CN101174912B - Self-adapting clock method based on time stamp facing Ethernet circuit simulation service - Google Patents

Abstract

The present invention relates to a clock synchronization method in the Ethernet circuit simulation business, in specific to an adaptive clock method based on the time stamp which faces to the Ethernet simulation operation. The method utilizes the local operation clock and counter to obtain the time stamp information (local time stamp) which represents the remote terminal and proximate operation clocks; the time stamp information is converted into random sequence which meets the kalman process and a measurement equation and the kalman filtering algorithm is used to filter noise from the sequence to obtain the real time stamp; the obtained real time stamp is used for regulating a voltage control oscillator to make local operation clock to track the remote terminal operation clock. The method adopts the kalman algorithm which makes the loop filtration can choose hardware to or software to realize; the application is more flexible; the frequency difference is monotonically decreasing basically without vibration and the lock anchor can be reached fast. Compared with the ordinary buffer zone method and time stamp method, the comprehensive function of the present invention is stronger.

Description

A kind of towards the self-adapted clock method of ethernet circuit artificial service based on time stamp

Technical field

The present invention relates to the clock synchronizing method in a kind of ethernet circuit artificial service, be specifically related to a kind of towards the self-adapted clock method of ethernet circuit artificial service based on time stamp.

Background technology

Consistently in the industry think that next generation network is based on the network of high-speed packet technology, and possess multiple services abilities such as voice-bearer, video and data simultaneously.Largest Packet Based Network Internet will be the main body of next generation network, and main flow group technologys such as IP and Ethernet are to realize the key technology of computer internet, traditional electrical phone network and the cable television network integration of three networks.

With regard to ethernet technology, present this technology is occupied the application share more than 90% in the access aspect, and all advantages of in local area network (LAN), showing of ethernet technology, makes it become the bearer network in metropolitan area network even the wide area network gradually.Most widely used procotol IP is also many simultaneously carries on Ethernet.The generation of carrier class Ethernet promotes Ethernet especially to the development of darker wider aspect.Ethernet technology will be one of mainstream technology of Packet Based Network of future generation.

On the other hand, developing into next generation network from existing network, is the process of an evolution, rather than the revolution negating fully to existing network and technology.SDH technology such as (Synchronous Digital Hierarchy) based on TDM (TimeDivision Multiplexing) has had extensive application on existing network.

Therefore, be the TDM network of representative and be that the packet network of representative is with long-term co-existence with IP and Ethernet etc. with SDH etc. in the next generation network evolution process.SDH and Ethernet also will merge each other.On SDH, transmit the existing extensive application of Ethernet at present; But when Packet Based Network is in leading position; TDM business such as Packet Based Network transmission SDH with such as Ethernet also are necessary, and the seamless link that how to realize Ethernet and existing TDM network is the important problem in network research field.

Transmit the existing relevant criterion of technology of TDM about Packet Based Network ^[1]-[5]Having done to specify and regulation, is pith wherein synchronously.Transmit synchronous TDM business based on the asynchronous Ethernet of statistics and certainly exist the bi-directional conversion between synchronous TDM code stream and the asynchronous Ethernet point data bag, TDM two ends clock be the major issue that at first will solve synchronously.From relevant criterion ^{[6] [7] [8]}Can know that TDM is to there is high requirement synchronously.

The two ends clock synchronization mainly contains three class methods such as network clocking method, difference method and adaptive clock recovery method ^{[5] [6]}Preceding two class methods all require at the TDM two ends same external reference clock to be provided, and this is the implacable requirement of Ethernet of current widespread usage.The adaptive clock recovery method is then recovered business clock according to the transmission situation that Ethernet data wraps on the network.So be suitable for existing Ethernet.

The adaptive clock recovery method is as shown in Figure 1; Basically be similar to the functional structure of PLL (Phase-LockLoop); Obtain error signal according to packet from the arrival situation that far-end is sent to near-end, control VCO (Voltage-Controlled Oscillator) or DCO (Digital-Controlled Oscillator) eliminates two ends PE (Provide Edge) or CE (Customer Edge) with error signal ^[3]-[5]The business clock frequency difference.Can be divided into based on reception buffering area filling level (hereinafter to be referred as buffer zone method) with based on time stamp (hereinafter to be referred as the time stamp method) two class methods according to the information of using.In Fig. 1, if input phase demodulation logic is the time stamp difference at two ends, then be the time stamp method, if input phase demodulation logic is the buffer state data, then be buffer zone method.The square that has letter " R " among the figure is represented clock recovery module, and its function is to make fs2 final sum fs1 frequency identical.

Buffer zone method is the difference that obtains the two ends business clock according to the operating position that receives buffering area (jitter-removing buffer), utilizes this difference to regulate the local service clock simultaneously and makes that two ends business clock trend is consistent ^{[9] [10] [11]}Be characterized in stable, realize simple relatively, but the locking (the two ends clock frequency the is equal basically) time longer.

The time stamp method then is: when the Ethernet data bag of TDM information is carried in transmission, obtain the transmission time stamp of this packet, and this time stab information is encapsulated into packet.When receiving packet, generate and receive time stamp.And utilize the difference adjusting local service clock that receives time stamp and transmission time stamp to make the two ends business clock reach unanimity ^[12]Be characterized in that locking is very fast, require two ends can both generate time stamp, need take certain network bandwidth and transmit time stamp, and receive the delay variation of network to influence bigger.

Summary of the invention

The objective of the invention is to provides a kind of adaptive clock recovery method based on time stamp of novelty to present development of internet technology situation.This method need be in packet with respect to common time stamp method explicit transmission time stab information, and more stable; With respect to ordinary buffer district method, the locking time of this method is shorter, uses more flexible.

Technical scheme of the present invention is following: a kind of towards the self-adapted clock method of ethernet circuit artificial service based on time stamp, comprise the steps:

(1) when the Ethernet side is received packet, the counter that a high frequency clock drives is sampled and conduct reception time stamp RX _mDeposit the time stamp buffering area in; When the TDM side begins to send a TDM unit, equally this counter is sampled, and with sampled value TX _mDeposit in the time stamp buffering area as sending time stamp;

When (2) business clock was unequal at two ends, adjacent two received time stamp RX _mDifference DELTA RX _mWith adjacent two transmission time stamp TX _mDifference DELTA TX _mBetween difference be dRX _m, use kalman filter method and from one group of distortion time stamp that causes because of network delay, extract real dRX _m

(3) with real dRX _mControl error signal controlling VCO sends PE and the business clock frequency difference that receives PE thereby eliminate.

In general, necessarily there is a controllable part VCO inside, supposes that this controllable part is a VVC voltage variable capacitance, with dRX _m, amplifier is input to D/A converting circuit after amplifying, from the voltage control variable capacitance of D/A converting circuit output, thus the output of control VCO.

Aforesaid towards the self-adapted clock method of ethernet circuit artificial service based on time stamp, wherein, the employed process of kalman filter method with the measurement equation is in the step (2):

G _k＝P _k-1/(P _k-1+R)

DRX _k＝DRX _k-1+G _k(M _k-DRX _k-1)

P _k＝(1-G _k)P _k-1

Order

$M_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{Z}_m-\mathrm{n\μ}$

${\mathrm{DRX}}_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{\mathrm{dRx}}_m$

Wherein, Z _mBe dRX _mMeasured value,

Information according in the time stamp buffering area obtains P ₀And R, utilize iterative algorithm to obtain DRX _k, further obtain dRX _m, P ₀Be the posteriority estimate variance of initial condition, R is the mean square deviation that is the measurement noise that just too distributes.

The present invention utilizes local service clock sum counter to obtain representing the time stab information (local time stamp) of far-end and near-end business clock; Time stab information is transformed into the random sequence that satisfies Kalman's process and measurement equation through statistical calculation; And utilize Kalman filtering algorithm filtered noise from this random sequence; Obtain true time stamp, make local service clock tracing far-end business clock thereby utilize the time stab information that obtains to regulate voltage controlled oscillator.This method adopts Kalman filtering algorithm to make loop filtering can select to realize or realize by means of software with hardware, should use more flexibly, and the frequency difference monotone decreasing does not have the generation of shake basically, comparatively fast reaches keyed end.It is long and need deficiency such as additional networks bandwidth that the present invention can overcome locking time, has good performance, and with respect to ordinary buffer district method and common time stamp method, combination property of the present invention is stronger.

Description of drawings

Fig. 1 is the illustrative view of functional configuration of adaptive clock recovery method.

Fig. 2 is the illustrative view of functional configuration of KBLTS method of the present invention.

Fig. 3 is for using the hardware structure diagram of emulator of the present invention.

Fig. 4 is the emulation testing topological diagram of emulator.

Fig. 5 is the frequency difference of the present invention sketch map of situation of change in time.

Embodiment

Below in conjunction with accompanying drawing and embodiment the present invention is carried out detailed description.

A kind of as shown in Figure 2 based on the functional structure of the self-adapted clock method (KBLTS) of time stamp towards the ethernet circuit artificial service.

Receiving PE (for explaining conveniently; The PE that the present invention will implement the self-adaptive recovery clock is called reception PE; The PE that communicates with it is called transmission PE; In fact all PE send PE, also can be to receive PE), receive buffering area and be used for TDM information that packet that buffer memory arrives from the Ethernet side carries to absorb the delay variation that network causes.The Ethernet side periodically writes data to jitter-removing buffer, and the TDM side constantly from the jitter-removing buffer reading of data to recover the TDM code stream.When the business clock at two ends there are differences; Cause the speed of the read-write operation of jitter-removing buffer there are differences; This species diversity accumulation will cause two kinds of possible results: buffering area completely causes loss of data, and buffer empty causes the space occurring between the TDM code stream, finally all shows as error code.The final function of adaptive clock recovery is exactly to make the business clock frequency at two ends identical, thereby avoids above-mentioned risk and other possible mistake.

Receive the VCO output high frequency clock (more than the 50MHz) of PE, it is as the input of counter, simultaneously behind the frequency division as the business clock f2 of TDM.The TDM load of here an Ethernet data bag being carried is called TDM unit (can be one or more TDM time slots).When the Ethernet side is received packet, counter is sampled and conduct reception time stamp RX _m(m=1,2... descend together) deposit the time stamp buffering area in.On the other hand, when the TDM side begins to send a TDM unit, equally counter is sampled, and with sampled value TX _mDeposit in the time stamp buffering area as sending time stamp.

Output at VCO is under the situation of stable state, and one group is sent time stamp and constitutes an arithmetic progression.That is to say Δ TX _m=TX _m-TX _M-1It is constant (TX representes with Δ).

Consider ideal situation earlier, in the ideal case, network delay is 0, and f2 and f1 can be regarded as constant.

A group of received time stamp constitutes an arithmetic progression equally so.Difference then is that transmitting terminal sends the result of the cycle of two adjacent data bags with the output tolerance of receiving terminal VCO.In other words, Δ RX _m=RX _m-RX _M-1Also be that a constant (RX representes with Δ) it and transmitting terminal business clock frequency f 1 are proportional.

When business clock is unequal at two ends, there are differences between two differences, that is to say dRX _m=Δ RX _m-Δ TX _m≠ 0, but can regard it as constant (representing) with dRX.With dRX _mControl VCO then can make f2=f1.In other words, under the ideal situation, receive two packets, the deviation of utilizing the time stamp of these two packets promptly can correct the two ends business clock.

Situation for real network is discussed again, can be known that by document [13] [14] long-term frequency departure of clock is minimum, therefore in a relatively short time, f2 and f1 still can be regarded as constant.But network delay then can not be ignored, and has randomness based on the network delay of the asynchronous Ethernet of statistic multiplexing, and the reception time stamp that therefore obtains at receiving terminal is distortion.The business clock information that can not directly reflect transmitting terminal.With Δ R ' X _mExpression Δ RX in this case _m, then can obtain:

AR′X _m＝ΔRX _m+d _m (1)

(1) d in the formula _mThe expression network delay is to Δ RX _mInfluence.

With d ' RX _mExpression dRX in this case _m, then

d’RX _m＝dRX _m+d _m (2)

Therefore real time stab information need from the time stab information of a series of distortions, be recovered, just according to d ' RX _mObtain dRX _m, the Kalman filter among Fig. 2 is just for realizing that this function designs.Basically equate with dRX from the signal of Kalman (Kalman) filter output, then can reach the target of f2=f1 with this signal controlling VCO.

● the introduction of Kalman Filtering for Discrete model is following:

The subject matter that Kalman filter solves is: attempt to estimate the quantity of state x in the discrete time control procedure.

This discrete time control procedure can use following difference equation to describe.

y _k＝Ay _k-1+Bu _k+w _k-1 (3)

The observed quantity z of this discrete time control procedure can represent with following formula.

z _k＝Hy _k+v _k (4)

(3)-(4) in the formula, y _kBe the n n dimensional vector n, z _kIt is the m n dimensional vector n.A is n * n matrix.Optional controlled quentity controlled variable u _kIt is the l dimensional vector.B is n * l matrix.H is m * n matrix.

Stochastic variable w _k(n n dimensional vector n), v _k(m n dimensional vector n) is respectively process and measurement noise.They obey the normal distribution that (5) formula and (6) formula are represented respectively.

p(w)～N(0，Q) (5)

p(v)～N(0，R) (6)

The time renewal equation that obtains Kalman Filtering for Discrete based on (3)-(4) formula is shown in (7) formula and (8) formula.

${\stackrel{\′}{y}}_k^{-}=A{\stackrel{\′}{y}}_{k-1}+{\mathrm{Bu}}_k---\left(7\right)$

$P_k^{-}={\mathrm{AP}}_{k-1}{A}^T+Q---\left(8\right)$

(7)-; (8)

and

is respectively the priori estimates and the posterior estimate in k-1 step in k step in the formula.P _k ^-And P _K-1, be respectively the prior estimate variance and the k-1 step posteriority estimate variance in k step.

The measurement renewal equation that obtains Kalman Filtering for Discrete based on (3)-(4) formula equally is shown in the formula of (9)-(11).

$g_k=P_k^{\′}{H}^T{({\mathrm{HP}}_k^{\′}{H}^T+R)}^{-1}---\left(9\right)$

$P_k=(1-g_{k}H)P_k^{-}---\left(11\right)$

(9)-(11) n n dimensional vector n g in the formula _kBe kalman gain, it can make the variance of posterior estimate minimum as regulatory factor.Based on time renewal equation and measurement renewal equation, use iterative algorithm can make posterior estimate

progressively approach its actual value.Document [15] is described in detail and analyzes Kalman filter.

● the method for application card Kalman Filtering is following in the KBLTS method:

Discussion by the front can know that the key of KBLTS method is from the time stamp of one group of distortion, to extract real dRX _m(12)-(13) two formulas can be described the character that receives time stamp:

dRX _m＝dRX _m-1 (12)

Z _m＝d′RX _m＝dRX _m+d _m (13)

Z in the formula _mThe dRX that can regard as _mMeasured value.

Discussion by the front can know that (12) formula can be regarded A=1 as, B=0, Kalman's process equation of Q=0.Measurement equation when (13) formula is also with H=1 is very similar.But stochastic variable d _m(0, R) distributional assumption is then not necessarily set up to defer to N.

Can know by central-limit theorem: if stochastic variable X ₁, X ₂... X _nSeparate, obey same distribution, and the mathematic expectaion of each stochastic variable and variance are respectively μ and σ ², so:

$\mathrm{Yn}=\underset{n\→\∞}{\lim }\left(\frac{\underset{m=0}{\overset{n}{\mathrm{\Σ}}}{X}_m-\mathrm{n\μ}}{\sqrt{n}\mathrm{\σ}}\right)~N\left(\mathrm{0,1}\right)---\left(14\right)$

Set up, that is to say, when n-＞∞, Y _nObey (0,1) normal distribution.

Since stochastic variable d1, d2...d _mBe the propagation delay time in the same time period on consolidated network, can regard independent identically distributed stochastic variable as.According to law of great number, when n was big, its mathematic expectaion and variance can be by the following formula approximate representations:

$\mathrm{\μ}\≈\left(\underset{m=0}{\overset{n}{\mathrm{\Σ}}}{d}_m\right)/n---\left(15\right)$

${\mathrm{\σ}}^2\≈\left(\underset{m=0}{\overset{n}{\mathrm{\Σ}}}{(d_m-\mathrm{\μ})}^2\right)/n---\left(16\right)$

By (14)-(16) Shi Kede, when n was enough big, following formula is approximate to be set up:

$j_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}\mathrm{dm}-\mathrm{n\μ}~N(0,R)---\left(17\right)$

Order

$M_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{Z}_m-\mathrm{n\μ}---\left(18\right)$

${\mathrm{DRX}}_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{\mathrm{dRx}}_m---\left(19\right)$

Obtain on this basis:

DRX _k＝DRX _k-1 (20)

M _k＝DRX _k+j _k (21)

(20) formula meets the condition of Kalman's process equation, and (21) formula also can be regarded H=1, j as _k(0, Kalman R) measures equation to ~ N.Therefore can use Kalman filtering algorithm.

So it is following to obtain the time renewal equation of Kalman filtering of KBLTS method:

DRX′ _k＝DRX _k-1 (22)

P′ _k＝P _k-1 (23)

The measurement renewal equation of Kalman filtering that obtains the KBLTS method is following:

G _k＝P′ _k/(P′ _k+R) (24)

DRX _k＝DRX′ _k+G _k(M _k-DRX′ _k)?(25)

P _k＝(1-G _k)P′ _k (26)

Obtain by the simplification of (22)-(26) formula:

G _k＝P _k-1/(P _k-1+R) (27)

DRX _k＝DRX _k-1+G _k(M _k-DRX _k-1)?(28)

P _k＝(1-G _k)P _k-1 (29)

Information according in the time stamp buffering area obtains P ₀And R, then utilize iterative algorithm can obtain DRX _k, further can obtain dRX _m, then be exactly the information that we need obtain.

In the present invention, definition P _kBe respectively k step posteriority estimate variance, P so ₀Be the posteriority estimate variance of initial condition, R is the mean square deviation that is the measurement noise that just too distributes.P ₀Can not directly obtain with R, need obtain through statistical computation based on the measured value in early stage.Such as, can make algorithm begin to carry out from obtaining n sample, the individual sample of 0--(n-1) can be as calculating P so ₀Value with R.

● the realization of KBLTS method and testing evaluation method are following:

Being without loss of generality, is representative design and simulation testing equipment on the 100 m ethernet network, to transmit E1 here, and the main hardware structure of emulator is as shown in Figure 3.In emulator, E1 line interface unit (LIU, Line Interface Unit) is connected with FPGA (Field-programmable gate array), and FPGA is connected to Ethernet switching chip through 100,000,000 PHY, is connected to CPU through pci bus simultaneously.Ethernet switching chip also is connected with CPU by pci bus, is connected with optical module by 100,000,000 PHY simultaneously.

FPGA realizes two-way mapping function, communication function, time stamp systematic function and the clock complex functionality etc. of pci interface function, TDM and Ethernet data bag.

FPGA receives TDM information and is mapped to the Ethernet data bag from LIU, and through Ethernet switching chip packet is sent on the ethernet network.Simultaneously, FPGA receives the Ethernet data bag from ethernet network, from the Ethernet data bag, extracts tdm data and deposits jitter-removing buffer in, the recorder time stab information.Fetch data from jitter-removing buffer simultaneously, utilize the local service clock to send to LIU, the corresponding time stamp that sends of record.Sending time stamp obtains by the local VCO actuation counter of 100M with the reception time stamp.

CPU then constantly reads from FPGA and receives time stamp and send time stab information, utilizes software to extract real dRX based on kalman filter method, and CPU writes dRX in the current time stamp difference register of FPGA then, and kind is used when supplying FPGA to regulate.

The dRX that FPGA writes back according to CPU combines digital-to-analogue conversion and gain logic, control VCO output local service clock.

For the self-adapted clock method with other compares, realized a kind of common buffer zone method and a kind of common time stamp method simultaneously, these two kinds of methods all adopt hardware low pass filter to come filtered noise.

The used network topology of emulation testing is as shown in Figure 4.

Emulator among Fig. 4 has functional structure shown in Figure 3.Import the time delay distribution of background traffic to obtain being similar at random with Smartbits during test with real network.Two E1 interfaces of test instrumentation ANT-20E are connected respectively to the E1 interface of two ends emulator.

When ANT-20E sends data, observe and write down measurement and the record of ANT-20E to receive clock.Successively emulator is arranged to ordinary buffer district method, common time stamp method and KBLTS method and logging test results, can obtains Fig. 5 according to test result.

Transverse axis express time (second) among Fig. 5, the longitudinal axis is represented time dependent two ends frequency difference.Can find out that from Fig. 5 for the KBLTS method, when clock original frequency difference was near 35ppm at two ends, after about 10 seconds, the two ends clock frequency was basic identical.And ordinary buffer district method is more much longer than KBLTS method locking time, and common time stamp method time of locking is basic suitable with the KBLTS method, but it is stable to be not so good as the KBLTS method.

KBLTS adopts local high frequency clock sampling is obtained receiving and sending time stamp respectively, thereby making to not be used in transmits time stab information and saved the use to the network bandwidth on the network.Adopt Kalman filtering algorithm to make loop filtering can select to realize or realize, should use more flexible by means of software with hardware.Can know from Fig. 5, adopt the KBLTS method, the frequency difference monotone decreasing does not have shake basically and takes place, and comparatively fast reaches keyed end.Obviously, it is long and need deficiency such as additional networks bandwidth that the KBLTS method can overcome locking time, has good performance.With respect to ordinary buffer district method and common time stamp method, the combination property of KBLTS method is better.

List of references

[1]Metro?Ethernet?Forum，MEF?3，Circuit?Emulation?Service?Definitions，Framework?andRequirements?in?Metro?Ethernet?Networks，2004.

[2]Metro?Ethernet?Forum，MEF?8，Implementation?Agreement?for?the?Emulation?of?PDH?Circuitsover?Metro?Ethernet?Networks，2004

[3]X.Xiao，D.McPherson，P.Pate，IETF?Rfc?3916，Requirements?for?Pseudo-Wire?EmulationEdge-to-Edge(PWE3)，2004

[4]S.Bryant，P.Pate，IETF?Rfc?3985，Pseudo?Wire?Emulation?Edge-to-Edge(PWE3)Architecture，2005

[5]M.Riegel，IETF?Rfc?4197，Requirements?for?Edge-to-Edge?Emulation?of?Time?DivisionMultiplexed(TDM)Circuits?over?Packet?Switching?Networks，2005

[6]INTERNATIONAL?TELECOMMUNICATION?UNION，ITU-T?G826.1/Y.136l，Ethernetover?Transport?aspects-Quality?and?availability?targets：Timing?and?synchronization?aspects?inPacket?Networks，2006

[7]INTERNATIONAL?TELECOMMUNICATION?UNION，ITU-TG.823，The?control?ofjitter?and?wander?within?digital?networks?which?are?based?on?the?2048kbit/s?hierarchy，2000

[8]INTERNATIONAL?TELECOMMUNICATION?UNION，ITU-T?G.824，The?control?ofjitter?and?wander?within?digital?networks?which?are?based?on?the?1544?kbit/s?hierarchy，2000

[9]Luis?Castel-Branco，Mario?Serafim?Nunes.Clock?recovery?for?circuit?emulation?servicesover?ATM?Broadband?Communications，1996.Global?Infrastructure?for?the?Information?Age.Proceedings?of?the?International?IFIP-IEEE?Conference?on?23-25?April?1996?Page(s)：617-625

[10]Fukada，Y.，Yasuda，T.，Komatsu，S.et?al.Adaptive?Clock?Recovery?Method?UtilizingProportional-Integral-Derivative(PID)Control?for?Circuit?Emulation，Communications，2005Asia-Pacific?Conference?on?03-05?Oct.2005?Page(s)：19-23

[11]Fujikawa，F.，Serizawa，Y.Experimental?considerations?on?adaptive?clock?method?fortransmitting?microprocessor-based?current?differential?teleprotection?signals?in?ATM?networksPower?Engineering?Society?1999?Winter?Meeting，IEEE?Volume?1，31?Jan.-4?Feb.1999Page(s)：229-234?vol.1?Digital?Object?Identifier?10.1109/PESW.1999.747459

[12]James?Aweya，Delfin?Y.Montuno，Michel?Ouellette，et?al.Clock?synchronization?forpacket?networks?using?a?weighted?least-squares?error?filtering?technique?and?enabling?circuitemulation?service，International?Journal?of?Communication?Systems?Early?View

[13]INTERNATIONAL?TELECOMMUNICATION?UNION，ITU-T?G.811，Timingcharacteristics?of?primary?reference?clocks，1997

[14]INTERNATIONAL TELECOMMUNICATION?UNION，ITU-T?G.812，Timingrequirements?of?slave?clocks?suitable?for?use?as?node?clocks?in?synchronization?networks，1998

[15]J.Huston McCulloch，The?Kalman?Foundations?of?Adaptive?Least?Squares，http://econ.ohio-state.edu/jhm/papers/KalmanAL.pdf

Claims (2)

1. one kind towards the self-adapted clock method of ethernet circuit artificial service based on time stamp, comprises the steps:

(1) when the Ethernet side is received packet, the counter that a high frequency clock drives is sampled and conduct reception time stamp RX _mDeposit the time stamp buffering area in; When time division multiplexing TDM side begins to send a TDM unit, equally this counter is sampled, and with sampled value TX _mDeposit in the time stamp buffering area as sending time stamp;

When (2) business clock was unequal at two ends, adjacent two received time stamp RX _mDifference DELTA RX _mWith adjacent two transmission time stamp TX _mDifference DELTA TX _mBetween difference be dRX _m, use kalman filter method and from one group of distortion time stamp that causes because of network delay, extract real dRX _mSaid two ends refer to two network equipments that the needs that connect through ethernet network carry out clock synchronization, promptly need connect the sending ending equipment and the receiving device of transceive data through ethernet network;

(3) with real control error signal dRX _mControl voltage controlled oscillator VCO, thereby the business clock frequency difference of elimination transmission PE and reception PE, PE is meant provider edge equipment, and the PE that implements the self-adaptive recovery clock is called reception PE, and the PE that communicates with reception PE is called transmission PE.

2. as claimed in claim 1 towards the self-adapted clock method of ethernet circuit artificial service based on time stamp, it is characterized in that: the employed measurement equation of kalman filter method is in the step (2):

G _k＝P _k-1/(P _k-1+R)

DRX _k＝DRX _k-1+G _k(M _k-DRX _k-1)

P _k＝(1-G _k)P _k-1

Kalman gain G _KBe an intermediate quantity in the computational process, the purpose that calculates this intermediate quantity is in order to calculate DRXk and P _k, G _kCalculating and P _kIt is relevant with the value of R,

M _kBe when calculating for the k time n measured value with respect to the sum of errors of mathematic expectaion, DRX _kN dRX measured value sum when being the k time calculating,

Definition P _kBe respectively k step posteriority estimate variance, P so ₀Be the posteriority estimate variance of initial condition, R is the mean square deviation that is the measurement noise of normal distribution, P ₀Can not directly obtain with R, need obtain through statistical computation based on the measured value in early stage,

Order

$M_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{Z}_m-\mathrm{n\μ}$

${\mathrm{DRX}}_k=\underset{m=\mathrm{kn}}{\overset{(k+1)n}{\mathrm{\Σ}}}{\mathrm{dRx}}_m$

Wherein, Z _mBe dRX _mMeasured value, the value of n, k is the positive integer of non-zero, u is the mathematic expectaion of stochastic variable,

Information according in the time stamp buffering area obtains P ₀And R, utilize iterative algorithm to obtain DRX _k, further obtain dRX _m, P ₀Be the posteriority estimate variance of initial condition, R is the mean square deviation that is the measurement noise of normal distribution.

Images