Time Interval Error Jitter Extraction Method Based on Hilbert Transform

Abstract

The invention discloses a time interval error jitter extraction method based on Hilbert transform, and relates to the field of high-speed data signal processing. After obtaining the measured data signal, firstly, a level histogram is used for statistics, and a corresponding level judgment reference is determined; secondly, the measured data signal is subjected to Hilbert transform to obtain the time information of the transition edge of the measured data signal; a "coarse frequency estimation + fine frequency estimation" method is used to extract the accurate symbol rate of the measured data signal, and a high-precision constant frequency clock is recovered; and high-precision time interval error is extracted according to the above-mentioned recovered clock. The time interval error jitter extraction method based on Hilbert transform provided by the invention provides a basis for high-precision time interval error jitter, and accurately performs jitter measurement on the measured data signal.

Description

Technical Field

[0001] The invention relates to the field of high-speed data signal processing, and in particular to a method for extracting time interval error jitter based on Hilbert transform. Background Art

[0002] The high-speed serial data link SerDes system consists of a transmitter, a channel, and a receiver, wherein the channel may include any physical transmission medium, including copper wire, coaxial cable, optical fiber, and other media.

[0003] Jitter is a core indicator for measuring the stability and reliability of data transmission. It plays an important role in the test and analysis of very large-scale integrated circuits, high-speed serial links, and cloud computing. Time Interval Error (TIE) was originally defined in the G.810 Recommendation issued by the Telecommunication Standardization Department of the International Telecommunication Union, and was subsequently widely used in test instruments. TIE is used as the basic unit for analyzing data jitter. In serial data systems, TIE jitter refers to the phase difference between the edge of the data signal and the edge of the clock signal.

[0004] Since high-speed serial data signals generally use embedded clock transmission, after receiving the data signal at the receiving end, it is necessary to extract the clock signal embedded in the transmitted data and perform subsequent data recovery, jitter decomposition measurement and other signal processing on the data signal according to the extracted clock signal. Usually, there are two ways to perform clock recovery at the receiving end, namely, constant frequency clock recovery and phase-locked loop-based clock recovery.

[0005] The clock recovery method based on the phase-locked loop can follow or track the low-frequency jitter changes in its input signal. Usually, in order to analyze all jitter components in the data signal, such as random jitter and data-related jitter, the constant-frequency clock recovery method is usually used to extract TIE. Therefore, extracting accurate time interval error TIE from the data signal is the premise of jitter analysis measurement technology, and the accuracy of constant-frequency clock recovery also determines the accuracy of jitter analysis measurement results. High-precision TIE extraction is an important part of accurately evaluating system performance, locating jitter, and improving system design.

[0006] Application number 202011468743.3 published: "A PAM4 signal clock data recovery method based on software phase-locked loop" is based on the edge crossing method, and the phase-locked loop-based clock recovery is realized by extracting the synchronous clock from the data signal. Under normal circumstances, the TIE extracted by the phase-locked loop-based clock recovery will track the low-frequency jitter of the data signal, that is, ignore the low-frequency jitter in the data signal, and cannot extract all the jitter information contained in the data signal. In particular, when the data signal has severe inter-symbol interference, crosstalk and other influences, the clock frequency obtained based on the edge crossing method has a large deviation.

[0007] Application No. 201610287412.7 published: "Improved clock recovery for data signals", based on the technology of spectral line method, determines edge information by taking derivative of data signal, and applies a band-pass filter with flat passband to data signal with SSC spread spectrum clock modulation, so as to improve the eye diagram of data signal after the recovered clock and increase the eye opening. When the data signal has a low signal-to-noise ratio, there is a large error in determining the data conversion edge by taking derivative.

[0008] Existing clock recovery methods are basically based on phase-locked loop clock recovery methods, and do not mention constant frequency clock recovery methods, as well as the extraction of time interval error TIE for analyzing various components of jitter.

[0009] Since the channel generally exhibits a low-pass characteristic, the inter-symbol interference caused by channel insertion loss and reflection, or the signal degradation caused by crosstalk from adjacent channels and noise from other sources becomes more obvious. Some points in the edge crossing will be lost in the received signal, thereby introducing errors to the traditional clock recovery based on edge crossing time. The time interval error extracted by the clock recovered by this means will have a greater impact on the jitter decomposition and measurement based on the time interval error. Summary of the invention

[0010] The purpose of the present invention is to: in view of the above-mentioned problems, the present invention provides a time interval error jitter extraction method based on Hilbert transform, performs high-precision constant frequency clock recovery on the measured data signal, and obtains time interval error information that retains all jitter components.

[0011] The technical solution adopted by the present invention is as follows:

[0012] The time interval error jitter extraction method based on Hilbert transform includes:

[0013] Perform level histogram statistics on the measured data signal to determine the decision level;

[0014] Performing Hilbert transform on the measured data signal to obtain the pulse width of the measured data signal;

[0015] By performing histogram statistics on the pulse width, the frequency of the recovered clock is roughly estimated to obtain a rough frequency value of the recovered clock;

[0016] Determining an accurate frequency estimate of the recovered clock using a coarse frequency value of the recovered clock;

[0017] The constant frequency clock is recovered by using the accurate frequency estimation value of the recovered clock to obtain the constant frequency clock information, and the time interval error is extracted using the constant frequency clock information.

[0018] Furthermore, the acquisition of pulse width includes: performing Hilbert transform on the measured data signal, and differentiating the imaginary part of the Hilbert transform to obtain multiple extreme points, the horizontal coordinates of the multiple extreme points are the time information of the transition edges of the measured data signal, and the interval between adjacent transition edges is the pulse width of the current measured data signal.

[0019] Furthermore, the pulse width is subjected to histogram statistics to obtain a pulse width histogram, and the pulse width histogram is used to roughly estimate the recovery clock rate to obtain a rough frequency value of the recovery clock.

[0020] Furthermore, the rough estimation includes making a rough estimation of the recovered clock rate by a weighted averaging method.

[0021] Furthermore, the data symbol rate is accurately estimated based on the coarse frequency value of the recovered clock to obtain an accurate frequency estimation value of the recovered clock.

[0022] Furthermore, obtaining the accurate frequency estimation value of the recovered clock includes:

[0023] The reciprocal of the coarse frequency value of the recovered clock is calculated to obtain the period of the recovered clock, the number of clocks in each pulse width is estimated, that is, the number of code elements contained in each pulse width, and all pulse widths are traversed to obtain the total number of code elements;

[0024] The accurate frequency estimation value of the recovered clock is calculated based on the difference between the last time jump information and the first time jump information of the measured data signal.

[0025] Furthermore, obtaining the time interval error includes:

[0026] The cross point of the recovered clock and the decision level is solved, and the cross point of the measured data signal and the decision level is solved, and the difference between the corresponding cross points is the time interval error TIE.

[0027] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0028] When the present invention adopts the method based on Hilbert transform to perform clock recovery, the phenomenon of edge crossing point missing can be effectively solved, and accurate data symbol conversion time can be obtained.

[0029] The present invention adopts the method of performing Hilbert transform on the data, and uses the imaginary part of the Hilbert transform to contain the amplitude and phase information of the data signal, and its extreme point represents the time information of the jump of the data signal, so as to obtain a more accurate data code element width;

[0030] The present invention proposes a method of "coarse frequency estimation + fine frequency estimation" for data symbol rate, which accurately estimates the data symbol rate, provides a basis for extracting high-precision time interval error jitter, and improves the accuracy of jitter measurement and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a functional module block diagram of the time interval error jitter extraction method based on Hilbert transform of the present invention;

[0032] Figure 2 It is a flow chart of a method for extracting time interval error jitter based on Hilbert transform of the present invention;

[0033] Figure 3 is a diagram of the measured data signal collected in Example 2 of the present invention;

[0034] Figure 4 is a diagram showing a data signal and an imaginary part of a Hilbert transform in Embodiment 2 of the present invention;

[0035] Figure 5 It is the pulse width statistical histogram of the traditional method and the method of the present invention;

[0036] Figure 6 It is a closed data eye diagram of a constant frequency clock recovered based on traditional histogram pulse width statistics according to embodiment 2 of the present invention;

[0037] Figure 7 It is an opened data eye diagram of the constant frequency clock recovery based on the “coarse frequency estimation + fine frequency estimation” method in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0039] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0040] Example 1

[0041] This embodiment provides a method for extracting time interval error jitter based on Hilbert transform, including:

[0042] Acquiring a data signal includes acquiring the measured data signal from an acquisition device, such as a real-time oscilloscope or an ADC acquisition module, performing level histogram statistics on the measured data signal PAM-N, and determining basic information such as a corresponding decision level.

[0043] Hilbert transform, the imaginary part of the Hilbert transform contains the amplitude and phase information of the measured data signal, and its extreme point represents the time information when the data signal jumps; the time information of all extreme points is extracted, and the time information of adjacent extreme points is the pulse width of the current data.

[0044] Perform histogram statistics on all pulse widths, and use the weighted average method to roughly estimate the frequency of the recovered clock, that is, search the statistical pulse histogram to obtain the first maximum pulse width, and take its derivative as the rough frequency estimate of the measured data signal, as shown in the following formula:

[0045] (1)

[0046] In the formula, To recover a rough estimate of the clock rate, is the pulse width, is the statistical probability at the normalized minimum pulse width histogram, i is a natural number, is the number of boxes divided at the minimum pulse width histogram;

[0047] Furthermore, the coarse frequency value of the recovered clock is inversely calculated to obtain the period of the recovered clock, and the number of clocks in each pulse width, that is, the number of code elements contained in each pulse width, is estimated. All pulse widths are traversed to obtain the total number of code elements. According to the difference between the last time jump information and the first time jump information of the measured data signal, the precise frequency estimation value of the recovered clock is calculated, as shown in the following formula:

[0048] (2)

[0049] In the formula, To recover an accurate frequency estimate of the clock signal, , are respectively the last and first data signal transition times obtained based on Hilbert transform;

[0050] The constant frequency clock is extracted according to the accurate estimated value of the recovered clock frequency, and the intersection points of the recovered clock and the decision level, and the intersection points of the measured data signal and the decision level are obtained by using the interpolation algorithm. The interpolation of the corresponding intersection points is the time interval error.

[0051] According to the above-extracted constant frequency clock information, the data symbols are cut into symbols and the eye diagram is displayed, and the relevant eye diagram information is calculated, including obtaining basic information such as the data signal eye width, eye height, signal-to-noise ratio, etc.

[0052] Example 2

[0053] like Figure 1 FIG. 1 is a functional modular block diagram of a time interval error jitter extraction method based on Hilbert transform in this embodiment. In this embodiment, at the data acquisition module 100, a measured data signal is acquired, for example, Figure 3 The waveform shown depicts the acquired measured data signal; through the clock recovery unit module 105, the high-precision clock recovery method proposed in the method of this embodiment can be implemented to extract the code element rate in the measured data signal, that is, the coarse / precise frequency of the recovered clock. Further, the measured data signal can be cut into code elements through the high-precision recovered clock, and the measured data signal can be recovered and decoded through the data recovery module 110, and the eye diagram can also be displayed and measured through the eye diagram measurement module 115; through the TIE extraction unit module 120, the data TIE can be extracted based on the high-precision recovered clock, and the jitter measurement analysis containing all jitter components can be performed.

[0054] like Figure 2 The flowchart of the specific implementation of the method of this embodiment is shown. After the measured data signal is obtained, the corresponding level judgment benchmark and other basic information are first determined through level histogram statistics; secondly, the measured data signal is Hilbert transformed to obtain the time information of the transition edge of the measured data signal; the "coarse frequency estimation + fine frequency estimation" method is used to extract the accurate code element rate of the measured data signal, and high-precision constant frequency clock recovery is performed; high-precision TIE extraction is performed based on the above-mentioned recovered clock, and further, the rising and falling edge information of the recovered clock can be used to perform code element cutting on the measured data signal, thereby obtaining basic information such as the data signal eye width, eye height, and signal-to-noise ratio.

[0055] like Figure 3 The figure depicts the characteristics of a signal with high intersymbol interference. Due to the low-pass filtering characteristics of the transmission line, the signal cannot reach a stable final value in a short time, causing the amplitude of the high-speed data signal to change with the bit sequence pattern. When there is a long series of "1", the amplitude of the data signal will eventually rise to a steady-state high level, such as Figure 3 As shown in point A in the figure, when the subsequent data state switches to "0", as shown in Figure 3As shown in point B, because the previous long string of "1" makes the initial level of the current "0" relatively high, "0" will experience a relatively long switching time to reach the threshold value, and it will take a relatively long time to reach its true low level. Therefore, the time for the "0" to cross the threshold lags behind the ideal moment, and the amplitude is also relatively high. The next "1" is as follows Figure 3 Point C makes the "0" signal at point B switch back again before reaching its ideal low level, so the "1" signal is very close to its threshold value, and the time it crosses the threshold must be earlier than the ideal moment. Therefore, the traditional pulse width statistics based on the edge crossing method will have problems such as edge missing and pulse width statistics being too small, resulting in serious distortion of the data symbol rate estimation; the method of this embodiment uses the Hilbert transform method with edge information recognition to realize the statistics of data pulse width, such as Figure 4 As shown, the imaginary part of the Hilbert transform can well identify the time information of the edge transition of the data signal and obtain more accurate pulse width statistics; compared with the traditional spectral line method of obtaining the edge transition of the data signal, this method has better noise tolerance.

[0056] In order to obtain the symbol rate of the measured data signal, traditional methods such as Figure 5 As shown in the figure, the position of the first peak in the histogram is the width of 1 UI, the position of the second peak is the width of 2 UI, and so on. Therefore, the symbol width of the data signal is T=L1+L2 / 2+L3 / 3+L4 / 4. The data symbol rate estimated by this method has a large error. When the constant frequency clock restored according to this rate is used for symbol cutting, it will seriously cause the data eye diagram to be closed, as shown in the figure. Figure 6 As shown. Furthermore, if the TIE extracted by the constant frequency clock according to this recovery method will also cause a large error, it will be impossible to perform correct jitter measurement and analysis; the method of this embodiment proposes a "coarse frequency estimation + fine frequency estimation" method to estimate the symbol rate, obtain a high-precision constant frequency recovery clock, and extract accurate TIE data for jitter measurement and analysis. When the constant frequency recovery clock obtained by the method of this embodiment is used to cut the data symbol, the eye diagram obtained is obviously open, as shown in Figure 7 As shown, more accurate basic information such as data signal eye width, eye height, signal-to-noise ratio, etc. can be obtained.

[0057] In summary, the advantage of using Hilbert transform in the present invention is that when the quality of the transmission channel is poor, there may be high ISI and noise in the data signal, resulting in the data being converted to the next symbol codeword level before it passes through the level decision reference. When the edge crossing method is used for clock recovery, the edge crossing point missing phenomenon will occur, and the codeword width obtained based on the edge crossing method is significantly smaller than the nominal codeword width of the data. When the Hilbert transform-based method is used for clock recovery, the edge crossing point missing phenomenon can be effectively solved, and the accurate data codeword conversion time can be obtained.

[0058] When the data signal has high noise, the traditional spectrum line method based on the derivative method is noise-sensitive and will amplify the high-frequency components and noise in the signal. Even slight noise may lead to erroneous edge detection or misdetection. If there is jitter in the signal, even a slight change in frequency or phase will cause edge information to be lost or blurred. The present invention adopts the method of performing Hilbert transform on the data, and uses the imaginary part of the Hilbert transform to contain the amplitude and phase information of the data signal. Its extreme point represents the time information of the jump of the data signal, which can obtain a more accurate data codeword width;

[0059] Traditional estimation based on data symbol width adopts the method of finding the maximum value of pulse width histogram statistics, which will have a large error when estimating the accurate data symbol rate, not only causing symbol cutting errors based on the recovered clock and eye diagram closure, but also causing large errors in jitter decomposition measurement based on TIE, which may cause directional errors in high-speed serial link feedback design; the present invention proposes a method of "coarse frequency estimation + fine frequency estimation" for data symbol rate, accurately estimating the data symbol rate, providing a basis for extracting high-precision time interval error jitter, and improving the accuracy of jitter measurement analysis.

[0060] The principles and implementation methods of the present invention are described in this article using specific embodiments. The description of the above embodiments is only used to help understand the method and core ideas of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A time interval error jitter extraction method based on Hilbert transform, characterized in that: include: Perform level histogram statistics on the measured data signal to determine the decision level; Performing Hilbert transform on the measured data signal to obtain the pulse width of the measured data signal; By performing histogram statistics on the pulse width, the frequency of the recovered clock is roughly estimated to obtain a rough frequency value of the recovered clock; Determining an accurate frequency estimate of the recovered clock using a coarse frequency value of the recovered clock; The constant frequency clock is recovered by using the accurate frequency estimation value of the recovered clock to obtain the constant frequency clock information, and the time interval error is extracted using the constant frequency clock information.

2. The method for extracting time interval error jitter based on Hilbert transform according to claim 1, characterized in that: The acquisition of the pulse width includes: performing a Hilbert transform on the measured data signal, and differentiating the imaginary part of the Hilbert transform to obtain multiple extreme points, the horizontal coordinates of the multiple extreme points are the time information of the transition edges of the measured data signal, and the interval between adjacent transition edges is the pulse width of the current measured data signal.

3. The method for extracting time interval error jitter based on Hilbert transform according to claim 2, characterized in that: The pulse width is subjected to histogram statistics to obtain a pulse width histogram, and the pulse width histogram is used to roughly estimate the recovery clock rate to obtain a rough frequency value of the recovery clock.

4. The method for extracting time interval error jitter based on Hilbert transform according to claim 3, characterized in that: The rough estimation includes making a rough estimation of the recovered clock rate by a weighted averaging method.

5. The method for extracting time interval error jitter based on Hilbert transform according to claim 1, characterized in that: Based on the rough frequency value of the recovered clock, the data symbol rate is accurately estimated to obtain the accurate frequency estimation value of the recovered clock.

6. The method for extracting time interval error jitter based on Hilbert transform according to claim 5, characterized in that: The acquisition of the accurate frequency estimation value of the recovered clock includes: The reciprocal of the coarse frequency value of the recovered clock is calculated to obtain the period of the recovered clock, the number of clocks in each pulse width is estimated, that is, the number of code elements contained in each pulse width, and all pulse widths are traversed to obtain the total number of code elements; The accurate frequency estimation value of the recovered clock is calculated based on the difference between the last time jump information and the first time jump information of the measured data signal.

7. The method for extracting time interval error jitter based on Hilbert transform according to claim 1, characterized in that: The time interval error acquisition comprises: The cross point of the recovered clock and the decision level is solved, and the cross point of the measured data signal and the decision level is solved, and the difference between the corresponding cross points is the time interval error TIE.

Images