Feed-forward equalizer performance optimization method

Abstract

The invention relates to a feed-forward equalizer performance optimization method. The method comprises the following steps: S1, acquiring an input signal; s2, inputting the input signal into a feed-forward equalizer, and obtaining an output signal of the feed-forward equalizer; s3, the output signals are processed, noise signals are obtained and grouped, each group of noise signals are weighted and summed, then symbols are taken, and the symbols of the noise weighted sums are obtained; and S4, judging whether a symbol of the noise weighted sum needs to be corrected or not, if the symbol needs to be corrected, correcting an output signal of the feed-forward equalizer according to a correction coefficient and then outputting the corrected output signal, if the symbol does not need to be corrected, outputting the output signal of the feed-forward equalizer, returning the corrected signal or the uncorrected signal as a next-level output signal to the step S3, and if the symbol does not need to be corrected, returning the corrected signal or the uncorrected signal to the step S3. And S4 is repeated until the preset number of iterations is reached. According to the invention, the performance of the FFE which is low in complexity and widely applied is improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and particularly to a method for optimizing the performance of a feedforward equalizer. Background Art

[0002] With the development of communication technologies and the popularity of applications such as large models and cloud services, the capacity of the next-generation communication network needs to be increased by multiples. However, the improvement of the bandwidth of physical devices cannot keep up with the demand for rate improvement. Therefore, it is a great challenge to perform high-quality communication transmission in a bandwidth-limited system. Currently, in a bandwidth-limited system, the most widely used equalizer is the feedforward equalizer (FFE). However, the frequency response of the feedforward equalizer is the inverse response of the channel frequency response. Therefore, in a bandwidth-limited system, it exhibits a high-pass characteristic, which enhances the high-frequency part of the noise in the communication system, resulting in a higher bit error rate. The decision feedback equalizer (DFE) solves the problem of enhanced high-frequency noise. However, due to its feedback structure, it causes the problem of error propagation and is difficult to implement in a circuit. The pre-whitening filter and the post filter-maximum likelihood sequence estimation equalizer (PF-MLSE) solve the problems of the above two equalizers and have powerful performance. However, the MLSE structure has a very high complexity, increasing a large amount of digital signal processing overhead and high cost. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for optimizing the performance of a feedforward equalizer, which improves the performance of the widely used FFE with low complexity.

[0004] To achieve the above purpose, the present invention provides the following solutions:

[0005] A method for optimizing the performance of a feedforward equalizer, comprising:

[0006] S1. Obtain an input signal;

[0007] S2. Input the input signal into a feedforward equalizer to obtain an output signal of the feedforward equalizer;

[0008] S3. Process the output signal to obtain a noise signal and group it. After performing weighted summation on each group of noise signals and taking the sign, obtain the sign of the weighted sum of the noise;

[0009] S4. Determine whether correction is needed based on the sign of the noise weighted sum. If correction is needed, correct the output signal of the feedforward equalizer according to the correction coefficient and then output it. If correction is not needed, output the output signal of the feedforward equalizer, and return the corrected signal or the uncorrected signal as the next-stage output signal to S3 until the preset number of iterations is reached.

[0010] Optionally, processing the output signal to obtain a noise signal and grouping it includes:

[0011] Subtract the decision output signal from the output signal of the feedforward equalizer to obtain the noise signal, where the decision output signal is obtained by inputting the output signal into a decision device;

[0012] Group the noise signals, where the number of noise signals in each group is 2N + 1, N = 1, 2, 3,..., and includes the current noise signal and the N noise signals before and after it.

[0013] Optionally, obtaining the sign of the noise weighted sum by taking the sign after weighting and summing each group of noise signals includes:

[0014]

[0015] E s = sign(E)

[0016] where E represents the weighted sum of the grouped noise signals, the sign function represents taking the sign of the signal, and E s is the sign of the noise weighted sum, α(i) represents the correction coefficient corresponding to the i-th noise signal before and after, N represents the N noise signals before and after the current noise signal, e(n - i) is the error value at the (n - i)-th position, and e(n + i) is the error value at the (n + i)-th position.

[0017] Optionally, determining whether correction is needed based on the sign of the noise weighted sum includes:

[0018] If the signs of the weighted sums of the first N noise signals and the last N noise signals in each group are the same as the sign of the current noise signal, correction is performed; if they are different, no correction is performed.

[0019] Optionally, the method of performing correction is:

[0020]

[0021] Among them, w(n) represents the algorithm output, y represents the algorithm input, that is, the output of the feedforward equalizer, E represents the weighted sum of the grouped noise signals, α(i) in E represents the correction coefficient corresponding to the i-th noise signal before and after, e(n - i) is the error value of the (n - i)-th bit, e(n + i) is the error value of the (n + i)-th bit, n is the index of the current noise signal, and i is the index of the signals before and after the current noise signal.

[0022] Optionally, obtaining the correction coefficient includes:

[0023] Obtain the correction coefficient corresponding to the 1st noise signal before and after the current noise signal, divide the correction coefficient corresponding to the 1st noise signal before and after by 2 as the initial correction coefficient corresponding to the 2nd noise signal before and after, and take a number of correction coefficients corresponding to the 2nd noise signal before and after at the left and right step lengths of the initial correction coefficient corresponding to the 2nd noise signal before and after, obtain the correction coefficient corresponding to the 2nd noise signal before and after with the optimal bit error rate as the final correction coefficient corresponding to the 2nd noise signal before and after, until the correction coefficient corresponding to the i-th noise signal is obtained;

[0024] Or, obtain a number of groups of correction coefficients at a preset step length within a preset interval, and obtain the correction coefficient corresponding to the optimal bit error rate as the correction coefficient corresponding to the i-th noise signal before and after the current noise signal.

[0025] The beneficial effects of the present invention are as follows: The present invention improves the performance of the widely used FFE with low complexity, and at the same time avoids the feedback structure of the DFE. Under the premise of only increasing a very small complexity, it obtains a performance far superior to that of the DFE and is close to the performance of PF-MLSE. Description of the Drawings

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0027] Figure 1 It is the bit error rate result graph of the back-to-back transmission experiment of the embodiment of the present invention;

[0028] Figure 2 It is the bit error rate result graph of the 20 km transmission experiment of the embodiment of the present invention;

[0029] Figure 3 It is the flowchart of a method for optimizing the performance of a feedforward equalizer according to an embodiment of the present invention;

[0030] Figure 4Signal flow graph of a method for optimizing the performance of a feed - forward equalizer according to an embodiment of the present invention;

[0031] Figure 5 It is a diagram of the experimental setup according to an embodiment of the present invention. Among them, (a) is a block diagram of the experimental setup, and (b) is a diagram of the frequency response result of the experimental system. Specific embodiments

[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0033] To make the above - mentioned objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0034] This embodiment provides a method for optimizing the performance of a feed - forward equalizer, which processes the output of the feed - forward equalizer, including:

[0035] S1. Obtain an input signal;

[0036] S2. Input the input signal into the feed - forward equalizer to obtain the output signal of the feed - forward equalizer;

[0037] S3. Process the output signal to obtain noise signals and group them. After performing weighted summation on each group of noise signals and then taking the sign, obtain the sign of the weighted sum of the noise;

[0038] S4. Determine whether correction is needed according to the sign of the weighted sum of the noise signals. If correction is needed, correct the output signal of the feed - forward equalizer according to the correction coefficient and then output it. If correction is not needed, output the output signal of the feed - forward equalizer, and use the corrected signal or the uncorrected signal as the next - level output signal and return to S3 until the preset number of iterations is reached.

[0039] Specifically, as Figure 3 shown, it represents the process of signal flow and the way of processing signals in the SND (selective noise decorrelation) selective noise decorrelation algorithm, where T represents a delay of one cycle. Represents a decision. The meaning of a decision is: in a communication system, a decision refers to the process by which the receiving end identifies and judges the received signal. Its core purpose is to restore the signal affected by noise or interference to the original discrete symbols (such as 0 / 1, letters, or specific codes) transmitted. For example, in digital communication, the receiver will convert the analog waveform signal into the corresponding digital bit stream through methods such as threshold comparison and maximum likelihood criterion. The accuracy of the decision directly affects the bit error rate and is a key link to ensure communication quality. Sign represents taking the sign operation on the signal.

[0040] Furthermore, processing the output signal to obtain the noise signal and grouping includes: subtracting the output signal of the feed-forward equalizer from the decision output signal to obtain the noise signal, where the decision output signal is obtained by inputting the output signal into a decision maker; grouping the noise signal, where the number of noise signals in each group is 2N + 1, N = 1, 2, 3,..., including the current noise signal and its N-bit noise signals before and after.

[0041] Specifically, the output of the feed-forward equalizer is:

[0042]

[0043] where w is the filter tap of the feed-forward equalizer, r is the received sequence to be equalized, and N is the one-sided tap length, that is, the N-bit noise signals before and after the current noise signal.

[0044] Perform the following processing on the output of the feed-forward equalizer to obtain the noise signal:

[0045] e = y - D(y)

[0046] where y is the output of the equalizer and D represents the decision process.

[0047] Group the noise signal. The number of noise signals in each group is 2N + 1 (N = 1, 2, 3,...), including the current noise signal and its N-bit noise signals before and after. Calculate the weighted sum of the noise signals in each group, and then take the sign of it to obtain the sign of the weighted sum of the noise signals.

[0048] Specifically, the weighted sum of the noise signals is:

[0049]

[0050] The sign of the weighted sum of the noise signals is:

[0051] E s = sign(E)

[0052] where E is the weighted sum of the grouped noise signals characterized, the sign function represents taking the sign of the signal, and E s is the sign of the weighted sum of the noise signals.

[0053] Further, judging whether correction is needed according to the weighted sum of noise signals includes:

[0054] Judge the sign of the weighted sum of noise signals. If the sign of the weighted sum of the first N noise signals and the last N noise signals in each group is the same as the sign of the current noise signal, correction is performed; if not, no correction is made;

[0055] Specifically, if the sign of the noise weighted sum is + and the sign of the current noise signal is also +, correction is needed; if the sign of the noise weighted sum is - and the sign of the current noise signal is +, no correction is needed.

[0056] The correction formula is:

[0057]

[0058] Among them, n represents the index of the current noise signal, z(n) represents the algorithm output, y represents the input of the algorithm, that is, the output of the feedforward equalizer, E represents the weighted sum of grouped noise signals, α(i) in E represents the correction coefficient corresponding to the i-th noise signal before and after, e(n - i) is the error value of the (n - i)-th bit, e(n + i) is the error value of the (n + i)-th bit, and i is the index of the signals before and after the current noise signal.

[0059] Further, obtaining the correction coefficient corresponding to the i-th noise signal before and after includes:

[0060] Obtain the correction coefficient corresponding to the first noise signal before and after the current noise signal, divide the correction coefficient corresponding to the first noise signal before and after by 2 as the initial correction coefficient corresponding to the second noise signal before and after, and take a number of correction coefficients corresponding to the second noise signal before and after at the left and right step lengths of the initial correction coefficient corresponding to the second noise signal before and after, obtain the correction coefficient corresponding to the second noise signal before and after with the optimal bit error rate as the final correction coefficient corresponding to the second noise signal before and after, divide the correction coefficient corresponding to the second noise signal before and after by 2 as the initial correction coefficient corresponding to the third noise signal before and after, and take a number of correction coefficients corresponding to the third noise signal before and after at the left and right step lengths of the initial correction coefficient corresponding to the third noise signal before and after, obtain the correction coefficient corresponding to the third noise signal before and after with the optimal bit error rate as the final correction coefficient corresponding to the third noise signal before and after, repeat the above steps until the correction coefficient corresponding to the i-th noise signal is obtained;

[0061] Or, obtain several groups of correction coefficients at a preset step length within a preset interval, and obtain the correction coefficient corresponding to the optimal bit error rate as the correction coefficient corresponding to the i-th noise signal before and after the current noise signal.

[0062] Specifically, the correction factors corresponding to the i-th noise signals before and after the two methods include that if N = 2, there are two correction factors: alpha1 and alpha2. The first way to obtain the specific alpha is to first obtain alpha1 (refer to the case of N = 1), and then use alpha1 / 2 as the initial value for alpha2. Five alpha2 values are taken with a left and right step size of 0.02, and then the corresponding alpha2 value with the optimal bit error rate is found; the second way to obtain it is to obtain several possible combinations with a certain step size, such as a step size of 0.01, within a preset interval from 0.1 to 0.5. The value corresponding to the [alpha1, alpha2] combination with the optimal bit error rate is the value to be found. If N is larger, continue to expand.

[0063] Taking simulation as an example, with N = 1, that is, three noise signal groups as an example, the method, feedback equalizer, and PF-MLSE equalizer of this embodiment are used to equalize the signal and perform performance comparison:

[0064] Step 1:

[0065] Build a simulation platform:

[0066] (1) Generate a PAM4 modulation signal

[0067] (2) The signal passes through a low-pass channel, and the channel uses [1, α], where the value of α can be adjusted, and the channel α is set to α0.

[0068] (3) The signal is distorted by additive white Gaussian noise, and the signal-to-noise ratio SNR is used as an adjustable parameter.

[0069] (4) Use a feed-forward equalizer to equalize the signal and obtain the equalized output of the feed-forward equalizer.

[0070] (5) At the same time, use a feedback equalizer and a PF-MLSE equalizer for equalization for performance comparison.

[0071] Step 2: Specifically judge by the probability of the output signal sampling points being incorrect, and then decorrelate the noise of the possible incorrect sampling points, so as to correct the noise and achieve the effect of correcting the incorrect decision of the sampling points. As Figure 4 shown in the processing of the feed-forward equalizer output signal, the specific process is as follows:

[0072] (1) Make a decision on the feed-forward equalizer output signal. Use the feed-forward equalizer output signal minus the decision output signal to represent the noise signal. Divide the noise symbols into groups of three, for example, -++-+, then there are three groups: -++, ++-, +-+. The decision signal is the result of the output signal passing through the decision device, Figure 4 where Q represents the decision.

[0073] (2) Perform weighted summation on the noise signals one before and one after the current noise signal.

[0074] (3) Take the sign of the weighted sum of the noise signals to obtain the sign of the weighted sum of the noise signals.

[0075] (4) Enter the judgment. If the sign E s of the weighted sum of the noise signals is the same as the sign of the current noise signal, then perform correction. The correction formula is:

[0076]

[0077] If they are not the same, then do not correct, that is, w(n) = y(n). w(n) is the output signal through the equalization method.

[0078] Taking N = 1, that is, taking three noise signals as a group as an example, the correction coefficients corresponding to the first noise signals before and after are:

[0079] α = α0 / (1 + α0)

[0080] where α0 is the parameter adopted for the response of the channel, and α is the correction coefficient of the equalizer.

[0081] Step 3:

[0082] Obtain the optimal performance:

[0083] (1) Set the initial α value for the equalizer, generally α = α0 / (1 + α0), where α0 is the parameter adopted for the response of the channel.

[0084] (2) Set the multi-level structure. One cycle is one level. For multiple levels, the output signal of the equalization method of the previous level is input as the input of the next level and enters the judgment and correction loop multiple times. Generally, repeat once, that is, perform two judgment and correction processes, and the α value is the same for each correction process.

[0085] (3) Taking 0.05 as the step size, take 5 α values on both the left and right sides of the α in (1) as the central value for equalization to obtain the relationship between the α value and the bit error rate, and find the α value corresponding to the lowest bit error rate. Obtaining this value is for optimization. After obtaining the best α, use this α optimal value in subsequent processing to ensure continuous acquisition of the optimal bit error performance.

[0086] Step 4:

[0087] Compare the algorithm performance and plot the bit error rate curve.

[0088] As Figure 1 shown in the bit error rate result graph of the back-to-back transmission experiment, it can be seen from Figure 1 that under the KP4-FEC standard, SND has achieved a 0.9dB performance improvement compared to DFE.

[0089] Figure 2 This is the bit error rate result graph for the 20 km transmission experiment. As can be seen from Figure 2 it, under the KP4-FEC standard, SND achieved a 0.9 dB performance improvement compared to DFE.

[0090] Specifically, back-to-back and 20-km transmission experiments were conducted on 32 GBaud PAM4 signals. The signals were generated by a digital-to-analog converter (DAC) with a sampling rate of 64 Gsample / s and a bandwidth of 25 GHz, and were limited in the digital domain with a normalized cutoff bandwidth of approximately 0.5 (6 dB). The experimental setup is as shown in Figure 5 (a). The tunable laser used in the experiment had a central wavelength of 1545 nm and an output power of 11 dBm. To simulate a typical O-band transmission scenario in data center interconnects, a dispersion compensation fiber (DCF) that could compensate for the dispersion of 20 km of standard single-mode fiber (SSMF) was used. At the receiving end, a variable optical attenuator (VOA) was used to control the received optical power. A photodetector (PD) with a transimpedance amplifier (TIA) was employed. The output signal of the PD was captured by a real-time oscilloscope operating at 80 Gsample / s, and then resampled to 2 samples per symbol (SPS) for digital signal processing (DSP), as shown in Figure 5 (b) which shows the frequency response of the experimental system. Here, the bit error rate (BER) performance of the proposed SND algorithm was evaluated against that of traditional FFE, DFE, PNC, and FFE-PF-MLSE. For a fair comparison, the number of linear taps of all equalizers was set to 21. After equalization, the BER was estimated by directly counting approximately 10 million bits.

[0091] The 3 dB bandwidth of the end-to-end channel response of the system was measured to be approximately 17 GHz, mainly limited by the DAC and the electrical amplifier. The performance of FFE, DFE, SND, PNC, and PF-MLSE in the back-to-back (B2B) case was evaluated, as shown in Figure 1 it. Each equalizer was optimized to its best performance to ensure a fair comparison. DFE has one feedback tap, and adding more taps does not contribute significantly to improving the BER performance. The α values in PNC, SND, and PF-MLSE were optimized to the minimum BER values. As shown in Figure 1 it, due to the enhancement of high-frequency noise, FFE performed the worst. The performance of DFE was better than that of FFE, especially at higher ROPs. The proposed SND scheme outperformed PNC at all ROP values and was comparable to the performance of PF-MLSE. At the KP4 FEC threshold, the received sensitivity of SND was improved by approximately 1.4 dB compared to FFE. The BER performance after 20-km SSMF transmission was also evaluated, and the results are as shown in Figure 2As shown, it can be observed that the performance difference is negligible.

[0092] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for optimizing the performance of a feed-forward equalizer, characterized in that, Including: S1. Obtain an input signal; S2. Input the input signal into a feed-forward equalizer to obtain an output signal of the feed-forward equalizer; S3. Process the output signal to obtain a noise signal and group it, sum the weighted values of each group of noise signals and then take the sign to obtain the sign of the weighted sum of the noise; S4. Determine whether correction is needed according to the sign of the weighted sum of the noise. If correction is needed, correct the output signal of the feed-forward equalizer according to a correction coefficient and then output it. If correction is not needed, output the output signal of the feed-forward equalizer, and use the corrected signal or the uncorrected signal as the next-stage output signal and return to S3 until a preset number of iterations is reached.

2. The method for optimizing the performance of the feedforward equalizer according to claim 1, wherein Processing the output signal to obtain a noise signal and grouping it includes: Subtract the decision output signal from the output signal of the feed-forward equalizer to obtain the noise signal, where the decision output signal is obtained by inputting the output signal into a decision device; Group the noise signals, where the number of each group of noise signals is 2N + 1, N = 1, 2, 3,... and includes the current noise signal and the N noise signals before and after it.

3. The method for optimizing the performance of the feedforward equalizer according to claim 1, wherein Summing the weighted values of each group of noise signals and then taking the sign to obtain the sign of the weighted sum of the noise includes: E s = sign(E) Among them, E is the weighted sum of the characterized grouped noise signals, the sign function represents taking the sign of the signal, and E s is the sign of the weighted sum of the noise signals, α(i) represents the correction coefficient corresponding to the i-th noise signal before and after, N represents the N noise signals before and after the current noise signal, e(n - i) is the error value of the (n - i)-th bit, e(n + i) is the error value of the (n + i)-th bit, n is the index of the current noise signal, and i is the index of the signals before and after the current noise signal.

4. The method for optimizing the performance of the feedforward equalizer according to claim 3, characterized in that Determining whether correction is needed according to the sign of the weighted sum of the noise includes: If the signs of the weighted sums of the first N noise signals and the last N noise signals in each group are the same as the sign of the current noise signal, correction is performed; if they are different, no correction is performed.

5. The method for optimizing the performance of the feed-forward equalizer according to claim 1, characterized in that, The method of performing correction is: Where w(n) represents the output after algorithm correction, y represents the input of the algorithm, that is, the output of the feed-forward equalizer, E represents the weighted sum of the grouped noise signals, α(i) in E represents the correction coefficient corresponding to the i-th noise signal before and after, e(n - i) is the error value of the (n - i)-th bit, e(n + i) is the error value of the (n + i)-th bit, n is the index of the current noise signal, and i is the index of the signals before and after the current noise signal.

6. The method for optimizing the performance of the feedforward equalizer according to claim 3, characterized in that Obtaining the correction coefficient includes: Obtain the correction coefficient corresponding to the first noise signal before and after the current noise signal, divide the correction coefficient corresponding to the first noise signal before and after by 2 to obtain the initial correction coefficient corresponding to the second noise signal before and after, and take several correction coefficients corresponding to the second noise signal before and after at the left and right step lengths of the initial correction coefficient corresponding to the second noise signal before and after, and obtain the correction coefficient corresponding to the second noise signal before and after with the optimal bit error rate as the final correction coefficient corresponding to the second noise signal before and after, until the correction coefficient corresponding to the i-th noise signal is obtained; Or, obtain several groups of correction coefficients at a preset step length within a preset interval, and obtain the correction coefficient corresponding to the optimal bit error rate as the correction coefficient corresponding to the i-th noise signal before and after the current noise signal.

Images