E2 in the delegation 602 and the row D1.Can utilize the long-pending anti-value that the ISI degree is calculated as E2 and D1 with " high/low " value corresponding "+1/-1 " value.As illustrative in row 614 and 616, the ISI degree of " 1 " and undercompensation balanced intensity and the increase of equalizer compensation is associated.The ISI degree of "+1 " and overcompensation balanced intensity and to the equalizer compensation reduce be associated.Therefore, based on the ISI degree, carry out specific adaptive equalizer action.In alternative embodiment, received data and boundary value and predefined binarization mode can be compared, and these predefined binarization modes can be corresponding to specific adaptive control action.
Should be pointed out that in a particular embodiment, adaptive controller 102 can the acceptance sampling value stream and from these values, select suitable value (data value that for example, comprises 1.5 positions before boundary value between two data values of transformation and this boundary value).Adaptive controller 102 then can for example be used the inverse correlation function and derive the ISI degree according to the value that these are selected by the value that these are chosen.Adaptive controller 102 then can be based on the result of this inverse correlation function and is carried out suitable adaptive control action.Select as another kind, adaptive controller 102 can compare these sample values and predefined binarization mode (it is corresponding to specific adaptive control action).Based on the pairing specific predefined binarization mode of these sample values, adaptive controller 102 can carry out corresponding adaptive control action.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated or discrete to the parts of described system and method according to concrete needs.In addition, can by more, still less or miscellaneous part carry out the operation of described system and method.
In a particular embodiment, equalizer (as the equalizer 42 of Fig. 3) can be controlled an above independent parameter (as unaltered component, first order component and two order components of signal).The example of multi-parameter (multidimensional) equalizer comprises second dervative equalizer and 3 tap finite impulse response (FIR) (FIR) filters.As discussed above like that, if for example the gain of other parameters is regulated and is fixed in the gain that is applied at a parameter, then can be in the multidimensional equalizer using method 500.Select as another kind, if (but not regulating gain independently at each independent parameter) regulated in the gain that is applied according to the specific function that has merged a plurality of independent parameters, then can be in the multidimensional equalizer using method 500.In this alternative example, can regulate the gain that is applied independently at each parameter of independently controlling.Such as described further below, in 3 tap FIR filters, for example, can regulate the second and the 3rd tap coefficient independently, and these each in regulating can comprise the adjusting at the compensation of distortion.
Based on one or more data from the sample survey value and boundary value (it is comprising between the continuous data value of transformation) particular kind of relationship between the data from the sample survey value of 1.5 positions before, can be at each independent Control Parameter independently to regulating such as the compensation of gain.Particular kind of relationship for example can be corresponding to the intersymbol interference at the particular type of specific independent Control Parameter.When for example adaptive controller 102 (for example utilizing the predefined data binarization mode) detected this relation between a plurality of data from the sample survey values, adaptive controller 102 can carry out specific adaptive equalizer control action so that one or more specific independent Control Parameter is regulated.
Can be especially responsive by the predefined data binarization mode that adaptive controller 102 is used for comparing with the stream of the arrival of data from the sample survey value to intersymbol interference at specific independent Control Parameter.For example, based on the susceptibility of the boundary value between the data value that comprises transformation, can select these patterns to the independent Control Parameter that is being conditioned.Specifically, can be based on selecting these patterns with respect to the partial derivative of this independent Control Parameter symbol and the value of this partial derivative (for example, based on) through the channel impulse response of equilibrium.This is because equalizer output signal 46 is represented as the convolution that sends data sequence and equalization channel impulse response.
For example, in simulation, in the second dervative equalizer, can locate to bear with being assumed to be peak value after 1.5 and 2.5 positions with respect to the partial derivative of first derivative gain through the channel impulse response of equilibrium.Therefore, if the first derivative gain is too high, then the correlation between the data value of 1.5 and 2.5 positions may all be born before boundary value and this boundary value.On the other hand, if the first derivative gain is too low, then the correlation between the data value of 1.5 and 2.5 positions may all be positive before boundary value and this boundary value.This be because these data corresponding to the peak value of impulse response and this border corresponding to the afterbody after the peak value in this impulse response.Can with equalization channel impulse response with respect to the partial derivative of second dervative gain be assumed to be peak value after place, 1.5 positions be bear and to locate be positive in 2.5 positions after peak value.Therefore, if second dervative gain is too high, then the correlation between the data value of 1.5 positions may be born before boundary value and this boundary value, and boundary value and this boundary value before the correlation between the data value of 2.5 positions may be positive.On the other hand, if second dervative gain is too low, then the correlation between the data value of 1.5 positions may be positive before boundary value and this boundary value, and boundary value and this boundary value before the correlation between the data value of 2.5 positions may bear.Utilize these relations, can adopt various technology (for example, following method 700) that the gain of the first derivative component that imposes on input signal and the gain that imposes on the second dervative component of input signal are regulated.
As another example, main tapping is in 3 tap finite impulse response (FIR) (FIR) filter equalizers of first tap therein, can place, 1.5 and 2.5 positions just is after peak value with being assumed to be with respect to the partial derivative of second tap coefficient through the channel impulse response of equilibrium.Can be zero and to locate be positive in 2.5 positions peak value after with being assumed to be after peak value place, 1.5 positions with respect to the partial derivative of the 3rd tap coefficient through the channel impulse response of equilibrium.Utilize these relations, can adopt various technology (for example, following method 1000) that second tap coefficient and the 3rd tap coefficient are regulated.By this mode, the FIR filter equalizer can carry out the equilibrium of many taps FIR filter concurrently.Should be pointed out that and to utilize different relations at dissimilar equalizers.For example, can utilize different relations at many taps DFF of carrying out many taps decision feedback equalization concurrently.
Fig. 7 is the illustration flow chart that is used for exemplary method 700 that simulation second dervative equalizer is made an explanation at the output signal value of a plurality of independent Control Parameter according to a particular embodiment of the present invention.For simulation second dervative equalizer, can control 3 independent Control Parameter (for example comprising) to the gain that does not change gain that part applies, the part of the first derivative that is changed into input signal of input signal is applied of input signal and gain that the part of the second dervative that is changed into input signal of input signal is applied.For example, can be by simulation first derivative filter continuous time the balanced and/or simulation second dervative filter equilibrium continuous time compensation that walks abreast.
Method 700 begins at step 710 place.Step 710 to 770 can be identical with the step 510 to 570 in the said method 500, therefore no longer they is described in detail.Yet, should be pointed out that in step 710 to 770, can control first independent parameter at first path.For example, can utilize step 710 to 770 pair to impose on the gain that not change part of input signal in first path (as the path 101A of equalizer 42) regulates.In a particular embodiment, can the gain that part applies of not changing in first path increase equalizer compensation in first path to input signal by reducing, and can be by increasing input signal not changing gain that part applies and reduce equalizer compensation in first path in first path.This is because the equalizer compensation rate can depend on the relative gain of second path and Third Road footpath to first path, and therefore, the gain that increases first path will effectively reduce the relative gain to first path of second path and Third Road footpath.In step 780 to 850, can control second independent parameter and the 3rd independent parameter respectively at second path and Third Road footpath.For example, the gain that can utilize step 780 to 850 pair to impose on the part of the first derivative that be changed into input signal of input signal in second path (as the path 101B of equalizer 42) is regulated, and the gain of the part that imposes on the second dervative that be changed into input signal of input signal in Third Road footpath (as the path 101C of equalizer 42) is regulated.
In step 780 and 790, if comprise that the data value of 1.5 positions has different value (high value or low value) before boundary value and this boundary value between the data value of transformation, then whether the value of the data value of 2.5 positions is identical or determine on the contrary before to the value (high value or low value) of the data value of 1.5 positions before this boundary value and this boundary value.For example, can be by carrying out suitable computing or being undertaken by received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) are compared that this is definite.If these two values different (on the contrary), then method 700 proceeds to step 800, and reduces gain that input signal is applied at the second dervative component of Third Road in directly.If these two values are identical, then method 700 proceeds to step 810, and reduces gain that the first derivative component of input signal in second path applied.
At step 820 and 830 places, if comprise that the data value of 1.5 positions has identical value (high value or low value) before boundary value and this boundary value between the data value of transformation, then whether the value of the data value of 2.5 positions is identical or determine on the contrary before to the value (high value or low value) of the data value of 1.5 positions before this boundary value and this boundary value.Equally, can be for example by carrying out suitable computing or being undertaken by received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) are compared that this is definite.If these two values different (on the contrary), then method 700 proceeds to step 840, and increases the gain that the second dervative component of input signal in the Third Road footpath applied.If these two values are identical, then method 700 proceeds to step 850, and increases the gain that the first derivative component in second path is applied.
Should be pointed out that as above the quantity of independent Control Parameter can be identical with the quantity that is conditioned parameter in a particular embodiment as described in the method 700.In alternative embodiment, the quantity of independent Control Parameter can be lacked than the quantity that is conditioned parameter.For example, in simulation second dervative equalizer, can there be 2 independent Control Parameter (promptly, first gain that the first derivative component of input signal is applied and second gain that the second dervative component of input signal is applied), and can exist 3 to be conditioned parameter the 3rd gain (fixed gain) that component applies that do not change of input signal (that is, to).As another example, first independent control variable can be controlled the relation that imposes on the gain that does not change component and impose between the gain of first derivative component, and second control variables can be to imposing on the second dervative component and imposing on the gain that does not change component and the relation that imposes between the greater in the gain of first derivative component is controlled.If suitable, can amending method 700 to satisfy these different situations.
The table 900 of the exemplary gain controlling schemes that Fig. 8 is an illustration is associated with the method 700 of Fig. 7.Each row 902 is carried out specific adaptive equalizer control action all corresponding to the particular value pattern at this particular value pattern.Row 910 comprise the pattern of being made up of data from the sample survey and boundary value, and its intermediate value can have height (" 1 ") value or low (" 0 ") value.Row " D0 " comprise the 0th data from the sample survey value of output signal, row " D1 " comprise the first data from the sample survey value of output signal, row " D2 " comprise the second data from the sample survey value of output signal, row " D3 " comprise the 3rd data from the sample survey value of output signal, and row " E2 " comprise the boundary value between second data value and the 3rd data value.These values are similar to Fig. 4 A illustrative those values in the 4C.As can be observed, between the data value of row " D2 " and " D3 " transformation have appearred in each pattern.
Should be pointed out that can by sampler 104 sample out in each row 902 value and send it to adaptive controller 102.Adaptive controller 102 can compare value and one or more predetermined binarization mode of being sampled out.In a particular embodiment, when detecting coupling, adaptive controller 102 can be taked the associated group of one or more adaptive equalizer action.In such an embodiment, may (for example because using predetermined binarization mode and) know the particular kind of relationship between these values, so above-mentioned one or more step (for example, step 780,790,820 and 830) in needn't manner of execution 700.
What should also be pointed out that is, in a particular embodiment, adaptive controller 102 can receive the more value than illustrated value, for example comprises boundary value between the data value among row " D0 " and " D1 " and the boundary value between the data value among row " D1 " and " D2 ".Select as another kind, as discussed above, adaptive controller 102 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E2) (thereby can obstructed oversampling device 104 sample out) according to these data values and other phase informations.
Row 920 comprise the alternative boundary value that each row 902 is located at row " E2 ".Row 924 comprise at AD HOC and specific degree of compensation and the adaptive equalizer action that component is associated that do not change input signal.Can move as the above adaptive equalizer that discussion in the method 700, applies.Row 930 comprise specific degree of compensation and the adaptive equalizer action that is associated at first derivative component AD HOC and input signal.Also can move as above these adaptive equalizers that discussion in the method 700, apply.Row 940 comprise specific degree of compensation and the adaptive equalizer action that is associated at second dervative component AD HOC and input signal.Also can move as above these adaptive equalizers that discussion in the method 700, apply.
The flow chart of Fig. 9 is the illustration exemplary method 1000 that the output signal value that is used for a plurality of parametric equalizers of 3 tap FIR filters according to a particular embodiment of the present invention makes an explanation.Method 1000 begins at step 1010 place.Step 1010 to 1030 can be similar with the step 510 to 530 in the said method 500, therefore no longer they is described in detail.In a particular embodiment, first tap coefficient can be fixed, and does not regulate it by adaptive control.Step 1080 to 1170 can be regulated second tap coefficient and the 3rd tap coefficient.Should be pointed out that in 3 tap FIR filters, such as described further below, can regulate second tap coefficient and the 3rd tap coefficient independently, and these each in regulating can comprise the adjusting at the compensation of distortion.Usually, no matter under 3 tap FIR filter equalizers, simulation derivative filter equalizer still are the environment of any other suitable equalizer, even the part of described compensation is only carried out in each path and just occur described compensation in the time will combining from the output in all paths in assembly, also the action that each path is carried out can be called and use or regulate having carried out at the compensation of distortion.
In step 1080 and 1090, whether the value of the data value of 2.5 positions is identical or determine on the contrary before to the value of the data value of 1.5 positions before the boundary value that comprises between the data value of transformation (high value or low value) and this boundary value.For example, can be by carrying out suitable computing or being undertaken by received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) are compared that this is definite.If these two values different (on the contrary), then method 1000 proceeds to step 1100.If these two values are identical, then method 1000 proceeds to step 1140.
At step 1100 and 1110 places, whether identical or determine on the contrary to the value of the value of the data value of 2.5 positions before the boundary value that comprises between the data value of transformation (high value or low value) and this boundary value.For example, can be by carrying out suitable computing or being undertaken by received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) are compared that this is definite.If these two values different (on the contrary), then method 1000 proceeds to step 1120, and increases the 3rd tap coefficient.If these two values are identical, then method 1000 proceeds to step 1130, and reduces the 3rd tap coefficient.
At step 1140 and 1150 places, whether identical or determine on the contrary to the value of the value of the data value of 2.5 positions before the boundary value that comprises between the data value of this transformation (high value or low value) and this boundary value.Equally, can be by carrying out suitable computing or being undertaken by received data value and boundary value and predefined binarization mode (it is corresponding to specific adaptive control action) are compared that this is definite.If these two values different (on the contrary), then method 1000 proceeds to step 1160, and increases the second and the 3rd tap coefficient.If these two values are identical, then method 1000 proceeds to step 1170, and reduces second tap coefficient and the 3rd tap coefficient.
The table 1200 of the exemplary gain controlling schemes that Figure 10 is an illustration is associated with the method for Fig. 9.Each row 1202 is carried out specific adaptive equalizer control action all corresponding to the particular value pattern at this particular value pattern.Row 1210 comprise the pattern by data from the sample survey and boundary value, and its intermediate value can be high (" 1 ") value or low (" 0 ") value.Row " D0 " comprise the 0th data from the sample survey value of output signal, row " D1 " comprise the first data from the sample survey value of output signal, row " D2 " comprise the second data from the sample survey value of output signal, row " D3 " comprise the 3rd data from the sample survey value of output signal, and row " E2 " comprise the boundary value between second data value and the 3rd data value.These values are similar to Fig. 4 A illustrative those values in the 4C.As can be observed, in each pattern, between the data value of row " D2 " and " D3 " transformation have appearred.
Should be pointed out that sampler 104 can sample out the value in each row 1202 pattern and send it to adaptive controller 102.Adaptive controller 102 can compare value and one or more predetermined value pattern of being sampled out.In a particular embodiment, when detecting coupling, adaptive controller 102 can be taked the associated group of one or more adaptive equalizer action.In such an embodiment, may (for example because using predetermined binarization mode and) know the particular kind of relationship between these values, so above-mentioned one or more step (for example, step 1080) in needn't manner of execution 1000.
What should also be pointed out that is, in a particular embodiment, adaptive controller 102 can receive the more value than illustrated value, for example comprises boundary value between the data value among row " D0 " and " D1 " and the boundary value between the data value among row " D1 " and " D2 ".Select as another kind, as discussed above such, adaptive controller 102 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E2) (thereby can obstructed oversampling device 104 sample out) according to these data values and other phase informations.
Row 1220 comprise the alternative boundary value that each row 1202 is located at row " E2 ".Row 1230 comprise at the particular factor level that is associated with second tap coefficient of AD HOC and adaptive equalizer action.Also can move as above these adaptive equalizers that discussion in the method 1000, apply.Row 1240 comprise at the particular factor level that is associated with the 3rd tap coefficient of AD HOC and adaptive equalizer action.Also can move as above these adaptive equalizers that discussion in the method 1000, apply.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
As above discuss in conjunction with Fig. 5, the relation between the data value of 1.5 positions before boundary value and this boundary value of comprising between the continuous data value of transformation can be relevant with the specific balanced intensity of signal.This correlation can be especially accurate for the signal with abundant randomized data sequence.Yet if signal has periodically or the quasi periodicity data sequence, this correlation can be subjected to periodically influencing of sequence.Specifically, if there are Duty Cycle Distortion in the signal or the clock signal that arrive, this correlation even can be subjected to periodically more seriously influencing of sequence then.
Usually, there are strong correlation in periodicity or quasi periodicity data sequence between a plurality of data values such as the adjacent data value, thereby can influence the frequency spectrum of signal.For example, if the adjacent data value more may be identical value rather than different values, then this signal is a low frequency tendency, and if the adjacent data value more may be different values rather than identical value, then this signal is the high frequency tendency.What this distortion meeting of signal spectrum influenced adaptive equalizer makes the signal spectrum smooth ability that flattens.Usually, even there is not Duty Cycle Distortion, also there are this negative effect to the adaptive gain control of equalizer in periodicity or quasi periodicity data sequence.
This negative effect that Duty Cycle Distortion can strengthen periodically and the quasi periodicity sequence is controlled the adaptive gain of equalizer.For example, suppose and the data that arrive can be labeled as even data and odd data successively.Border between the data can also be labeled as successively even border and strange border.Here, even border can be illustrated in after the even data and the border before odd data, and strange border can be illustrated in after the odd data and the border before even data.Duty Cycle Distortion can cause the even boundary value of receiver logic and strange boundary value seriously to depart to " early " phase place (that is, with the identical phase place of last data value) or " evening " phase place (that is, phase place) identical with next data value.For example, even boundary value can be to " early " phase deviation, and strange boundary value can be to " evening " phase deviation.
If the cycle of periodicity or quasi periodicity data sequence is the multiple of two data values, the transformation quantity of so even boundary also can depart from the transformation quantity of strange boundary.For example, periodically or in the quasi periodicity data sequence can occur more continually changing at strange boundary at even boundary ratio.Equalizer control can be subjected to periodically or in the quasi periodicity data sequence how the boundary value that changes reigning even border or strange boundary being departed from (promptly, be " early " phase place or " evening " phase place) influence, this depart from be since Duty Cycle Distortion caused.Lock onto in clock recovered before the data sequence of arrival of the periodicity of this arrival or quasi periodicity, this influence can not determine because the boundary value that has " early or evening " phase place owing to Duty Cycle Distortion depart from because the correspondence that periodicity or quasi periodicity data sequence had between the departing from of border of " mastery or non-mastery " transformation may depend on how the recovered clock that has Duty Cycle Distortion locks onto the periodicity or the quasi periodicity data sequence of this arrival.
Since this departing from (" early " phase place or " evening " phase place) of boundary value, reigning to changing " idol " or " very " boundary in periodicity or quasi periodicity data sequence, and the equalizer control action also can exist and depart from.Depart to the imbalance of particular equalizer control action and can make equalizer control produce unacceptable result.
Figure 11 illustration the example boundary information 1300 that influenced by Duty Cycle Distortion.Exist four phase half rate clocks of Duty Cycle Distortion that this signal 1310 is sampled in receiver logic place received signal 1310 and utilization.The even data value of can sampling in the rising edge of clock A (CLKA) 1320, the odd data value of can sampling in the rising edge of clock C (CLKC) 1340, can sample in the rising edge of clock B (CLKB) the 1330 even boundary value and the strange boundary value of can sampling in the rising edge of clock D (CLKD) 1350.In this example, the duty ratio of clock B 1330 surpasses 50%, and the duty ratio of clock D 1350 is less than 50%.As a result, the even boundary value of sampling out in the rising edge of clock B 1330 is seriously to " early " phase deviation, and the strange boundary value of sampling out in the rising edge of clock D1350 is seriously to " evening " phase deviation.If get the average of " early " counting and " evening " counting, then the clock recovery circulation can be locked in this phase position.If the existence between even border and strange border of periodicity that arrives or quasi-cycling signal departs from transformation, then to changing departing from phase place adaptive control action being departed from of reigning boundary.For example, if having than the strange border of sampling out in the rising edge of clock D (CLKD) 1350, the even border of sampling out in the rising edge of clock B (CLKB) 1330 more changes, then can be to " early " phase deviation at even boundary adaptive control action.Do not depart from if do not consider these in the adaptive gain control unit of equalizer, then the result of the self adaptation of equalizer operation understands influenced.
Consider that the mode that departs from the contrary in even boundary value and the strange boundary value is to make these depart from balance.Can just apply adaptive equalizer when being distributed in two specific data pattern (having probability of occurrence about equally) in the phase place about equally and move and realize this balance by only detecting at controller.The weighing apparatus that flattens that departs from that can use these specific data pattern (can be referred to as filter mode) to make in (standard) periodic data sequence, to produce, these specific data pattern also can be used with abundant randomization data sequence, and do not have what problem.By this mode, can make the characteristic of adaptive equalizer control consistent at (standard) periodic data sequence and abundant randomization data sequence.If there is independent Control Parameter more than, then adaptive equalizer control can use the specific filter pattern to control the gain that imposes on each specific independent Control Parameter.In a particular embodiment, discuss like that in conjunction with Fig. 7 to 10 as above, can be from the modal sets that occurs about equally basis select the filter mode controlled in order to the gain that imposes on this parameter at the partial differential gradient through the channel impulse response of equilibrium of specific independent Control Parameter.
Figure 12 is the illustration flow chart that is used for the selective filter pattern with the exemplary method 1400 of the negative effect that reduces Duty Cycle Distortion according to a particular embodiment of the present invention.Can using method 1400 select with periodically or the filter mode used of quasi periodicity data sequence (as 8B10B idle sequence and 8B10B CJPAT cycle tests) by Institute for Electrical and Electronics Engineers 802.3ae standard definition.
Method 1400 begins at step 1410 place, and at step 1410 place, antithesis data sequence and odd data sequence set are monitored.The even data sequence set comprises from idol (that is, even phase place place) data and begins, then is the data sequence of odd data, even data, odd data or the like.The odd data sequence set comprises from strange (that is, strange phase place place) data and begins, then is the data sequence of even data, odd data, even data or the like.At step 1420 place, determine the distribution of the data pattern in this even data sequence and odd data sequence.In a particular embodiment, this data pattern can comprise 6 positions.In alternative embodiment, this data pattern can comprise 5 positions.In an alternative embodiment again, this data pattern can comprise the figure place that any other is suitable.At step 1430 and 1440 places, only all observed those patterns in even data sequence and odd data sequence are further analyzed.Do not select not in two sequences all observed those patterns as filter mode at step 1500 place.
At step 1450 and 1460 places, for all observed those patterns in the data pattern of even data sequence and odd data sequence distributes, to any pattern in these patterns whether equally (or about equally) be distributed in even data sequence and the odd data sequence and determine.As discussed above, the data pattern of selecting to distribute about equally can be offset owing to departing from that Duty Cycle Distortion produces as filter mode.Those patterns that (or about equally) distributes for equally, method 1400 proceeds to step 1470.Any pattern of not selecting at step 1500 place not to be to distribute about equally is as filter mode.
At step 1470 and 1480 places, in the end whether exist transformation to determine between two positions to any pattern in all the other patterns.As discussed above, utilize before boundary value and this boundary value between the continuous data value comprise transformation the relation between one or more data value to determine that the adaptive equalizer that will apply moves.Therefore, specific embodiment can only be chosen in and have those data patterns that change between latter two.In such an embodiment, at step 1500 place thereby be not chosen in and do not exist any pattern of transformation as filter mode between latter two.
At step 1490 place, be chosen in all observed in two sequences, (roughly) equally be distributed in two sequences and in the end have those patterns that change between two positions, and they are chosen as filter mode.Yet, should be pointed out that select these patterns as filter mode before in addition can further analyze these patterns.For example, such as described further below, if specific (a plurality of) pattern in these patterns is more suitable for being used to control specific independent Control Parameter, then can select this specific (a plurality of) pattern to control this specific independent Control Parameter.Such as discussed here, then can use these filter mode to carry out the equalizer regulating and controlling.
The table of the example distribution 1600 of six bit data patterns in even 8B10B idle data sequence that Figure 13 has been an illustration and the strange 8B10B idle data sequence.Row 1610 comprise a plurality of six bit data patterns, and row 1620 are included in the probability of observing AD HOC in the even 8B10B idle data sequence, and row 1630 are included in the probability of observing AD HOC in the strange 8B10B idle data sequence.Here, the even data sequence is meant with even data and begins, then is the data sequence of odd data, even data, odd data or the like.The odd data sequence is meant with odd data and begins, then is the data sequence of even data, odd data, even data or the like.For each data pattern in the row 1610, position early is on the left side of later position.It is zero that blank cell in row 1620 or the row 1630 is illustrated in the probability of observing the data pattern that is associated in the data sequence that is associated.
As illustrative in the distribution 1600, in even 8B10B idle data sequence and strange 8B10B idle data sequence, all observe 4 data sequences: 000010,111010,000101 and 111101.And these 4 patterns equally are distributed in two sequences, have time of 4.796% to observe each pattern.In addition, in the end there is transformation in these 4 patterns between two positions.Therefore, in a particular embodiment, can utilize method 1400 to select these data sequences, and can only when observing these filter mode, apply adaptive control action as filter mode.Utilize these filter mode, can reduce the negative effect of Duty Cycle Distortion.In addition, can be consistent at the control behavior of 8B10B idle data sequence with control behavior at abundant randomization data sequence.
Should be pointed out that in a particular embodiment the single order that (for example, in the second dervative equalizer) can utilize a plurality of independent Control Parameter to regulate to impose on input signal and the gain of second dervative component.In this case, when receiving 8B10B idle data sequence, for these independent Control Parameter, the specific filter pattern in the viewed and filter mode that equally distribute may be more suitable for.For example, the good filter mode that is fit to can comprise such filter mode: it has the boundary value between the continuous data value that comprises transformation, and this boundary value is responsive relatively to the gain that imposes on single order or second dervative component.Therefore, in this example, the good filter mode that is fit to comprises the single order that makes adaptive controller make signal effectively or those filter mode of second dervative component equilibrium.
For 8B10B idle data sequence, the good filter mode that is fit to can comprise and be used for the first derivative component is applied 000010 and 111101 of gain.Therefore in a particular embodiment, can be only when the data value observed at the equalizer place corresponding to these filter mode ability the first derivative component is applied adaptive control action.By this mode, can make the first derivative component equilibrium of signal effectively.For (under the environment of 8B10B idle data sequence) applies gain to the second dervative component, the good filter mode that is fit to can comprise 000101 and 111010.Therefore in a particular embodiment, can be only when the data value observed at the equalizer place corresponding to these filter mode ability the second dervative component is applied adaptive control action.By this mode, can make the second dervative component equilibrium of signal effectively.
The table 1700 of the exemplary gain controlling schemes that the gain that does not change component, first derivative component and second dervative component that Figure 14 is an illustration with use the example filter pattern of deriving from the table of Figure 13 to regulate imposes on input signal is associated.Input signal for example can be 8B10B idle data sequence, another (standard) cyclical signal (wherein the filter mode of Figure 13 distributes about equally) or good randomized signal.Each row 1702 is carried out specific adaptive equalizer control action all corresponding to the particular value pattern at this particular value pattern.Each data pattern in these row 1702 is all corresponding to one in the above filter mode of describing in conjunction with Figure 13 of selecting.
Row 1710 comprise the pattern of being made up of data from the sample survey and boundary value, and its intermediate value can be high (" 1 ") value or low (" 0 ") value.Row " D0 " comprise the 0th data from the sample survey value of output signal, row " D1 " comprise the first data from the sample survey value of output signal, row " D2 " comprise the second data from the sample survey value of output signal, row " D3 " comprise the 3rd data from the sample survey value of output signal, row " D4 " comprise the 4th data from the sample survey value of output signal, row " D5 " comprise the 5th data from the sample survey value of output signal, and row " E4 " comprise the boundary value between the 4th data value and the 5th data value.As can be observed, in each pattern, between the data value of row " D4 " and " D5 " transformation have appearred.
Should be pointed out that can by sampler 104 sample out the value in each row 1702 pattern and send it to adaptive controller 102.Adaptive controller 102 can compare value and one or more predetermined filters pattern of being sampled out.In a particular embodiment, when detecting coupling, as described herein, one group of one or more adaptive equalizer that adaptive controller 102 can be taked to be associated moves.
It should also be noted that in a particular embodiment adaptive controller 102 can receive the more value than illustrated value, for example comprise that row " D0 " arrive the boundary value between the data value in " D4 ".Select as another kind, as discussed above such, adaptive controller 102 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value the row E4) (thereby can obstructed oversampling device 104 sample out) from these data values and other phase informations.
Row 1720 comprise the alternative boundary value that each row 1702 is located at row " E4 ".For AD HOC, row 1724 comprise specific degree of compensation and the adaptive equalizer action that component is associated that do not change with input signal.Can (for example comprise) in any suitable way that action is controlled to adaptive equalizer by particular value is analyzed.For AD HOC, row 1730 comprise specific degree of compensation and the adaptive equalizer action that is associated with the first derivative component of input signal.Also can (for example comprise) in any suitable way these adaptive equalizer actions are controlled by particular value is analyzed.For AD HOC, row 1740 comprise specific degree of compensation and the adaptive equalizer action that is associated with the second dervative component of input signal.Also can (for example comprise) in any suitable way these adaptive equalizer actions are controlled by particular value is analyzed.
The table of the example distribution 1800 of six bit data patterns that Figure 15 has been an illustration in even 8B10B CJPAT data sequence and strange 8B10B CJPAT data sequence.Row 1810 comprise a plurality of six bit data patterns, and row 1820 are included in the probability of observing AD HOC in the even 8B10B CJPAT data sequence, and row 1830 are included in the probability of observing AD HOC in the strange 8B10B CJPAT data sequence.Here, the even data sequence is meant with even data and begins, then is the data sequence of odd data, even data, odd data or the like.The odd data sequence is meant with odd data and begins, then is the data sequence of even data, odd data, even data or the like.For each data pattern in the row 1810, position early is on the left side of later position.It is zero that blank cell in row 1820 or the row 1830 is illustrated in the probability of observing the data pattern that is associated in the data sequence that is associated.Should be pointed out that distribution 1800 is distributions of the data pattern in " class CJPAT type " data sequence.Except the lane to lane (lane-to-lane) difference (as startup, introduction, CRC and IPG sequence), class CJPAT type data sequence is identical with the 8B10B CJPAT data sequence of IEEE 802.3ae standard.The overall characteristic of class CJPAT type data sequence is similar to actual 8B10B CJPAT data sequence comparatively speaking.
As illustrative in the distribution 1800, in even 8B10B CJPAT cycle tests and strange 8B 10B CJPAT cycle tests, all observe pattern 001110 and 110001.In addition, these 2 patterns also are distributed in two sequences about equally.The probability of observing pattern 001110 in even sequence is 7.340%, is 7.394% in strange pattern.The probability of observing pattern 110001 in even sequence is 7.394%, is 7.394% in strange pattern.In addition, in the end there is transformation in these two patterns between two positions.Therefore, in a particular embodiment, can utilize method 1400 to select these data sequences, and can only when observing these filter mode, apply adaptive control action as filter mode.Utilize these filter mode, can reduce the negative effect of Duty Cycle Distortion.In addition, can be consistent at the control behavior of 8B10B CJPAT data sequence with control behavior at abundant randomization data sequence.
Should be noted that, in a particular embodiment, the equalizer that receives 8B10B idle data sequence, 8B10BCJPAT data sequence and random sequence can use simultaneously at 8B10B idle data sequence with at the selected filter mode of 8B10B CJPAT data sequence.As can observed in the distribution 1800 of 8B10B CJPAT data sequence, the idle filter mode of 8B10B in even sequence and odd sequence, disproportionately occurring: 000010,111101,000101 and 111010.Yet, owing in the process that receives 8B10B CJPAT data sequence, just observe these uneven filter mode low relatively probability, therefore in the process that receives 8B10B CJPAT data sequence, use them can not make adaptive control produce bad result usually.As can 8B10B idle data sequence distribute in 1600 observed, use to 8B10B CJPAT filter mode 001110 and 110001 during reception 8B10B idle data sequence can not make adaptive control produce bad result, because never observe these filter mode in 8B10B idle data sequence.
It should also be noted that the single order that (for example, in the second dervative equalizer) can utilize a plurality of independent Control Parameter to regulate to impose on input signal and the gain of second dervative component.In this case, when receiving 8B10B CJPAT data sequence, for these independent Control Parameter, the specific filter pattern in the viewed and filter mode that equally distribute may be more suitable for.
For 8B10B CJPAT data sequence, the good filter mode that is fit to that is used for the first derivative component is applied gain can comprise 110001 and 001110.Therefore in a particular embodiment, can be only when the data value observed at the equalizer place corresponding to these filter mode ability the first derivative component is applied adaptive control action.By this mode, can make the first derivative component equilibrium of signal effectively.In the aforesaid embodiment that also uses the idle filter mode of 8B10B, the good filter mode that is fit to that is used for the first derivative component is applied gain can also comprise 000010 and 111101.
For (under the environment of 8B10B CJPAT data sequence) applies gain to the second dervative component, the good filter mode that is fit to can comprise 000101 and 111010.Should be pointed out that these filter mode are identical with those filter mode that are used for the second dervative component in the 8B10B idle data sequence is applied gain by equalizer.It should also be noted that in even data sequence in 8B10B CJPAT data sequence and the odd data sequence and this two filter mode disproportionately occur.Yet,, therefore during receiving 8B10B CJPAT data sequence, use them can not make adaptive control produce bad result owing to just observe these uneven filter mode low relatively probability.
The table 1900 of the exemplary gain controlling schemes that the gain that does not change component, first derivative component and second dervative component that Figure 16 is an illustration with use the example filter pattern of deriving from Figure 15 to regulate imposes on input signal is associated.Input signal for example can be 8B10B CJPAT data-signal, another (standard) cyclical signal (wherein the filter mode of Figure 15 distributes about equally) or good randomized signal.Each row 1902 is carried out specific adaptive equalizer control action all corresponding to the particular value pattern at this particular value pattern.Each data pattern in these row 1902 is all corresponding to one in the above filter mode of describing in conjunction with Figure 15 of selecting.
Row 1910 comprise the pattern about data from the sample survey and boundary value, and its intermediate value can be high (" 1 ") value or low (" 0 ") value.Row " D0 " comprise the 0th data from the sample survey value of output signal, row " D1 " comprise the first data from the sample survey value of output signal, row " D2 " comprise the second data from the sample survey value of output signal, row " D3 " comprise the 3rd data from the sample survey value of output signal, row " D4 " comprise the 4th data from the sample survey value of output signal, row " D5 " comprise the 5th data from the sample survey value of output signal, and row " E4 " comprise the boundary value between the 4th data value and the 5th data value.As can be observed, in each pattern, between the data value of row " D4 " and " D5 " transformation have appearred.
Should be pointed out that and to sample and send it to adaptive controller 102 by the pattern of the value in 104 pairs of each row 1902 of sampler.Adaptive controller 102 can compare value and one or more predetermined filters pattern of being sampled out.In a particular embodiment, when detecting coupling, as described herein, one group of one or more adaptive equalizer that adaptive controller 102 can be taked to be associated moves.
It should also be noted that in a particular embodiment adaptive controller 102 can receive the more value than illustrated value, for example comprise that row " D0 " arrive the boundary value between the data value in " D4 ".Select as another kind, as discussed above such, adaptive controller 102 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value the row E4) (thereby can obstructed oversampling device 104 sample out) from these data values and other phase informations.
Row 1920 comprise the alternative boundary value that each row 1902 is located at row " E4 ".For AD HOC, row 1924 comprise specific degree of compensation and the adaptive equalizer action that component is associated that do not change with input signal.Can (for example comprise) in any suitable way that action is controlled to adaptive equalizer by particular value is analyzed.For AD HOC, row 1930 comprise specific degree of compensation and the adaptive equalizer action that is associated with the first derivative component of input signal.Also can (for example comprise) in any suitable way these adaptive equalizer actions are controlled by particular value is analyzed.For AD HOC, row 1940 comprise specific degree of compensation and the adaptive equalizer action that is associated with the second dervative component of input signal.Also can (for example comprise) in any suitable way these adaptive equalizer actions are controlled by particular value is analyzed.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
Can utilize filter mode to reduce the negative effect of Duty Cycle Distortion and specific (standard) cyclical signal in a particular embodiment.As discussed above, can come the selective filter pattern based on the even sequence of one or more (standard) cyclical signal and the well balanced performance of the pattern in the odd sequence group.In a particular embodiment, can be at specific predefine signal (as 8B10B idle signal or 8B10B CJPAT signal) selective filter pattern particularly.Then can utilize these filter mode to make adaptive control only when detecting these filter mode, just carry out action.Receive in the process of specific predefined, the cyclical signal selected its filter mode particularly at equalizer, can reduce the negative effect of Duty Cycle Distortion to the use of filter mode.Yet, receive in the process of other cyclical signals of not selecting its filter mode particularly at equalizer, these filter mode can make the unacceptable result that causes of adaptive equalizer control to the use of filter mode, because may not be well balanced for these other cyclical signals.
To a solution of limited this problem of applicability of selected filter mode is adaptive equalizer to be controlled and freeze (stop or not using) when signal that can not determine filter mode and arrival is compatible.For example, not good randomized or be not the data sequence of arrival that is used for selecting (standard) periodic data sequence of (a plurality of) filter mode, adaptive equalizer can be controlled and freeze for being identified as.For not being identified as the data sequence of perhaps inconsistent arrival mutually, can additionally or alternatively adaptive equalizer be freezed.For example, if can not determine in seeming good randomized sequence, whether to comprise incompatible (standard) periodic sequence, then seem that at this good randomized sequence can freeze adaptive equalizer.
It is acceptable that adaptive equalizer control is temporarily freezed, because the characteristic of channel unlikely changes in a short time.Yet it may be disadvantageous that the long relatively period is freezed in equalizer control, can change because influence the characteristic of channel of intersymbol interference.For example, may the make a difference environmental change of moving such as temperature drift or cable of intersymbol interference.If the characteristic of channel changes thereby influenced intersymbol interference, may need the signal attenuation after the adaptive control of equalizer compensates variation so.Therefore, under specific circumstances adaptive control is freezed long duration and waited for that simultaneously the data sequence compatible with selected filter mode may be disadvantageous.
Second kind of solution is to select so useful filter mode group (or tabulation): when applying this useful filter mode group (or tabulation) by equalizer control by balance mode, it is compatible with any (standard) periodic data signal.In a particular embodiment, these filter mode needn't depend on their distributions in specific (standard) cyclical signal.Therefore, one or more filter mode in this tabulation may appear in the even sequence and odd sequence of specific (standard) periodic data signal unequally; Yet, can balance out departing from of the filter mode that distributes unequally to the balance application of the various filter mode that may distribute unequally.The balance application can cause observing about equally even pattern and strange pattern and work about equally, and the self adaptation of offsetting them departs from and reduce the negative effect of (standard) cyclical signal of the arrival of Duty Cycle Distortion and any kind.
Except offsetting the self adaptation that causes owing to Duty Cycle Distortion departs from, depart from by offsetting the various self adaptations in (standard) periodic data signal, may occur, the balance of filter mode is used the result that can provide at the unanimity of the adaptive control of any (standard) periodic data or good randomization data.In other words, if the application to filter mode is unbalanced, the adaptive control result can depart to the mastery filter mode that may change in various (standard) periodic data sequence or good randomization data sequence consumingly, therefore, adaptive control result can depend on the data sequence of arrival.For example, if the data that arrive are low frequency tendencies, then the adaptive control meeting departs to low frequency mode, thereby the adaptive control the possibility of result is the high frequency tendency.If the data that arrive are high frequency tendencies, then the adaptive control meeting departs to high frequency mode, thereby the adaptive control the possibility of result is the low frequency tendency.If the application to filter mode is a balance, then adaptive control is taked action by probability about equally at each filter mode, therefore, the adaptive control result can become consistent for any (standard) periodic data or good randomization data.
In a particular embodiment,, can when initial, generate useful filter mode tabulation, remain unchanged (that is, " the fixing ") of then should useful filter mode tabulating in the operating period of equalizer adaptation control.If this tabulation comprises a plurality of six filter mode, then this tabulation can comprise all possible modification of six bit patterns.Select as another kind, this tabulation can comprise all possible modification of six bit patterns that have the transformation of data between some continuous two data bit (as last two data bit).In an alternative embodiment again, this tabulation can include only the subclass of all possible modification of six bit patterns, and the generation person of this tabulation can determine that this particular subset is useful.Under any circumstance, adaptive control can circulate in whole useful filter mode fixed list in any suitable way.For example, adaptive control can circulate in whole fixed list as discussing below in conjunction with Figure 20 to 22.
Although can use fixed list in a particular embodiment, in alternative embodiment, the dynamic listing of the sequence that use can be adapted to arrive may be favourable.In a particular embodiment, the dynamic listing of the sequence that use can be adapted to arrive can improve the frequency of self application control action, solves the characteristic of channel that changes more quickly.This adaptive control can circulate in whole dynamic listing in any suitable way.For example, this adaptive control can circulate in whole fixed list as discussing below in conjunction with Figure 20 to 22.Should be pointed out that in a particular embodiment, when using fixing or dynamic listing, can enable this equalizer adaptation control consistently.It should also be noted that if there is independent Control Parameter more than, then can use independent (fixing or dynamic) useful filter mode tabulation at each independent Control Parameter in a particular embodiment.
On the other hand, consider the consistency of the adaptive control result in various (standard) periodic datas or the good randomization data, fixed list may be more favourable than dynamic listing, this is because dynamic listing may change certain consistency that the filter mode tabulation damages the adaptive control result owing to dynamic, and fixed list can not damaged any consistency of adaptive control result owing to adhere to the fixed filters mode list.
Figure 17 is the illustration flow chart that is used for dynamically generating the exemplary method 2000 of useful filter mode tabulation according to a particular embodiment of the present invention.Can manner of execution 2000, so that it is included in observed useful pattern in the sequence of arrival, and remove the pattern of dying on for example useful filter mode tabulation is upgraded.Method 2000 begins at step 2010 place, uses new useful filter mode tabulation at step 2010 place.In a particular embodiment, as discussing, can use new useful filter mode tabulation by balance mode below in conjunction with Figure 20 and 21.In addition, as discussing, can skip undetected filter mode below in conjunction with Figure 22.
At step 2020 place, the data pattern in the sequence of (a plurality of) arrival is monitored, thereby can detect useful data pattern and hash pattern.In a particular embodiment, only the data pattern of (with the filter mode that is applied big or small corresponding) certain size is monitored.The useful data pattern for example can comprise in the sequence that (a plurality of) arrive observed continually data pattern and comprise the data pattern of at least one transformation between the continuous data value in this pattern.In a particular embodiment, also can be only when having occurred between certain two data value in the data pattern changing such as last two data values this pattern be only useful.If for example data pattern has improved the Be Controlled parameter to the sensitivity of boundary value then this data pattern also can be useful.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useful, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data patterns different with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useful.Useless pattern may comprise not to be observed or seldom observed pattern, does not comprise the pattern of at least one transformation between the continuous data value in pattern or reduced the pattern of Be Controlled parameter to the sensitivity of boundary value.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data patterns different with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless.Because these patterns can not improve the frequency (this is to use the purpose of dynamic listing) of adaptive control action or adaptive control is not effectively contributed, so these patterns may be useless.After being detected, useful pattern can be collected in the tabulation, perhaps store the pattern of using.In a particular embodiment, also useless pattern can be collected in the tabulation, perhaps store useless pattern.
In step 2030, determine whether having monitored the sequence that enough (a plurality of) arrive.If do not monitor the sequence that enough (a plurality of) arrive as yet, then this method turns back to step 2020.If monitored the sequence that enough (a plurality of) arrive, then this method proceeds to step 2040.For example, after detecting the useful data pattern of some or type or after passing through the certain hour amount, may monitor the sequence that enough (a plurality of) arrive.
At step 2040 place, for example, utilize the detected useful mode list of having compiled in the data sequence that (a plurality of) arrive, useful filter mode tabulation is upgraded.In a particular embodiment, can join in the useful filter mode tabulation or this useful filter mode tabulation of replacement with detecting one or more (or owning) useful pattern in the data sequence that arrives (a plurality of).Select as another kind, this useful filter mode tabulation may comprise detected pattern, therefore needn't revise this useful filter mode and tabulate and comprise detected pattern.In either case, can from useful filter mode tabulation, delete useless filter mode.That in a particular embodiment, also can delete detected useless pattern compiles tabulation.After having upgraded useful filter mode tabulation, this method turns back to step 2010, uses new useful filter mode tabulation at step 2010 place.
Figure 18 is the illustration flow chart that is used for dynamically generating another exemplary method 2100 of useful filter mode tabulation according to a particular embodiment of the present invention.Similar with method 2000, can manner of execution 2100 so that useful filter mode tabulation is upgraded so that it is included in observed useful pattern in the sequence of arrival, and remove the pattern of dying on.Method 2100 can also be created independently dynamic listing for even data sequence and odd data sequence.Method 2100 can also be edited these tabulations alternatively, makes even sequence and odd sequence not play dominating role to adaptive control.Do the negative effect that can reduce Duty Cycle Distortion like this.
Method 2100 begins at step 2110 place, uses new useful filter mode tabulation at step 2110 place.In a particular embodiment, as discussing, can use new useful filter mode tabulation by balance mode below in conjunction with Figure 20 and 21.As discussing, also can use new filter mode tabulation, after certain period, to skip undetected filter mode in conjunction with overtime detection below in conjunction with Figure 22.
At step 2120 place, antithesis data sequence and odd data sequence (its respectively or with idol position beginning, then be strange position, idol position, strange position or the like, with strange position beginning, then be even position, strange position, idol position or the like) in data pattern monitor, and detect useful and hash pattern in each data sequence.The useful data pattern for example can comprise in the data sequence that arrives observed continually data pattern and comprise the data pattern of at least one transformation between the continuous data value in this pattern.In a particular embodiment, also can be only when having occurred between certain two data value in the data pattern changing this pattern be only useful.If for example data pattern has improved the Be Controlled parameter to the sensitivity of boundary value then this data pattern also can be useful.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useful, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data patterns different with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useful.Useless pattern may comprise not to be observed or seldom observed pattern, does not comprise the pattern of at least one transformation between the continuous data value in pattern or reduced the pattern of Be Controlled parameter to the sensitivity of boundary value.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data patterns different with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless.Detected useful pattern can be collected in respectively in the discrete tabulation of even data sequence and odd data sequence or the storage this useful pattern.In a particular embodiment, also useless pattern can be collected in respectively at this useless pattern of storage in the discrete tabulation of even data sequence and odd data sequence or discretely.
At step 2130 place, the even sequence and the odd sequence of whether having monitored enough arrivals are determined.If do not monitor enough sequences as yet, then this method turns back to step 2120.If monitored enough sequences, then this method proceeds to step 2140.For example, after detecting the useful data pattern of some or type or after passing through the certain hour amount, may monitor the sequence of enough arrivals.
At step 2140 place, pattern in detected those useful patterns, that only appear among in even sequence and the odd sequence one is neglected.In a particular embodiment, detected useful pattern in the even data sequence and detected useful pattern in the odd data sequence can be compared, and only among in even sequence and odd sequence observed those patterns from limit of consideration, get rid of as the filter mode the new filter mode tabulation.For example, can not consider these patterns by from the idol tabulation of the detected useful pattern compiled or strange tabulation, removing them.In a particular embodiment, these patterns can be placed in the idol tabulation or strange tabulation of the detected useless pattern of compiling.
At step 2150 place, useful filter mode tabulation is upgraded.In a particular embodiment, after those patterns of from the detecting pattern tabulation that is compiled, having removed in one that only appears in even sequence and the odd sequence, one or more (or owning) the useful pattern in the detecting pattern tabulation that is compiled can be joined useful filter mode tabulation or replace this useful filter mode tabulation.Select as another kind, this tabulation may comprise detected pattern, therefore needn't revise this tabulation to comprise detected pattern.In either case, can from useful filter mode tabulation, delete useless filter mode.In a particular embodiment, also can delete the idol tabulation or the strange tabulation of the useless detecting pattern that is compiled.After having upgraded useful filter mode tabulation, this method turns back to step 2110, uses new useful filter mode tabulation at step 2110 place.
Figure 19 has been the illustration flow chart of an exemplary method 2200 again that is used for dynamically generating useful filter mode tabulation according to a particular embodiment of the present invention.Similar with method 2000 and 2100, method 2200 can be upgraded so that it has comprised observed useful pattern in the sequence that arrives useful filter mode tabulation, and removes the pattern of dying on.Similar with method 2100, method 2200 can also be for creating independent dynamic listing (promptly with the data sequence of idol position and the beginning of strange position, idol tabulation and strange tabulation), these tabulations are edited, make even sequence and odd sequence not play dominating role to adaptive control.Method 2200 can be done like this by following processing: the pattern quantity in pattern quantity in the antithesis tabulation and the strange tabulation is counted, these two quantity are compared, and from having than removal pattern the tabulation of multi-mode, the pattern quantity in the idol tabulation equals the pattern quantity in the strange tabulation.By this mode, can reduce the influence of Duty Cycle Distortion.And the dependence of the data sequence of 2200 pairs of arrivals of method can be less relatively.
Method 2200 begins at step 2210 place, uses new useful filter mode tabulation at step 2210 place.In a particular embodiment, as discussing, can use this new useful filter mode tabulation by balance mode below in conjunction with Figure 20 and 21.As discussing, also can use new filter mode tabulation, after certain period, to skip undetected filter mode in conjunction with overtime detection below in conjunction with Figure 22.
At step 2220 place, data pattern in data pattern in the antithesis data sequence (its with idol position beginning, then be strange position, idol position, strange position or the like) and the odd data sequence (its with the beginning of strange position, then be even position, strange position, idol position or the like) is monitored, and detects the useful and useless pattern in each data sequence.The useful data pattern for example can comprise in the data sequence that arrives observed continually data pattern and comprise the data pattern of at least one transformation between the continuous data value in this pattern.In a particular embodiment, also can be only when changing has appearred in (between last two data values) between two particular data value in the data pattern this pattern be only useful.If for example data pattern has improved the Be Controlled parameter to the sensitivity of boundary value then this data pattern also can be useful.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useful, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data patterns different with the data value of 2.5 positions before the border at the data value of 1.5 positions before the border may be useful.Useless pattern may comprise not to be observed or seldom observed pattern, does not comprise the pattern of at least one transformation between the continuous data value in pattern or reduced the pattern of Be Controlled parameter to the sensitivity of boundary value.For example, simulate the gain of the first derivative component of second dervative equalizer for control, the data patterns different with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless, and for the gain of controlling the second dervative component of simulating the second dervative equalizer, the data pattern identical with the data value of 2.5 positions before this border at the data value of 1.5 positions before the border may be useless.Detected useful pattern can be collected in the discrete tabulation (" idol " tabulation and " very " tabulation) at even data sequence and odd data sequence respectively or otherwise this useful pattern of storage.In a particular embodiment, also useless pattern can be collected in respectively in the discrete tabulation of even data sequence and odd data sequence or the storage this useless pattern.
At step 2230 place, the even sequence and the odd sequence of whether having monitored enough arrivals are determined.If do not monitor enough sequences as yet, then this method turns back to step 2220.If monitored enough sequences, then this method proceeds to step 2240.For example, after having detected the useful data pattern of some or type or after passing through the certain hour amount, monitored the sequence of enough arrivals.
At step 2240 place, the detecting pattern in idol tabulation and the strange tabulation is compared, quantity that only appears at the pattern in the idol tabulation and the quantity that only appears at the pattern in the strange tabulation are counted.At step 2250 place, determine whether the quantity that only appears at the pattern in the idol tabulation and the quantity of pattern in only appearing at strange tabulation is identical.If these quantity differences, then this method proceeds to step 2260.If these quantity are identical, then this method proceeds to step 2270.
At step 2260 place, if the quantity that only appears at the pattern in the idol tabulation is different with the quantity of pattern in only appearing at strange tabulation, then from having than pattern of removal the tabulation of multi-mode.In a particular embodiment, can be from having than removing any suitable pattern the tabulation of multi-mode in any suitable way.For example, in a particular embodiment, can remove the pattern that appears at the most continually in even data sequence or the odd data sequence.Removed this pattern from have the tabulation than multi-mode after, this method turns back to step 2240, and the quantity that only appears at the pattern in the idol tabulation and the quantity that only appears at the pattern in the strange tabulation are counted and compared.
At step 2270 place,, then use the detecting pattern tabulation of for example being edited that useful filter mode is tabulated and upgrade if the quantity that only appears at the pattern in the idol tabulation is identical with the quantity of pattern in only appearing at strange tabulation.In a particular embodiment, can join one or more (or all) the useful pattern in the tabulation of the detecting pattern edited in the useful filter mode tabulation or replace useful filter mode tabulation.Select as another kind, this useful filter mode tabulation may comprise detected pattern, therefore needn't revise this tabulation to comprise detected pattern.In either case, can from useful filter mode tabulation, delete useless filter mode.In a particular embodiment, also can delete the idol tabulation or the strange tabulation of the useless detecting pattern that is compiled.After having upgraded useful filter mode tabulation, this method turns back to step 2210, uses new useful filter mode tabulation at this step place.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described method.
As discussed above, useful filter mode tabulation can be fixing or dynamic.Equalizer control can use filter mode in the tabulation of arbitrary type to reduce the negative effect of Duty Cycle Distortion by balance mode.Use filter mode to typically refer to equally to use each filter mode in the filter mode tabulation or give equal weight or selection probability by balance mode each filter mode in this tabulation.
Figure 20 has been the illustration flow chart that is used for using the exemplary method 2300 of filter mode according to a particular embodiment of the present invention by balance mode.Method 2300 begins at step 2310 place, selects a filter mode at this step place from useful filter mode tabulation.This useful filter mode tabulation can be fixing or dynamic.Can be in any suitable way from useful filter mode tabulation the selective filter pattern.In a particular embodiment, can be from tabulation selective filter pattern sequentially.In alternative embodiment, can from tabulation, press equal probability selective filter pattern randomly.Should be pointed out that filter mode can for example comprise six bit patterns, and can change such as existing between latter two specific two positions.
At step 2320 place, can monitor (a plurality of) data sequence of the signal of arrival at selected filter mode.At step 2330 place,, then can continue (a plurality of) data sequence of the signal that arrives is monitored if do not detect selected filter mode.If detected selected filter mode, then this method proceeds to step 2340.
At step 2340 place, take suitable control action with the control parametric equalizer.In a particular embodiment, as discussed above, can analyze and take suitable control action to the data and the boundary value information of detected pattern.The above discussion in conjunction with Fig. 5,7 and 9 is used for exemplary method that output signal value is made an explanation.After output signal value carried out analyzing, equalizer can apply suitable control action to signal.In another embodiment of the present invention, the adaptive control action of taking at step 2340 place can be any suitable adaptive control action that adopts various conventional adaptive control algorithms.For example, the adaptive control action at step 2340 place can be based on the conventional adaptive control algorithm such as lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm, ZF (ZF) algorithm or the like.The data that these conventional adaptive control algorithms require to arrive usually are well randomized, and are may produce unacceptable result under the situation of (standard) periodic data in the data that arrive.In a particular embodiment, the balance of the adaptive control action that utilizes filter mode is used made that these conventional adaptive control algorithms can be in various (standard) periodic data sequence and well provide consistent self adaptation result in the randomization data sequence.Because these conventional adaptive control algorithms not necessarily need transformation of data to take adaptive control action, therefore filter mode needn't comprise transformation of data in a particular embodiment.After having applied control action, this method turns back to step 2310, selects new filter mode at this step place.By this mode, press balance mode and use filter mode, reduced Duty Cycle Distortion negative effect and various (standard) periodic data sequence with well provide consistent self adaptation result in the randomization data sequence.
Figure 21 has been the illustration flow chart that is used for using another exemplary method 2400 of filter mode according to a particular embodiment of the present invention by balance mode.In method 2400, monitor at the filter mode in the tabulation simultaneously, with detected and taked those filter mode of self adaptation action to carry out mark or identification to it, no longer these filter mode are monitored then, and when having detected all filter mode, remove these marks.By this mode, press balance mode and use filter mode, reduced the negative effect of Duty Cycle Distortion.
Method 2400 begins at step 2410 place.At step 2410 place, the mark that clears all makes the neither one pattern be labeled.The expression that do not exist of the mark that is associated with pattern can be monitored (because not detecting this pattern as yet) to the data sequence that (a plurality of) arrive at this pattern.Mark is represented can be no longer at the associative mode of this mark the data sequence that (a plurality of) arrive to be monitored (because having detected this pattern).Although should be pointed out that and used mark in the present embodiment, in alternative embodiment, can realize being used to represent to detect when any suitable technique of certain pattern.It should also be noted that useful filter mode tabulation can be fixing or dynamic.
At step 2420 place, the data sequence that (a plurality of) arrive is monitored at unlabelled those filter mode still.Therefore, be right after having removed underlined after, at all filter mode in the tabulation of useful filter mode the data sequence that (a plurality of) arrive is monitored.In a particular embodiment, can monitor all unmarked filter mode simultaneously.Detected along with filter mode, work and be labeled, at filter mode (unmarked filter mode) still less the data sequence that (a plurality of) arrive is monitored.
At step 2430 place, determine whether having detected any monitored filter mode in the data sequence that arrives (a plurality of).If do not detect a filter mode, then can continue the data sequence that (a plurality of) arrive to be monitored, and the correspondence markings of each filter mode all keeps not choosing at these filter mode.If detect a filter mode, then this method proceeds to step 2440.
At step 2440 place, take suitable control action with the control parametric equalizer.In a particular embodiment, can analyze and take control action to the data and the boundary value information of detected pattern.The above discussion in conjunction with Fig. 5,7 and 9 is used for exemplary method that output signal value is made an explanation.After output signal value carried out analyzing, equalizer can apply suitable control action to signal.In alternative embodiment, the adaptive control action of taking at step 2440 place can be any suitable adaptive control action that adopts various conventional adaptive control algorithms.For example, the adaptive control action at step 2440 place can be based on such as lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm or ZF (ZF) algorithm or the like.The data that these conventional adaptive control algorithms require to arrive usually are well randomized, and are the results that may not can produce under the situation of (standard) periodic data in the data that arrive.The balance of the adaptive control action that utilizes filter mode used make that these conventional adaptive control algorithms can be in various (standard) periodic data sequence and well provide consistent self adaptation result in the randomization data sequence.After detecting filter mode (and alternatively after having taked control action), the detected filter mode of mark.
At step 2450 place, to whether mark all filter mode in the tabulation of useful filter mode determine.If not, then this method turns back to step 2420, and at unlabelled those filter mode still the data sequence that (a plurality of) arrive is monitored.If mark all filter mode in the tabulation of useful filter mode, then this method proceeds to step 2410, remove institute at this step place underlined.By this mode, press balance mode and use filter mode, reduced Duty Cycle Distortion negative effect and various (standard) periodic data sequence with well provide consistent self adaptation result in the randomization data sequence.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
As discussed above, can be by using filter mode with the negative effect that reduces Duty Cycle Distortion and between various (standard) periodic data sequence and good randomization data sequence, provide consistent self adaptation result by balance mode, and the exemplary method that is used for using by balance mode filter mode has been discussed in conjunction with Figure 20 and 21.Yet, when in the step 2430 of the step 2330 of method 2300 or method 2400, carrying out and whether to have detected filter mode relevant really regularly, these methods can pause (stall) under the situation of the filter mode that does not for example detect expectation.In other words, if do not detect the specific filter pattern, then can not take other self adaptation actions.It may not be disadvantageous by this mode adaptive control for example being freezed in short time, because the characteristic of channel unlikely changes.Yet even under the situation of short-term and especially under long-term situation, forcing this method to skip undetected filter mode after certain period can be so that equalizer be adapted to the condition that changes more quickly.Can carry out more frequent adaptive control action by making, skip undetected filter mode and can prevent self-adaptive controlled pause when the characteristic of channel changes.
Figure 22 is the illustration flow chart that is used for after certain time period, skipping the exemplary method 2500 of undetected filter mode according to a particular embodiment of the present invention.In method 2500, after certain period that in the data sequence that (a plurality of) arrive, does not detect filter mode, can detect overtime.Detect this overtime after, can skip its (for example, in step 2430 of above method 2400) of this filter mode (for example, in the step 2330 of method 2300) or mark.
Should be pointed out that to tabulate in conjunction with fixing or dynamic useful filter mode uses overtime detection.Yet, overtime detection still less can appear when using fixed list when using dynamic listing, this be because the filter mode in the dynamic listing by based on they in the sequence that arrives frequency and upgrade.In other words, if will observed continually pattern be included in the useful filter mode tabulation and remove not observed continually pattern in the sequence that arrives, then detecting overtime chance will tail off.Still can be in conjunction with dynamic listing using method 2500, after having changed the sequence that arrives, to compensate any delay to the renewal process of dynamic listing.
Method 2500 begins at step 2510 place, at this step place, timer is resetted.Associated methods 2300 can reset timer after next filter mode has been selected at step 2310 place.Associated methods 2400, can be for example after step 2410 place is initially with the underlined removing of institute and/or after step 2430 place detects filter mode, timer is resetted.Timer can be set at such period, that is, after this period, will skip this filter mode (perhaps in method 2400, skipping a plurality of filter mode).The period of being set by timer can comprise any suitable period.
At step 2520 place, determine whether having detected filter mode in the data sequence that arrives (a plurality of).If detected filter mode, then this method turns back to step 2510 and timer is resetted.If do not detect filter mode as yet, then this method proceeds to step 2530.
At step 2530 place, overtimely determine whether having taken place.Overtimely be meant the time of setting by timer of having used up.If do not take place overtimely as yet, then this method turns back to step 2520.If taken place overtimely, then this method proceeds to step 2540.
After generation is overtime, skip undetected (a plurality of) filter mode at step 2540 place.Associated methods 2300 is skipped this filter mode and is selected next filter mode at step 2310 place.Associated methods 2400 is skipped all and is remained unmarked filter mode (for example, can with they whole marks) and remove institute at step 2410 place underlined, makes and can restart this processing.In a particular embodiment, can from useful filter mode tabulation, remove any (a plurality of) pattern of having skipped ((a plurality of) pattern of being skipped to prevent makes adaptive equalizer pause once more).Method 2500 then turns back to step 2510, at this step place timer is resetted.By this mode, can skip undetected filter mode, and can take adaptive control action more continually, prevent that adaptive control from pausing at undetected filter mode.
On the other hand, in a particular embodiment, consistency for adaptive control result in various (standard) periodic datas or good randomization data, not carrying out may be more favourable to overtime detection, because may damage some consistency of adaptive control result owing to skipping undetected filter mode, even and do not carry out this detection owing to take place to pause and also to adhere to all filter mode and can not damage any consistency of adaptive control result to overtime detection.In a particular embodiment, pause may not be a problem, perhaps for possibility or even most popular scheme such as some data sequence of continuous 0101 data sequence, because the periodic data sequence of this height lacks frequency spectrum thereby may not comprise the enough information that is used to carry out adaptive control in frequency domain.If adaptive control is not at such as continuous 0101 height periodic data sequence and pause, Control Parameter may float to very poor value so.Therefore, in a particular embodiment, pausing for this height periodic data sequence may be most popular scheme.Do not carry out and to allow to pause at this height periodic data sequence to overtime detection.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
A kind of distorted signals that is called as residual intersymbol interference is paid close attention in below very most of discussion.The distorted signals of another type that produces in telecommunication is the residual direct current offset in the signal.If do not offset (that is, compensation) residual direct current offset, it can reduce the input sensitivity at receiver place so.Therefore, the residual direct current offset at counteracting receiver place is useful.If receiver has such as the analog front circuit of equalizer or before its decision circuit and has limiting amplifier, then offsetting residual direct current offset may be especially useful, because these components can add skew in the signal.
Referring again to Fig. 1 to 3, except intersymbol interference, the signal that sends by communication channel 30 also can stand the direct current offset distortion, and this direct current offset distortion meeting further strengthens at receiver balance device 42 places.Receiver balance device 42 can apply DC-offset compensation (that is, proofread and correct) to received input signal with direct current offset that offseting signal was shown.Receiver logic 47 then can be analyzed the output signal after regulating at residual direct current offset.Specifically, sampler 104 can receive equalizer output signal 46 (output signal after the adjusting) and clock signal.Sampler 104 then can be sampled to this output signal at the specified point place that is determined by clock signal, to generate data value and boundary value.Sampler 104 can be transmitted to offset controller 106 with these data and boundary value and carry out suitable analysis with (as described below).Analyze based on this, the offset controller 106 of receiver logic 47 and equalizer 42 can be proofreaied and correct the direct current offset distortion in the input signal that arrives.
A plurality of examples of clock signal that Figure 23 A, 23B and 23C and the equalizer output signal 46 that has showed polytype residual direct current offset have contrasted the ground illustration.In a particular embodiment, sampler 104 can receive the equalizer output signal 46 such as illustrative those signals in these figure, and according to 2x oversampling clock and data recovery (CDR) scheme equalizer output signal 46 is sampled.By this scheme, sampler 104 can be sampled twice to equalizer output signal 46 every data bit period (it can be limited by clock signal).For a data bit period, sampler 104 can be in equalizer output signal 46 should be to equalizer output signal 46 sampling once corresponding to the some place of data value, and in equalizer output signal 46 should be to equalizer output signal 46 sampling once corresponding to the some place of boundary value.Sampler 104 then can be transmitted to offset controller 106 with these data values and boundary value.Based on the analysis to particular data value and boundary value, such as described further below, offset controller 106 can be regulated the DC-offset compensation that imposes on the signal that is received by equalizer 42.Should be noted that, in a particular embodiment, as mentioned above, same data value and the boundary value information that is transmitted to offset controller 106 also can be transmitted to adaptive controller 102, and the combining adaptive gain controlling is used described data value and boundary value information.
Clock signal example 2600 that Figure 23 A and the equalizer output signal 46 that does not show residual direct current offset have contrasted the ground illustration.Example 2600 is similar to above example 200 in conjunction with Fig. 4 A description, therefore no longer it is described in detail.Yet, should be noted that, in the signal that does not show residual direct current offset (as in example 2600), as discussed above, comprising that each boundary value (for example boundary value at E2, E3 and E4 place) between two continuous data values of transformation comprises high value randomly or comprises low value (being illustrated as " X ").For sort signal, because the direct current offset distortion is by full remuneration or do not exist, so the DC-offset compensation that offset controller 106 can impose on input signal is regulated randomly up or down.If equate basically with the number of times of regulating downwards to adjusted, the DC-offset compensation that then imposes on input signal is said the level that keeps identical on average.If is not about equally to adjusted with the number of times of regulating downwards, the DC-offset compensation that then imposes on input signal can drift about slightly from initial level.This drift of DC-offset compensation level can produce slight direct current offset distortion.As following illustrative, the equalizer receiver can detect this distortion and this DC-offset compensation is corrected back to average initial level.
Clock signal example 2650 that Figure 23 B and the equalizer output signal 46 that shows just residual direct current offset have contrasted the ground illustration.With respect to the signal that does not show residual direct current offset, show the upwards drift (as in this exemplary plot) of equalizer output signal 46 of just residual direct current offset.And, may be identical in high impulse (for example pulse at the D3 place) boundary value (for example E2) or the value of boundary value afterwards (for example E3) before with the data value (for example D3) of this pulse place.May different with the data value of this pulse place (on the contrary) before low pulse with the value of afterwards boundary value.Therefore, such as described further below, when to particular data value and (a plurality of) when boundary value is analyzed, offset controller 106 can reduce impose on input signal DC-offset compensation to offset just residual direct current offset.Should be pointed out that in a particular embodiment and as described below that offset controller 106 is occurring may not offsetting the just residual direct current offset that output signal shows before (for example between D2 and the D3) transformation.
Clock signal example 2700 that Figure 23 C and the equalizer output signal 46 that shows negative residual direct current offset have contrasted the ground illustration.With respect to the signal that does not show residual direct current offset, the equalizer output signal 46 that shows negative residual direct current offset has drifted about downwards (as in this exemplary plot).And, may different with the data value of this pulse place (on the contrary) in high impulse (for example pulse at the D3 place) boundary value (for example E2) or the value of boundary value afterwards (for example E3) before.Value with afterwards boundary value before low pulse may be identical with the data value of this pulse place.Therefore, such as described further below, when to particular data value and (a plurality of) when boundary value is analyzed, offset controller 106 can increase impose on input signal DC-offset compensation to offset negative residual direct current offset.Should be pointed out that in a particular embodiment and as described below that offset controller 106 is occurring may not offsetting the negative residual direct current offset that output signal shows before (for example between D2 and the D3) transformation.
Figure 24 is an illustration, and according to a particular embodiment of the present invention being used for makes an explanation with the flow chart of method 2800 that residual direct current offset is offset to output signal value.Method 2800 begins at step 2810 place, at this step place, uses clock signal that output signal 46 is sampled.Such as above described in conjunction with Figure 3, this output signal 46 can be the output of equalizer, and can sample to this output signal according to clock signal.
In a particular embodiment, can sample to output signal at the base value strong point and the boundary point place that determine by clock signal.Select as another kind, can output signal not sampled, and can derive and the corresponding boundary value of these non-sample points at the boundary point place.In a particular embodiment, offset controller 106 can be derived boundary value according to the data value of sampling out and other phase informations (that is, the phase place of output signal is morning or late).For example, if the phase place of output signal early, then can to determine the high value or the low value of boundary value identical with the high value or the low value that are right after the data value before this boundary value for offset controller 106.If the phase place evening of output signal, then can to determine the high value or the low value of boundary value identical with the high value or the low value that are right after the data value after this boundary value for offset controller 106.
At step 2820 place, after to the output signal sampling, can analyze the data value of being sampled out, to determine in these values, whether having occurred transformation.At step 2830 place, if do not detect transformation, then this method turns back to step 2820.If detect transformation between continuous data value, then this method proceeds to step 2840.Should be pointed out that in a particular embodiment, can received data value relatively be detected transformation mutually by direct.In alternative embodiment, can detect transformation by received data value and boundary value and the predefine binarization mode that comprises transformation (and corresponding to particular offset offset action) are compared.It should also be noted that in a particular embodiment, can after only detecting a transformation, carry out the offset cancellation action.
If detect transformation, then the value (high or low) to the boundary value between the data value that comprises this transformation is discerned at step 2840 place.At step 2850 place, if this boundary value height, then this method proceeds to step 2860, and takes negative bias to move and offset action signal is regulated (because residual direct current offset is positive) downwards.If this boundary value is low, then this method proceeds to step 2870, and take the action of positive offset cancellation with signal to adjusted (because residual direct current offset is born).In a particular embodiment, can discern and take the offset cancellation action boundary value by boundary value and predefined pattern (it offsets action corresponding to particular offset) are compared.
Should be pointed out that in a particular embodiment, can be by offset controller 106 execution in step 2820 to 2870, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.And, in a particular embodiment, if an above signal path (for example to the path in the example equalizer 42 101) is applied DC-offset compensation, then can in a path, regulate and fix the DC-offset compensation in other paths to the DC-offset compensation that is applied.In alternative embodiment, can utilize specific function that independent control variable is mapped to a plurality of paths, and can apply DC-offset compensation to these paths according to this mapping process.Select as another kind, as further discussing, can regulate DC-offset compensation independently at each path below in conjunction with Figure 30 to 40.
The table of the example direct current offset controlling schemes 2900 that Figure 25 is an illustration is associated with the method 2800 of Figure 24.Each row 2902 is carried out the action of particular offset Canceller all corresponding to the particular value pattern at this particular value pattern.Row 2910 comprise the data of being sampled out and the high value and the low value (" 1 " or " 0 ") of each data in the boundary value series and boundary value.Row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D1 " and " D2 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 2902 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value the row E1) (thereby can can't help sampler 104 sample out) from these data values and phase information.
Row 2912 comprise the alternative boundary value that each row 2902 is located at row " E1 ".For AD HOC, row 2914 comprise specific residual direct current offset level.For AD HOC, row 2916 comprise that specific action to the offset cancellation setting is to compensate specific residual direct current offset level.Can discussing in the method 2800, apply the action that is provided with at offset cancellation as above.
Should be pointed out that in a particular embodiment the stream that offset controller 106 can the acceptance sampling value is also selected suitable value (for example, comprise between two data values of transformation boundary value) from these values.Offset controller 106 then can carry out suitable offset cancellation action based on this boundary value.Select as another kind, offset controller 106 can compare these sample values and predefine binarization mode (it offsets action corresponding to particular offset).Based on the pairing specific predefine binarization mode of these sample values, offset controller 106 can apply corresponding offset cancellation action.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
Utilizing output signal value that residual direct current offset is assessed a challenge that can cause is the mistake lockout issue.When clock signal and offset cancellation device interact improperly and when having produced the mistake lock-out state, the mistake lockout issue can occur.For example, under the same high situation of initial residual skew, can produce the mistake lock-out state with input signal amplitude.Under the mistake lock-out state, the sample phases of boundary value and data value is exchanged.Described skew is also removed from real center to a great extent, and the border sampling is locked in the above or following place, crosspoint of true eye pattern aperture (eye opening).The mistake lockout issue can twist the assessment that utilizes output signal value that residual direct current offset is carried out.
Figure 26 has been the illustration flow chart that is used for method 3000 that the mistake locking that takes place in the process of offsetting residual direct current offset is corrected according to a particular embodiment of the present invention.Method 3000 is corrected the mistake lockout issue by based on the value (high or low) of each boundary value DC-offset compensation being regulated.Therefore, different with method 2800, (except based on the boundary value between the data value that comprises transformation, also) can regulate DC-offset compensation based on the boundary value between the data value with identical value.As a result, before regulating DC-offset compensation, in method 3000, needn't discern transformation.By regulating DC-offset compensation by this mode, method 3000 can be touched the signal that is in the mistake lock-out state (nudge) and be gone out the mistake lock-out state.Should be pointed out that method 3000 can be similar to the said method 2800 of Figure 24 (except that step 2820 and 2830).
Method 3000 begins at step 3010 place, uses clock signal that output signal is sampled at this step place.Such as above described in conjunction with Figure 3, this output signal can be the output of equalizer, and can sample to this output signal according to clock signal.In a particular embodiment, can sample to output signal at the base value strong point and the boundary point place that determine by clock signal.Select as another kind, can output signal not sampled, and can derive and the corresponding boundary value of these non-sample points at the boundary point place.Should be pointed out that this moment, signal may be in the mistake lock-out state.
After to the output signal sampling,, can discern the value (high or low) of boundary value at step 3020 place.At step 3030 place, if the boundary value height, then this method proceeds to step 3040, and takes negative bias to move and offset action so that signal is regulated downwards.If this boundary value is low, then this method proceeds to step 3050, and take the action of positive offset cancellation with signal to adjusted.In a particular embodiment, can discern and take the offset cancellation action boundary value by boundary value and predefined pattern (it offsets action corresponding to particular offset) are compared.
By take the offset cancellation action based on any boundary value (that is, no matter whether having identified transformation), the signal that is in the mistake lock-out state can be touched out the mistake lock-out state.In addition, because employed some boundary value may appear between the continuous data value that comprises transformation, therefore carrying out offset adjusted also can offset residual direct current offset.
The table 3100 of the example skew controlling schemes that Figure 27 is an illustration is associated with the method 3000 of Figure 26.Each row 3102 is carried out the action of particular offset Canceller all corresponding to the particular value pattern at this particular value pattern.Row 3110 comprise the data of being sampled out and the high value or the low value (" 1 " or " 0 ") of each data in the boundary value series and boundary value.Row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, transformation not necessarily can appear between the data value of row " D1 " and " D2 " in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 3102 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 3112 comprise the alternative boundary value that each row 3102 is located at row " E1 ".For AD HOC, row 3114 comprise specific residual direct current offset level.For AD HOC, row 3116 comprise the specific action that offset cancellation is provided with.Can be as the above action of discussing in the method 3000, carrying out at the offset cancellation setting.
Should be pointed out that the certain edges thereof dividing value in the row 3112 is placed in the bracket.Boundary value in the bracket is to have the high value different with high value that is right after two values around this boundary value or low value or those boundary values of low value.This in a particular embodiment situation may be atypical.Yet, can take the action of offset cancellation device based on the high value or the low value of boundary value in a particular embodiment.
Especially when method 3000 is applied to random signal or do not adopt the double sampling technology (by than data rate the sampling techniques of lower speed) situation under, method 3000 can provide the effective solution at the mistake lockout issue.Yet, if the signal that arrives is (standard) cyclical signal and has adopted the double sampling technology that method 3000 may not can prevent to occur system's residual bias in signal in some cases.For example, being locked in the double sampling cycle may be so under the situation in cycle of cyclical signal.Therefore, under specific circumstances, can under the situation that adopts the double sampling technology, correct the mistake lockout issue and may be useful the offset cancellation device that the residual direct current offset of (standard) cyclical signal is suitably regulated.
Figure 28 has been the illustration flow chart that is used for other method 3200 that the mistake locking that takes place in the process of offsetting residual direct current offset is corrected according to a particular embodiment of the present invention.Method 3200 is at first monitored the mistake lockout issue of correcting in the cyclical signal to data direct current imbalance (synonym of mistake lockout issue) by utilizing output signal value.When detecting imbalance, DC-offset compensation is regulated based on detected imbalance.If the imbalance of not detecting is then only regulated DC-offset compensation (being similar to said method 2800) based on the value (high or low) of those boundary values between the continuous data value that comprises transformation.Utilize method 3200,, also data direct current imbalance can be controlled in the tolerance interval even adopt the double sampling technology that (standard) periodic data signal is sampled.
Method 3200 begins at step 3210 place, uses clock signal that output signal is sampled at this step place.Such as above described in conjunction with Figure 3, this output signal can be the output of equalizer, and can sample to this output signal according to clock signal.In a particular embodiment, can sample to output signal at the base value strong point and the boundary point place that determine by clock signal.Select as another kind, can output signal not sampled, and can derive and the corresponding boundary value of these non-sample points at the boundary point place.In a particular embodiment, offset controller 106 can be derived boundary value according to data from the sample survey value and other phase informations (for example, phase place is morning or late).
At step 3220 place, on one side output signal is sampled, on one side the quantity of low data value (for example " 0 ") and the quantity of high data value (for example " 1 ") are counted, and when signal is sampled, upgraded this counting.In a particular embodiment, the quantity of the data value (low and high) in each counting can include only the previous observed data value of some.In the embodiment of alternative, the quantity of data value in each counting can only be included in previous observed those data values during certain period.Any suitable (a plurality of) counter can be stored the quantity of viewed high data value and the quantity of low data value, and can this (a plurality of) counter be upgraded based on the high value or the low value of the data value that arrives.
At step 3230 place, to compare the counting of high data value and counting low data value, and whether much frequent as to be observed and to determine (whether be in definite signal and miss lock-out state) to the data value of a type than the data value of another type.If the data value of a type is much more frequent that ground (only as example, more than 3 times more continually) is observed than the data value of another type, then method 3200 proceeds to step 3240.If the data value of neither one type is much more frequent as to be observed than the data value of another type, then method 3200 proceeds to step 3270.In a particular embodiment, the ratio of the quantity of the difference of the quantity of all types of data values or all types of data value can be compared with predetermined quantity or ratio respectively.
At step 3240 place, much more frequent as to have observed high data value and determine than low data value to whether.If much more frequent as to have observed high data value than low data value, then method 3200 proceeds to step 3250, and takes negative bias to move and offset action so that signal is regulated downwards.If much more frequent as to have observed low data value than high data value, then method 3200 proceeds to step 3260, and take the action of positive offset cancellation with signal to adjusted.By the DC-offset compensation that is applied being regulated, the signal that is in the mistake lock-out state can be touched out the mistake lock-out state by this mode.
Should be noted that, although method 3200 is by counting output data value and relatively coming Monitoring Data direct current imbalance, but in alternative embodiment, can count output data value and/or boundary value and compare by similar mode, with Monitoring Data direct current imbalance.Also can analyze the counting of data value and/or boundary value, so that the migration that imposes on input data signal is regulated by similar mode.
If the data value in step 3230 place neither one type is much more frequent as to be observed than the data value of another type, then method 3200 proceeds to step 3270.At step 3270 place, can analyze to determine in these values, whether having occurred transformation the data from the sample survey value.At step 3280 place, if do not detect transformation, then this method turns back to step 3210, and output signal is sampled.If detect transformation between the continuous data value, then this method proceeds to step 3290.Should be pointed out that in a particular embodiment, can received data value relatively be detected transformation mutually by direct.In alternative embodiment, can detect transformation by received data value and boundary value and the predefine binarization mode that comprises transformation (and corresponding to particular offset offset action) are compared.It should also be noted that in a particular embodiment, can after only detecting a transformation, carry out the offset cancellation action.
If detect transformation, then the value (high or low) to the boundary value between the data value that comprises this transformation is discerned at step 3290 place.At step 3300 place, if this boundary value height, then this method proceeds to step 3250, and takes negative bias to move and offset action signal is regulated (because residual direct current offset is born) downwards.If this boundary value is low, then this method proceeds to step 3260, and take the action of positive offset cancellation with signal to adjusted (because residual direct current offset is born).In a particular embodiment, can discern and take the offset cancellation action boundary value by boundary value and predefined pattern (it offsets action corresponding to particular offset) are compared.Method 3200 then turn back to step 3210 and 3220 in output signal is sampled and to low data value and the quantity of high data value upgrade.
Should be pointed out that in a particular embodiment, can be by offset controller 106 execution in step 3220 to 3300, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.And, in a particular embodiment, if an above signal path (for example to the path in the example equalizer 42 101) is applied DC-offset compensation, then can in a path, regulate and fix the DC-offset compensation in other paths to the DC-offset compensation that is applied.In alternative embodiment, can utilize specific function that independent control variable is mapped to a plurality of paths, and can apply DC-offset compensation to these paths according to this mapping process.Select as another kind, as further discussing, can regulate DC-offset compensation independently at each path below in conjunction with Figure 30 to 40.
As can be observed, the mode that method 3200 will be regulated DC-offset compensation based on the relative frequency of observing particular data value be divided into two kinds.If much more frequent as to have observed the data value of another type than one type data value, then 3200 pairs of methods might be owing to this disequilibrium that the mistake locking causes be proofreaied and correct.If the data value of neither one type is much more frequent as to be observed than the data value of another type, then method 3200 these signals of supposition are not in the mistake lock-out state and only those boundary values between the continuous data value that comprises transformation are analyzed to proofread and correct residual direct current offset.By this mode, method 3200 can be so that any data direct current imbalance remains in the tolerance interval.In a particular embodiment, though the signal that arrives be (standard) cyclical signal and adopted the double sampling technology, data direct current imbalance is remained in the tolerance interval.
The table 3400 of the example skew controlling schemes that Figure 29 is an illustration is associated with the method 3200 of Figure 28.Each row 3402 is carried out the action of particular offset Canceller all corresponding to the particular value pattern at this particular value pattern.Row 3410 comprise the data of being sampled out and the high value or the low value (" 1 " or " 0 ") of each data in the boundary value series and boundary value.Row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, transformation not necessarily can appear between the data value of row " D1 " and " D2 " in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 3402 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 3412 comprise the alternative boundary value that capable 3402a and 3402b locate at row " E1 ".For row 3402c and 3402d, comprise in row 3412 " X " that this is because for the continuous data value with identical value, do not consider boundary value when taking the action of offset cancellation device.On the contrary, as described in the method 3200, only consider the relative frequency of high data value or low data value as above.For AD HOC, row 3414 comprise specific residual direct current offset level.For AD HOC, row 3416 comprise the specific action that offset cancellation is provided with.Can discussing in the method 3200, apply the action that is provided with at offset cancellation as above.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
In a particular embodiment, equalizer (as the equalizer 42 of Fig. 3) can apply direct current offset to an above signal path (as not change (direct current) component, first order component and two order components of signal) and offset action.Can apply direct current offset to an above signal path offsets the example of the equalizer of action and comprises the second dervative equalizer.In some equalizer, can apply direct current offset to a plurality of signal paths by the mode of setting and offset action.Apply direct current offset counteracting action by the mode of setting and be included in the DC-offset compensation of regulating the direct current offset gain in the path and fixing other paths.Select as another kind, apply direct current offset by the mode of setting and offset action and comprise and utilize specific function that independent control variable is mapped to a plurality of paths, and the DC-offset compensation in these paths is regulated according to this mapping.
It may be disadvantageous by the mode of setting a plurality of signal paths being applied direct current offset counteracting action under specific circumstances.For example, if the residual direct current offset of a plurality of signal paths (promptly in the opposite direction, make at combiner 118 places when these residual direct current offsets are merged, they can be cancelled out each other), then equalizer can not proofreaied and correct the residual direct current offset in each signal path, thereby direct current offset can make the signal component in one or more path saturated.Allow residual direct current offset to persist in the performance degradation that can make equalizer in one or more signal path, for example limit the linear operation scope of equalizer.
Figure 30 and clock signal have contrasted the ground illustration DC path output 3510 that shows negative residual direct current offset in the example first derivative equalizer, the first derivative path output 3520 that shows just residual direct current offset and the example 3500 that mainly shows the equalizer output signal 3530 of zero residual direct current offset.Should be noted that, although with clock signal contrasted ground illustration DC path output 3510 and first derivative path output 3520, receiver can individually not monitor or utilize clock signal to sample to exporting 3510 and 3520 to exporting 3510 and 3520.In illustrated example, equalizer output 3530 be DC path output 3510 and first derivative path output 3520 and.As can be observed, although equalizer output 3530 mainly shows zero overall direct current offset, DC path output 3510 be saturated and shows negative bias and move, and first derivative path output 3520 just shows and is offset.Therefore, by the mode of setting the path is applied direct current offset and offset the equalizer of action and may not can proofread and correct residual bias in each path, thereby residual bias can make the equalizer performance deterioration.Therefore, the equalizer of respectively forming the residual bias in the path that under specific circumstances can offseting signal is favourable.
Should be noted that, equalizer output 3530 may mainly show at border E3 and border E4 place and obtain the overall offset offset, but it may show slight just skew at border E2 place, and this skew with first derivative path output 3520 is in identical polarity.This may be saturated the causing owing to DC path output 3510.In other words, if DC path output 3510 has the polarity phase identical residual bias of value on the contrary with first derivative path output 3520, export 3530 places at equalizer so, they can be located to cancel out each other fully in the border (as at E3 and E4 place) after a plurality of continuous transformations, and this may not be saturated because of DC path output 3510 after a plurality of continuous transformations.On the other hand, even DC path output 3510 has the polarity phase identical residual bias of value on the contrary with first derivative path output 3520, export 3530 places at equalizer, they can be on the border after having several data bit of identical value yet (as at the E2 place) locate to cancel out each other fully, and equalizer output 3530 may be tending towards having with the first derivative path exports 3520 identical polarity skews, this be because DC path output 3510 because the saturation effect (it may occur having continuous a plurality of data bit of identical value after) of DC path output 3510 and may have than first derivative path and export 3520 slightly little offset values.By this mode, can be by selecting boundary value according to the data pattern before the border and from overall equalizer output 3530, detecting such as the residual bias in the single path of first derivative path output 3520.
Figure 31 is the illustration flow chart that is used for exemplary method 3600 that the residual direct current offset of first derivative analog equalizer is offset according to a particular embodiment of the present invention.Under specific circumstances, by one or two path that does not change in DC path and the first derivative path in the first derivative equalizer is applied the offset cancellation action, method 3600 can be offset the residual direct current offset in these two paths.
For example, when before changing, observing continuous data value, can only apply the offset cancellation action to the first derivative path with identical high value or low value.Continuous data value with identical value shows that DC path is saturated (overall offset of signal has roughly been offset in supposition).Therefore, comprise that the value (high or low) of the boundary value between the data value of transformation is probably corresponding to the residual direct current offset in the first derivative path.When before changing, observing continuous a plurality of data value, can apply the offset cancellation action to first derivative path and DC path both with different high values or low value.By applying offset cancellation action in such a way, even do not change DC path and the first derivative routing table reveals rightabout residual bias, method 3600 also can be proofreaied and correct the residual bias in each signal path.
Method 3600 begins at step 3610 place, uses clock signal that output signal is sampled at this step place.Such as above described in conjunction with Figure 3, this output signal can be the output of equalizer, and can sample to this output signal according to clock signal.In a particular embodiment, can sample to output signal at the base value strong point and the boundary point place that determine by clock signal.Select as another kind, can output signal not sampled, and can derive and the corresponding boundary value of these non-sample points at the boundary point place.In a particular embodiment, offset controller 106 can be derived boundary value from data from the sample survey value and other phase informations (for example, the phase place of output signal is morning or late).
At step 3620 place, after to the output signal sampling, can analyze the data value of being sampled out, to determine in these data values, whether having occurred transformation.At step 3630 place, if do not detect transformation, then this method turns back to step 3620.If detect transformation between continuous data value, then this method proceeds to step 3640.Should be pointed out that in a particular embodiment, can received data value relatively be detected transformation mutually by direct.In alternative embodiment, can detect transformation by received data value and boundary value and the predefine binarization mode that comprises transformation (and corresponding to particular offset offset action) are compared.It should also be noted that in a particular embodiment, can after detecting single transformation, carry out the offset cancellation action.
If detect transformation, then the value of the boundary value between the continuous data value that comprises this transformation is discerned at step 3640 place.At step 3650 place, if this boundary value height, then this method proceeds to step 3660.If this boundary value is low, then this method proceeds to step 3690.
At step 3660 place, to before the boundary value 0.5 whether identically with the data value of 1.5 positions determine.If identical, then DC path may be saturated, and the value of this boundary value may reflect the residual direct current offset in the first derivative path.Therefore, if 0.5 data value with 1.5 positions is identical before the boundary value, then method 3600 proceeds to step 3670, and the first derivative path is applied negative bias moves and offset action signal is regulated (is positive because of residual first derivative path offset) downwards.If 0.5 data value with 1.5 positions is different before the boundary value, then method 3600 proceeds to step 3680, and applies negative bias and move and offset action signal is regulated (because the skew of residual equalizer is positive) downwards not changing DC path and first derivative path.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
If boundary value is low at step 3650 place, then method 3600 proceeds to step 3690.At step 3690 place, to before the boundary value 0.5 whether identically with the data value of 1.5 positions determine.If identical, then DC path may be saturated, and the value of this boundary value may reflect the residual direct current offset in the first derivative path.Therefore, if 0.5 data value with 1.5 positions is identical before the boundary value, then method 3600 proceeds to step 3700, and to the first derivative path apply the action of positive offset cancellation with signal to adjusted (bearing) because of residual first derivative path offset.If 0.5 data value with 1.5 positions is different before the boundary value, then method 3600 proceeds to step 3710, and to do not change DC path and first derivative path apply the action of positive offset cancellation with signal to adjusted (because the skew of residual equalizer is born).Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
Should be pointed out that in a particular embodiment, can be by offset controller 106 execution in step 3620 to 3710, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.What should also be pointed out that is, determine to apply the group of paths of migration although utilize relation between the data value of 0.5 and 1.5 position before the boundary value, but also can utilize the relation (for example, the data value of 2.5 positions before the boundary value being taken into account) between any suitable data value.It should also be noted that the equalizer that method 3600 can be promoted to be applied to be associated with the signal path of any suitable quantity.
The table 3800 of the example skew controlling schemes that Figure 32 is an illustration is associated with the method 3600 of Figure 31.Each row 3802 is all corresponding to the particular value pattern, at this particular value pattern (to the first derivative path otherwise to the first derivative path and do not change DC path) carry out the action of particular offset Canceller.Row 3810 comprise each data value in the series of the data value of being sampled out and boundary value and the high value or the low value (" 1 " or " 0 ") of boundary value.Row " D0 " comprise the 0th data from the sample survey value of output signal, and row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D1 " and " D2 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 3802 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 3812 comprise the alternative boundary value that each row 3802 is located at row " E1 ".For AD HOC, row 3814 comprise specific residual direct current offset degree and the direct current offset counteracting action that DC path is associated that do not change with input signal.Can offset action as the above direct current offset that discussion in the method 3600, carries out.For AD HOC, row 3816 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the first derivative path of input signal.Can offset action as the above direct current offset that discussion in the method 3600, carries out.
Figure 33 is the illustration flow chart that is used for another exemplary method 3900 that the residual direct current offset of first derivative analog equalizer is offset according to a particular embodiment of the present invention.Similar with the method 3600 among Figure 31, under specific circumstances, method 3900 can apply offset cancellation and move and offset residual direct current offset in these paths by the arbitrary path in DC path and the first derivative path of not changing to the first derivative equalizer.
In method 3900, when before changing, observing continuous data value, can only apply the offset cancellation action to the first derivative path with identical high value or low value.Continuous data value with identical value shows that DC path is saturated (overall offset of signal has roughly been offset in supposition).Therefore, comprise that the value (high or low) of the boundary value between a plurality of data values of transformation is probably corresponding to the residual direct current offset in the first derivative path.When before changing, observing continuous data value, can only not apply the offset cancellation action to changing DC path with different high values or low value.By applying offset cancellation action in such a way, even do not change DC path and the first derivative routing table reveals rightabout residual bias, method 3900 also can be proofreaied and correct the residual bias in each signal path.
Method 3900 begins at step 3910 place, at this step place, uses clock signal that output signal is sampled.Because step 3910 to 3960 and 3990 can be identical with step 3610 to 3660 and 3690 respectively, therefore step 3910 to 3960 and 3990 is not described in detail.Before step 3960 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 3900 proceeds to step 3970 under the identical situation of these values, and proceeds to step 3980 under the different situation of these values.At step 3970 place, the first derivative path is applied negative bias move the counteracting action signal is regulated (because residual first derivative path offset is positive) downwards.At step 3980 place,, then do not apply negative bias and move and offset action signal is regulated (is positive because residual equalizer is offset) downwards to changing DC path if 0.5 data value with 1.5 positions is different before the boundary value.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
Before step 3990 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 3900 proceeds to step 4000 under the identical situation of these values, and proceeds to step 4010 under the different situation of these values.At step 4000 place, to the first derivative path apply the action of positive offset cancellation with signal to adjusted (because residual first derivative path offset is born).At step 4010 place, if 0.5 data value with 1.5 positions is different before the boundary value, then to do not change DC path apply the action of positive offset cancellation with signal to adjusted (bearing) because residual equalizer is offset.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
It should also be noted that in a particular embodiment, can be by offset controller 106 execution in step 3920 to 4010, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.What should also be pointed out that is, determine to be applied in the group of paths of migration although utilize relation between the data value of 0.5 and 1.5 position before the boundary value, but also can utilize relation between any suitable a plurality of data values (for example, taking into account) with changing before the boundary value data value of 2.5 positions.It should also be noted that the equalizer that method 3900 can be promoted to be applied to be associated with the signal path of any suitable quantity.
The table 4100 of the example skew controlling schemes that Figure 34 is an illustration is associated with the method 3900 of Figure 33.Each row 4102 is carried out the action of particular offset Canceller all corresponding to the particular value pattern at this particular value pattern (to the first derivative path or to not changing DC path).Row 4110 comprise each data value in the series of the data value of being sampled out and boundary value and the high value or the low value (" 1 " or " 0 ") of boundary value.Row " D0 " comprise the 0th data from the sample survey value of output signal, and row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D1 " and " D2 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 4102 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 4112 comprise the alternative boundary value that each row 4102 is located at row " E1 ".For AD HOC, row 4114 comprise specific residual direct current offset degree and the direct current offset counteracting action that DC path is associated that do not change with input signal.Can offset action as the above direct current offset that discussion in the method 3900, applies.For AD HOC, row 4116 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the first derivative path of input signal.Can offset action as the above direct current offset that discussion in the method 3900, applies.
Figure 35 is an illustration the flow chart of an exemplary method 4200 again that the residual direct current offset of first derivative analog equalizer is offset of being used for according to a particular embodiment of the present invention.Similar with the method 3600 among Figure 31, in some cases, method 4200 can apply the offset cancellation action to offset the residual bias in these paths in DC path and the first derivative path one or two of not changing of first derivative analog equalizer.Method 4200 can be done like this by following processing: when observing the continuous data value with identical high value or low value before changing, all do not apply the offset cancellation action to changing DC path and first derivative path.When before changing, observing continuous data value, can only not apply the offset cancellation action to changing DC path with different high values or low value.By applying the offset cancellation action in such a way, method 4200 can be proofreaied and correct the residual bias in each signal path.
Method 4200 begins at step 4210 place, uses clock signal that output signal is sampled at this step place.Because step 4210 to 4260 and 4290 can be identical with step 3610 to 3660 and 3690 respectively, therefore step 4210 to 4260 and 4290 is not described in detail.Before step 4260 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 4200 proceeds to step 4270 under the identical situation of these values, and proceeds to step 4280 under the different situation of these values.At step 4270 place, do not apply the action of positive offset cancellation will not changing DC path to changing DC path, and the first derivative path is applied negative bias move and offset action the first derivative path is regulated (because residual first derivative path offset is positive) downwards to adjusted.Can carry out to adjusted the skew that does not change DC path by the mode opposite, remain on the same level, simultaneously the skew in first derivative path be regulated downwards so that overall offset is proofreaied and correct with the first derivative path.At step 4280 place,, then do not apply negative bias and move and offset action signal is regulated (is positive because residual equalizer is offset) downwards to changing DC path if 0.5 data value with 1.5 positions is different before the boundary value.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
Before step 4290 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 4200 proceeds to step 4300 under the identical situation of these values, and proceeds to step 4310 under the different situation of these values.At step 4300 place, do not apply negative bias and move and offset action and do not regulate downwards changing DC path will change DC path, and to the first derivative path apply positive offset cancellation action with the first derivative path to adjusted (because residual first derivative path offset is born).Can regulate downwards the skew that does not change DC path by the mode opposite, remain on the same level so that overall offset is proofreaied and correct with the first derivative path, simultaneously with the skew in first derivative path to adjusted.At step 4310 place,, then do not apply positive offset cancellation action will not change DC path to adjusted (because the skew of residual equalizer is born) to changing DC path if 0.5 data value with 1.5 positions is different before the boundary value.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
It should also be noted that in a particular embodiment, can be by offset controller 106 execution in step 4220 to 4310, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.What should also be pointed out that is, determine to be applied in the group of paths of migration although utilize relation between the data value of 0.5 and 1.5 position before the boundary value, but also can utilize the relation (for example, the data value of 2.5 positions before the transformation boundary value being taken into account) between any suitable data value.It should also be noted that the equalizer that method 4200 can be promoted to be applied to be associated with the signal path of any suitable quantity.
The table 4400 of the example skew controlling schemes that Figure 36 is an illustration is associated with the method 4200 of Figure 35.Each row 4402 is all corresponding to the particular value pattern, carries out the action of particular offset Canceller at this particular value pattern (to first derivative path and DC path or only to DC path).Row 4410 comprise the data value of being sampled out and the high value or the low value (" 1 " or " 0 ") of data value in the boundary value series and boundary value.Row " D0 " comprise the 0th data from the sample survey value of output signal, and row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D1 " and " D2 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 4402 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 4412 comprise the alternative boundary value that each row 4402 is located at row " E1 ".For AD HOC, row 4414 comprise specific residual direct current offset degree and the direct current offset counteracting action that DC path is associated that do not change with input signal.Can offset action as the above direct current offset that discussion in the method 4200, applies.For AD HOC, row 4416 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the first derivative path of input signal.Can offset action as the above direct current offset that discussion in the method 4200, applies.
Figure 37 is an illustration the flow chart of an exemplary method 4500 again that the residual direct current offset of first derivative analog equalizer is offset of being used for according to a particular embodiment of the present invention.Similar with the method 3600 among Figure 31, in some cases, method 4500 can all apply the offset cancellation action to offset the residual bias in these paths to do not change DC path and the first derivative path of first derivative analog equalizer by bias mode.Method 4500 can be done like this by following processing: when observing the continuous data value with identical high value or low value before changing, DC path and first derivative path are all applied the offset cancellation action, setover on the first derivative path.When before changing, observing continuous data value, can all apply the offset cancellation action, on DC path, setover DC path and first derivative path with different high values or low value.By applying offset cancellation action by this bias mode, even do not change DC path and the first derivative routing table reveals rightabout residual bias, method 4500 also can be proofreaied and correct the residual bias in each signal path.
Method 4500 begins at step 4510 place, uses clock signal that output signal is sampled at this step place.Because step 4510 to 4560 and 4590 can be identical with step 3610 to 3660 and 3690 respectively, therefore step 4510 to 4560 and 4590 is not described in detail.Before step 4560 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 4500 proceeds to step 4570 under the identical situation of these values, and proceeds to step 4580 under the different situation of these values.At step 4570 place, do not apply negative bias and move and offset action and do not regulate downwards changing DC path, and the first derivative path is applied bigger negative bias move and offset action the first derivative path is regulated (because residual first derivative path offset is positive) downwards will change DC path.At step 4580 place, if 0.5 data value with 1.5 positions is different before the boundary value, then the first derivative path is applied negative bias and move and offset action so that the first derivative path is regulated downwards, and apply bigger negative bias and move and offset action and do not regulate (because the skew of residual equalizer is positive) will change DC path downwards not changing DC path.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
Before step 4590 place is to boundary value 0.5 whether identical with the data value of 1.5 positions carried out determining after, method 4500 proceeds to step 4600 under the identical situation of these values, and proceeds to step 4610 under the different situation of these values.At step 4600 place, do not apply the action of positive offset cancellation will not changing DC path to changing DC path to adjusted, and to the first derivative path apply bigger positive offset cancellation action with the first derivative path to adjusted (because residual first derivative path offset is born).At step 4610 place, if 0.5 data value with 1.5 positions is different before the boundary value, then to the first derivative path apply the action of positive offset cancellation with the first derivative path to adjusted, and to do not change DC path apply bigger positive offset cancellation action with DC path to adjusted (because the skew of residual equalizer is born).Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
It should also be noted that in a particular embodiment, can be by offset controller 106 execution in step 4520 to 4610, and can for example use variable gain amplifier 116 to carry out direct current offset and offset action.What should also be pointed out that is, determine to be applied in the group of paths of migration although utilize relation between the data value of 0.5 and 1.5 position before the boundary value, but also can utilize the relation (for example, the data value of 2.5 positions before the boundary value being taken into account) between any suitable data value.It should also be noted that the equalizer that method 4500 can be promoted to be applied to be associated with the signal path of any suitable quantity.
The table 4700 of the example skew controlling schemes that Figure 38 is an illustration is associated with the method 4500 of Figure 37.Each row 4702 is carried out the action of particular offset Canceller all corresponding to the particular value pattern at this particular value pattern (to first derivative path and DC path).Row 4710 comprise the data value of being sampled out and the high value or the low value (" 1 " or " 0 ") of each data value in the boundary value series and boundary value.Row " D0 " comprise the 0th data from the sample survey value of output signal, and row " D1 " comprise the first data from the sample survey value of output signal, and row " D2 " comprise the second data from the sample survey value of output signal, and row " E1 " comprise the boundary value between first data value and second data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D1 " and " D2 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in 104 pairs of each row 4702 of sampler.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E1) (thereby can can't help sampler 104 sample out) according to these data values and phase information.
Row 4712 comprise the alternative boundary value that each row 4702 is located at row " E1 ".For AD HOC, row 4714 comprise specific residual direct current offset degree and the direct current offset counteracting action that DC path is associated that do not change with input signal.Can offset action as the above direct current offset that discussion in the method 4500, applies.For AD HOC, row 4716 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the first derivative path of input signal.Can offset action as the above direct current offset that discussion in the method 4500, applies.
Figure 39 is the illustration flow chart that is used for exemplary method 4800 that the residual direct current offset of second dervative analog equalizer is offset according to a particular embodiment of the present invention.In some cases, method 4800 can apply the offset cancellation action in DC path, first derivative path and the second dervative path of second dervative analog equalizer one or more, to offset the residual bias in these paths.Method 4800 can be done like this by following processing: when observing 3 continuous data values with identical high value or low value before changing, only the second dervative path is applied the offset cancellation action.When observing 2 continuous data values with identical high value or low value before changing, method 4800 can all apply the offset cancellation action to single order and second dervative path.When observing 2 continuous data values with opposite high value or low value before changing, method 4800 can all apply the offset cancellation action in these 3 paths to all.By applying the offset cancellation action in such a way, even the residual bias in these 3 paths is cancelled out each other, method 4800 also can be proofreaied and correct the residual bias in each signal path.
Method 4800 begins at step 4810 place, uses clock signal that output signal is sampled at this step place.Because step 4810 to 4850 can be identical with step 3610 to 3650 respectively, therefore step 4810 to 4850 is not described in detail.Whether is that method 4800 proceeded to step 4860 under the high situation of boundary value, and proceeds to step 4910 under boundary value low (that is, equal'sing " 0 ") situation after high (that is, whether equaling " 1 ") carried out determining at step 4850 place to boundary value.
At step 4860 place, determine whether the data value of 0.5,1.5 and 2.5 position before the boundary value is identical.If these values are identical, then method 4800 proceeds to step 4870, and applies negative bias and move and offset action the second dervative path is regulated (because residual second dervative path offset is positive) downwards.If determine that at step 4860 place these values are inequality, then method 4800 proceeds to step 4880.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
At step 4880 place, determine whether the data value of 0.5,1.5 and 2.5 position before the boundary value is identical.Method 4800 proceeds to step 4890 under the identical situation of these values, perhaps proceed to step 4900 under these values situation inequality.At step 4890 place, in single order and the second dervative path each is all applied negative bias move and offset action so that in single order and the second dervative path each is all regulated (because residual single order and/or second dervative path offset are positive) downwards.At step 4900 place,, then in 3 paths each is all applied negative bias and move and offset action so that in 3 paths each is all regulated (because the skew of residual equalizer is positive) downwards if 0.5 different before the boundary value with the data value of 1.5 positions.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
If determining boundary value at step 4850 place is low, then method 4800 proceeds to step 4910.At step 4910 place, determine whether the data value of 0.5,1.5 and 2.5 position before boundary value is identical.If these values are identical, then method 4800 proceeds to step 4920, and the second dervative path is applied the action of positive offset cancellation, with signal to adjusted (because residual second dervative path offset is born).If it is inequality to determine described data value in step 4910, then method 4800 proceeds to step 4930.Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
At step 4930 place, to before boundary value, 0.5 whether identically with the data value of 1.5 positions determining.If these values are identical, then method 4800 proceeds to step 4940.If these values are inequality, then method 4800 proceeds to step 4950.At step 4940 place, in single order and the second dervative path each all apply positive offset cancellation action with in single order and the second dervative path each all to adjusted (because residual single order and/or second dervative path offset are born).At step 4950 place, if 0.5 different before the boundary value with the data value of 1.5 positions, then in 3 paths each all apply positive offset cancellation action with in 3 paths each all to adjusted (because the skew of residual equalizer is born).Should be pointed out that in a particular embodiment, can discern boundary value by boundary value and data value and predefined pattern (it offsets action corresponding to particular offset) are compared, data value is compared and take the offset cancellation action.
It should also be noted that in a particular embodiment, can be by offset controller 106 execution in step 4820 to 4950, and can for example use variable gain amplifier 116 to apply direct current offset and offset action.What should also be pointed out that is, determine to be applied in the group of paths of migration although utilize relation between the data value of 0.5,1.5 and 2.5 position before the boundary value, but also can utilize the relation (for example, the data value of 3.5 positions before the boundary value being taken into account) between any suitable data value.It should also be noted that the equalizer that method 4800 can be promoted to be applied to be associated with the signal path of any suitable quantity.
The table 5000 of the example skew controlling schemes that Figure 40 is an illustration is associated with the method 4800 of Figure 39.Each row 5002 is all corresponding to the particular value pattern, carries out the action of particular offset Canceller at this particular value pattern (to one or more the group in DC path, first derivative path and the second dervative path).Row 5010 comprise the data value of being sampled out and the high value or the low value (" 1 " or " 0 ") of each data value in the boundary value series and boundary value." X " represents that this value can be " 0 " or " 1 ".Row " D0 " comprise the 0th data from the sample survey value of output signal, row " D1 " comprise the first data from the sample survey value of output signal, row " D2 " comprise the second data from the sample survey value of output signal, row " D3 " comprise the 3rd data from the sample survey value of output signal, and row " E2 " comprise the boundary value between second data value and the 3rd data value.These values are similar to Figure 23 A illustrative those values in the 23C.As can be observed, between the data value of row " D2 " and " D3 " transformation have appearred in each pattern.
Should be pointed out that and to sample and send it to offset controller 106 by the pattern of the value in sampler 104 each row 5002.Select as another kind, offset controller 106 can receive only data from the sample survey value and other phase informations, and can derive certain edges thereof dividing value (for example comprising the boundary value among the row E2) (thereby can can't help sampler 104 sample) according to these data values and phase information.
Row 5012 comprise the alternative boundary value that each row 5002 is located at row " E2 ".For AD HOC, row 5014 comprise specific residual direct current offset degree and the direct current offset counteracting action that DC path is associated that do not change with input signal.Can offset action as the above direct current offset that discussion in the method 4800, applies.For AD HOC, row 5016 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the first derivative path of input signal.Can offset action as the above direct current offset that discussion in the method 4800, applies.For AD HOC, row 5018 comprise specific residual direct current offset degree and the direct current offset counteracting action that is associated with the second dervative path of input signal.Can offset action as the above direct current offset that discussion in the method 4800, applies.
Should be pointed out that in a particular embodiment can be with illustrative embodiment merger in Figure 30 to 40 together.For example, in a particular embodiment, the step 4280 in the method 4200 and 4310 can be replaced to the step 3680 and 3710 in the method 3600 respectively.In other embodiments, by identical migration being imposed on second dervative path and first derivative path, method 3600,3900,4200 or 4500 can be applied to the second dervative analog equalizer, this is because may be difficult to employing method 4800 is distinguished each path effectively between first derivative path and second dervative path residual bias, and may become out of control at the independent offset control in single order and second dervative path.In such an embodiment, if the residual bias in first derivative path and the second dervative path has opposite polarity, then these residual bias may not offset fully.Yet in a particular embodiment, and employing method 4800 allows their change at random and owing to can not effectively detecting the wild phase ratio that becomes to each residual bias between first derivative path and the second dervative path, they are bound may be more useful.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
Discuss like that in conjunction with Figure 11 as above, Duty Cycle Distortion may influence at periodically or the adaptive gain of quasi periodicity data sequence controls.Duty Cycle Distortion also may influence the offset cancellation control at periodicity or quasi periodicity data sequence.If the cycle of data is even number data values, then may be owing to Duty Cycle Distortion seriously to " early " or " evening " phase deviation to the transformation in the data sequence reigning " idol " or " very " boundary value.Because this departs from, the equalizer compensation may also can might make adaptive gain control and skew control exceed acceptable operating condition for example towards the deviation in driction that increases or reduce the gain or the skew of signal.
Discuss like that in conjunction with Figure 12 to 22 as above, can select AD HOC as filter mode with negative effect that reduces Duty Cycle Distortion and the result that unanimity is provided between (standard) cyclical signal.In a particular embodiment, these filter mode may be special-purpose for specific (standard) cyclical signal.As discussed above, use specific (standard) cyclical signal a shortcoming of special-purpose filter mode be that their applicability is restricted.For example, in other (standard) cyclical signals, these filter mode may not be to be distributed in about equally between even data sequence and the odd data sequence, if therefore use them may cause unacceptable operating condition at these other (standard) cyclical signal
Use specific (standard) cyclical signal another latent defect of special-purpose filter mode be that these filter mode may not be to be distributed in about equally between even sequence and the odd sequence when adopting the double sampling technology.Here, the double sampling technology is meant " vectorial double sampling technology ", it begins to carry out " vector sampling " termly by the speed littler than naked channel speed, but each " by double sampling " vector representation by naked channel speed continuously to one section length that sampling is identical with filter mode of data and boundary value.Notice that if " double sampling " period is shorter than the length of filter mode, then " by double sampling " vector can overlap each other.When carrying out double sampling with mutually harmonious cycle of (standard) cyclical signal, pattern can occur by the different probability of the probability that can occur in whole (standard) cyclical signal with pattern.
The above certain methods that the distortion that solves these types has been discussed in conjunction with Figure 17 to 22.For example, can be from static or dynamic useful filter mode tabulation selective filter pattern and use these filter mode sequentially, randomly or side by side by balance mode, rather than select them particularly at specific (standard) cyclical signal.Yet, the additive method that has the negative effect be used to reduce Duty Cycle Distortion and/or double sampling and (standard) cyclical signal, these methods can be used as the alternate technologies of filter mode technology discussed above, perhaps can use these methods together in conjunction with filter mode technology discussed above.
The flow chart of the exemplary method 5100 of the influence that is used to reduce Duty Cycle Distortion according to a particular embodiment of the present invention that Figure 41 is an illustration.Method 5100 is by monitoring and take adaptive gain control action and/or offset cancellation to move to reduce by balance mode the negative effect of Duty Cycle Distortion and/or double sampling and (standard) cyclical signal between even data sequence (it begins with even data, then is odd data, even data, odd data or the like) and odd data sequence (it begins with odd data, then is even data, odd data, even data or the like).Can be combined in the above filter mode technology of discussing in conjunction with Figure 17 to 22 using method 5100 together, perhaps method 5100 can be used as the alternate technologies of these filter mode technology.
Method 5100 begins at step 5110 place.At step 5110 place, logic (for example, receiver logic 47) receives the signal of the arrival that comprises even data and odd data successively.The signal of Dao Laiing can be (standard) cyclical signal in a particular embodiment.At the condition that will take control action, this logic is monitored the even data sequence that begins with even data (rather than the odd data sequence that begins with odd data).In a particular embodiment, this logic can utilize one or more filter mode that the even data sequence that begins with even data is monitored.Although should be pointed out that in method 5100 at first the antithesis data sequence is monitored, in alternative embodiment, can at first monitor to the odd data sequence.
At step 5120 place, to take the condition of control action to determine to whether having detected.Select as another kind, can determine whether having detected corresponding to the specific data pattern of certain filter mode.If do not detect this condition as yet, then method 5100 turns back to step 5110, and this logic continues at this condition the even data sequence that begins with even data to be monitored then.If detected this condition, then method 5100 proceeds to step 5130.At step 5130 place, take first control action.This first control action can be adaptive gain control action and/or offset cancellation action.First control action can perhaps can adopt other suitable technique such as the conventional adaptive control algorithm that comprises lowest mean square (LMS) algorithm, symbol one symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm for example based on as above sampling boundary value and/or one or more data values in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 descriptions.
After having taked first control action, method 5100 proceeds to step 5140.At step 5140 place, at the condition that will take control action, this logic is monitored the odd data sequence that begins with odd data (rather than the even data sequence that begins with even data).In a particular embodiment, this logic can utilize one or more filter mode that the odd data sequence that begins with odd data is monitored.At step 5150 place, to take the condition of control action to determine to whether having detected.Select as another kind, can determine whether having detected corresponding to the specific data pattern of certain filter mode.If do not detect this condition as yet, then method 5100 turns back to step 5140, and this logic continues at this condition the odd data sequence that begins with odd data to be monitored then.If detected this condition, then method 5100 proceeds to step 5160.At step 5160 place, take second control action.This second control action can be adaptive gain control action and/or offset cancellation action.Second control action can perhaps can adopt other suitable technique such as the conventional adaptive control algorithm that comprises lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm for example based on as above sampling boundary value and/or one or more data values in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 descriptions.After having taked second control action, method 5100 turns back to step 5110.Monitor by antithesis data sequence and odd data sequence successively, method 5100 makes " early " or " evening " the phase deviation balance that causes owing to Duty Cycle Distortion, and reduces the negative effect of Duty Cycle Distortion and/or double sampling (standard) cyclical signal.
In a particular embodiment, can arrive the cycle of (standard) cyclical signal to avoid phase locking in conjunction with randomizer technology using method together 5100, thereby avoid other possible distortions.The randomizer technology for example comprises stochastic filtering device model selection embodiment and method discussed below 5300 and 5400 of method 2300 discussed above.And, as discussed above, can the combining adaptive gain controlling and/or offset cancellation control using method 5100 together, and can using method 5100 as the alternate technologies of filter mode technology discussed above or in conjunction with this filter mode technology using method 5100 together.
The flow chart of another exemplary method 5200 of the influence that is used to reduce Duty Cycle Distortion according to a particular embodiment of the present invention that Figure 42 is an illustration.Method 5200 begins at step 5210 place.At step 5210 place, logic (for example, receiver logic 47) receives the signal of the arrival that comprises even data and odd data successively.The signal of Dao Laiing can be (standard) cyclical signal in a particular embodiment.This logic selects even data sequence (it begins, then is odd data, even data, odd data or the like with even data) or odd data sequence (it begins, then is even data, odd data, even data or the like with odd data) to monitor by the probability that equates randomly.In a particular embodiment, this logic for example can generate a random number (for example " 1 " or " 0 ") and select even data sequence or odd data sequence based on the value of this random number.
At step 5220 place, still be that the odd data sequence is determined to having selected the even data sequence.For example, can carry out this based on the value of the random number that is generated determines.If determine to have selected the even data sequence, then method 5200 proceeds to step 5230.If determine to have selected the odd data sequence, then method 5200 proceeds to step 5260.
If determine to have selected the even data sequence, then at the condition that will take control action the even data sequence that begins at the even data that receives (rather than the odd data sequence that begins at the odd data that receives) is monitored in this logic of step 5230 place.In a particular embodiment, this logic can utilize one or more filter mode that the even data sequence that begins with even data is monitored.
At step 5240 place, to take the condition of control action to determine to whether having detected.Select as another kind, can determine whether having detected corresponding to the specific data pattern of certain filter mode.If do not detect this condition as yet, then method 5200 turns back to step 5230, and this logic continues at this condition the even data sequence that begins with even data to be monitored then.If detected this condition, then method 5200 proceeds to step 5250.At step 5250 place, take control action.This control action can be adaptive gain control action and/or offset cancellation action.This control action can perhaps can adopt other suitable technique such as the conventional adaptive control algorithm that comprises lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm for example based on as above sampling boundary value and/or one or more data values in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 descriptions.After having taked control action, method 5200 turns back to step 5210.
If selected the odd data sequence at step 5220 place, then method 5200 proceeds to step 5260.At step 5260 place, this logic is monitored the odd data sequence that begins with odd data (rather than the even data sequence that begins with even data) at the condition that will take control action.In a particular embodiment, this logic can utilize one or more filter mode that the odd data sequence that begins with odd data is monitored.
At step 5270 place, to take the condition of control action to determine to whether having detected.Select as another kind, can determine whether having detected corresponding to the specific data pattern of certain filter mode.If do not detect this condition as yet, then method 5200 turns back to step 5260, and this logic continues at this condition the odd data sequence that begins with odd data to be monitored then.If detected this condition, then method 5200 proceeds to step 5280.At step 5280 place, take control action.This control action can be adaptive gain control action and/or offset cancellation action.This control action can perhaps can adopt other suitable technique such as the conventional adaptive control algorithm that comprises lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm for example based on as above sampling boundary value and/or one or more data values in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 descriptions.After having taked control action, method 5200 turns back to step 5210.
By one in the data sequence of selecting to begin with random fashion with even data or odd data by the probability that equates, method 5200 can make " early " or " evening " phase deviation that causes owing to Duty Cycle Distortion weighing apparatus that flattens, and (especially under long-term behavior) reduces the negative effect of Duty Cycle Distortion and/or double sampling and (standard) cyclical signal.Yet, under short-term conditions, method 5200 is being not so good as method 5100 effective (for example, because employing method 5200 may be chosen same even data sequence or odd data sequence continuously at random) aspect the negative effect that reduces Duty Cycle Distortion and/or double sampling and (standard) cyclical signal.Yet the advantage that method 5200 surpasses method 5100 is that the selection at random of 5200 pairs of data sequences of method avoided the cycle of phase locking to (standard) cyclical signal, thereby has reduced other possible distortions.Should be pointed out that can the combining adaptive gain controlling and/or offset cancellation control using method 5200 together.In addition, can using method 5200 as the alternate technologies of filter mode technology discussed above, perhaps in conjunction with these filter mode technology using method 5200 together.
Should be noted that, except based on as the adaptive gain control and/or offset cancellation control of the above sampling boundary values of describing in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 and/or one or more data value, method 5100 and 5200 can also be applied to utilize any other suitable control system of sampler output, to reduce the negative effect of Duty Cycle Distortion and/or double sampling and (standard) cyclical signal.For example, in a particular embodiment, these methods can be applied to based on conventional adaptive equalizer control such as the conventional algorithm of lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm.In a particular embodiment, these methods can also be applied to export clock and data recovery (CDR) system of metering needle to the recovered clock of sampler based on sampler.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
As discussed above, the double sampling technology is meant " vectorial double sampling technology ", it begins to carry out " vector sampling " termly by the speed littler than naked channel speed, but each " by double sampling " one piece of data and boundary value that vector representation samples continuously by naked channel speed, the length of this section is identical with the length of filter mode.Because the double sampling cycle may be locked in by the cycle of (standard) cyclical signal of double sampling, therefore observed data pattern may be different with the data pattern in whole (standard) cyclical signal in the data that double sampling goes out.For example, if be that 320 cyclical signal is carried out double sampling to the cycle, will in each cycle, repeatedly carry out double sampling to this cyclical signal so, and carry out double sampling at other 310 some places never at 10 identical some places by 1/32 speed.Only sample and may make the control action of equalizer execution become deflection in the part point place in the whole points in (standard) cyclical signal.
The solution that locking took place between cycle of double sampling cycle and (standard) cyclical signal of contending with is to change the point that (standard) cyclical signal is carried out double sampling in the cycle at each double sampling.For example, if carry out double sampling by 1/32 speed, then the double sampling cycle is 32 thereby has 32 possible points that (standard) cyclical signal carried out double sampling in the cycle at each double sampling.Can select so that it changed with each 32 double sampling cycle the point that double sampling takes place.
Should be pointed out that before the double sampling that double sampling carried out in the cycle can comprise the specified point in this cycle for example, afterwards or the processing of sampling of a plurality of data bit on every side.For example, if adopting 1/32 double sampling speed and double sampling to handle to comprise carries out double sampling to 6 positions, then can be in each 32 bit period before specified point, 6 positions of sampling afterwards or on every side.Below in conjunction with Figure 43 to 45 pair of exemplary method that is used for changing the point that double sampling takes place in each double sampling cycle.
The alternative solution that locking took place between cycle of double sampling cycle and (standard) cyclical signal of contending with is to select (standard) cyclical signal is carried out the next point of double sampling under the situation that is not limited to the double sampling cycle of having set at random.For example, selecting randomly and sampling out after first, selecting randomly and sample down a bit, this point might not be limited in double sampling in the cycle, and the rest may be inferred.In a particular embodiment, can use pseudorandom number generator, and can be weighted this pseudorandom number generator in any suitable way, make to produce average double sampling speed.For example, in a particular embodiment, can make the point of next selected at random can not exceed certain units of lighting in the past to this pseudorandom number generator capping.In such an embodiment, this upper limit can play the effect that limits average double sampling speed.In alternative embodiment, can under selecting under the various constraints a bit make double sampling speed always less than certain maximum double sampling speed.
Double sampling cycle and (standard) of contending with an alternative solution again that locking takes place between cycle of cyclical signal is based on that the random number that generates at each sample point is fixed by certain or but changeable probability is taked control action randomly, rather than double sampling cycle of selecting double sampling point or double sampling to circulate at random.In a particular embodiment, can use pseudorandom number generator, and and if only if the pseudo random number that is generated just adopts sampling when dropping in certain scope.The probability of taking control action can be fixing or variable.In a particular embodiment, in case adopted sampling, but then can be in certain period or before certain time point changeable probability be set at zero, being less than certain value with maximum double sampling rate limit.In other embodiments, when not adopting sampling, but can be in time through increasing changeable probability gradually, but and in case carried out control action just with the changeable probability reset-to-zero or be fixed to little numerical value, so that mean sample speed and/or minimum sampling rate are restricted to greater than certain value.
Figure 43 has been the illustration flow chart that is used for changing the exemplary method 5300 of the point that double sampling takes place according to a particular embodiment of the present invention in each double sampling circulation.Method 5300 selects double sampling point to change this point by (pressing the probability that equates usually) randomly at each double sampling circulation.Method 5300 begins at step 5310 place, and the double sampling point is selected in circulation randomly at double sampling at this step place.Can for example use pseudorandom number generator to select this double sampling point randomly.If for example carry out double sampling by 1/32 speed, then this double sampling of selecting at random point can be any one in 32 points in this double sampling circulation.
At step 5320 place, signal is sampled at the double sampling point place that is chosen.As discussed above, the processing of sampling at the double sampling point place that chooses can comprise to before the point of being chosen, afterwards or the processing of sampling of a plurality of data bit on every side.For example, in a particular embodiment, 6 positions of can sampling, and first can be corresponding to the point of being chosen.Should be pointed out that the sampling of carrying out at the double sampling point place that is chosen can also can not make control action come into force.Appearance transformation in this sampling for example,, control action come into force, and if then can not make control action come into force if in this sampling, transformation occurred.As another example, in the specific embodiment that has used filter mode,, then can not take control action if in this sampling, do not observe the suitable filters pattern.If in this sampling, observe certain filter mode, then can take control action.The double sampling of choosing some place in circulation has carried out after the sampling signal, and method 5300 turns back to step 5310, and selects new double sampling point at random at next circulation.By this mode, can avoid double sampling cycle and the cycle of (standard) cyclical signal that any locking takes place.
Figure 44 has been the illustration flow chart that is used for changing another exemplary method 5400 of the point that double sampling takes place according to a particular embodiment of the present invention in each double sampling circulation.At being connected on the double sampling circulation double sampling circulation afterwards of having taked control action, method 5400 is by selecting the double sampling point to change the double sampling point randomly by the probability that equates usually.Method 5400 begins at step 5410 place, and the double sampling point is selected in circulation randomly at double sampling at this step place.Can for example use pseudorandom number generator to select this double sampling point randomly.If for example carry out double sampling by 1/32 speed, then this double sampling of selecting at random point can be any one in 32 points in this double sampling circulation.
At step 5420 place, signal is sampled at the double sampling point place that is chosen.As discussed above, the processing of sampling at the double sampling point place that chooses can comprise to before the point of being chosen, afterwards or the processing of sampling of a plurality of data bit on every side.The sampling of carrying out at the double sampling point place that is chosen can also can not make control action come into force.Appearance transformation in this sampling for example,, control action come into force, and if then can not make control action come into force if in this sampling, transformation occurred.As another example, in the specific embodiment that has used filter mode,, then can not take control action if in this sampling, do not observe the suitable filters pattern.If in this sampling, observe certain filter mode, then can take control action.
At step 5430 place, to whether having taked control action to determine.If not, then method 5400 turns back to step 5420, and samples to signal in the double sampling of the being chosen some place in next circulation.If determine at step 5430 place and to take control action (in the circulation of first or double sampling subsequently), then method 5400 turns back to step 5410, and selects new double sampling point randomly.By this mode, can avoid double sampling cycle and the cycle of (standard) cyclical signal that any locking takes place.
Should be pointed out that in a particular embodiment,, can take control action by different number of times for gain controlling and offset cancellation.Take control action can cause selecting the speed of double sampling point different by different number of times for gain controlling and offset cancellation.In a particular embodiment, when having regulated gain or skew, can reset the double sampling point of choosing at gain controlling and offset cancellation.
Figure 45 has been the illustration flow chart of an exemplary method 5500 again that is used for changing the point that double sampling takes place according to a particular embodiment of the present invention in each double sampling circulation.At being connected on the double sampling circulation double sampling circulation afterwards of having taked control action, method 5500 is passed whole double sampling point tabulation change double sampling point by circulation sequentially.Method 5500 begins at step 5510 place, next the double sampling point in this tabulation of selection is circulated at this step place at double sampling.Notice that this double sampling point tabulation is not necessarily sequenced.For example, if be arranged in the 13rd the some place that this double sampling circulates, then can select next double sampling point at the 4th some place in this double sampling circulation at step 5510 place by 1/32 speed execution double sampling and last double sampling point.
At step 5520 place, signal is sampled at the double sampling point place that is chosen.As discussed above, sample at the double sampling point place that chooses can comprise to before the point of being chosen, afterwards or the processing of sampling of a plurality of data bit on every side.The sampling of carrying out at the double sampling point place that is chosen can also can not make control action come into force.Appearance transformation in this sampling for example,, control action come into force, and if then can not make control action come into force if in this sampling, transformation occurred.As another example, in the specific embodiment that has used filter mode,, then can not take control action if in this sampling, do not observe the suitable filters pattern.If in this sampling, observe certain filter mode, then can take control action.
At step 5530 place, to whether having taked control action to determine.If not, then method 5500 turns back to step 5520, and samples to signal in the double sampling of the being chosen some place in next circulation.If determine at step 5530 place and to have taked control action (in the circulation of first or double sampling subsequently), then method 5500 turns back to step 5510, and selects next double sampling point.The circulation that tabulation is carried out to whole double sampling point may be introduced the locking of another level: travel through between cycle of whole cycle of this double sampling tabulation and (standard) cyclical signal and lock.Yet,, therefore can reduce the possibility that double sampling is handled and take place between the cycle of (standard) cyclical signal to lock effectively owing to the whole period ratio double sampling cycle of this tabulation of traversal is much longer.
Should be pointed out that in a particular embodiment,, can take control action by different number of times for gain controlling and offset cancellation.Take control action can cause selecting the speed of double sampling point different by different number of times for gain controlling and offset cancellation.In a particular embodiment, when having regulated gain or skew, can reset the double sampling point of choosing at gain controlling and offset cancellation.
It should also be noted that method 5100 comprises the specific embodiment of method 5500 (wherein carrying out double sampling by 1/2 speed), and method 5200 comprises the specific embodiment of method 5400 (wherein carrying out double sampling by 1/2 speed).
It should also be noted that in a particular embodiment, can method 5300,5400 and 5500 be combined by various forms.For example, if carry out double sampling, then 32 possible double sampling points minute level ground can be divided into 8 groups (each group has 4 double sampling points) by 1/32 speed.In a particular embodiment, can adopt method 5300 to select in 8 possible groups of double sampling point one, and employing method 5400 is selected one in 4 possible double sampling points in each group.
What should also be pointed out that is, except based on as the adaptive gain control action and/or offset cancellation control of the above sampling boundary values of describing in conjunction with Fig. 5 to 10,24 to 29 and 31 to 40 and/or one or more data value, in a particular embodiment, method 5300,5400 and 5500 can be applied to utilize any other suitable control system of sampler output, with prevent or alleviate that double sampling is handled and (standard) cyclical signal between locking relation.For example, in a particular embodiment, these methods can be applied to based on conventional adaptive equalizer control such as the conventional algorithm of lowest mean square (LMS) algorithm, symbol-symbol lowest mean square (SS-LMS) algorithm and ZF (ZF) algorithm.In a particular embodiment, these methods can also be applied to export clock and data recovery (CDR) system of metering needle to the recovered clock of sampler based on sampler.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
As discussed above, equalizer can utilize two or more control loops simultaneously so that signal equalization.For example, equalizer can utilize adaptive equalizer to control to regulate gain and reduce residual intersymbol interference.Equalizer can also use the offset cancellation device to regulate skew simultaneously and offset residual bias.
A challenge adopting a plurality of control loops to cause simultaneously is the contingent coupling between these a plurality of control loops.Coupling between a plurality of control loops may make convergence time postpone or even make the control loop instability.For example, if gain is optimum optimum but residual bias is owed, then boundary value might depart to high value or low value (for example,, then be the high value if residual bias is positive, and if residual bias is born, then be low value).The equalizer gain controlling may become overcompensation or undercompensation situation with boundary value and/or other information distortions that departs from.If at the high data value of equalizer gain controlling and the counting of low data value is unbalanced, then the misunderstanding to overcompensation or undercompensation situation also is unbalanced, and the equalizer gain is shifted to owing optimum from optimum.By similar mode, optimum if gain is owed, then residual bias can shift to owing optimum from optimum.
In a particular embodiment, insensitive each other by making a plurality of control loops, can be with these loop decoupling zeros.For example, can make adaptive equalizer control insensitive, and can make the offset cancellation device insensitive residual intersymbol interference to residual bias.In order to make adaptive equalizer control insensitive mutually with the offset cancellation device, in a particular embodiment, adaptive equalizer control and offset cancellation device can use two groups of complementary data patterns by unbalanced mode.The certain bits place that the complementary data pattern can for example be included in the data pattern has those patterns of the data value (for example, " 0 " or " 1 ") of different value.
For example, before control action is based on the boundary value between the continuous data value that comprises transformation and this boundary value under the situation of the comparison of the data value of 1.5 positions, the quantity of the control action of taking when adaptive equalizer control and offset cancellation device can make the data value of 1.5 positions before this boundary value high or low weighing apparatus that flattens.In a particular embodiment, adaptive equalizer control and offset cancellation device can be done like this by following processing: alternately use to have the high data value that was right after before comprising the data value of transformation or the filter mode of low data value in filter mode.By this mode, adaptive equalizer control is become to residual bias insensitive (or more insensitive), and the offset cancellation device is become to residual intersymbol interference insensitive (or more insensitive), thereby with a plurality of control loop decoupling zeros.Further describe specific embodiment below in conjunction with Figure 46 and 47 with a plurality of control loop decoupling zeros.
Figure 46 has been the illustration flow chart that is used for the exemplary method 5600 of a plurality of control loop decoupling zeros according to a particular embodiment of the present invention.Method 5600 can be for example by using two groups of complementary data patterns with adaptive equalization controller and the decoupling zero of offset cancellation device by balance mode.Method 5600 can be used two groups of complementary data patterns by balance mode by replacing between following two processing: the signal that arrives is monitored and take control action based on this group at a group; And then the signal that arrives is monitored and take control action based on this another group at another group.Wherein, for example, control action is based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value, first group of data pattern can be included in before this boundary value those data patterns that 1.5 positions have a high data value (promptly, those data patterns that have the high data value that was right after before comprising the data value of described transformation), 1.5 positions had those data patterns (vice versa) of low data value before second group of data pattern can be included in this boundary value.
Method 5600 begins at step 5610 place.At this step place, logic (for example, receiver logic 47) is monitored the signal that arrives at first group of data pattern.For example, this logic can be monitored the signal that arrives at the continuous data value that comprises transformation (data value that wherein was right after before this continuous data value comprises low value).In a particular embodiment, this logic can use filter mode to carry out this monitoring.The suitable filters pattern for example can comprise above AD HOC in conjunction with Fig. 6,8,10,14,16,32,34,36,38 and 40 illustrations and description.At step 5620 place, if detect first group of data pattern in the data pattern, then method 5600 proceeds to step 5630.If do not detect first group of data pattern in the data pattern, then method 5600 turns back to step 5610, and this logic continues at first group of data pattern the signal that arrives to be monitored then.
At step 5630 place, after detecting first group of data pattern in the data pattern, take suitable control action.In a particular embodiment, this control action can be for example based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value.This control action can be adaptive equalizer action and/or offset cancellation action.This control action can also be based on conventional adaptive control algorithm in a particular embodiment.After having taked control action, method 5600 proceeds to step 5640.
At step 5640 place, this logic is monitored the signal that arrives at second group of data pattern.For example, this logic can be monitored the signal that arrives at the continuous data value that comprises transformation (data value that wherein was right after before this continuous data value comprises high value).In a particular embodiment, this logic can use filter mode to carry out this monitoring.The suitable filters pattern for example can comprise above AD HOC in conjunction with Fig. 6,8,10,14,16,32,34,36,38 and 40 illustrations and description.At step 5650 place, if detect second group of data pattern in the data pattern, then method 5600 proceeds to step 5660.If do not detect second group of data pattern in the data pattern, then method 5600 turns back to step 5640, and this logic continues at second group of data pattern the signal that arrives to be monitored then.
At step 5660 place, after detecting second group of data pattern in the data pattern, take suitable control action.In a particular embodiment, this control action can be for example based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value.This control action can be adaptive equalizer action and/or offset cancellation action.This control action can also be based on conventional adaptive control algorithm in a particular embodiment.After having taked control action, method 5600 turns back to step 5610.By use two groups of complementary data patterns by balance mode, method 5600 can be with adaptive equalizer control and the decoupling zero of offset cancellation device.Should be pointed out that for fear of the alternate cycles of two groups of data patterns of use and the cycle of (standard) cyclical signal to lock, can use randomization balancer (for example, seeing above method 5300 to 5400) simultaneously.
Figure 47 has been the illustration flow chart that is used for another exemplary method 5700 of a plurality of control loop decoupling zeros according to a particular embodiment of the present invention.Method 5700 can be for example by using two groups of complementary data patterns with adaptive equalization control and the decoupling zero of offset cancellation device by balance mode.Method 5700 can be by using this two groups of complementary data patterns in following processing and by balance mode: select in described two groups one group randomly by the probability that equates, at the group of being chosen the signal that arrives is monitored, take control action based on the group of being chosen, then select in described two groups one group randomly by the probability that equates again.Wherein, for example, control action is based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value, first group of data pattern can be included in before this boundary value those data patterns that 1.5 positions have a high data value (promptly, those data patterns that have the high data value that was right after before comprising the data value of transformation), 1.5 positions had those data patterns (vice versa) of low data value before second group of data pattern can be included in this boundary value.
Method 5700 begins at step 5710 place, and at this step place, logic (for example, receiver logic 47) is usually selected in two groups of complementary data patterns one by the probability that equates.This logic can be for example by using pseudorandom number generator and one in a plurality of numbers that generated being associated with a group and in a plurality of numbers that generated another is associated with another group, select in these two groups randomly.After one in having selected these two groups of complementary data patterns, method 5700 proceeds to step 5720.
At step 5720 place, if the data pattern group of being chosen is first group (for example, comprising the data pattern that is right after the low data value before comprising the data value of transformation), then method 5700 proceeds to step 5730.If the data pattern group of being chosen is second group (for example, comprising the data pattern that is right after the high data value before comprising the data value of transformation), then method 5700 proceeds to step 5760.
At step 5730 place, this logic is monitored the signal that arrives at first group of data pattern.For example, this logic can be monitored the signal that arrives at the continuous data value that comprises transformation (data value that wherein was right after before this continuous data value comprises low value).In a particular embodiment, this logic can use filter mode to carry out this monitoring.The suitable filters pattern for example can comprise above AD HOC in conjunction with Fig. 6,8,10,14,16,32,34,36,38 and 40 illustrations and description.
At step 5740 place, if detect first group of data pattern in the data pattern, then method 5700 proceeds to step 5750.If do not detect first group of data pattern in the data pattern, then method 5700 turns back to step 5730, and this logic continues at first group of data pattern the signal that arrives to be monitored then.At step 5750 place, after detecting first group of data pattern in the data pattern, take suitable control action.In a particular embodiment, this control action can be for example based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value.This control action can be adaptive equalizer action and/or offset cancellation action.In a particular embodiment, this control action can also be based on conventional adaptive control algorithm.After having taked control action, method 5700 turns back to step 5710, and selects in two groups of complementary data patterns one group randomly.
If the data pattern group of choosing in step 5710 and 5720 places is second group (for example, comprising the data pattern that is right after the high data value before comprising the data value of described transformation), then method 5700 proceeds to step 5760.At step 5760 place, this logic is monitored the signal that arrives at second group of data pattern.For example, this logic can be monitored the signal that arrives at the continuous data value that comprises transformation (data value that wherein was right after before this continuous data value comprises high value).In a particular embodiment, equalizer can use filter mode to carry out this monitoring.The suitable filters pattern for example can comprise above AD HOC in conjunction with Fig. 6,8,10,14,16,32,34,36,38 and 40 illustrations and description.
At step 5770 place, if detect second group of data pattern in the data pattern, then method 5700 proceeds to step 5780.If do not detect second group of data pattern in the data pattern, then method 5700 turns back to step 5760, continues then at second group of data pattern the signal that arrives to be monitored.At step 5780 place, after detecting second group of data pattern in the data pattern, take suitable control action.In a particular embodiment, this control action can be for example based on the comparison to the data value of 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value.This control action can be adaptive equalizer action and/or offset cancellation action.After having taked control action, method 5700 turns back to step 5710, and selects in two complementary data patterns one randomly.By use two groups of complementary data patterns by balance mode, method 5700 can be with adaptive equalizer control and the decoupling zero of offset cancellation device.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated or discrete to the part of described system and method according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
As discussed above, logic (for example receiver logic 47) can be regulated gain and/or the skew that the signal that arrives is applied after detecting particular data and boundary value.In a particular embodiment, can apply gain and/or skew according to the start stop mode controlling schemes.
In the start stop mode controlling schemes, based on get in two states " height " or " low " one binary target variable (for example, binary form ISI degree, EQ degree or residual bias in more than describing, or the residual amplitude error of the binary form in automatic gain control (AGC) system) (the " high " state that wherein may cause this target variable owing to " height " value of control variables, and the " low " state that may cause this target variable) regulate control variables (for example, gain or skew) owing to " low " value of control variables.In this case,, then can reduce control variables if target variable is represented " high " state, and if target variable is represented " low " state, then can increase control variables.
In conventional start stop mode control system (for example automatic gain control (AGC) system), by symmetric mode to control variables (for example amplifier gain) upgrade, make the increase of control variables have the identical value that reduces with control variables, this is because binary target variable (for example residual amplitude error of binary form) is only carried qualitative information.Therefore, when control variables being applied the increase of equal number and reducing, control variables can remain on identical level on average, and control system can reach poised state.Should be noted that, if (pressing and the above identical mode of ISI level in conjunction with Fig. 6 description) (for example distributes numerical value to the binary target variable, with " height " and " low " state corresponding "+1 " and " 1 "), then in conventional start stop mode control system under poised state the mean value of this binary target variable can converge to zero.
The binary target variable that has zero mean in particular system (for example AGC system) under poised state may be desirable.For example, if binary target variable (for example residual amplitude error of binary form) is to derive out according to the output of the comparator that directly simulated target variable (for example amplifier output amplitude) and controlled target (for example target level of amplitude) is compared, then the optimum mean value of this binary target variable under poised state may be exactly zero naturally, can generate equal number "+1 " and " 1 " output because expect this comparator most when the simulated target variable approaches controlled target.This situation is quite general in the application of conventional start stop mode controlling schemes, and therefore, conventional start stop mode controlling schemes is utilized simply the symmetry of control variables is upgraded.
On the other hand, it is may be not necessarily optimum to have a binary target variable of zero mean in particular system under poised state.The optimum mean value of the ISI degree of describing in conjunction with Fig. 6 for example, may be greater than or less than zero according to various conditions (for example signal of channel loss and arrival itself).In a particular embodiment, optimum average ISI degree may be for high loss channel height, and low for the low-loss channel.And only as example, optimum average ISI degree may from-0.6 to+0.5 variation.For another example, under the optimal conditions that depends on the various systematic errors (for example unmatched skew in border sampler rather than sampling of data device) of residual bias in measuring, more than the residual bias described in conjunction with Figure 25 may on statistics, tend to plus or minus.Therefore, the non-zero mean of binary target variable under poised state may be useful under specific circumstances.
In a particular embodiment, by as illustrative in the following formula group, control variables being introduced asymmetric renewal, the mean value of binary target variable (for example ISI degree) under poised state in the start stop mode control system is become be not equal to zero.Yet, should be pointed out that and can use asymmetric start stop mode controlling schemes down at any suitable control variables (for example, if best residual bias is not zero, at skew control) and at any suitable environment (and being not only described environment).In following formula, K pAnd K nBe respectively the control step value that increases and reduce control variables (for example equalizer gain), N pAnd N nBe respectively that time per unit is to the number of times of action up and down of control variables under poised state, A is the mean value of binary target variable under poised state.The binary target variable is got "+1 " or " 1 " value respectively when here, supposing in control variables to be " height " or " low ".Should be pointed out that N pAnd N nAlso be respectively the quantity that time per unit has the binary target variable of " low " and " high " state.As can be observed, K under long-term behavior pWith N pThe long-pending K that equals nWith N nLong-pending, this is because control variables should not change under poised state.It should also be noted that by making K pWith K nDifference can make A be not equal to zero.For example, in a particular embodiment, work as K pBe 0.3 and K nBe 0.2 o'clock, A can be 0.2.Should be pointed out that A can have from-1 (works as K p=0 and K n＞0 o'clock) (work as K to+1 p＞0 and K n=0 o'clock) any value.It should also be noted that and work as K p=K nA becomes zero during＞0 (this is the situation of conventional start stop mode controlling schemes).
K p×N p-K n×N n＝0
N n÷N p＝K p÷K n
A = N n - N p N n + N p = K p - K n K p + K n
The flow chart of the exemplary method 5800 of the certain average value that is used to generate the binary target variable (for example, ISI degree, EQ degree or residual bias) under the poised state according to a particular embodiment of the present invention that Figure 48 is an illustration.This method begins at step 5810 place, at this step place, for example uses adaptive controller 102 to check that in any suitable way target variable is a height or low.For example, in a particular embodiment, apply inverse correlation function (or correlation function) or XOR (or XNOR) operation by data value to 1.5 positions before the boundary value between the continuous data value that comprises transformation and this boundary value, can check that the ISI degree is "+1 " or " 1 ", can check that perhaps the EQ degree is " height " or " low ".In alternative embodiment, can use filter mode (as the AD HOC in the above patterns of describing in conjunction with Fig. 6,8,10,14 and 16) to check.For another example, utilize above tables in conjunction with Figure 25,27,29,32,34,36,38 and 40 descriptions, can check that residual bias is " just " or " bearing ".
Determine increasing or reduce control variables (for example, equalizer gain) based on the high value or the low value of the target variable of checking out at step 5810 place in step 5820 and 5830 places.For example, in the specific embodiment at the equalizer gain controlling, if be that " 1 " or EQ degree are " low " in the ISI of step 5810 place degree, then determine to increase the equalizer gain, this method proceeds to step 5840 then.If in the ISI of step 5810 place degree is that "+1 " or EQ degree are " height ", then determine to reduce the equalizer gain, this method proceeds to step 5850 then.In the alternative embodiment of using filter mode, detected specific filter pattern can determine to reduce or to increase the equalizer gain.As another example, in specific embodiment,, then determine to increase equalizer and be offset if be " bear " in step 5810 place residual bias at equalizer skew control, this method proceeds to step 5840 then.If in step 5810 place residual bias is " just ", then determine to reduce the equalizer skew, this method proceeds to step 5850 then.In the alternative embodiment of using filter mode, detected specific filter pattern can determine to reduce or to increase the equalizer skew.
At step 5840 place, after determining to increase control variables, control variables is increased Kp.For example, in specific embodiment, the equalizer gain is increased Kp at the equalizer gain controlling.As another example, in specific embodiment, the equalizer skew is increased Kp at equalizer skew control.After having increased control variables, method 5800 turns back to step 5810.
At step 5850 place, after having determined to reduce control variables, control variables is reduced Kn.For example, in specific embodiment, with equalizer gain reducing Kn at the equalizer gain controlling.As another example, in specific embodiment, with equalizer skew reducing Kn at equalizer skew control.After having reduced control variables, method 5800 turns back to step 5810.
Different with conventional start stop mode control system, K pNeed not be equal to K nOn the contrary, can utilize parameter T (it is to the controlled target of binary target variable under poised state) to make K based on following formula group pWith K nDifferent.Can select from-1 to+1 any suitable value (not necessarily zero) as T, and in a particular embodiment, the value of T can depend on the various conditions that are associated with for example bit error rate and can be corresponding to optimal objective value under these conditions.In alternative embodiment, the value of T can be fixed.In a particular embodiment T can with K pWith K nBetween ratio or difference be associated.For example, in a particular embodiment, control target T can be included in the target proportion (fixing or variable) of the frequency that under second state and first state target variable is detected.
K is at the increase of control variables and the public loop constants that reduces in following formula, and is defined as K pAnd K nArithmetic average.As can be observed, work as K p=K * (1+T) and K nDuring=K * (1-T), the mean value A of binary target variable will converge to T under poised state.
K p＝K×(1+T)
K n＝K×(1-T)
K p + K n 2 = K × ( 1 + T ) + K × ( 1 - T ) 2 = 2 K 2 = K
A = K p - K n K p + K n = K × ( 1 + T ) - K × ( 1 - T ) K × ( 1 + T ) + K × ( 1 - T ) = 2 KT 2 K = T
Need not be equal to zero controlled target T by employing, method 5800 can be so that the mean value of binary target variable (for example, ISI degree, balanced intensity, residual bias or other suitable target variables) converges to more suitably the point corresponding to specified conditions.As discussed above, in a particular embodiment, controlled target T can fix.In alternative embodiment, controlled target T can be used as the function of one or more specific variable and dynamically changes.
Figure 49 is the illustration flow chart that is used for dynamically generating at the exemplary method 5900 of the controlled target of the mean value of binary target variable under the poised state (for example, ISI degree) according to a particular embodiment of the present invention.As discussed above, in a particular embodiment, optimum average ISI degree may be high for high loss channel, and may be low for the low-loss channel.Therefore, in a particular embodiment, the controlled target of the mean value of such binary target variable (for example ISI degree or other suitable balanced intensities) may be favourable, the dynamic change with the value (for example the equalizer gain is provided with) of control variables of this controlled target.
Method 5900 begins at step 5910 place, at this step place, for example uses adaptive controller 102 to check that in any suitable way target variable is a height or low.Step 5910 to 5950 can be identical with above-mentioned steps 5810 to 5850, therefore no longer they is described.The place increases K with control variables in step 5940 pOr it is reduced K at step 5950 place nAfterwards, method 5900 proceeds to step 5960.At step 5960 place, the value of control variables (for example equalizer gain) is discerned.At step 5970 place, come controlled target T and the control step value K of metering needle to the mean value of binary target variable (for example ISI degree or other suitable balanced intensities) based on discre value to control variables (for example equalizer gain) pAnd K n
For example, in a particular embodiment, controlled target T can be adjusted in fixed range and change with control variables (for example equalizer gain).Only as example, when the value of control variables is high relatively, controlled target T can be set at+0.4, when the value of control variables is low relatively, controlled target T can be set at-0.4.When the value of control variables between high relatively value and the relative low value time, can be the interpolate value between+0.4 and-0.4 with controlled target T.By this mode, can dynamically generate K according to control variables (for example equalizer gain) pAnd K nThereby, based on the optimum mean value of bit error rate generation binary target variable (for example ISI degree).
In a particular embodiment, can utilize following formula group controlled target T dynamically to be calculated as the function of the currency of control variables (for example equalizer gain code):
T ( G ) = T H &times; G G C + T L &times; G C - G G C . . . G < G C
T(G)＝T H ...G≥G C
Wherein, G (for example 0 to 126) is the currency of control variables (for example equalizer gain code, the amount of frequency compensation that its expression is applied by equalizer).In addition, T H(for example-1.0 to+1.0) are the T values of high relatively control variables value, T L(for example-1.0 to+1.0) are the T values of low relatively control variables value, G C(for example 0,1,2,4,8,16,32,64) is that promptly, this function has smooth T value when G is above this turning in the value of the G of the corner of such function.Adaptive controller (as adaptive controller 102) can utilize the T of dynamic calculation and (above-mentioned) loop constants K to generate at K pAnd K nUpdating value.Upgrading T, K pAnd K nAfterwards, method 5900 turns back to step 5910.
Figure 50 is an illustration uses example controlled target formula dynamically to generate the curve chart 6000 at the result of the example controlled target of the mean value of binary target variable under poised state in equalizer gain controlling according to a particular embodiment of the present invention.This example controlled target formula is the above formula of listing at T (G).As can be observed, when the currency G of equalizer gain code equalled zero, controlled target T equals low relatively yield value T LWhen G 0 and G CBetween the time, controlled target T can be set at T LWith relative high yield value T HBetween interpolate value.When G greater than G CThe time, controlled target T equals T HShould be pointed out that in alternative embodiment, can adopt different controlled target formula, obtain with curve chart 6000 in illustrative different curve chart.Described controlled target although it should also be noted that this discussion, can be described balanced intensity and be target (and might not follow the tracks of average ISI degree) by other suitable manner with it at average ISI degree.
In a particular embodiment, high-frequency gain code G can be separated in two or more paths (for example, with upper pathway 101A to 101C) of multidimensional equalizer.For example, adaptive controller 102 can convert high-frequency gain code G to DC path gain code and single order path gain code.Can be in any suitable way (as following as described in Figure 51 and 52) high-frequency gain code G is changed.
Figure 51 has been the illustration table that is used for the high-frequency gain code is converted to the exemplary scenario 6100 of DC path gain code and single order path gain code according to a particular embodiment of the present invention.The row 6110 of scheme 6100 comprise the value of high-frequency gain code G, and high-frequency gain code G will be converted into DC path gain code and single order path gain code.Row 6120 comprise DC path gain code G 0Value, row 6130 comprise single order path gain code G 1Value.Each row 6140 all comprises high-frequency gain code (or scope of high-frequency gain code) and corresponding DC path and single order path gain code.Should be pointed out that in scheme 6100 G 0MAX is the maximum of the DC path gain code of appointment in the register in adaptive controller 102.In addition, the maximum of single order path gain code is 63 in a particular embodiment.
The result's of the exemplary scenario 6100 of using Figure 51 in order to convert the high-frequency gain code to DC path gain code and single order path gain code according to a particular embodiment of the present invention that Figure 52 A and 52B are illustrations curve chart.Figure 52 A is an illustration as the curve chart 6200 of the DC path gain code of the function of high-frequency gain code.Figure 52 B is an illustration as the curve chart 6300 of the single order path gain code of the function of high-frequency gain code.Should be pointed out that in alternative embodiment, can adopt the conversion plan different, obtain and curve chart 6200 and the different curve chart of 6300 illustrated curve charts with scheme 6100.
Can make amendment, add or omit described system and method without departing from the scope of the invention.Can carry out integrated to the part of described system and method or separate according to concrete needs.In addition, can by more, still less or other parts carry out the operation of described system and method.
Although invention has been described according to several embodiment, but those skilled in the art can propose many variations, modification, change, conversion and modification, the present invention is intended to contain this variation, modification, change, conversion and modification, as long as they fall within the scope of the appended claims.For example, although under the linguistic context of gain controlling illustration and described several embodiment, as long as suitable, just can skew control or the linguistic context of any other suitable Control Parameter under additionally or alternatively realize alternative embodiment.Although illustration and described several embodiment under the linguistic context of skew control, it is suitable to need only, and just can additionally or alternatively realize alternative embodiment under the linguistic context of gain controlling or any other suitable Control Parameter.
It is the priority of the U.S. Provisional Application 35U.S.C. § 119 (e) (sequence number 60/803,451) of " Adaptive Equalizer " that the application requires in the title that on May 30th, 2006 submitted to.

Claims (12)

1. method that is used for conditioning signal, this method may further comprise the steps:
To take place signal before or after the distortion apply at the compensating for loss and damage of frequency dependent distortion and in the migration of DC bias distortion at least one to generate output signal;
Use clock signal, described output signal is sampled to generate a plurality of data values and boundary value;
At first group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in described first group of data pattern;
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected described first group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated;
After in described compensating for loss and damage and the described migration at least one having been carried out regulate, at second group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in described second group of data pattern; And
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected described second group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated,
Wherein, data pattern is represented predefined binarization mode,
And wherein:
Each data pattern in described first group of data pattern comprises the high data value at the certain bits place that is arranged in this data pattern; And
Each data pattern in described second group of data pattern comprises the low data value at the certain bits place that is arranged in this data pattern, and the described certain bits that wherein comprises described high data value in described first group of data pattern is corresponding to the described certain bits that comprises described low data value in described second group of data pattern.
2. method according to claim 1, wherein:
Each data pattern in described first group of data pattern comprises the transformation and the high data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value; And
Each data pattern in described second group of data pattern comprises the transformation and the low data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value.
3. method according to claim 1, wherein, described clock signal is associated with described output signal.
4. method that is used for conditioning signal, this method may further comprise the steps:
To take place signal before or after the distortion apply at the compensating for loss and damage of frequency dependent distortion and in the migration of DC bias distortion at least one to generate output signal;
Use clock signal, described output signal is sampled to generate a plurality of data values and boundary value;
Select one of first group of data pattern and second group of data pattern with random fashion;
At selected one group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in selected one group of data pattern;
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected selected one group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated; And
After in described compensating for loss and damage and the described migration at least one having been carried out regulate, select one of described first group of data pattern and described second group of data pattern with random fashion, and repeat the step of monitoring, detection and the adjusting of front,
Wherein, data pattern is represented predefined binarization mode,
And wherein:
Each data pattern in described first group of data pattern comprises the high data value at the certain bits place that is arranged in this data pattern; And
Each data pattern in described second group of data pattern comprises the low data value at the certain bits place that is arranged in this data pattern, and the described certain bits that wherein comprises described high data value in described first group of data pattern is corresponding to the described certain bits that comprises described low data value in described second group of data pattern.
5. method according to claim 4, wherein:
Each data pattern in described first group of data pattern comprises the transformation and the high data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value; And
Each data pattern in described second group of data pattern comprises the transformation and the low data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value.
6. method according to claim 4, wherein, described clock signal is associated with described output signal.
7. adaptive equalizer, this adaptive equalizer comprises:
Equalizer, its be configured to take place signal before or after the distortion apply at the compensating for loss and damage of frequency dependent distortion and in the migration of DC bias distortion at least one to generate output signal;
Sampler, it is configured to:
Receive described output signal and clock signal;
Use described clock signal, described output signal is sampled to generate a plurality of data values and boundary value; With
Controller, it is configured to:
At first group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in described first group of data pattern;
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected described first group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated;
After in described compensating for loss and damage and the described migration at least one having been carried out regulate, at second group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in described second group of data pattern; And
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected described second group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated,
Wherein, data pattern is represented predefined binarization mode,
And wherein:
Each data pattern in described first group of data pattern comprises the high data value at the certain bits place that is arranged in this data pattern; And
Each data pattern in described second group of data pattern comprises the low data value at the certain bits place that is arranged in this data pattern, and the described certain bits that wherein comprises described high data value in described first group of data pattern is corresponding to the described certain bits that comprises described low data value in described second group of data pattern.
8. adaptive equalizer according to claim 7, wherein:
Each data pattern in described first group of data pattern comprises the transformation and the high data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value; And
Each data pattern in described second group of data pattern comprises the transformation and the low data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value.
9. adaptive equalizer according to claim 7, wherein, described clock signal is associated with described output signal.
10. adaptive equalizer, this adaptive equalizer comprises:
Equalizer, its be configured to take place signal before or after the distortion apply at the compensating for loss and damage of frequency dependent distortion and in the migration of DC bias distortion at least one to generate output signal;
Sampler, it is configured to:
Receive described output signal and clock signal;
Use described clock signal, described output signal is sampled to generate a plurality of data values and boundary value; With
Controller, it is configured to:
Select one of first group of data pattern and second group of data pattern with random fashion;
At selected one group of data pattern, the value that monitoring is sampled;
In the value of being sampled, detect the data pattern in selected one group of data pattern;
Based on one or more data value of sampling and boundary value of being associated with data pattern in detected selected one group of data pattern, the described compensating for loss and damage that imposes on described signal and at least one in the described migration are regulated; And
After in described compensating for loss and damage and the described migration at least one having been carried out regulate, select one of described first group of data pattern and described second group of data pattern with random fashion, and repeat the process of monitoring, detection and the adjusting of front,
Wherein, data pattern is represented predefined binarization mode,
And wherein:
Each data pattern in described first group of data pattern comprises the high data value at the certain bits place that is arranged in this data pattern; And
Each data pattern in described second group of data pattern comprises the low data value at the certain bits place that is arranged in this data pattern, and the described certain bits that wherein comprises described high data value in described first group of data pattern is corresponding to the described certain bits that comprises described low data value in described second group of data pattern.
11. adaptive equalizer according to claim 10, wherein:
Each data pattern in described first group of data pattern comprises the transformation and the high data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value; And
Each data pattern in described second group of data pattern comprises the transformation and the low data value before 1.5 positions of the boundary value between two continuous data values of the transformation that comprises this value of value.
12. adaptive equalizer according to claim 10, wherein, described clock signal is associated with described output signal.
CN 200710109830 2006-05-30 2007-05-30 System and method for performing decoupling to a plurality of control loop Expired - Fee Related CN101083637B (en)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US80345106P 2006-05-30 2006-05-30
US60/803,451 2006-05-30
US11/753,164 2007-05-24
US11/753,164 US7804921B2 (en) 2006-05-30 2007-05-24 System and method for decoupling multiple control loops

Publications (2)

Publication Number Publication Date
CN101083637A CN101083637A (en) 2007-12-05
CN101083637B true CN101083637B (en) 2011-06-29

Family

ID=38912792

Family Applications (9)

Application Number Title Priority Date Filing Date
CN 200710109832 Expired - Fee Related CN101083638B (en) 2006-05-30 2007-05-30 System and method for adjusting offset compensation applied to a signal
CN 200710109830 Expired - Fee Related CN101083637B (en) 2006-05-30 2007-05-30 System and method for performing decoupling to a plurality of control loop
CN 200710109829 Expired - Fee Related CN101083474B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal using filter patterns
CN 200710109833 Expired - Fee Related CN101083642B (en) 2006-05-30 2007-05-30 System and method for adjusting offset compensation applied to a signal
CN 200710109826 Expired - Fee Related CN101094003B (en) 2006-05-30 2007-05-30 System and method for independently adjusting multiple offset compensations applied to a signal
CN 200710109827 Expired - Fee Related CN101083472B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal
CN 200710109828 Expired - Fee Related CN101083473B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal using filter patterns
CN 200710109831 Expired - Fee Related CN101083641B (en) 2006-05-30 2007-05-30 Input signal regulation method, self-adaptive balancer and method for controlling variable regualtion
CN 200710109834 Expired - Fee Related CN101083475B (en) 2006-05-30 2007-05-30 System and method for independently adjusting multiple offset compensations applied to a signal

Family Applications Before (1)

Application Number Title Priority Date Filing Date
CN 200710109832 Expired - Fee Related CN101083638B (en) 2006-05-30 2007-05-30 System and method for adjusting offset compensation applied to a signal

Family Applications After (7)

Application Number Title Priority Date Filing Date
CN 200710109829 Expired - Fee Related CN101083474B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal using filter patterns
CN 200710109833 Expired - Fee Related CN101083642B (en) 2006-05-30 2007-05-30 System and method for adjusting offset compensation applied to a signal
CN 200710109826 Expired - Fee Related CN101094003B (en) 2006-05-30 2007-05-30 System and method for independently adjusting multiple offset compensations applied to a signal
CN 200710109827 Expired - Fee Related CN101083472B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal
CN 200710109828 Expired - Fee Related CN101083473B (en) 2006-05-30 2007-05-30 System and method for adjusting compensation applied to a signal using filter patterns
CN 200710109831 Expired - Fee Related CN101083641B (en) 2006-05-30 2007-05-30 Input signal regulation method, self-adaptive balancer and method for controlling variable regualtion
CN 200710109834 Expired - Fee Related CN101083475B (en) 2006-05-30 2007-05-30 System and method for independently adjusting multiple offset compensations applied to a signal

Country Status (1)

Country Link
CN (9) CN101083638B (en)

Families Citing this family (29)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101459632B (en) * 2007-12-12 2012-12-05 林武 Adaptive equalizing circuit and method
CN101562717B (en) * 2008-04-18 2011-04-20 中兴通讯股份有限公司 Mobile TV automatic gain control device and dynamic step length adjusting method thereof
CN101340408B (en) * 2008-08-08 2010-12-08 无锡辐导微电子有限公司 Analog determining feedback equalizer used for high-speed serial interface
CN102339245B (en) * 2010-07-22 2014-08-20 鸿富锦精密工业（深圳）有限公司 System and method for compensating circuit direct-current (DC) offset
CN103053140B (en) * 2010-08-03 2015-01-28 松下电器产业株式会社 Adaptive receiver system and adaptive transceiver system
US9071478B2 (en) * 2011-05-24 2015-06-30 Mediatek Inc. Methods for performing adaptive equalization and associated apparatus
TWI433480B (en) * 2011-07-19 2014-04-01 Raydium Semiconductor Corp Adaptive equalizer and operating method thereof
JP5612626B2 (en) * 2012-03-26 2014-10-22 アンリツ株式会社 Emphasis optimization device and emphasis optimization method
US8861583B2 (en) * 2012-12-14 2014-10-14 Altera Corporation Apparatus and methods for equalizer adaptation
CN103399418B (en) * 2013-07-23 2016-01-20 清华大学 Compensate electroabsorption modulator non-linear method and device
US9755818B2 (en) 2013-10-03 2017-09-05 Qualcomm Incorporated Method to enhance MIPI D-PHY link rate with minimal PHY changes and no protocol changes
US9735948B2 (en) 2013-10-03 2017-08-15 Qualcomm Incorporated Multi-lane N-factorial (N!) and other multi-wire communication systems
CN106063181B (en) * 2014-03-06 2018-03-13 高通股份有限公司 Receiver circuit and the method operated on receiver circuit
US9479365B2 (en) * 2014-08-28 2016-10-25 Mediatek Inc. Method for performing loop unrolled decision feedback equalization in an electronic device with aid of voltage feedforward, and associated apparatus
US9634624B2 (en) * 2014-12-24 2017-04-25 Stmicroelectronics S.R.L. Method of operating digital-to-analog processing chains, corresponding device, apparatus and computer program product
CN107241160B (en) * 2016-03-28 2021-03-23 华为技术有限公司 Method and device for determining parameters
CN108289043B (en) * 2017-01-10 2020-06-05 工业和信息化部电信研究院 Data consistency detection method and device
US10128826B2 (en) * 2017-01-18 2018-11-13 Microsemi Semiconductor Ulc Clock synthesizer with integral non-linear interpolation (INL) distortion compensation
US11374613B2 (en) * 2018-03-29 2022-06-28 U-Blox Ag Active interference cancellation apparatus, signal isolation control apparatus and method of actively cancelling interference
CN108716887B (en) * 2018-05-18 2024-09-13 重庆天成集创科技有限公司 Differential displacement sensor and measuring method thereof
CN110620635A (en) * 2018-06-20 2019-12-27 深圳市华星光电技术有限公司 Decoding method, apparatus and readable storage medium
US10784845B2 (en) 2018-09-27 2020-09-22 Macom Technology Solutions Holdings, Inc. Error detection and compensation for a multiplexing transmitter
KR102646911B1 (en) * 2019-03-14 2024-03-14 삼성디스플레이 주식회사 Display device
CN110832452A (en) * 2019-09-20 2020-02-21 焦旭 Random number generation method and generation device
US10972249B1 (en) * 2020-04-20 2021-04-06 Microchip Technology Inc. System and method for data sampler drift compensation
CN112835315B (en) * 2021-01-06 2022-03-08 苏州维嘉科技股份有限公司 Device and method for controlling drilling depth of circuit board
CN113163197B (en) * 2021-04-26 2023-02-21 北京欧铼德微电子技术有限公司 Data transmission method, circuit, device, equipment and storage medium
CN116094682B (en) * 2023-02-13 2023-11-07 南京金阵微电子技术有限公司 Data receiving device, data receiving method, storage medium, and electronic device
CN118509071A (en) * 2024-07-17 2024-08-16 厦门市合佳兴电子有限公司 Method for adjusting transmission power of mobile radio system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4873702A (en) * 1988-10-20 1989-10-10 Chiu Ran Fun Method and apparatus for DC restoration in digital receivers
CN1148759A (en) * 1995-07-12 1997-04-30 三星电子株式会社 Equalizer circuit of data communication terminal equipment and method thereof
CN1316136A (en) * 1999-07-07 2001-10-03 三菱电机株式会社 Receiver and Adaptive Equalization Processing Method

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5159547A (en) * 1990-10-16 1992-10-27 Rockwell International Corporation Self-monitoring tuner for feedback controller
US5245611A (en) * 1991-05-31 1993-09-14 Motorola, Inc. Method and apparatus for providing carrier frequency offset compensation in a tdma communication system
US5659581A (en) * 1994-12-13 1997-08-19 Paradyne Corporation Modem receiver pre-emphasis
JPH09199961A (en) * 1996-01-22 1997-07-31 Oki Electric Ind Co Ltd Agc equipment
CN1103135C (en) * 1997-09-30 2003-03-12 夸尔柯姆股份有限公司 Method and device for increasing anti-interference of receiver
US6496548B1 (en) * 1998-10-09 2002-12-17 Harris Corporation Apparatus and method for decoding asynchronous data using derivative calculation
US6735259B1 (en) * 1999-12-20 2004-05-11 Nortel Networks Limited Method and apparatus for optimization of a data communications system using sacrificial bits
KR100708070B1 (en) * 2000-01-10 2007-04-17 삼성전자주식회사 Data reproducing apparatus and method of improving detection performance by adjusting decision level used in data detector
JP3420228B2 (en) * 2001-09-07 2003-06-23 松下電器産業株式会社 Interference wave power measurement device, interference wave power measurement method, transmission power control device, and transmission power control method
GB2397980B (en) * 2003-01-28 2005-11-23 Phyworks Ltd Receiver
US7277516B2 (en) * 2003-10-14 2007-10-02 Realtek Semiconductor Corp Adaptive equalization system for a signal receiver
CN1332062C (en) * 2004-05-14 2007-08-15 中国科学院半导体研究所 Method for depositing thin oxide coating through oxygen ion beam in low energy and auxiliary impulse laser

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4873702A (en) * 1988-10-20 1989-10-10 Chiu Ran Fun Method and apparatus for DC restoration in digital receivers
CN1148759A (en) * 1995-07-12 1997-04-30 三星电子株式会社 Equalizer circuit of data communication terminal equipment and method thereof
CN1316136A (en) * 1999-07-07 2001-10-03 三菱电机株式会社 Receiver and Adaptive Equalization Processing Method

Also Published As

Publication number Publication date
CN101083472B (en) 2011-07-20
CN101083641A (en) 2007-12-05
CN101083474A (en) 2007-12-05
CN101083472A (en) 2007-12-05
CN101083475A (en) 2007-12-05
CN101083638B (en) 2010-07-28
CN101094003A (en) 2007-12-26
CN101094003B (en) 2013-08-21
CN101083641B (en) 2010-06-09
CN101083638A (en) 2007-12-05
CN101083642B (en) 2010-10-20
CN101083475B (en) 2013-01-16
CN101083474B (en) 2010-11-03
CN101083637A (en) 2007-12-05
CN101083642A (en) 2007-12-05
CN101083473B (en) 2010-11-03
CN101083473A (en) 2007-12-05

Similar Documents

Publication Publication Date Title
CN101083637B (en) System and method for performing decoupling to a plurality of control loop
EP1862777B1 (en) System and method for adjusting offset compensation applied to a signal
EP1863247B1 (en) System and method for the adjustment of compensation applied to a signal using filter patterns
EP1862778B1 (en) System and method for the adjustment of offset compensation applied to a signal
EP1863245B1 (en) System and method for adjusting compensation applied to a signal using filter patterns
EP1862779B1 (en) System and method for the non-linear adjustment of compensation applied to a signal
EP1863246B1 (en) System and method for independently adjusting multiple offset compensations applied to a signal
EP1862766B1 (en) System and method for asymmetrically adjusting compensation applied to a signal
US7760798B2 (en) System and method for adjusting compensation applied to a signal
EP1862776B1 (en) System and method for decoupling multiple control loops
EP1863244A2 (en) System and method for independently adjusting multiple offset compensations applied to a signal
US7839958B2 (en) System and method for the adjustment of compensation applied to a signal

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20110629

Termination date: 20200530