CN114070229A

CN114070229A - Decision feedback equalizer and related control method

Info

Publication number: CN114070229A
Application number: CN202010762606.4A
Authority: CN
Inventors: 吴宗振; 黄亮维
Original assignee: Realtek Semiconductor Corp
Current assignee: Realtek Semiconductor Corp
Priority date: 2020-07-31
Filing date: 2020-07-31
Publication date: 2022-02-18

Abstract

Embodiments of the present disclosure relate to decision feedback equalizers and related control methods. A decision feedback equalizer for generating a decision output signal based on an input signal, comprising: a feedforward equalizer, a feedback equalizer and a weight coefficient control unit. The feedforward equalizer comprises a plurality of tapped delay lines and is controlled by a group of first weight coefficients. The feedback equalizer includes a plurality of tapped delay lines and is controlled by a set of second weight coefficients. The weight coefficient control unit is used for selectively adjusting at least one of the set of first weight coefficients and determining a set of first boundary values for at least one of the set of second weight coefficients. The weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

Description

Decision feedback equalizer and related control method

Technical Field

The present invention relates to a decision feedback equalizer, and more particularly, to a decision feedback equalizer and a related control method for avoiding error propagation by adjusting weight coefficients of the equalizer in stages.

Background

During transmission of a signal through a channel, the signal is often distorted due to the Time Dispersion (Time Dispersion) effect of the channel. The main reason is that when the channel frequency response is non-constant amplitude and non-linear phase, the amplitude and phase of the signal are distorted by the channel response, which results in Inter Symbol Interference (ISI), which makes the receiving end unable to correctly recognize the signal.

Decision Feedback equalizer (Decision Feedback Equa)lizer) may be used to remove the above-mentioned channel distortion. The decision Feedback Equalizer includes a Feed-forward Equalizer (Feed-forward Equalizer), a decision unit and a Feedback Equalizer (Feedback Equalizer). A feed forward Equalizer, also known as a forced-Zero Equalizer (Zero-Forcing Equalizer), allows the impulse responses of the channel and Equalizer to have values in only one place after they are convolved with each other. The feedforward equalizer has the advantage of simple architecture, but has the disadvantage of amplifying noise, resulting in decision error (decision error). Therefore, a feedback equalizer is needed to solve this problem. The feedback equalizer uses the detected Symbol d under the assumption that the Symbol (Symbol) detected by the decision feedback equalizer is correct₀As an input, thereby canceling the ISI of the channel. Thus, the feedback equalizer does not have the problem of amplifying noise.

In general, decision feedback equalizers typically rely on the use of a Least Mean Square (LMS) algorithm to determine the weight coefficients for each tapped delay line (tapped delay line) in the feed forward equalizer and feedback equalizer. By repeatedly adjusting the weight coefficients of the feedforward equalizer and the feedback equalizer, the least square solution obtained by the least mean square algorithm is approximated, thereby converging to a high signal-to-noise ratio. However, in some cases, when a decision unit in the decision feedback equalizer has a decision error, the error is input to the feedback equalizer. The decision error at this time is also fed back to the whole decision feedback equalizer by the output end of the feedback equalizer. If the error is large, it may cause a large error loop between the feedback equalizer and the decision device, which is called error propagation. When the error propagation situation is too serious, the whole decision feedback equalizer system will not converge, resulting in systematic collapse of the whole decision feedback equalizer.

Disclosure of Invention

To avoid the occurrence of error propagation phenomena, the present invention provides a mechanism for controlling a decision feedback equalizer. In the control mechanism of the invention, the weight coefficient in the feedback equalizer is limited at the initial stage of convergence of the least mean square algorithm, so as to limit the weight energy of the feedback equalizer, thus avoiding error propagation caused by error amplification when a decision error occurs, and further ensuring that the least mean square algorithm can not be converged. In addition, since the decision error is considerable at the initial stage of convergence compared to when the convergence is stable, suppressing the weight energy of the feedback equalizer contributes to improving the stability at the initial stage of convergence. Furthermore, when the least mean square algorithm tends to converge stably, the control mechanism of the present invention relaxes the limitation on the weight coefficient of the feedback equalizer, so as to gradually make the weight coefficient approach the least square solution in the stage of stable convergence, thereby improving the signal-to-noise ratio of the signal. Therefore, the stability and the good signal-to-noise ratio of the least mean square algorithm can be considered.

An embodiment of the present invention provides a dfe equalizer for generating a dfe output signal according to an input signal, comprising: a feedforward equalizer, a feedback equalizer and a weight coefficient control unit. The feedforward equalizer comprises a plurality of tapped delay lines and is controlled by a group of first weight coefficients. The feedback equalizer includes a plurality of tapped delay lines and is controlled by a set of second weight coefficients. The weight coefficient control unit is used for selectively adjusting at least one of the first weight coefficients and determining a first boundary value of at least one of the second weight coefficients. The weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

An embodiment of the present invention provides a method for controlling a decision feedback equalizer. The method is used for generating a decision output signal according to an input signal, wherein the decision feedback equalizer comprises a feedforward equalizer and a feedback equalizer. The method comprises the following steps: at least one of a set of first weight coefficients corresponding to a plurality of tapped delay lines of the feed forward equalizer is selectively adjusted. Determining a set of first boundary values corresponding to at least one of a set of second weight coefficients of a plurality of tapped delay lines of the feedback equalizer; and the weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

Drawings

FIG. 1 is a functional block diagram of an embodiment of a DFE equalizer of the present invention.

Fig. 2A and fig. 2B are schematic diagrams illustrating an implementation architecture of a feedforward equalizer and a feedback equalizer in a decision feedback equalizer according to the present invention.

FIG. 3 is a flowchart illustrating a preliminary adjustment phase according to an embodiment of the control method of the present invention.

FIG. 4 is a flowchart illustrating a fine tuning phase according to an embodiment of the control method of the present invention.

FIG. 5 is a simplified flow chart of an embodiment of the control method of the present invention.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of embodiments of the invention to the reader. However, those skilled in the art will understand how to implement the invention without one or more of the specific details, or with other methods or components or materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the invention.

Reference in the specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. Thus, the appearances of the phrase "in one embodiment" appearing in various places throughout the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics described above may be combined in any suitable manner in one or more embodiments.

Referring to fig. 1, a functional block diagram of a dfe according to an embodiment of the invention is shown. As shown, the dfe 100 includes a feedforward equalizer 110, a feedback equalizer 120, a decision unit 130, and a weight coefficient control unit 140. The dfe 100 is used for receiving an input signal r from a channel₀And generates a decision output signal d₀. Wherein the feedforward equalizer 110 is used for generating the feedforward equalizer according to the input signal r₀Generating a feedforward output signal FF₀Thereby eliminating pre-cursor (pre-cursor) interference and post-cursor (post-cursor) interference in ISI caused by the channel. And the feedback equalizer 120 outputs a signal d according to the decision₀Generating a feedforward output signal FB₀Thereby eliminating post-cursor interference in the ISI. The decision unit 130 generates a feedforward output signal FF according to the feedforward equalizer 110₀And a feed-forward output signal FB generated by the feedback equalizer 120₀Generating a decision output signal d₀。

Fig. 2A and 2B are schematic diagrams illustrating the architecture of the feedforward equalizer 110 and the feedback equalizer 120, respectively. The feedforward equalizer 110 and the feedback equalizer 120 respectively have a plurality of tapped delay-lines (tapped delay-lines) 112_1 to 112_ N and 122_1 to 122_ N. The feedforward equalizer 110 and the feedback equalizer 120 are controlled by a set of weight coefficients cf respectively₁～cf_NAnd a set of weighting coefficients cb₁～cb_N. Wherein, the weight coefficient control unit 140 may control the weight coefficient cf during convergence of the decision feedback equalizer 100₁～cf_NAnd cb₁～cb_N. In the control mechanism of the present invention, the weight coefficient control unit 140 will perform the weight coefficients cf for the convergence process of the decision feedback equalizer 100₁～cf_NAnd cb₁～cb_NAnd (4) adjusting. Wherein, in the early stage of convergence, the weight coefficient control unit 140 of the present invention conservatively adjusts the weight coefficient cf according to the least mean square algorithm and according to a certain limit₁～cf_NAnd cb₁～cb_NTherefore, the signal-to-noise ratio of the system is improved, and meanwhile, the stability of convergence is guaranteed. And upon entering the convergence stabilization phase, the weight coefficient control unit 140 targets one or more weight coefficients cf in the feedforward equalizer 110_KMaking adjustments to observe one or more weight coefficients cb in feedback equalizer 120_KAccordingly, the weight coefficient cf is gradually given back_KAnd a weight coefficient cb_KApproximating the least squares solution found by the least mean square algorithm. In the early stage of convergence, the present invention will pair the weight coefficients cf_KAnd cb_KMake a limit, which gives the weight coefficient cf_KIs smaller. While in the convergence stable stage, the weighting coefficient cb is relaxed_KBy a weight coefficient cf_KHas a larger adjustment range, thereby leading to the weight coefficient cf_KAnd cb_KAnd in the convergence stable stage, the method can be more approximate to the least square solution so as to improve the signal-to-noise ratio of the system.

Please refer to fig. 3, which is a flowchart illustrating a preliminary adjustment phase according to an embodiment of the control method of the present invention. First, in step 310, the decision feedback equalizer is still in the early stage of convergence, so that one or more weighting coefficients cb in the feedback equalizer 110 are used_KAnd cf_KLimiting, setting each or more weight coefficients cb_KIts corresponding upper boundary value cb _ max_KAnd a lower boundary value cb _ min_KRequiring each or a plurality of weight coefficients cb_KMust not exceed the corresponding upper boundary value cb _ max_KAnd a lower boundary value cb _ min_KAnd setting each or a plurality of weight coefficients cf_KIts corresponding upper boundary value cf _ max_KAnd a lower boundary value cf _ min_KRequiring each or a plurality of weight coefficients cf_KMust not exceed the corresponding upper boundary value cf _ max_KAnd a lower boundary value cf _ min_K. Wherein, in different embodiments, different weight coefficients cb_KCorresponding upper/lower boundary values cb _ max_KAnd cb _ min_KPossibly the same or different, and different weight coefficients cf_KCorresponding upper/lower boundary values cf _ max_KAnd cf _ min_KAnd may be the same or different. However, the upper boundary value cb _ max_KAnd a lower boundary value cb _ min_KMust be smaller than the absolute value of the minimum square solution (MMSE solution) in order to avoid error propagation.

Further, in step 320, a least mean square algorithm is used to step the decision feedback equalizer into a convergence state, which includes weighting coefficients cf for one or more of the feedforward equalizer 120_KAnd for one or more weight coefficients cb in the feedback equalizer 110_KAnd ensures one or more weight coefficients cb_KWill not exceed the corresponding upper/lower edgesLimit value cb _ max_KAnd cb _ min_KOne or more weight coefficients cf_KWill not exceed the corresponding upper/lower boundary values cf _ max_KAnd cf _ min_K。

In step 330, it is checked whether a predetermined time T _ avg has elapsed and the snr of the dfe 100 meets the minimum requirement for convergence stability. The minimum requirement for stable convergence can be determined by calculating the SNR of the dfe 100 and comparing it with a threshold SNR _ stable, which can be determined by performing a repeated test on the dfe 100 in advance and then according to the SNR when error propagation occurs.

If in the above phase it has been ensured that the decision feedback equalizer 100 enters a stable converged state, the control mechanism of the present invention enters a fine-tuning phase (fine-tuning). At this point, the control mechanism of the present invention will relax one or more of the weighting coefficients cb in feedback equalizer 120_KIn pursuit of a better signal-to-noise ratio.

Please refer to the flowchart shown in fig. 4, which is a flowchart of the fine tuning stage of the control method according to the embodiment of the present invention. First, in step 410, the control mechanism of the present invention first checks the current snr of the system, and if the snr of the system is high enough, the fine tuning process is terminated to avoid the weight coefficient cb of the feedback equalizer 120_KToo high, may pose a risk of error propagation. However, if there is still room for the system signal-to-noise ratio to continue to improve, the fine tuning phase will continue. The system SNR is compared with a target SNR _ target, and the fine tuning process continues only if the system SNR is less than the target SNR _ target.

In step 420, the weight coefficient control unit 140 will relax the weighting coefficient cb_KI.e. let one or more weight coefficients cb in the feedback equalizer 120_KThe least square solution obtained by the least mean square algorithm can be more approximate to the preliminary adjustment stage. Wherein the weight coefficient control unit 140 sets each or a plurality of weight coefficients cb_KCorresponding new upper boundary value cb _ max _ tgt_KAnd a new lower boundary value cb _ min _ tgt_KAnd requires each or a plurality of weight coefficients cb_KMust not exceed the corresponding upper boundary value cb _ max _ tgt_KAnd a lower boundary value cb _ min _ tgt_K. And the boundary value cb _ max _ tgt set in the fine adjustment stage_KAnd a lower boundary value cb _ min _ tgt_KWill be smaller than the boundary value cb _ max at the initial stage of convergence shown in FIG. 3_KAnd a lower boundary value cb _ min_KAnd is larger. By relaxing one or more of the weighting coefficients cb_KLet one or more weight coefficients cb_KThe system signal-to-noise ratio can be further improved by being closer to the least square solution.

In step 430, the weight coefficient control unit 140 examines one or more weight coefficients cb in the feedback equalizer 120_KWhether or not the boundary value set in step 420 is not exceeded. If the weight coefficient cb_KIf the boundary value set in step 420 has been exceeded, the fine tuning phase (step 480) is also ended. On the other hand, in the weight coefficient cb_KIn the case that the boundary value is not exceeded, the weight coefficient control unit 140 gradually increases one or more weight coefficients cf in the feedforward equalizer 110_K. Due to the weight coefficient cb_KAnd a weight coefficient cf_KThere is a certain dependency between them, so the weight coefficient cb_KAnd will also be lifted accordingly.

In the control mechanism of the present invention, the weight coefficient control unit 140 will wait for a predetermined time T _ thd before increasing the weight coefficient cf_K(step 440). When the predetermined time T _ thd has elapsed, step 450 is performed to set the weighting factor cf_KIf not, go to step 460 to maintain the weighting factor cf_KDoes not change and continues to wait (step 465).

In the fine tuning phase, the weight coefficient cf is increased every time_KThen, the weighting factor cf is increased again after waiting for a predetermined time_K. The purpose of the design latency is to ensure the stability of the system. Further, if the weight coefficient cf is suddenly adjusted_KIt may cause transient errors, and when the errors are too large, it may also destroy the decisionThe convergence stability of feedback equalizer 100. Thus, slowly adjusting helps to avoid the occurrence of such errors. On the other hand, the length of the waiting time may vary according to the characteristics of the system, and in an extreme case, it may be considered to ignore the waiting time or use a long waiting time.

In step 450, the weight coefficient control unit 140 increases one or more weight coefficients cf in the feedforward equalizer by an equal amount (e.g., step size cf _ step)_KAnd its corresponding upper/lower boundary values cf _ max, respectively_KAnd cf _ min_K. For example, each time step 450 is entered, the weighting coefficients cf are simultaneously assigned_KAnd upper/lower boundary values cf _ max_KAnd cf _ min_KThe step size cf _ step is increased by one unit. Wherein the step size cf step is also selected in relation to the system characteristics. The control mechanism of the present invention will then pair the weight coefficients cf at a time_KAnd their respective corresponding upper/lower boundary values cf _ max_KAnd cf _ min_KAfter the adjustment, one or more weight coefficients cb are confirmed again_KWhether or not it has reached its corresponding limit value cb _ max _ tgt_KAnd cb _ min _ tgt_K(return to step 430). If so, ending the flow of the fine tuning stage; otherwise, the weighting coefficient cf is continuously matched_KAnd its upper/lower boundary value cf _ max_KAnd cf _ min_KThe next adjustment is made.

It should be noted that in some embodiments, the control flow of the present invention may also only include the fine tuning phase shown in fig. 4, i.e., the fine tuning phase is expanded once the dfe 100 is confirmed to enter the stable convergence phase, but the previous control adjustment mechanism may be different from the preliminary tuning phase shown in fig. 3. The control method of the present invention can also be simplified to the steps shown in fig. 5:

step 510: selectively adjusting at least one of a set of first weight coefficients corresponding to a plurality of tapped delay lines of a feed-forward equalizer;

step 520: determining a set of first boundary values corresponding to at least one of a set of second weight coefficients for a plurality of tapped delay lines of a feedback equalizer; and

step 530: the weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

The above steps can be modified and adjusted to obtain the same/similar processes and effects as the previous embodiments.

The invention comprises the following features: first, at the initial stage of convergence using the least mean square algorithm, the weight coefficient cb of the equalizer 120 is fed back_KA tighter constraint is imposed to avoid too close to the least squares solution, thereby limiting the energy of the weights of the feedback equalizer 120, and thus avoiding the occurrence of error propagation. The weighting factor cb only after convergence into a steady state_KThe constraint of (c) is relaxed so that it approaches the least squares solution, allowing the dfb 100 to further improve the snr to a desired target. Furthermore, after entering the steady state, the weight coefficient cf is more actively and actively adjusted_K(periodically increasing the fixed stepping amount), the signal-to-noise ratio of the system can be increased more quickly, and meanwhile, transient errors caused by the increase of the signal-to-noise ratio are avoided through the setting of the waiting time. Therefore, the method and the device have the advantages of taking the stability of the least mean square algorithm and good signal-to-noise ratio into consideration.

Embodiments of the invention may be implemented using hardware, software, firmware, and combinations thereof. Embodiments of the invention may be implemented using software or firmware stored in a memory with appropriate instruction execution systems. In terms of hardware, this can be accomplished using any or a combination of the following techniques: individual arithmetic logic devices having logic gates that perform logic functions in accordance with data signals, Application Specific Integrated Circuits (ASICs) having suitable combinational logic gates, Programmable Gate Arrays (PGAs), Field Programmable Gate Arrays (FPGAs), etc.

The flowcharts and blocks in the flowcharts within this specification illustrate the architecture, functionality, and operation of what may be implemented by systems, methods and computer software products according to various embodiments of the present invention. In this regard, each block in the flowchart or functional block diagrams may represent a module, segment, or portion of program code, which comprises one or more executable instructions for implementing the specified logical function(s). In addition, each block of the functional block diagrams and/or flowchart illustrations, and combinations of blocks, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer program instructions. These computer program instructions may also be stored in a computer-readable medium that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable medium implement the function/act specified in the flowchart and/or block diagram block or blocks. The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.

[ notation ] to show

100 decision feedback equalizer

110 feedforward equalizer

120 feedback equalizer

130 decision unit

140 weight control unit

112_1 to 112_ N, 122_1 to 122_ N tapped delay line

310 to 340, 410 to 480

Claims

1. A decision feedback equalizer for generating a decision output signal based on an input signal, comprising:

a feed forward equalizer including a plurality of tapped delay lines and controlled by a set of first weight coefficients;

a feedback equalizer including a plurality of tapped delay lines and controlled by a set of second weight coefficients; and

a weight coefficient control unit to selectively adjust at least one of the set of first weight coefficients and determine a set of first boundary values for at least one of the set of second weight coefficients; the weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

2. The decision feedback equalizer of claim 1, wherein the weight coefficient limiting unit does not increment at least one of the set of first weight coefficients when at least one of the set of second weight coefficients is equal to the set of first boundary values.

3. The decision feedback equalizer of claim 1 wherein the weight coefficient limiting unit experiences a predetermined time interval between consecutive increments of at least one of the set of first weight coefficients.

4. The dfe of claim 1, wherein when at least one of the set of second weighting coefficients does not exceed the set of first boundary values, and the weighting coefficient control unit increments at least one of the set of first weighting coefficients while also incrementing a set of second boundary values to which at least one of the set of first weighting coefficients corresponds.

5. The decision feedback equalizer of claim 1, wherein the weight coefficient control unit does not adjust any of the set of first weight coefficients if a signal-to-noise ratio of the decision output signal is greater than or equal to a target signal-to-noise ratio prior to the weight coefficient control unit adjusting the set of first weight coefficients.

6. The decision feedback equalizer of claim 1, wherein the weight coefficient control unit further determines a set of third boundary values for at least one of the set of second weight coefficients, wherein the set of third boundary values is less than the set of first boundary values and the set of second boundary values is less than a least-squares solution based on a least-mean-square algorithm; wherein, when the signal-to-noise ratio of the decision output signal is less than a stable signal-to-noise ratio, the weight coefficient control unit adjusts at least one of the set of first weight coefficients and at least one of the set of second weight coefficients according to the least mean square algorithm until the signal-to-noise ratio is not less than the stable signal-to-noise ratio.

7. A method for controlling a decision feedback equalizer to generate a decision output signal from an input signal, wherein the decision feedback equalizer has a feed forward equalizer and a feedback equalizer, the method comprising:

selectively adjusting at least one of a set of first weight coefficients corresponding to a plurality of tapped delay lines of the feed-forward equalizer;

determining a set of first boundary values corresponding to at least one of a set of second weight coefficients for a plurality of tapped delay lines of the feedback equalizer; and

the weight coefficient control unit increments at least one of the set of first weight coefficients when at least one of the set of second weight coefficients does not exceed the set of first boundary values.

8. The method of claim 7, wherein the step of selectively adjusting at least one of the set of first weight coefficients comprises:

when at least one of the set of second weight coefficients is equal to the set of first boundary values, at least one of the set of first weight coefficients is not incremented.

9. The method of claim 7, wherein the step of selectively adjusting at least one of the set of first weight coefficients comprises:

a predetermined time interval elapses between successive increments of at least one of the set of first weight coefficients.

10. The method of claim 7, wherein the step of selectively adjusting at least one of the set of first weight coefficients comprises:

when at least one of the set of second weight coefficients does not exceed the first boundary value, and the weight coefficient control unit increments at least one of the set of first weight coefficients while also incrementing a set of second boundary values to which at least one of the set of first weight coefficients corresponds.