Balancing method and device
By introducing a sequence estimator and a feedforward equalizer into the signal receiver, combining a delay unit and a hard decision decoder, the problem of high computational complexity in the prior art is solved, and the processing of ISI at low complexity is realized. Improve the transmission performance of optical communication system.
Patent Information
- Application Number
- CN202080101402.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-06-03
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2040-06-03
AI Technical Summary
When existing signal receivers handle intersymbol interference (ISI), especially in optical communication systems, have high computational complexity and are difficult to effectively apply in power-sensitive systems, limiting transmission performance.
A signal receiver is designed, including an input terminal, a sequence estimator, a first feedforward equalizer, a second feedforward equalizer and a summing unit, and the signal is processed by filtering and synchronizing the signal to form a sum signal for estimating the symbol stream , and adjust it through the delay unit and hard decision decoder, reducing system complexity.
It realizes effective processing of ISI at low complexity, improves the performance of signal receivers, is suitable for power-sensitive systems, and has a performance close to leading systems, suitable for optical communication applications.
Smart Images

Figure CN115699595B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to equalization in a signal receiver, for example for use in optical communication systems. Background Art
[0002] Inter-symbol-interference (ISI) exists in most communication systems. ISI can lead to degradation of signal quality at the receiver side. Taking optical communication systems as an example, as bandwidth requirements continue to increase, the bit rate of optical signals is being pushed to the bandwidth limit of optoelectronic devices and components, or even beyond the bandwidth limit of optoelectronic devices and components. Therefore, it is common to have moderate or even strong ISI in optical communication systems, which may significantly limit transmission performance.
[0003] In order to combat ISI, various equalization techniques have been proposed, such as feed-forward equalizer (FFE), decision feedback equalizer (DFE) and maximum likelihood sequence estimator (MLSE). Among these methods, MLSE is the most powerful or optimal in minimizing the probability of sequence error. MLSE usually adopts a hard-output Viterbi algorithm and can therefore only be cascaded by a hard decision feed forward error correction (FEC) decoder, as shown in FIG1 . The output signal 101 from the transmitter 102 (Tx) is a sequence of pulse amplitude modulation (PAM) or quadrature amplitude modulation (QAM) symbols, which is then distorted in the channel by adding inter-symbol-interference (ISI) in module 103. The ISI signal 104 is contaminated by an additive white Gaussian noise (AWGN) module 105 before entering the receiver. The receiver in FIG1 includes only the functional blocks relevant to this problem, namely, MLSE 106, deinterleaver (π –1) 107 and a hard decision decoder (HDD) 108. Other functional blocks in the receiver, such as phase recovery and clock recovery, are not shown. In FIG1 , the received symbol sequence is equalized at 109 by the maximum likelihood sequence estimation (MLSE) block 106 and converted to a bit sequence at 110. The equalized bit sequence 110 is then converted to a bit sequence by π –1 Deinterleaving, the π –1 Involves a permutation of bit positions. The permuted bit sequence at 111 is now suitable for decoding.
[0004] Soft decision decoding may bring about 1dB to 2dB of additional net coding gain (NCG) over hard decision decoding. A soft decision decoder requires soft bit information instead of a hard decision bit sequence.
[0005] In order to take advantage of the benefits of soft decision decoding, the soft output Viterbi algorithm (SOVA) and the Bahl-Coke-Jelinek-Raviv (BCJR) algorithm are proposed. The soft output Viterbi algorithm (SOVA) and the Bahl-Coke-Jelinek-Raviv (BCJR) algorithm as described in https: / / en.wikipedia.org / wiki / Soft_output_Viterbi_algorithm are two widely used equalization methods for compensating for ISI and generating soft information.
[0006] The use of these algorithms by a signal receiver is shown in FIG2. Much of FIG2 is the same as FIG1, except that the MLSE block 106 is replaced by a soft output block 202. Block 202 receives a noisy ISI signal 201. Block 202 typically uses a SOVA or BCJR algorithm to generate soft information, such as a Log-Likelihood Ratio (LLR) 203. The subsequent π –1 204 and a soft decision decoder (SDD) 205 work on the LLR 203 and its permuted version 206 accordingly.
[0007] However, the computational complexity of SOVA and BCJR is much higher than that of MLSE, prohibiting the application of SOVA and BCJR in power-sensitive systems.
[0008] BCJR is a maximum a posterior (MAP) detection algorithm, which is generally considered to be the best algorithm. It involves calculating forward and reverse probabilities based on a grid, so it can be very complex.
[0009] SOVA differs from the classic Viterbi algorithm in that it uses a modified path metric that takes into account the prior probabilities of the input symbols and produces a soft output that indicates the reliability of the decision. Like the hard-output Viterbi algorithm used by MLSE, SOVA only includes forward path metric calculations according to the trellis, rather than bidirectional calculations like BCJR. However, SOVA needs to store soft information for each bit and update the soft information at each step forward along the trellis. Therefore, the complexity of SOVA is much higher than MLSE, prohibiting the application of SOVA in systems that require low power consumption.
[0010] It would be desirable to develop a low complexity alternative to the above-described signal receiver techniques that overcome these problems. Summary of the invention
[0011] According to a first aspect, a signal receiver is provided, which comprises: an input end for receiving an input signal including multiple symbols; a sequence estimator for receiving the input signal and forming an estimate of the most likely content of the input signal; a first feedforward equalizer for receiving the estimate or a signal derived therefrom, and filtering it to form a first equalizer output, wherein the first feedforward equalizer is used to make its center tap effectively 0; a second feedforward equalizer for receiving the input signal and filtering it to form a second equalizer output; a summing unit for summing the first equalizer output and the second equalizer output to form a sum signal, wherein the receiver is used to make the summing unit synchronously sum the first equalizer output and the second equalizer output relative to the dependence of the equalizer output on the input signal; and a symbol estimator for estimating a symbol stream based on the sum signal.
[0012] In some implementations, the signal receivers described herein can perform operations close to those of leading systems while having lower complexity, which can facilitate their application in power-sensitive systems.
[0013] The signal receiver may include a delay unit for delaying the input signal between the input terminal and the second feedforward equalizer so that the first equalizer output and the second equalizer output are provided to the summing unit synchronously with respect to their dependence on the input signal. Thus, the delay unit can align symbol sequences in the outputs from the first feedforward equalizer and the second feedforward equalizer.
[0014] The sequence estimator may be a maximum likelihood sequence estimator. This may be a convenient implementation.
[0015] The sequence estimator may be a feed-forward equalizer, an infinite impulse response filter or a decision feedback equalizer. This may further reduce the complexity of the proposed equalization scheme.
[0016] The sequence estimator may generate hard outputs. The sequence estimator may output hard-determined symbols instead of bits.
[0017] The second feed-forward equalizer may be used to act on m symbols before the reference symbol and n symbols after the reference symbol, and the first feed-forward equalizer may be used to act on m+n symbols before the reference symbol and m+n symbols after the reference symbol. This may result in improved performance because the scattered energy of the symbol of interest (SOI) can be obtained from adjacent symbols.
[0018] The signal receiver may be configured to adjust the tap coefficients of the first equalizer and / or the second equalizer according to the output of the summing unit.Thus, the number of taps of the first FFE and the second FFE may depend on the number of taps of the channel response.
[0019] The sequence estimator may be used to form an estimate of the most likely symbol content of the input signal and the first equalizer may be used to receive the estimate as input.Thus, the sequence estimator and the first equalizer may be comprised in a lower branch of the signal receiver.
[0020] The sequence estimator may be used to form an estimate of the most likely bit content of the input signal, the receiver may include a bit-to-symbol converter for forming a symbol stream based on the estimate, the first equalizer being used to receive the symbol stream as input. Thus, the sequence estimator may estimate the bit content or symbol content of the input signal.
[0021] The signal receiver may include a hard decision decoder for performing a hard decision decoding operation on the estimate to form a hard decision stream, and the bit-to-symbol converter is used to form the symbol stream from the hard decision stream. Including the HDD unit can more accurately reconstruct the ISI.
[0022] The signal receiver may include a deinterleaver for deinterleaving the estimated output from the sequence estimator to form an input to the hard decision decoder, and an interleaver for interleaving the output of the hard decision decoder to form an input to the bit to symbol converter. This may produce a sequence suitable for decoding.
[0023] The symbol estimator may be used to estimate the symbol stream by log likelihood ratio (LLR) or hard decision. This may be a convenient implementation.
[0024] According to a second aspect, a communication terminal is provided, comprising: a signal receiver according to any one of the above claims; and an optical signal detector for detecting an externally generated optical signal and forming the input signal therefrom. Therefore, the signal receiver can be used in optical communication applications.
[0025] According to a third aspect, there is provided a communication terminal comprising a plurality of signal receivers, each signal receiver being arranged in series as described above. If further iterations are used, the performance may be further improved.
[0026] According to a fourth aspect, a method for decoding a signal is provided, which comprises: receiving an input signal comprising a plurality of symbols; forming an estimate of the most likely content of the input signal; filtering the estimate or a signal derived therefrom by a first feedforward equalizer to form a first equalizer output, wherein the first feedforward equalizer is used to make its center tap effectively zero; filtering the input signal by a second feedforward equalizer to form a second equalizer output; summing the first equalizer output and the second equalizer output to form a sum signal, wherein the first equalizer output and the second equalizer output are summed synchronously with respect to their dependence on the input signal; and estimating a symbol stream based on the sum signal.
[0027] In some implementations, the methods described herein can perform operations close to leading methods while having lower complexity, which can facilitate their application in power-sensitive systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] The present invention will now be described by way of examples with reference to the accompanying drawings.
[0029] In the attached picture:
[0030] FIG. 1 schematically shows a signal receiver with MLSE and hard decision FEC.
[0031] FIG. 2 schematically shows a signal receiver with SOVA / BCJR and soft-decision FEC.
[0032] Figure 3 A first example of a signal receiver comprising a soft output equalizer is schematically shown.
[0033] Figure 4 A second example of a signal receiver comprising a soft output equalizer is schematically shown.
[0034] Figure 5 Shows Figure 3 A low complexity variant of the signal receiver is shown in FIG. 3 , in which the MLSE 305 is replaced by a simpler equalizer 305 a (eg, FFE, IIR, or DFE) and a clipper 305 b.
[0035] Figure 6 Shows Figure 4 A low complexity variant of the signal receiver is shown in FIG. 4 , in which the MLSE 404 is replaced by a simpler equalizer 404a (eg, FFE, IIR, or DFE) and a symbol-to-bits block 404b.
[0036] Figure 7 A cascade scheme is shown.
[0037] Figure 8 Shows Figure 7 The building blocks used in
[0038] Fig. 9 A reference system is shown.
[0039] Fig.10 Simulation results using PAM4 signals on 1+0.5D channels are shown.
[0040] Fig.11 A flow chart showing an example of a method for decoding a signal. DETAILED DESCRIPTION
[0041] In the example given below, the system is used to implement a soft output equalizer. The number of output bits of the equalizer can be reduced to the number of available quanta (e.g., symbols / bit) at each quantization so that the equalizer performs as a hard output equalizer or hard decision decoder (HDD).
[0042] Figure 3An example of a training receiver 300 is shown. A noisy ISI signal 301 is received at the input of the receiver and fed to two branches. A delay module 302 in the upper branch aligns a symbol sequence 303 with an output symbol sequence 304 of an MLSE block 305 in the lower branch. Unlike the MLSE 106 in FIG. 1 , the MLSE 305 outputs hard decision symbols instead of bits. Then, the signals 303 and 304 are each filtered by an FFE: FFE 306 in the upper branch and FFE 307 in the lower branch. The FFE 307 at the lower branch has no center tap, or the center tap coefficient can be forced to be approximately or exactly 0. If the channel impulse response is known (an example is given in the embodiment described below), the tap coefficients of the two FFEs 306, 307 can be derived, or a constant modulus algorithm (CMA), data aided least mean square (LMS), decision directed least mean square (DD-LMS) or recursive least square (RLS) can be used.
[0043] The two filtered signals 308 and 309 are then combined in an adder 310. The adder output 311 is fed to an LLR calculator 312 to generate soft information 313 for processing by a subsequent deinterleaver and SDD.
[0044] The MLSE output may be more reliable than the hard decisions of the symbols output by the channel and thus may be used to reconstruct the ISI, which may be inferred from the channel output. Preferably, the signal at 311 may be close to an ISI-free symbol with only additive noise, from which high quality soft information used by the FEC decoder may be easily computed.
[0045] Even though conventionally, in known embodiments, ISI falling into the SOI may be counteracted with the help of MLSE, the performance may still suffer from power loss since the energy of the SOI is spread over adjacent time slots due to the ISI channel.
[0046] In the receiver described herein, Figure 3 The FFE 306 in the upper branch is used to collect the scattered SOI energy. In order to obtain the energy of the SOI from the adjacent symbols, the help from the lower branch is used to cancel the adjacent symbols themselves, as well as the ISI from other symbols (except the SOI).
[0047] The number of taps of the upper FFE and the lower FFE may depend on the number of taps of the channel response. Assume that the response of the ISI channel has Npre taps before SOI and Npost taps after SOI, which means that SOI will impose ISI on symbols before Npre and after Npost. Then, the upper FFE has Npre+1+Npost taps and the lower FFE has Npre+Npost+0+Npre+Npost taps, where the center 0 indicates that the taps of the lower FFE corresponding to SOI are forced to 0.
[0048] Thus, here, the center tap is one of the taps aligned with the SOI (eg, the number of taps is an even number).
[0049] Thus, in general, the signal receiver 300 comprises an input for receiving an input signal 301 comprising a plurality of symbols. The sequence estimator 300 is configured to receive the input signal 301 and form an estimate (in the MLSE 305) of the most probable content of the input signal 301. The first feedforward equalizer 307 is configured to receive the estimate or a signal derived therefrom and filter it to form a first equalizer output 309. The first feedforward equalizer 307 is configured to cause its center tap to be effectively zero. The receiver further comprises a second feedforward equalizer 306 for receiving the input signal and filtering it to form a second equalizer output 308. The receiver further comprises a summing unit 310 for summing the first equalizer output 309 and the second equalizer output 308 to form a sum signal 311. The signal receiver is configured to cause the summing unit 310 to sum the first equalizer output 309 and the second equalizer output 308 synchronously with respect to the dependence of the equalizer output on the input signal. The signal receiver further comprises a symbol estimator 312 for estimating the symbol stream from the sum signal.
[0050] Figure 4 A second example of a soft output equalizer is shown, which can improve the performance of a signal receiver due to the HDD included in the lower branch. With the help of the HDD, the ISI can be reconstructed more accurately and the performance of the equivalent channel may be closer to a pure AWGN channel.
[0051] Similar to Figure 3 ,exist Figure 4 In FIG. 4 , a signal receiver 400 receives a noisy ISI signal 401 fed to two branches at the input of the receiver. A delay module 402 in the upper branch aligns a symbol sequence 403 with an output symbol sequence 411 from the block in the lower branch.
[0052] To facilitate HDD, MLSE 404 at the lower branch generates hard decision bits 405 instead of symbols as in MLSE 305. The bit sequence 405 is then processed by deinterleaver 406, HDD 407, and interleaver 408. With the help of HDD 407, the bit sequence 409 may be much more reliable than 405. After bit-to-symbol block 410, signal 409 is converted into symbol sequence 411. Figure 3 The signal 304 in is more reliable. Figure 4 There are more processing blocks at the lower branch in , so a larger delay module 402 is needed at the upper branch to align the symbol sequences 403 and 411. The following operations of 403 and 411 are similar to those of reference Figure 3 The operations described are the same.
[0053] Signals 403 and 411 are each filtered by an FFE: FFE 412 in the upper branch, FFE 413 in the lower branch. The FFE 413 at the lower branch has no center tap, or the center tap coefficients can be forced to be approximately or exactly 0. If the channel impulse response is known (an example is given in the embodiment described below), the tap coefficients of the two FFEs 412, 413 can be derived, or these coefficients can be adaptively updated using algorithms such as CMA, LMS, DD-LMS or RLS, as described above.
[0054] The two filtered signals 414 and 415 are then combined in an adder 416. The adder output 417 is fed to an LLR calculator 418 to generate soft information 419 for subsequent processing by a deinterleaver and SDD.
[0055] In order to further reduce the complexity of the proposed equalization scheme, Figure 3 and Figure 4 The MLSE module in can be replaced by other simpler equalizers, such as FFE filter, infinite impulse response (IIR) filter or decision feedback equalizer (DFE) or other suitable equalizers. Figure 5 and Figure 6 shown.
[0056] Figure 5 Shows Figure 3 A low complexity variant of the signal receiver is shown in FIG. 3 , in which the MLSE 305 is replaced by a simpler equalizer 305 a (eg, FFE, IIR, or DFE) and a clipper 305 b.
[0057] Figure 6 Shows Figure 4A low complexity variant of the signal receiver is shown in FIG. 4 , in which the MLSE 404 is replaced by a simpler equalizer 404a (eg, FFE, IIR, or DFE) and a symbol-to-bits block 404b.
[0058] like Figure 7 As shown, the performance can be further improved by introducing further iterations or by cascading multiple equalizers in the signal receiver 700. The first stage uses one of the soft output equalizers described previously, such as 300, 400, 500 or 600. The details of 800 are as shown in FIG. Figure 8 Stage 800 uses two input signals: one is a replica of the noisy input signal 701 ( Figure 8 801a), and the other is the signal from the previous stage ( Figure 8 801b).
[0059] exist Figure 8 In the embodiment, stage 800 receives input signals 801a and 801b in two branches. A delay module 802 in the upper branch aligns a symbol sequence 803 with an output symbol sequence 811 from a block in the lower branch.
[0060] To facilitate HDD, hard decision 804 in the lower branch generates hard decision bits 805. The bit sequence 805 is then processed by deinterleaver 806, HDD 807, and interleaver 808 to give a bit sequence 809. After bit-to-symbol block 810, signal 809 is converted to symbol sequence 811. Delay module 802 aligns signal 803 with 811.
[0061] Signals 803 and 811 are each filtered by an FFE: FFE 812 in the upper branch, and FFE 813 in the lower branch. The FFE 813 in the lower branch has no center tap, or the center tap coefficients can be forced to be approximately or exactly 0. If the channel impulse response is known, the tap coefficients of the two FFEs 812, 813 can be derived, or these parameters can be adaptively updated using algorithms such as CMA, LMS, DD-LMS or RLS, as described above.
[0062] The two filtered signals 814 and 815 are then combined in an adder 816. The adder output 817 is fed to an LLR calculator 818 to generate soft information 819.
[0063] Figure 7 The final output 702 in is fed to the deinterleaver and SDD.
[0064] In an alternative configuration, the signal receiver includes a hard output equalizer, the output may include hard information and may be fed to a HDD instead of an SDD.
[0065] In summary, the equalizer described in this article includes two branches. The first branch includes a delay module and an FFE. The delay module aligns the first branch with the second branch. The FFE collects the energy of the desired symbol in the relevant time slot.
[0066] The second branch includes MLSE and FFE without the center tap (current symbol) that reconstructs the interference to the desired symbol in the relevant time slot. The tap coefficients of FFE can be trained by various control algorithms (e.g., LMS or RLS) with or without a training sequence.
[0067] A hard decision decoder may optionally be included in the lower branch to more accurately reconstruct the ISI.
[0068] If further iterations are used, performance may be further improved.
[0069] In the following simulation, Fig. 9 The system 900 shown in FIG. 9 is used as a reference (MLSE+HDD). Here, PAM4 is used as the modulation format. The ISI channel has two taps 1 and α, where α≤1, or equivalently, the impulse response of the channel is [1,α], i.e., the amplitude of the SOI is 1, and an ISI with an amplitude of α is applied to the next symbol. To illustrate the performance of FEC, a BCH code with t=2 is used in the simulation.
[0070] Since the channel tap coefficients are known, the tap coefficients of the FFE in the upper and lower branches in the previously described system can be calculated.
[0071] Assume r -1 , r0, r1 are three consecutive symbols in the channel, is the corresponding symbol output by the bit-to-symbol module after the hard decision decoder in the MLSE or lower branch. The output symbol in the adder after the two FFEs can be expressed as:
[0072]
[0073] According to this formula, the impulse responses of the two FFEs can be derived.
[0074] The upper FFE has an impulse response:
[0075]
[0076] The following FFE has an impulse response:
[0077]
[0078] Replace r0=s0+αs -1 +n0 and r1=s1+αs0+n1 is z0, assuming It turns out that:
[0079] z0=[(s0+n0)+α(αs0+n1)] / (1+α 2 )
[0080] =s0+(n0+αn1) / (1+α 2 )
[0081] Therefore, the same AWGN performance can be obtained without ISI.
[0082] Fig.10 The simulation results obtained using a PAM4 signal on a 1+0.5D channel are shown. If we compare BER = 1e -5 The signal-to-noise ratio (SNR) when Figure 3 ) exceeds the reference (MLSE+HDD) by 0.75 dB, while the proposed exemplary solution 2 (i.e., using Figure 4 The device shown in FIG) provides an SNR gain of up to 0.5 dB.
[0083] Fig.11 A flow chart of an example of a method 1100 for decoding a signal described herein is shown. In step 1101, the method includes receiving an input signal, the input signal including a plurality of symbols. In step 1102, the method includes forming an estimate of the most likely content of the input signal. In step 1103, the method includes filtering the estimate by a first feedforward equalizer to form a first equalizer output, the first feedforward equalizer being used to make its center tap effectively 0. In step 1104, the method includes filtering the input signal by a second feedforward equalizer to form a second equalizer output. In step 1105, the method includes adding the first equalizer output and the second equalizer output to form a sum signal, the first equalizer output and the second equalizer output being synchronously summed relative to their dependence on the input signal. In step 1106, the method includes estimating a symbol stream according to the sum signal.
[0084] To obtain coding gain from a soft-decision FEC decoder, this paper describes a low-complexity soft-output equalizer and its high-performance variant. As mentioned earlier, according to alternative configurations, the equalizer and its high-performance variant can provide hard outputs to accommodate subsequent HDDs.
[0085] In some implementations, the performance of the equalizer has been shown to be better than previously proposed solutions, and in some implementations, high-performance variants have been shown to perform close to SOVA with much lower complexity.
[0086] Low-cost short-distance optical transmission systems can only accommodate low-power DSPs. Traditionally, only hard-output equalizers such as MLSE and hard-decision FEC are considered because traditional soft-output equalizers such as SOVA and BCJR consume too much power. The device and method described in this article can be used in systems that require low power consumption, and can therefore be effectively used in next-generation short-distance systems.
[0087] Applicants hereby disclose separately each individual feature described herein and any combination of two or more such features. With the common knowledge of those skilled in the art, such features or combinations can be implemented as a whole based on this specification, without considering whether such features or combinations of features can solve any problems disclosed herein, and without limiting the scope of the claims. This application shows that various aspects of the present invention can be composed of any such individual features or combinations of features. In view of the above description, it will be obvious to those skilled in the art that various modifications can be made within the scope of the present invention.
Claims
1. A signal receiver, characterized in that: include: An input terminal for receiving an input signal comprising a plurality of symbols; a sequence estimator for receiving the input signal and forming an estimate of the most probable content of the input signal; a first feedforward equalizer configured to receive the estimate or a signal derived therefrom and filter the estimate to form a first equalizer output, wherein the first feedforward equalizer is configured to have a center tap thereof effectively zero; a second feedforward equalizer for receiving the input signal and filtering it to form a second equalizer output; a summing unit configured to sum the first equalizer output and the second equalizer output to form a sum signal, wherein the signal receiver is configured to cause the summing unit to sum the first equalizer output and the second equalizer output synchronously based on the dependency of the first equalizer output on the input signal and the dependency of the second equalizer output on the input signal; A symbol estimator is used to estimate a symbol stream according to the sum signal.
2. The signal receiver according to claim 1, wherein: A delay unit is included for delaying the input signal between the input terminal and the second feedforward equalizer so that the first equalizer output and the second equalizer output are provided to the summing unit synchronously with respect to their dependence on the input signal.
3. The signal receiver according to claim 1 or 2, characterized in that: The sequence estimator is a maximum likelihood sequence estimator.
4. The signal receiver according to claim 1 or 2, characterized in that: The sequence estimator is a feedforward equalizer, an infinite impulse response filter or a decision feedback equalizer.
5. The signal receiver according to claim 1 or 2, characterized in that: The sequence estimator produces hard outputs.
6. The signal receiver according to claim 1 or 2, characterized in that: The second feedforward equalizer is used to act on m symbols before the reference symbol and n symbols after the reference symbol, and the first feedforward equalizer is used to act on m+n symbols before the reference symbol and m+n symbols after the reference symbol.
7. The signal receiver according to any one of claims 1 to 6, characterized in that: The signal receiver is configured to adjust tap coefficients of the first feedforward equalizer and / or the second feedforward equalizer according to an output of the summing unit.
8. The signal receiver according to claim 1 or 2, characterized in that: The sequence estimator is configured to form an estimate of the most probable symbol content of the input signal, and the first feedforward equalizer is configured to receive as input the estimate of the most probable symbol content of the input signal.
9. The signal receiver according to claim 1 or 2, characterized in that: The sequence estimator is used to form an estimate of the most likely bit content of the input signal, the signal receiver includes a bit-to-symbol converter for forming a symbol stream based on the estimate of the most likely bit content of the input signal, and the first feedforward equalizer is used to receive the symbol stream formed by the estimate of the most likely bit content of the input signal as input.
10. The signal receiver according to claim 9, characterized in that A hard decision decoder is included for performing a hard decision decoding operation on the estimate to form a hard decision stream, and the bit-to-symbol converter is used to form a symbol stream of the hard decision stream based on the hard decision stream.
11. The signal receiver according to claim 10, characterized in that: include: a deinterleaver for deinterleaving the estimated output from the sequence estimator to form an input to the hard decision decoder; An interleaver is configured to interleave the output of the hard decision decoder to form an input to the bit-to-symbol converter.
12. The signal receiver according to claim 1 or 2, characterized in that: The symbol estimator is configured to estimate the symbol stream by using a log-likelihood ratio.
13. A communication terminal, characterized in that: include: The signal receiver according to any one of claims 1 to 12; The optical signal detector is used to detect an externally generated optical signal and form the input signal accordingly.
14. A communication terminal, characterized in that: The invention comprises a plurality of signal receivers, wherein the signal receiver comprises the signal receiver according to any one of claims 1 to 12; the plurality of signal receivers are arranged in series.
15. A method for decoding a signal, characterized in that include: receiving an input signal comprising a plurality of symbols; forming an estimate of the most likely content of the input signal; filtering the estimate or a signal derived therefrom through a first feedforward equalizer to form a first equalizer output, wherein the first feedforward equalizer is configured to have a center tap thereof effectively zero; filtering the input signal through a second feedforward equalizer to form a second equalizer output; summing the first equalizer output and the second equalizer output to form a sum signal, wherein the dependence of the first equalizer output with respect to the input signal and the dependence of the second equalizer output with respect to the input signal are summed synchronously; A symbol stream is estimated based on the sum signal.
