A method of adjusting power consumption of a SerDes system

Abstract

The application discloses a method for adjusting power consumption of a SerDes system and relates to the technical field of high-speed serial communication, and comprises the following steps: obtaining an average amplitude of a service modulation signal and a service baud rate, generating a plurality of single-frequency amplitude modulation signals and superimposing the signals to the service modulation signal, calibrating a frequency domain amplitude of a sending end, extracting a frequency domain amplitude of a receiving end, combining a continuous-time linear equalizer amplitude-frequency response reverse compensation, determining a half-baud rate insertion loss and writing the insertion loss into a power consumption gate register. In the application, the amplitude-consistent multi-frequency point detection information is embedded in the service data, and the equalization response of the frequency domain amplitude of the receiving end is stripped, so that the channel insertion loss can be directly represented by the amplitude drop of the detection frequency point, the deviation caused by the error code, the eye diagram or the lock loss state evaluation is reduced, the digital processing configuration of the receiving end is matched with the real channel loss, the power consumption occupation caused by the continuous opening of the redundant processing is reduced, and the accuracy and stability of the serial link power consumption adjustment are improved.

Description

Technical Field

[0001] This invention relates to the field of high-speed serial communication technology, and in particular to a method for adjusting the power consumption of a SerDes system. Background Technology

[0002] High-speed serial communication technology involves technologies such as high-speed serial data transmission, link establishment, link maintenance, transmission quality detection, channel loss assessment, and communication equipment operation configuration. It typically involves monitoring and managing the data transmission process in a high-speed serial link through methods such as transmitter data modulation, physical channel transmission, receiver clock data recovery, equalization processing, bit error rate detection, and transmission status determination. In traditional SerDes systems, power consumption adjustment refers to configuring and selecting digital processing modules such as CDR, RXFFE, DFE, and MLSD in the receiver's DSP based on the receiver's channel attenuation, insertion loss, and noise level in a high-speed serial transceiver link. Traditional methods typically estimate channel attenuation during the link training phase based on the bit error rate after CTLE, FFE, and DFE equalization, eye diagram quality, or CDR unlock status. Alternatively, the transmitter's FEC inserts error correction redundancy codes, and the receiver assesses signal quality by detecting the error correction codeword status. Based on the assessment results, the number of FFE taps, the DFE switch status, and the MLSD switch status are then selected.

[0003] However, in traditional SerDes systems operating on high-speed serial links, equalization processes such as CTLE, FFE, and DFE are typically performed at the receiver before link quality is assessed based on bit error rate, eye diagram quality, CDR lockout status, or FEC error correction codewords. These assessments are easily affected by channel attenuation, equalization compensation, noise disturbances, and codeword statistical errors, making it difficult to directly obtain frequency domain data characterizing the physical channel insertion loss during service transmission. This leads to inaccurate half-baud rate insertion loss assessments. Consequently, receiver digital processing resources tend to operate at high redundancy levels for extended periods, resulting in excessive receiver power consumption and hindering fine-tuning to match the actual channel attenuation. Summary of the Invention

[0004] The purpose of this invention is to provide a method for adjusting the power consumption of a SerDes system, enabling the receiver digital processing configuration to be matched and adjusted according to the channel insertion loss during service transmission, reducing judgment bias caused by indirect evaluation methods such as bit error rate, eye diagram quality, CDR unlock status, or FEC error correction codewords, and avoiding the receiver digital processing resources from being in a redundant enabled state for a long time, thereby achieving the adaptation of serial link power consumption configuration to the actual physical channel attenuation state.

[0005] To achieve the above objectives, the present invention provides a method for adjusting the power consumption of a SerDes system, comprising the following technical solutions: The process involves acquiring the average amplitude and baud rate of the service modulation signal, determining multiple probe frequencies based on the baud rate, generating multiple single-frequency amplitude modulation (AM) signals based on the average amplitude and the multiple probe frequencies, and superimposing these single-frequency AM signals onto the service modulation signal to form the final transmitted digital signal. A transmit-end frequency domain amplitude check is performed on the final transmitted digital signal, extracting the transmit-end frequency domain amplitude of each single-frequency AM signal at the corresponding probe frequency, determining the average transmit-end frequency domain amplitude based on these amplitudes, and adjusting the time-domain amplitude of each single-frequency AM signal according to its deviation from the average transmit-end frequency domain amplitude until the amplitude difference between each transmit-end frequency domain amplitude does not exceed the amplitude consistency benchmark generated from the average transmit-end frequency domain amplitude. The sampling amplitude of the final transmitted digital signal after passing through the physical transmission channel and continuous-time linear equalizer by the analog-to-digital converter at the receiving end is obtained. The low-speed frequency domain data formed by the sampling amplitude is analyzed, and the receiving end frequency domain amplitude of each single-frequency amplitude-modulated signal is extracted. The amplitude-frequency response corresponding to the continuous-time linear equalizer configuration is obtained, and inverse compensation is performed on the receiving end frequency domain amplitude based on the amplitude-frequency response to generate the probe frequency point compensation amplitude. The channel attenuation slope is determined based on the amplitude drop and frequency interval between the probe frequency point compensation amplitudes. The half-baud rate insertion loss is determined based on the channel attenuation slope, the service baud rate, and the probe frequency point compensation amplitude corresponding to the reference probe frequency point. The receiving end digital processing configuration corresponding to the half-baud rate insertion loss is written into the power consumption gating register.

[0006] Further, the process of forming the final transmitted digital signal includes: acquiring the time-domain sampling sequence of the service modulation signal within a preset observation window, calculating the absolute amplitude average of each sampling amplitude in the time-domain sampling sequence, and determining the absolute amplitude average as the average amplitude; determining multiple detection frequency points based on the service baud rate, such that each detection frequency point is located in a frequency range lower than the half baud rate corresponding to the service baud rate and a preset frequency interval is maintained between two adjacent detection frequency points; configuring the detection time-domain amplitude for each detection frequency point according to the average amplitude; generating each single-frequency amplitude modulation signal according to the detection frequency point and the detection time-domain amplitude; and superimposing each single-frequency amplitude modulation signal with the service modulation signal according to the sampling time to generate the final transmitted digital signal.

[0007] Furthermore, the process of checking the amplitude in the transmitting frequency domain and adjusting the amplitude in the time domain includes: acquiring the observation sampling sequence of the final transmitted digital signal in the transmitting digital domain; performing frequency domain transformation on the observation sampling sequence and locating the detection frequency points corresponding to each of the single-frequency amplitude-modulated signals; extracting the amplitude peak value at each of the detection frequency points as the transmitting frequency domain amplitude of each of the single-frequency amplitude-modulated signals; calculating the average value of each transmitting frequency domain amplitude to obtain the average transmitting frequency domain amplitude; adjusting the time domain amplitude of the corresponding single-frequency amplitude-modulated signal based on the deviation direction of each transmitting frequency domain amplitude relative to the average transmitting frequency domain amplitude; and outputting the adjusted final transmitted digital signal when the amplitude difference between each transmitting frequency domain amplitude does not exceed the amplitude consistency reference.

[0008] Preferably, the generation process of the amplitude consistency benchmark includes: obtaining the average amplitude value of the transmitting frequency domain and the amplitude consistency tolerance ratio; multiplying the average amplitude value of the transmitting frequency domain and the amplitude consistency tolerance ratio to obtain a first candidate consistency benchmark; obtaining the amplitude quantization step size corresponding to the frequency domain transformation; multiplying the amplitude quantization step size and a preset step size multiple to obtain a second candidate consistency benchmark; comparing the first candidate consistency benchmark and the second candidate consistency benchmark; and determining the candidate consistency benchmark with the larger value as the amplitude consistency benchmark.

[0009] Further, the receiving end frequency domain amplitude extraction process includes: acquiring the sampled amplitude sequence output by the receiving end analog-to-digital converter; determining a low-speed analysis window based on the service baud rate and each of the probe frequencies, such that the low-speed analysis window covers at least one complete probe cycle and avoids the update period of the receiving end digital processing configuration; performing decimation processing and frequency domain energy analysis on the sampled amplitude sequence within the low-speed analysis window; locating the frequency domain amplitude peak corresponding to each of the probe frequencies; and determining each of the frequency domain amplitude peaks as the receiving end frequency domain amplitude of each of the single-frequency amplitude-modulated signals.

[0010] Further, the process of determining the low-speed analysis window includes: acquiring the sampling clock information and probe frequency configuration table of the receiver analog-to-digital converter; determining the number of periodic samples for each probe frequency in the receiver sampling domain based on the sampling clock information; and determining the maximum value among the multiple number of periodic samples as the window length reference value; determining the window sampling length according to the window length reference value and the preset period coverage number; and when the window sampling length does not match the receiver buffer depth, truncating the window sampling length to the maximum analyzable length not exceeding the receiver buffer depth to generate the low-speed analysis window.

[0011] Further, the process of generating the compensation amplitude of the detection frequency point includes: obtaining the current configuration word of the continuous-time linear equalizer and each of the detection frequency points; reading the equalization gain value corresponding to each of the detection frequency points from the amplitude-frequency response configuration table based on the current configuration word; establishing a frequency point correspondence between each of the equalization gain values ​​and each of the receiver frequency domain amplitudes; performing gain inverse cancellation processing on each of the receiver frequency domain amplitudes according to the frequency point correspondence to obtain the compensated amplitude corresponding to each of the detection frequency points; and determining each of the compensated amplitudes as the compensation amplitude of the detection frequency point.

[0012] Preferably, the generation process of the amplitude-frequency response configuration table includes: obtaining the sweep calibration records of the continuous-time linear equalizer under multiple configuration levels; extracting the amplitude change between the output amplitude and the input amplitude at multiple calibration frequency points for each configuration level to obtain the level frequency response record; indexing and storing the level frequency response record according to the configuration level identifier and the calibration frequency point identifier; and determining the interpolation gain value based on the amplitude change between two adjacent calibration frequency points when any of the detection frequency points does not match any of the calibration frequency points to generate the amplitude-frequency response configuration table.

[0013] Further, the process of writing the channel attenuation slope, half-baud rate insertion loss, and power consumption gating register includes: arranging each of the detection frequency points and the corresponding compensation amplitude of the detection frequency points in ascending order of frequency; calculating the amplitude drop between two adjacent compensation amplitudes of the detection frequency points; and determining the frequency difference between two adjacent detection frequency points as the frequency interval; determining the channel attenuation slope based on multiple amplitude drop values ​​and the corresponding frequency intervals; obtaining the compensation amplitude of the detection frequency point corresponding to the lowest detection frequency point as the low-frequency reference amplitude; performing insertion loss extrapolation processing on the channel attenuation slope according to the half-baud rate corresponding to the service baud rate to obtain the extrapolated compensation amplitude corresponding to the half-baud rate; determining the amplitude difference between the low-frequency reference amplitude and the extrapolated compensation amplitude as the half-baud rate insertion loss; and reading the receiver digital processing configuration matching the half-baud rate insertion loss from the receiver digital processing configuration table and writing it into the power consumption gating register.

[0014] Further, the process of determining the channel attenuation slope includes: acquiring a first detection frequency point and a second detection frequency point that are adjacent in ascending order of frequency, and making the frequency of the first detection frequency point lower than the frequency of the second detection frequency point; subtracting the compensation amplitude of the second detection frequency point from the compensation amplitude of the detection frequency point corresponding to the first detection frequency point to obtain the segmented amplitude reduction amount; acquiring the segmented frequency interval between the first detection frequency point and the second detection frequency point; converting the segmented amplitude reduction amount to a unit frequency according to the segmented frequency interval to obtain the segmented attenuation slope; and averaging multiple segmented attenuation slopes to generate the channel attenuation slope.

[0015] The present invention has the following beneficial effects: 1. In this invention, by acquiring the average amplitude and baud rate of the service modulation signal, multiple single-frequency amplitude modulation signals are generated and superimposed on the service modulation signal, so that the detection information is transmitted synchronously with the service data, forming a detection basis characterizing the channel response at different frequency points.

[0016] 2. In this invention, by performing a frequency domain amplitude check at the transmitting end and calibrating the time domain amplitude based on the average frequency domain amplitude at the transmitting end, the amplitude of each detection frequency point is kept consistent before entering the channel, thereby reducing the impact of differences injected at the transmitting end on the amplitude analysis at the receiving end.

[0017] 3. In this invention, by extracting the frequency domain amplitude of the receiver and combining it with the inverse compensation of the amplitude-frequency response of the continuous-time linear equalizer, the influence of the equalization gain is removed. Based on the decrease in amplitude after compensation, the channel attenuation slope and half-baud rate insertion loss are determined, so that the digital processing configuration matches the actual channel loss and reduces redundant power consumption. Attached Figure Description

[0018] Figure 1 This is the main flowchart for adjusting the power consumption of the SerDes system in this invention; Figure 2 This is the data flow diagram of the final transmitted digital signal in this invention; Figure 3 This is a closed-loop flowchart of the frequency domain amplitude check and time domain amplitude adjustment at the beginning of this invention; Figure 4 This is a data flow diagram for frequency domain amplitude extraction at the receiving end of the present invention; Figure 5 This is a logic block diagram of the reverse compensation of the amplitude-frequency response of the continuous-time linear equalizer of the present invention. Figure 6 This is a feedback closed-loop diagram of the half-baud rate insertion loss determination and power consumption gating writing of the present invention. Detailed Implementation

[0019] The technical solution described in this invention will be explained below with reference to the accompanying drawings. This embodiment takes the service modulation signal in a SerDes serial link as the processing object. The service modulation signal, average amplitude, service baud rate, probe frequency, single-frequency amplitude modulation signal, final transmitted digital signal, transmitting-end frequency domain amplitude, receiving-end frequency domain amplitude, probe frequency compensation amplitude, channel attenuation slope, half-baud rate insertion loss, and power consumption gate register involved are all understood according to the technical meanings described in the application documents. Parameters for which no specific values ​​are limited in this embodiment can be set based on the link rate, sampling conditions, buffer depth, and the receiving-end digital processing configuration table.

[0020] Please see Figures 1 to 6 This embodiment provides a method for adjusting the power consumption of a SerDes system. The method superimposes multiple single-frequency amplitude-modulated signals onto the service modulation signal, and uses a transmitter frequency domain amplitude check to ensure that the multiple probe frequencies have a consistent transmitter amplitude basis. The receiver extracts the receiver frequency domain amplitude of each probe frequency from the low-speed frequency domain data and performs inverse compensation based on the amplitude-frequency response of a continuous-time linear equalizer. The compensated probe frequency amplitude is used to determine the channel attenuation slope and half-baud rate insertion loss, and then the half-baud rate insertion loss is matched to the receiver's digital processing configuration and written into the power consumption gating register.

[0021] S1 forms the final transmitted digital signal.

[0022] S101: Acquire the time-domain sampling sequence of the service modulation signal within a preset observation window. Take the absolute amplitude of each sample in the time-domain sampling sequence and average it to obtain the average amplitude. For example, in a SerDes link with a single-channel transmission rate of 112Gbps PAM4 (fourth-order pulse amplitude modulation), if the observation window is set to 1024 UI (unit intervals), the time-domain absolute average amplitude of the service modulation signal at this time is acquired to be 400mV. This average amplitude is used to subsequently configure the detection time-domain amplitude for the detection frequency point, ensuring that the amplitude source of the detected components corresponds to the current amplitude state of the service signal.

[0023] S102, acquire the service baud rate, and determine multiple probe frequencies based on the service baud rate. Each probe frequency is located within a frequency range lower than the half-baud rate corresponding to the service baud rate, and a preset frequency interval is maintained between adjacent probe frequencies. Specifically, when the service baud rate is 56 GBaud, its corresponding half-baud rate (Nyquist frequency) is 28 GHz. The system can select frequencies lower than this. , , The three probe frequencies maintain a strict 2GHz frequency interval between adjacent frequencies. These probe frequencies serve as a common frequency reference for subsequent signal generation, amplitude checking, receiver extraction, equalization compensation, and insertion loss extrapolation.

[0024] S103: Based on the average amplitude, configure the detection time-domain amplitude for each detection frequency point, and generate multiple single-frequency amplitude-modulated (AM) signals according to the detection frequency point and the detection time-domain amplitude. In this application scenario, based on an average amplitude of 400mV, configure a peak-to-peak detection time-domain amplitude of 12mV (amplitude percentage of approximately 3%) for the aforementioned 2GHz, 4GHz, and 6GHz detection frequencies, ensuring that the superposition of the single-frequency AM signals does not cause the eye diagram closure margin of the original PAM4 service signal to fall below the tolerance. Subsequently, align and superimpose each single-frequency AM signal with the service modulation signal according to the sampling time to form the final transmitted digital signal. The final transmitted digital signal carries both the service modulation signal and the detection frequency component used for channel loss characterization.

[0025] S2, perform a frequency domain amplitude check at the transmitting end and adjust the amplitude of the probe components.

[0026] S201: Obtain the observation sampling sequence of the final transmitted digital signal in the digital domain of the transmitting end, perform frequency domain transformation on the observation sampling sequence, and locate the detection frequency points corresponding to each single-frequency amplitude modulation signal. Extract the amplitude peak value at each detection frequency point as the transmitting frequency domain amplitude of the corresponding single-frequency amplitude modulation signal.

[0027] S202, calculate the average value of the frequency domain amplitude at each transmitting end to obtain the average value of the frequency domain amplitude at the transmitting end. Based on the deviation direction of each transmitting end's frequency domain amplitude from the average value of the frequency domain amplitude at the transmitting end, determine the adjustment direction of the time domain amplitude of the corresponding single-frequency amplitude modulated signal. When the transmitting end's frequency domain amplitude is higher than the average value of the frequency domain amplitude at the transmitting end, decrease the time domain amplitude of the corresponding single-frequency amplitude modulated signal; when the transmitting end's frequency domain amplitude is lower than the average value of the frequency domain amplitude at the transmitting end, increase the time domain amplitude of the corresponding single-frequency amplitude modulated signal.

[0028] S203, Generate an amplitude consistency benchmark. Specifically: Obtain the average amplitude value and amplitude consistency tolerance ratio in the transmitting frequency domain, and multiply them as the first candidate consistency benchmark; obtain the amplitude quantization step size corresponding to the frequency domain transformation, and multiply the amplitude quantization step size by a preset step size multiple as the second candidate consistency benchmark; compare the first and second candidate consistency benchmarks, and determine the candidate consistency benchmark with the larger value as the amplitude consistency benchmark. Assuming the average amplitude value in the transmitting frequency domain is 12mV, and the amplitude consistency tolerance ratio is set to 2.5%, then the first candidate consistency benchmark is 0.3mV; if the transmitting digital-to-analog converter (DAC) resolution is 8-bit, the amplitude quantization step size corresponding to a full-scale range of 800mV is approximately 3.125mV, and the preset step size multiple is 0.1, then the second candidate consistency benchmark is 0.3125mV. After comparison, the larger value of 0.3125mV is selected as the amplitude consistency benchmark.

[0029] S204: After completing one time-domain amplitude adjustment, a second transmit-end frequency-domain amplitude check is performed. If the amplitude difference between the frequency-domain amplitudes of each transmit end exceeds the amplitude consistency benchmark, the time-domain amplitude of the corresponding single-frequency amplitude-modulated signal continues to be adjusted according to the deviation relationship; if the amplitude difference between the frequency-domain amplitudes of each transmit end does not exceed the amplitude consistency benchmark, the adjusted final transmit digital signal is output. This ensures that multiple probe frequencies have a consistent frequency-domain amplitude base at the transmit end, facilitating subsequent characterization of channel attenuation based on amplitude changes at the receive end.

[0030] S3, extract the frequency domain amplitude of the receiver.

[0031] S301, the receiver's analog-to-digital converter acquires the sampling amplitude of the final transmitted digital signal after passing through the physical transmission channel and the continuous-time linear equalizer, forming a sampling amplitude sequence. This sampling amplitude sequence contains the service signal component and the probe frequency component, and is affected by both the physical transmission channel attenuation and the amplitude-frequency response of the continuous-time linear equalizer.

[0032] S302, the low-speed analysis window is determined based on the service baud rate and each probe frequency. Specifically: the sampling clock information of the receiver's analog-to-digital converter and the probe frequency configuration table are obtained. Based on the sampling clock information, the number of periodic samples for each probe frequency in the receiver's sampling domain is determined, and the maximum value among multiple periodic sample counts is used as the window length reference value. Then, the window sampling length is determined according to the window length reference value and the preset period coverage quantity. The low-speed analysis window covers at least one complete probe cycle and avoids the update period of the receiver's digital processing configuration. When the window sampling length exceeds the analysis length allowed by the receiver's buffer depth, the window sampling length is truncated to the maximum analyzable length not exceeding the receiver's buffer depth. For example, if the receiver's ADC sampling clock is 56GS / s, the number of single-cycle sampling points to cover the 2GHz probe frequency is 28; to improve the FFT frequency resolution, multiple cycles can be covered and the window length can be set to 28,000 sampling points. Considering that the maximum cache depth of the receiver's static random access memory (SRAM) for data capture is 32768 (32K) words, the final window sampling length is set to 32768, which satisfies the depth limit and avoids the adaptive algorithm configuration update period triggered every 10ms.

[0033] S303 performs decimation processing and frequency domain energy analysis on the sampled amplitude sequence within the low-speed analysis window, locates the frequency domain amplitude peak corresponding to each detection frequency point, and determines each frequency domain amplitude peak as the receiver frequency domain amplitude of each single-frequency amplitude-modulated signal. The receiver frequency domain amplitude corresponds one-to-one with the detection frequency point and serves as the input for subsequent amplitude-frequency response inverse compensation.

[0034] S4 generates the compensation amplitude for the detection frequency point.

[0035] S401, Obtain the current configuration word and each probe frequency point of the continuous-time linear equalizer. The current configuration word is used to determine the current configuration level of the continuous-time linear equalizer, and each probe frequency point is used to determine the frequency position to be read.

[0036] S402: Based on the current configuration word, read the equalization gain value corresponding to each probe frequency point from the amplitude-frequency response configuration table. The amplitude-frequency response configuration table is formed based on the sweep calibration records of the continuous-time linear equalizer at multiple configuration levels. It records the amplitude change between the output amplitude and the input amplitude at multiple calibration frequency points for each configuration level, and stores them indexed according to the configuration level identifier and calibration frequency point identifier. When the probe frequency point is inconsistent with the calibration frequency point, the interpolation gain value can be determined based on the amplitude change between two adjacent calibration frequency points.

[0037] S403 establishes a correspondence between each equalization gain value and the frequency domain amplitude of each receiver according to the detection frequency point, and performs inverse gain cancellation processing on the frequency domain amplitude of the receiver based on this correspondence to obtain the compensated amplitude corresponding to each detection frequency point. To accurately isolate the gain influence of the analog front-end, the inverse gain cancellation processing here is calculated using the following formula: ; In the formula, Representing the Each detection frequency point The corresponding compensation amplitude (unit: dBV). This represents the received frequency domain amplitude (unit: V) of the single-frequency amplitude-modulated signal extracted from the low-speed analysis window at the receiving end. Represents the current configuration word The continuous-time linear equalizer (CTLE) found in the amplitude-frequency response configuration table at the frequency point The equalization gain value provided at the location (unit: dB).

[0038] For example, for the 2GHz frequency point, the extracted The value is 6mV (equivalent to -44.4dBV). If the table shows that the current CTLE configuration at this frequency has a gain of 1.5dB, then the compensated amplitude after removing the CTLE effect is... The value is -45.9 dBV. The compensated amplitude is determined as the probe frequency point compensation amplitude, which is used to represent the frequency point amplitude state after stripping the amplitude-frequency response of the continuous-time linear equalizer.

[0039] S5 determines the half-baud rate insertion loss and writes it into the power gating register.

[0040] S501: Arrange the detection frequency points and their corresponding compensation amplitudes in ascending order of frequency to form an ordered frequency point amplitude sequence. For two adjacent detection frequency points, take the difference between the compensation amplitude of the detection frequency point corresponding to the low-frequency detection frequency point and the compensation amplitude of the detection frequency point corresponding to the high-frequency detection frequency point to obtain the segmented amplitude decrease; take the frequency difference between two adjacent detection frequency points to obtain the segmented frequency interval.

[0041] S502, the segmented amplitude decrease is converted to a unit frequency according to the segmented frequency interval to obtain the segmented attenuation slope. The segmented attenuation slopes formed by multiple adjacent detection frequency points are averaged to generate the channel attenuation slope. The channel attenuation slope is used to characterize the decreasing trend of the detection frequency point compensation amplitude as the frequency increases.

[0042] S503: Obtain the compensation amplitude of the detection frequency corresponding to the lowest detection frequency as the low-frequency reference amplitude, and perform insertion loss extrapolation processing on the channel attenuation slope according to the half-baud rate corresponding to the service baud rate to obtain the extrapolated compensation amplitude corresponding to the half-baud rate. The amplitude difference between the low-frequency reference amplitude and the extrapolated compensation amplitude is determined as the half-baud rate insertion loss. In this step, the specific extrapolation model of the half-baud rate insertion loss is supported by the following multi-frequency slope fitting formula: ; In the formula, This represents the insertion loss at the final calculated half-baud rate (in dB). This represents the total number of detection frequency points (as in the aforementioned scenario). ), and These represent the compensated amplitudes (unit: dBV) at low frequency and adjacent high frequency detection points, respectively. and This represents the corresponding detection frequency (unit: GHz). The half-baud rate corresponding to the service baud rate (e.g., 28GHz corresponding to 56GBaud). Represents the lowest detection frequency (i.e. ).

[0043] The summation term and division in the formula The portion referred to is the channel attenuation slope generated earlier. Assuming the average channel attenuation slope is calculated to be 0.8 dB / GHz based on the compensated amplitude, the half-baud rate insertion loss extrapolated from 2 GHz to 28 GHz is... .

[0044] S504 reads the receiver digital processing configuration matching the half-baud rate insertion loss from the receiver digital processing configuration table and writes this configuration to the power gating register. Taking the calculated half-baud rate insertion loss of 20.8dB as an example, since it is less than the severe attenuation threshold of 25dB, the system determines that the current channel loss is at a medium level. It then finds the matching "medium compensation mode" configuration from the configuration table and writes the hexadecimal control word 0x0F to the power gating register. This write operation, through a clock gating mechanism, dynamically disables the 9th to 16th order redundant tap logic in the decision feedback equalizer (DFE) used to handle long-tail reflections, significantly reducing the overall power consumption of the SerDes receiver DSP core from 850mW to approximately 720mW. This achieves significant power optimization while ensuring the bit error rate meets the target. After receiving the write result, the power gating register is used to adjust the on / off state or configuration level of the receiver digital processing resources. The next low-speed analysis window will avoid the period when the receiver's digital processing configuration is updated, so as to reduce the impact of configuration switching on the receiver's frequency domain amplitude extraction.

[0045] Through the above process, the amplitude of the probe frequency point of the final transmitted digital signal undergoes consistency calibration at the transmitting end. The frequency domain amplitude extracted at the receiving end is then compensated inversely by the amplitude-frequency response of a continuous-time linear equalizer. The compensated amplitude of the probe frequency point is thus used to determine the channel attenuation slope and half-baud rate insertion loss. The half-baud rate insertion loss is further used to match the receiver's digital processing configuration and written into the power consumption gating register, so that the configuration of the receiver's digital processing resources corresponds to the current channel attenuation state.

[0046] The above embodiments are used to illustrate the implementable process of the present invention and should not be construed as limiting the present invention to the specific processes, parameter selections, or table configurations described above. Without departing from the technical solutions described in the present invention and the scope supported by the application documents, any equivalent substitutions, modifications, combinations, order adjustments, equivalent transformations of field names, or equivalent acceptance of the executing entity that can be made by those skilled in the art should fall within the scope of protection of this patent; however, the substantive correspondence of the technical objects must not be altered through changes in name, number, or carrier.

Claims

1. A method for adjusting the power consumption of a SerDes system, characterized in that, include: The average amplitude and baud rate of the service modulation signal are obtained. Multiple probe frequencies are determined based on the service baud rate. Multiple single-frequency amplitude modulation signals are generated based on the average amplitude and multiple probe frequencies. The single-frequency amplitude modulation signals are superimposed on the service modulation signal to form the final transmitted digital signal. Perform a transmit frequency domain amplitude check on the final transmitted digital signal, extract the transmit frequency domain amplitude of each single-frequency amplitude modulated signal at the corresponding detection frequency point, determine the average transmit frequency domain amplitude based on each transmit frequency domain amplitude, and adjust the time domain amplitude of each single-frequency amplitude modulated signal according to the deviation of each transmit frequency domain amplitude from the average transmit frequency domain amplitude until the amplitude difference between each transmit frequency domain amplitude does not exceed the amplitude consistency benchmark generated by the average transmit frequency domain amplitude. The sampling amplitude of the final transmitted digital signal after passing through the physical transmission channel and continuous-time linear equalizer by the analog-to-digital converter at the receiving end is obtained. The low-speed frequency domain data formed by the sampling amplitude is analyzed, and the receiving frequency domain amplitude of each single-frequency amplitude-modulated signal is extracted. Obtain the amplitude-frequency response corresponding to the continuous-time linear equalizer configuration, perform reverse compensation on the frequency domain amplitude of the receiver based on the amplitude-frequency response, and generate the compensation amplitude of the probe frequency point. The channel attenuation slope is determined based on the amplitude drop between the compensation amplitudes of the probe frequency points and the frequency interval. The half-baud rate insertion loss is determined based on the channel attenuation slope, the service baud rate, and the compensation amplitude of the probe frequency point corresponding to the reference probe frequency point. The receiver digital processing configuration corresponding to the half-baud rate insertion loss is written into the power consumption gating register.

2. The method according to claim 1, characterized in that, The process of forming the final transmitted digital signal includes: Obtain the time-domain sampling sequence of the service modulation signal within a preset observation window, calculate the absolute amplitude average of each sampling amplitude in the time-domain sampling sequence, and determine the absolute amplitude average as the average amplitude; Multiple detection frequency points are determined based on the service baud rate, and each detection frequency point is located in a frequency range lower than the half baud rate corresponding to the service baud rate, with a preset frequency interval maintained between two adjacent detection frequency points. The detection time domain amplitude is configured for each detection frequency point according to the average amplitude. Each single-frequency amplitude modulation signal is generated according to the detection frequency point and the detection time domain amplitude. Each single-frequency amplitude modulation signal is aligned and superimposed with the service modulation signal according to the sampling time to generate the final transmitted digital signal.

3. The method according to claim 2, characterized in that, The process of frequency domain amplitude check and time domain amplitude adjustment at the origin includes: The observation sampling sequence of the final transmitted digital signal in the digital domain of the transmitting end is obtained, the frequency domain transformation is performed on the observation sampling sequence and the detection frequency point corresponding to each single-frequency amplitude modulation signal is located, and the amplitude peak value at each detection frequency point is extracted as the transmitting frequency domain amplitude of each single-frequency amplitude modulation signal. Calculate the average value of the frequency domain amplitude of each of the transmitting ends to obtain the average value of the frequency domain amplitude of the transmitting ends. Based on the deviation direction of each of the frequency domain amplitudes of the transmitting ends relative to the average value of the frequency domain amplitude of the transmitting ends, adjust the time domain amplitude of the corresponding single-frequency amplitude-modulated signal. When the amplitude difference between each of the frequency domain amplitudes of the transmitting ends does not exceed the amplitude consistency reference, output the adjusted final transmitted digital signal.

4. The method according to claim 3, characterized in that, The process of generating the amplitude consistency benchmark includes: The average amplitude of the transmitting frequency domain and the amplitude consistency tolerance ratio are obtained. The product of the average amplitude of the transmitting frequency domain and the amplitude consistency tolerance ratio is calculated to obtain the first candidate consistency benchmark. Obtain the amplitude quantization step size corresponding to the frequency domain transformation, multiply the amplitude quantization step size by a preset step size multiple to obtain a second candidate consensus benchmark, compare the first candidate consensus benchmark and the second candidate consensus benchmark, and determine the candidate consensus benchmark with the larger value as the amplitude consensus benchmark.

5. The method according to claim 3, characterized in that, The frequency domain amplitude extraction process at the receiving end includes: The sampling amplitude sequence output by the analog-to-digital converter at the receiving end is obtained, and a low-speed analysis window is determined based on the service baud rate and each of the detection frequency points, so that the low-speed analysis window covers at least one complete detection cycle and avoids the update period of the digital processing configuration at the receiving end. The sampled amplitude sequence is subjected to decimation processing and frequency domain energy analysis within the low-speed analysis window to locate the frequency domain amplitude peak corresponding to each of the detection frequency points, and each of the frequency domain amplitude peaks is determined as the receiving end frequency domain amplitude of each of the single-frequency amplitude-modulated signals.

6. The method according to claim 5, characterized in that, The process of determining the low-speed analysis window includes: Obtain the sampling clock information and probe frequency configuration table of the analog-to-digital converter at the receiving end, determine the number of periodic samples for each probe frequency in the sampling domain at the receiving end based on the sampling clock information, and determine the maximum value among the multiple number of periodic samples as the window length reference value; The window sampling length is determined according to the window length reference value and the preset period coverage quantity. When the window sampling length does not match the receiver buffer depth, the window sampling length is truncated to the maximum analyzable length not exceeding the receiver buffer depth to generate the low-speed analysis window.

7. The method according to claim 5, characterized in that, The process of generating the detection frequency compensation amplitude includes: Obtain the current configuration word of the continuous-time linear equalizer and each of the probe frequencies; based on the current configuration word, read the equalization gain value corresponding to each of the probe frequencies from the amplitude-frequency response configuration table; and establish a frequency point correspondence between each of the equalization gain values ​​and each of the receiver frequency domain amplitudes. Based on the frequency point correspondence, gain inverse cancellation processing is performed on the frequency domain amplitude of each receiver to obtain the compensated amplitude corresponding to each detection frequency point, and each compensated amplitude is determined as the compensation amplitude of the detection frequency point.

8. The method according to claim 7, characterized in that, The generation process of the amplitude-frequency response configuration table includes: Obtain the frequency sweep calibration record of the continuous-time linear equalizer under multiple configuration levels, and extract the amplitude change between the output amplitude and the input amplitude at multiple calibration frequency points for each configuration level to obtain the frequency response record of the level. The gear frequency response records are indexed and stored according to the configured gear identifier and calibration frequency point identifier. When any of the detection frequency points is not consistent with any of the calibration frequency points, the interpolation gain value is determined based on the amplitude change of two adjacent calibration frequency points, and the amplitude frequency response configuration table is generated.

9. The method according to claim 7, characterized in that, The process of writing the channel attenuation slope, half-baud rate insertion loss, and power consumption gate registers includes: Arrange the detection frequency points and their corresponding compensation amplitudes in order of frequency from low to high, calculate the amplitude drop between two adjacent detection frequency point compensation amplitudes, and determine the frequency difference between two adjacent detection frequency points as the frequency interval. The channel attenuation slope is determined based on multiple amplitude reduction amounts and the corresponding frequency intervals. The compensation amplitude of the detection frequency point corresponding to the lowest detection frequency point is obtained as the low-frequency reference amplitude. The channel attenuation slope is extrapolated according to the half-baud rate corresponding to the service baud rate to obtain the extrapolated compensation amplitude corresponding to the half-baud rate. The amplitude difference between the low-frequency reference amplitude and the extrapolated compensation amplitude is determined as the half-baud rate insertion loss. The receiver digital processing configuration that matches the half-baud rate insertion loss is read from the receiver digital processing configuration table and written into the power consumption gate register.

10. The method according to claim 9, characterized in that, The process of determining the channel attenuation slope includes: Obtain the first and second detection frequency points that are adjacent in order of frequency from low to high, and make the frequency of the first detection frequency point lower than the frequency of the second detection frequency point. Subtract the compensation amplitude of the second detection frequency point from the compensation amplitude of the detection frequency point corresponding to the first detection frequency point to obtain the segmented amplitude reduction amount. Obtain the segmented frequency interval between the first detection frequency point and the second detection frequency point, convert the segmented amplitude decrease amount according to the segmented frequency interval to a unit frequency to obtain the segmented attenuation slope, and average the multiple segmented attenuation slopes to generate the channel attenuation slope.

Images