Decision feedback equalizer and transmission channel reflection loss compensation method
By designing a feedback filter combining fixed and sliding taps in the decision feedback equalizer, the problem that traditional DFE cannot effectively compensate for the low frequency reflectance amount is solved, and higher signal transmission quality is achieved.
Patent Information
- Application Number
- CN202311803575.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-25
- Publication Date
- 2025-06-27
AI Technical Summary
Traditional decision feedback equalizers (DFEs) cannot effectively compensate when facing low-frequency reflections, resulting in a degradation of signal transmission quality.
A judgment feedback equalizer is designed, and a combination structure of feedback filter is adopted, where the first N1 taps are fixed taps, and the next N3 taps are sliding taps generated based on N2 virtual taps behind the first N1 taps. Through the adapter, N3 virtual taps and their compensation weights are selected from the last N2 virtual taps and mapped to the last N3 sliding taps to generate a compensation signal.
Without significantly increasing line power consumption and area, the compensation ability of the judgment feedback equalizer to the channel reflection components is improved, and the signal transmission quality is improved.
Smart Images
Figure CN120223194A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of circuit technologies, and particularly to a decision feedback equalizer and a method for compensating reflection loss of a transmission channel. Background Art
[0002] Field-programmable gate arrays (FPGAs) are widely used in fields such as signal processing, military, medical, and consumer electronics. As an important resource in FPGAs, high-speed serial interface (SerDes) technology can solve the bottleneck of the bandwidth of user application interfaces and greatly increase the application scenarios of FPGAs. In SerDes technology, with the continuous increase in the transmission data rate, signal integrity problems caused by dielectric loss, crosstalk, reflection, etc. become increasingly serious. For transceivers with high-speed data rate transmission, the broadband design of the data path faces increasing challenges, especially the broadband design of the receiver (RX) is particularly difficult. At the RX front end, a continuous-time linear equalizer (CTLE) is generally used to compensate for channel loss, but this structure will bring problems such as large area loss, phase distortion, and large trace parasitics. Generally, for transceivers with higher data rates and larger channel losses, a decision feedback equalizer (DFE) is used at the receiver to compensate for channel loss.
[0003] Existing DFE technologies are mainly used to eliminate inter-symbol interference (ISI), with fewer compensation taps. When there are non-ideal situations such as reflection during data transmission, the traditional DFE structure cannot compensate for the reflection amount with a lower frequency. Summary of the Invention
[0004] Embodiments of the present invention provide a decision feedback equalizer and a method for compensating reflection loss of a transmission channel, which can solve the problem that the traditional DFE structure cannot compensate for the reflection amount with a lower frequency without significantly increasing the line power consumption and area, and improve the signal transmission quality.
[0005] To this end, the present invention provides the following technical solutions:
[0006] On the one hand, an embodiment of the present invention provides a decision feedback equalizer, including: an adder, a sampler, a feedback filter, and an adaptation controller; the first N1 taps of the feedback filter are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, where N3 < N2;
[0007] The sampler is used to sample the output signal of the adder and output sampled data;
[0008] The feedback filter is used to output a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps according to the sampled data;
[0009] The adaptation controller is used to select N3 virtual taps and their corresponding compensation weights from the last N2 virtual taps and map them to the last N3 sliding taps;
[0010] The adder is used to compensate the input signal according to the first compensation signal and the second compensation signal and output the compensated signal.
[0011] Optionally, the adaptation controller is specifically configured to traverse the last N2 virtual taps, select N3 virtual taps with the largest compensation weights and their corresponding compensation weights, and map them to the last N3 sliding taps.
[0012] Optionally, the compensation weights corresponding to each tap are determined according to the LMS algorithm.
[0013] Optionally, the value of N3 corresponding to different signals is the same.
[0014] Optionally, the adaptation controller is specifically configured to traverse the last N2 virtual taps, select some or all of the virtual taps whose compensation weights are greater than a set threshold and their corresponding compensation weights, and map them to the last N3 sliding taps.
[0015] Optionally, the value of N3 corresponding to different signals is the same or different.
[0016] On the other hand, an embodiment of the present invention further provides a method for compensating the reflection loss of a transmission channel. The method includes:
[0017] Sampling the output signal of the transmission channel to obtain sampled data;
[0018] Nonlinearly filtering the sampled data through a feedback filter. The first N1 taps of the feedback filter are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, and N3 < N2; generating a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps;
[0019] Compensating the input signal of the transmission channel according to the first compensation signal and the second compensation signal.
[0020] Optionally, the nonlinearly filtering the sampled data through a feedback filter includes:
[0021] The sampled data is successively delayed to determine the output data and compensation weights corresponding to each tap, where the taps include the first N1 fixed taps and the subsequent N2 virtual taps;
[0022] According to the output data and compensation weights of each tap, the corresponding compensation signal of the tap is determined.
[0023] Optionally, generating the second compensation signal corresponding to the subsequent N3 sliding taps includes: traversing the subsequent N2 virtual taps, selecting N3 virtual taps with the largest compensation weights therefrom and their corresponding compensation weights are mapped to the subsequent N3 sliding taps to generate the second compensation signal.
[0024] Optionally, generating the second compensation signal corresponding to the subsequent N3 sliding taps includes: traversing the subsequent N2 taps, selecting some or all of the virtual taps whose compensation weights are greater than the set threshold and their corresponding compensation weights are mapped to the subsequent N3 sliding taps to generate the second compensation signal.
[0025] The decision feedback equalizer and the transmission channel reflection loss compensation method provided by the embodiments of the present invention design the feedback filter in a combination of fixed taps and sliding taps, which can significantly improve the compensation ability of the decision feedback equalizer for the channel reflection component without introducing a large area and power consumption, and improve the signal transmission quality.
[0026] Furthermore, the decision feedback equalizer provided by the embodiments of the present invention can reuse the existing DFE architecture and algorithm to compensate for the channel reflection, has better compatibility, and can select whether to perform the calculation of the sliding tap part according to the specific application environment requirements, increasing the flexibility of the solution and the adaptability to different environments.
[0027] Furthermore, the number of virtual taps can be flexibly configured according to user needs, so as to better meet the needs of different users and signal transmission quality. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 is the structural block diagram of a traditional DFE;
[0029] Figure 2 is the under-compensated code type schematic diagram of h2 tap in a traditional DFE;
[0030] Figure 3 is the over-compensated code type schematic diagram of h2 tap in a traditional DFE;
[0031] Figure 4 is the unit impulse response schematic diagram of a single-ended signal passing through a lossy channel in a traditional 2tap DFE;
[0032] Figure 5 is forFigure 4 The waveform diagram of the unit impulse signal after compensation through a lossy channel;
[0033] Figure 6 is the waveform diagram of the unit impulse signal after compensation through a lossy channel with reflection;
[0034] Figure 7 is the principle block diagram of the decision feedback equalizer provided by the embodiments of the present invention;
[0035] Figure 8 is a schematic diagram of a specific structure of the decision feedback equalizer provided by the embodiments of the present invention;
[0036] Figure 9 is a flowchart of a method for compensating the reflection loss of a transmission channel provided by the embodiments of the present invention. Detailed implementation manners
[0037] Next, the principles and spirits of the present invention will be described with reference to the exemplary embodiments shown in the accompanying drawings. It should be understood that the description of these embodiments is only for enabling those skilled in the art to better understand and then implement the present invention, rather than limiting the scope of the present invention in any way.
[0038] First, a brief description of the traditional DFE will be given below.
[0039] Refer to Figure 1 , Figure 1 is the structural block diagram of the traditional DFE, which mainly consists of an adder, a sampler, and a feedback filter.
[0040] Among them, the input signal y k and the output signal of the feedback filter are summed through an adder to obtain the signal Z k , and then the signal Z k is sampled through a sampler to obtain a digital output as the input of the feedback filter. Through the adaptive filter algorithm (Least mean square, LMS), an appropriate compensation amount can be obtained to compensate for the loss of the channel.
[0041] Figure 1 The equivalent formula of
[0042]
[0043] is as follows: n_next w n = w n + u n ×Δw n Δw ideal = VP real - VP
[0044] Among them, u n represents the step value adjusted by each tap. This value will vary slightly according to the convergence speed and stability for different taps; VP ideal refers to the ideal eye height of the transmitted data, and VP real refers to the actual eye height of the transmitted data.
[0045] Taking h2 as an example below, the convergence process of the LMS algorithm will be illustrated.
[0046] According to the above Figure 1 principle description, for the h2 tap, the following formula can be obtained:
[0047]
[0048] Among them, w 2_next = w2 + u2×Δw2; w2 = VP ideal -VP real ;
[0049] For the data pattern 0010, the current data Data[x] = 1, the value of its h2 tap data Data[x - 2] is 0, the compensation amount is W2, the eye height before compensation is VP ori , and the eye height after compensation is VP real , and the ideal eye height is VP ideal .
[0050] When the compensation is too small, the corresponding pattern diagram is as Figure 2 shown.
[0051] It can be seen from Figure 2 that at this time, VP real < VP ideal , and W 2_next increases, and VP real will increase and gradually converge to VP ideal .
[0052] When the compensation is too large, the corresponding pattern diagram is as Figure 3 shown.
[0053] It can be seen from Figure 3 that at this time, VP real > VP ideal , and W 2_next decreases, and VP real will decrease and gradually converge to VP ideal .
[0054] As the number of taps of the feedback filter in the DFE structure increases, the DFE can better compensate for the channel. However, the area and power consumption of the circuit will also increase continuously. Considering the trade-off of various performance aspects, the number of taps of the feedback filter in the traditional DFE structure is generally not too large, which leads to the problem that when there are non-ideal situations such as reflections during data transmission, the traditional DFE structure cannot compensate for the reflected amount with a relatively low frequency.
[0055] The following takes a traditional 2-tap DFE as an example to illustrate its deficiencies in the signal compensation mechanism. The unit impulse response of the single-ended signal passing through the channel normally is as Figure 4 shown. Among them, the a -1 component is the pre-cursor (also known as the Pre-shoot coefficient), the a0 component is the main-cursor (mainly affecting the output swing), and a1, a2, and a3 are the post-cursor (also known as the de-emphasis coefficient, de-emphasizing the output signal) amounts respectively.
[0056] The waveform of the unit impulse signal after passing through the lossy channel and compensated by the 2-tap DFE is as Figure 5 shown.
[0057] The waveform of the unit impulse signal after passing through the lossy channel with reflections and compensated by 2-tap ISI compensation and 1-tap reflection compensation is as Figure 6 shown. By subtracting the a1 and a2 post-cursor amounts through 2-tap ISI compensation, the 1-tap reflection compensation can eliminate the reflected influence amount a9 of the channel discontinuity. After compensation, the transmission signal quality can be greatly improved, and the signal's ability to pass through the channel can be enhanced.
[0058] Therefore, the embodiment of the present invention provides a decision feedback equalizer, which can solve the problem that the traditional DFE structure cannot compensate for the reflected amount with a relatively low frequency without significantly increasing the power consumption and area of the circuit, and improve the signal's ability to pass through the channel.
[0059] As Figure 7 shown, it is the principle block diagram of the decision feedback equalizer provided by the embodiment of the present invention.
[0060] The decision feedback equalizer includes: an adder 701, a sampler 702, a feedback filter 703, and an adaptation controller 704.
[0061] Among them, the first N1 taps of the feedback filter 703 are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, where N3 < N2; that is to say, the actual number of taps of the feedback filter 703 is N1 + N3, and N3 can be greater than, equal to, or less than N1. The embodiment of the present invention does not make any limitations in this regard.
[0062] Reference Figure 7 In this embodiment, the sampler 702 is used to sample the output signal of the adder 701 and output the sampled data to the feedback filter 703.
[0063] The feedback filter 703 outputs a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps according to the sampled data.
[0064] The above-mentioned second compensation signal corresponding to the last N3 sliding taps is not a certain or several fixed virtual taps, but is dynamically selected and determined according to the compensation weights corresponding to each virtual tap obtained through actual calculation. Specifically, the adaptation controller 704 selects N3 virtual taps and their corresponding compensation weights from the corresponding last N2 virtual taps and maps them to the last N3 sliding taps.
[0065] In a non-limiting embodiment, the adaptation controller 704 can traverse the last N2 virtual taps, select N3 virtual taps with the largest compensation weights and their corresponding compensation weights, and map them to the last N3 sliding taps. In this case, the value of N3 corresponding to different signals is the same, for example, N3 = 3.
[0066] In another non-limiting embodiment, the adaptation controller 704 can traverse the last N2 virtual taps, select some or all of the virtual taps whose compensation weights are greater than the set threshold and their corresponding compensation weights, and map them to the last N3 sliding taps. In this case, the value of N3 corresponding to different signals can be the same or different.
[0067] It should be noted that whether it is the first N1 fixed taps or the last N2 virtual taps, the same algorithm, such as the LMS algorithm, etc., is used to calculate the compensation weights corresponding to each tap.
[0068] In this embodiment, the adder 701 is used to compensate the input signal according to the first compensation signal and the second compensation signal, and output the compensated signal.
[0069] The decision feedback equalizer provided by the embodiment of the present invention designs the feedback filter in a combination of fixed taps and sliding taps, which can significantly improve the compensation ability of the decision feedback equalizer for the channel reflection component without introducing a large area and power consumption, and improve the signal transmission quality.
[0070] The decision feedback equalizer provided by the embodiment of the present invention can also reuse the existing DFE architecture and algorithm to compensate for channel reflection, and has better compatibility.
[0071] The following further gives examples to further elaborate on the decision feedback equalizer provided by the embodiment of the present invention.
[0072] like Figure 8 , which is a schematic diagram of a specific structure of a decision feedback equalizer provided by an embodiment of the present invention.
[0073] In this example, N1=7, N2=63-7, and N3=4.
[0074] Reference Figure 8 , the decision feedback equalizer of this embodiment is Figure 1 The line design shown in the dashed box is added to the traditional DFE structure. Since the low-frequency loss of the continuous channel is small, most DFE structures can meet the demand with a small number of compensation taps. Tap1 to tap7 are used in the example of the present invention. The reflection points in the channel are generally discontinuous quantities, and generally act on individual low-frequency taps, as described above. Traverse the low-frequency tap8 to tap63, and find the four taps with the largest losses for compensation. The invention only needs an additional 4 taps (tapa to tapd) to compensate for reflections in a wide frequency range. Reflections caused by factors such as channel discontinuity are mainly achieved through the newly added line structure in the dashed box. The working mechanism of the line in the dashed box is:
[0075] 1) Pass tap7 through the shift register to obtain tap8~tap63.
[0076] 2) Adaptation controller 704 controls d a ~d d The four taps traverse tap8 to tap63 in turn. The LMS equalization algorithm of each tap in this step is consistent with the traditional structure tap1 to tap7. Finally, the compensation weights w8 to w63 corresponding to tap8 to tap63 can be equivalently obtained. 63 ;
[0077] For example, the first time: d a =tap8,d b =tap9,d c =tap10,d d =tap11; second time: d a =tap12,d b =tap13,d c =tap14,d d =tap15,…, and so on, until tap63.
[0078] 3) The adaptation controller 704 converts w8 to w 63 Sort by size, take out the four largest values and assign them to w respectively a ~w d , and assign the tap numbers corresponding to the four largest values to
[0079] 4) Sum the compensation amounts of the four taps (tapa to tapd) with the largest losses among tap8 to tap63 obtained by the adaptive algorithm and the input signal, and compensate the input signal.
[0080] Figure 8 The corresponding equivalent formula is as follows:
[0081]
[0082] Among them, w n_next = w n + u n × △w n ;
[0083] △w n = VP ideal - VP real ;
[0084] Among them, u n represents the step value for adjusting each tap. This value will vary slightly for different taps according to the convergence speed and stability; generally, the tap with a larger loss will have a relatively larger step value, which is related to the actual application and is usually comprehensively considered according to the convergence speed and accuracy, etc. For example, in the embodiments of the present invention, the step value ratio of tap1: tap2: tap3: tap4: tap5: tap6: tap7 = 8:4:2:1:1:1:1 can be selected; there will be certain differences in the setting of the step value for different structures and applications, and the embodiments of the present invention do not limit this;
[0085] Among them, VP ideal refers to the ideal eye height of the transmitted data, and VP real refers to the actual eye height of the transmitted data.
[0086] It should be noted that the values of N1, N2, and N3 in the above examples are only illustrative. In specific applications, they can be determined according to needs, and the embodiments of the present invention do not limit this.
[0087] Correspondingly, the embodiments of the present invention also provide a method for compensating the reflection loss of a transmission channel. As Figure 9 shown, it is a flowchart of a method for compensating the reflection loss of a transmission channel provided by the embodiments of the present invention, including the following steps:
[0088] Step 901, sample the output signal of the transmission channel to obtain sampled data.
[0089] Step 902, non-linearly filter the sampled data through a feedback filter, where the first N1 taps of the feedback filter are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, with N3 < N2; generate a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps.
[0090] Specifically, delay the sampled data in sequence, determine the output data and compensation weights corresponding to each tap, where the taps include the first N1 fixed taps and the last N2 virtual taps; determine the compensation signals corresponding to the respective taps according to the output data and compensation weights of the respective taps.
[0091] In a non-limiting embodiment, the last N2 virtual taps can be traversed, and N3 virtual taps with the largest compensation weights and their corresponding compensation weights are selected and mapped to the last N3 sliding taps to generate a second compensation signal. In this case, N3 can be a fixed value, such as N3 = 4.
[0092] In another non-limiting embodiment, the last N2 taps can be traversed, and some or all of the virtual taps whose compensation weights are greater than a set threshold and their corresponding compensation weights are selected and mapped to the last N3 sliding taps to generate a second compensation signal. In this case, N3 can be a non-fixed value, but is dynamically determined according to the magnitudes of the compensation weights corresponding to the respective virtual taps actually determined. Of course, in order to reduce the complexity of the circuit structure, in this case, a maximum value of N3 can also be set.
[0093] It should be noted that for the traversal of the last N2 virtual taps above, each tap uses the same LMS algorithm as the first N1 fixed taps to calculate the compensation weights corresponding to the respective virtual taps.
[0094] Step 903, compensate the input signal of the transmission channel according to the first compensation signal and the second compensation signal.
[0095] The transmission channel reflection loss compensation method provided by the embodiments of the present invention can significantly improve the compensation ability of the decision feedback equalizer for the channel reflection component and improve the signal transmission quality by using a feedback filter with sliding taps without introducing a large area and power consumption.
[0096] In a specific implementation, regarding each module / unit included in each device and product described in the above embodiments, it can be a software module / unit, a hardware module / unit, or it can also be partially a software module / unit and partially a hardware module / unit.
[0097] For example, for each device or product applied to or integrated with a chip, each module / unit included therein can be implemented in the form of hardware such as a circuit, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the chip, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit; for each device or product applied to or integrated with a chip module, each module / unit included therein can be implemented in the form of hardware such as a circuit, and different modules / units can be located in the same component (such as a chip, a circuit module, etc.) or different components of the chip module, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the chip module, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit; for each device or product applied to or integrated with a terminal, each module / unit included therein can be implemented in the form of hardware such as a circuit, and different modules / units can be located in the same component (such as a chip, a circuit module, etc.) or different components inside the terminal, or at least some of the modules / units can be implemented in the form of a software program that runs on a processor integrated inside the terminal, and the remaining (if any) part of the modules / units can be implemented in the form of hardware such as a circuit.
[0098] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.
Claims
1. A decision feedback equalizer, characterized in that, including: an adder, a sampler, a feedback filter, and an adaptation controller; the first N1 taps of the feedback filter are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, where N3 < N2; the sampler is configured to sample the output signal of the adder and output sampled data; the feedback filter is configured to output a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps according to the sampled data; the adaptation controller is configured to select N3 virtual taps and their corresponding compensation weights from the corresponding last N2 virtual taps and map them to the last N3 sliding taps; the adder is configured to compensate the input signal according to the first compensation signal and the second compensation signal and output a compensated signal.
2. The decision feedback equalizer according to claim 1, wherein the adaptation controller is specifically configured to traverse the last N2 virtual taps, select N3 virtual taps with the largest compensation weights and their corresponding compensation weights, and map them to the last N3 sliding taps.
3. The decision feedback equalizer according to claim 2, characterized in that, The compensation weight corresponding to each tap is determined according to the LMS algorithm.
4. The decision feedback equalizer according to claim 2, wherein The value of N3 corresponding to different signals is the same.
5. The decision feedback equalizer according to claim 1, wherein the adaptation controller is specifically configured to traverse the last N2 virtual taps, select some or all of the virtual taps whose compensation weights are greater than a set threshold and their corresponding compensation weights, and map them to the last N3 sliding taps.
6. The decision feedback equalizer according to claim 5, characterized in that, The value of N3 corresponding to different signals is the same or different.
7. A method for compensating the reflection loss of a transmission channel, characterized in that, The method includes: sampling the output signal of the transmission channel to obtain sampled data; performing non-linear filtering on the sampled data through a feedback filter, where the first N1 taps of the feedback filter are fixed taps, and the last N3 taps are sliding taps generated based on N2 virtual taps after the first N1 taps, N3 < N2; generating a first compensation signal corresponding to the first N1 fixed taps and a second compensation signal corresponding to the last N3 sliding taps; compensating the input signal of the transmission channel according to the first compensation signal and the second compensation signal.
8. The transmission channel reflection loss compensation method according to claim 7, characterized in that The performing non-linear filtering on the sampled data through the feedback filter includes: delaying the sampled data in sequence to determine the output data and compensation weights corresponding to each tap, where the taps include the first N1 fixed taps and the last N2 virtual taps; determining the compensation signal corresponding to the corresponding tap according to the output data and compensation weight of each tap.
9. The transmission channel reflection loss compensation method according to claim 8, wherein The generating the second compensation signal corresponding to the last N3 sliding taps includes: traversing the last N2 virtual taps, selecting N3 virtual taps with the largest compensation weights and their corresponding compensation weights, mapping them to the last N3 sliding taps, and generating a second compensation signal.
10. The transmission channel reflection loss compensation method according to claim 8, wherein The generating the second compensation signal corresponding to the last N3 sliding taps includes: traversing the last N2 taps, selecting some or all of the virtual taps whose compensation weights are greater than a set threshold and their corresponding compensation weights, mapping them to the last N3 sliding taps, and generating a second compensation signal.