阅读说明：本技术 操作过程中的采样器偏移校准 (Sampler offset calibration during operation ) 是由 阿里·霍马提 于 2020-02-20 设计创作，主要内容包括：在所描述的方法和系统中：以数据采样器对数据信号进行采样,其中,所述数据采样器在具有与判定反馈均衡校正因子相关联的判定阈值的数据信号处理路径内操作；通过调节操作于所述数据信号处理路径外的备用采样器的判定阈值,测量所述数据信号的眼开度,以确定该备用采样器判定阈值的眼图中心值；根据所述眼图中心值和所述DFE校正因子,初始化所述备用采样器的判定阈值；在检测出预设数据模型后,生成所述备用采样器和数据采样器的相应的相位误差信号组；以及根据相应的所述相位误差信号组中相位误差信号的差异累加结果,更新所述数据采样器的判定阈值。(In the described method and system: sampling a data signal with a data sampler, wherein the data sampler operates within a data signal processing path having a decision threshold associated with a decision feedback equalization correction factor; measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler operating outside the data signal processing path to determine an eye center value of the spare sampler decision threshold; initializing a decision threshold of the backup sampler as a function of the eye center value and the DFE correction factor; after a preset data model is detected, generating corresponding phase error signal groups of the standby sampler and the data sampler; and updating the decision threshold of the data sampler according to the difference accumulation result of the phase error signals in the corresponding phase error signal group.)

1. A method, comprising:

sampling a data signal with a data sampler, wherein the data sampler operates within a data signal processing path having a decision threshold associated with a decision feedback equalization correction factor;

measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler to determine an eye center value of the decision threshold of the spare sampler, wherein the spare sampler operates outside the data signal processing path;

initializing a decision threshold of the standby sampler according to the eye diagram center value and the decision feedback equalization correction factor;

generating respective sets of phase error signals for the spare sampler and the data sampler in response to detection of a preset data pattern; and

and updating the decision threshold of the data sampler according to the difference accumulation result of the phase error signals in the corresponding phase error signal group.

2. The method of claim 1, further comprising: updating a decision threshold of the spare sampler in response to the difference accumulation result of phase error signals in the respective set of phase error signals.

3. The method of claim 2, wherein the decision threshold of the spare sampler is updated until the difference accumulation results for phase error signals in the corresponding set of phase error signals are within a threshold.

4. The method of claim 3, further comprising: determining an offset correction value between an updated decision threshold of the spare sampler and an initialized decision threshold of the spare sampler.

5. The method of claim 4, wherein the decision threshold of the data sampler is updated in accordance with the offset correction value.

6. The method of claim 5, wherein the decision threshold of the data sampler is updated only in the direction determined by the offset correction value.

7. The method of claim 5, wherein the decision threshold of the data sampler is updated according to the direction and magnitude determined by the offset correction value.

8. The method of any of claims 1 to 7, further comprising: determining the difference accumulation result between the respective sets of phase error signals by comparing the respective sets of phase error signals.

9. An apparatus, comprising:

a data sampler operable within a data signal processing path having a decision threshold associated with a decision feedback equalization correction factor, wherein the data sampler is to sample a data signal;

a backup sampler operative outside the data signal processing path, wherein the backup sampler is to measure an eye opening of the data signal by adjusting a decision threshold of the backup sampler to determine an eye center value of the decision threshold of the backup sampler;

a decision feedback equalization tap generator, wherein the decision feedback equalization tap generator is to provide the decision feedback equalization correction factor to an eye range sampler to initialize a decision threshold for the standby sampler,

the data sampler and the standby sampler are used for responding to the detection of a preset data model and generating a corresponding phase error signal group; and

an accumulator, wherein the accumulator is configured to update the decision threshold of the data sampler according to a difference accumulation result of the phase error signals in the corresponding phase error signal group.

10. The apparatus of claim 9, wherein decisions of the spare samplers are updated in response to the difference accumulation results for phase error signals in the respective set of phase error signals.

11. The apparatus of claim 10, wherein the decision threshold of the spare sampler is updated until the difference accumulation result of the phase error signals in the corresponding set of phase error signals is within a preset threshold.

12. The apparatus of claim 11, further comprising control logic to determine an offset correction value between an updated decision threshold of the spare sampler and an initialized decision threshold of the spare sampler.

13. The apparatus of claim 12, wherein a decision threshold of the data sampler is updated according to the offset correction value.

14. Apparatus according to any one of claims 9 to 13, wherein said accumulator comprises a register for holding said difference accumulated result between respective said sets of phase error signals.

15. The apparatus of claim 14, wherein the register is to update a least significant bit in response to a determination of a difference between a pair of phase error signals selected from the respective set of phase error signals.

Background

In modern digital systems, digital information must be processed efficiently and reliably. In this context, digital information must be understood as information contained within discrete values (i.e. non-continuous values). Digital information may be represented not only by bits and bit sets, but also by numbers within a limited set.

To increase the overall bandwidth, most inter-chip or inter-device communication systems employ multiple lines for communication. Each or each pair of these lines may be referred to as a channel or link, and the multiple channels make up a communication bus between the electronic devices. At the physical circuit level, buses within an inter-chip communication system are typically made up of encapsulated electrical conductors between the chip and the motherboard, on a Printed Circuit Board (PCB), or within inter-PCB cables and connectors. Furthermore, in high frequency applications, microstrip or strip PCB traces may also be employed.

Common bus line signaling methods include single-ended signaling and differential signaling. In applications requiring high-speed communication, these methods may be further optimized in terms of power consumption and pin utilization, particularly in high-speed communication. Recently, the vector signaling method proposed in Shokrollahi, etc. can achieve more optimized trade-offs in power consumption, pin utilization, and noise robustness of the inter-chip communication system. Such vector signaling systems convert the digital information of the transmitter into a different representation space of vector codewords, and select different vector codewords according to the characteristics of the transmission channel and the design constraints of the communication system to make a better trade-off between power consumption, pin utilization, and speed. This process is referred to as "encoding" in this application. The encoded codeword is transmitted from the transmitter to one or more receivers in the form of a set of signals. The receiver inverts the received signal corresponding to the codeword into the original digital information representation space. This process is referred to as "decoding" in this application.

Regardless of the encoding method employed, the signals received by the receiving device must be sampled at intervals (or otherwise have their signal values recorded), and the sampling intervals must be such that the sampled values represent the original transmitted values in an optimal manner, regardless of the delay, interference and noise conditions of the transmission channel. This Clock and Data Recovery (CDR) operation requires that not only the proper sampling time be determined, but that the proper sampling time be determined continuously to achieve dynamic compensation for changing signal propagation conditions. A communication receiver typically extracts a receive clock signal from a received data stream. To facilitate such Clock Data Recovery (CDR) operations, some communication protocols distinguish between clock-related and data-related signal components by imposing constraints on the communication signaling. Similarly, some communication receivers detect data by processing the received signal above a required minimum limit in order to provide additional information that facilitates clock recovery. For example, a so-called dual-bitrate receive sampler can measure the received signal level at twice the required data reception rate, thereby enabling independent detection of the received signal level corresponding to the data component and the received signal transition associated with the signal clock component and offset in time.

However, the transitions introduced into external communication protocols are known to cause limitations on the data communication rates that can be achieved. At the same time, receiving at a sampling rate higher than the transmit data rate is known to significantly increase the power consumption of the receiver.

Real-world communication channels are imperfect and can negatively affect both the amplitude (e.g., fading) and time (e.g., delay and pulse smearing) of the transmitted signal, and such effects can be eliminated by transmitter pre-compensation and/or receive equalization. Continuous Time Linear Equalization (CTLE) is a known frequency domain equalization method that can, for example, compensate for an increase in channel attenuation at high frequencies. In addition, time-domain guided equalization methods are used to compensate for the effects of intersymbol interference (ISI) on the received signal. Such ISI results because the residual electrical effects of the previously transmitted signal within the communication transmission medium have an effect on the amplitude and time of the current symbol interval. For example, the presence of one or more transmission line media with an impedance anomaly may cause signal reflections. As such, when the transmitted signal propagates within the medium, partial reflections will occur due to the one or more impedance anomalies, and such reflected signals will arrive at the receiver at some later point in time in a superimposed form with the normally propagating signal.

One data-dependent receive equalization method is Decision Feedback Equalization (DFE). In DFE, time-domain guided equalization is performed by: maintaining, by the receiver, a history of previously received data values; the history of received data values is then processed by the transmission line model to infer the expected impact of each historical data value on the currently received signal. The transmission line model may be pre-calculated, obtained by measurement, generated by trial and error, and may cover the effects of one or more previous data intervals. The predicted degree of influence for the one or more previous data intervals is collectively referred to as a DFE compensation amount, which at low or medium data rates may be calculated in time before a subsequent data sample value is detected, for application, e.g., explicit subtraction from the received data signal prior to a received sampling operation, or implicit subtraction by altering a reference level in a received data sampler or comparator that is the subject of comparison of the received data signal. However, at high data rates, the detection of the previous data bit and the calculation of the DFE compensation amount may not be completed in time before the next data sample value is detected, thus requiring the calculation of a so-called "expansion" DFE for the speculative or possible data values, rather than the known previous data values. For example, the expansion DFE processing stage may predict two different compensation values based on whether the result of the parsing of the determined data bits is "1" or "0", while the receiving detector performs a sampling (slicing) operation based on the two prediction results simultaneously, and retains the resulting multiple results until the parsing of the previous unit interval DFE decision is completed.

Disclosure of Invention

In the method and system described in the present application: sampling a data signal with a data sampler, wherein the data sampler operates within a data signal processing path having a decision threshold associated with a decision feedback equalization correction factor; measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler operating outside the data signal processing path to determine an eye center value of the spare sampler decision threshold; initializing a decision threshold of the backup sampler as a function of the eye center value and the DFE correction factor; after a preset data model is detected, generating corresponding phase error signal groups of the standby sampler and the data sampler; and updating the decision threshold of the data sampler according to the difference accumulation result of the phase error signals in the corresponding phase error signal group.

A digital receiver system samples a received signal in both amplitude and time to obtain sufficient information to enable accurate detection and decoding of transmitted data despite signal degradation effects produced by the communication medium. Signal amplification, frequency and time domain filtering, and fine adjustment of the sampling time and amplitude may be required to address issues with the specific characteristics of the communication medium. The method and system described herein allow for the setting and adjustment of the above-mentioned reception parameters during operation, thereby achieving accurate signal reception.

Drawings

Fig. 1 is a block diagram of a receiver according to some embodiments.

Fig. 2A and 2B show exemplary received signals in the form of eye diagrams.

FIG. 3 is a block diagram of one embodiment of a measurement system.

Fig. 4A is a block diagram of a data sampler and a backup sampler, according to some embodiments.

Fig. 4B and 4C are schematic diagrams of logic details of the control circuit of the system of fig. 3 according to some embodiments.

FIG. 5 is a schematic view of another embodiment.

Fig. 6 is a flow diagram of a method 600 according to some embodiments.

Detailed Description

In recent years, the signaling rate of high speed communication systems has reached several tens of gigabits per second, such that a single data unit interval is measured in picoseconds. An example of such a system is the Shokrollahi, which describes the transmission of traffic signaling codes over a very high bandwidth multi-wire data communication link disposed at an equal location between two integrated circuit devices within the system. Depending on the particular vector signaling code used, the number of channels making up a communication link may be as few as two and as many as eight or more, and one or more clock signals may be transmitted within a data channel or within an alternate communication channel.

In one embodiment employing vector signaling codes, the transmitter encodes a plurality of data bits into a vector signaling "codeword," i.e., a set of symbols to be transmitted substantially simultaneously over a plurality of lines or communication medium channels. Since each such line or channel may carry more than two possible values, each codeword symbol is taken from a set of signal value symbols. In the example given in [ Shokrollahi ], five data bits are encoded into a six symbol codeword using a four-valued symbol set and a ten-valued symbol set. The receiver determines the received codeword by detecting the multi-level line signal and then decodes it (e.g., by looking up a mapping table) into the received data.

In an alternative embodiment, it is noted that each vector signaling codeword is in the form of "subchannel" components superimposed on one another, each such subchannel being an orthogonal mode or modulation pattern of the corresponding line or channel. Thus, in the [ Shokrollahi ] example, the transmit signal may be the summation of five sub-channels, each modulated by one of five transmit data bits. Meanwhile, the vector signaling code receiver can directly detect the received line signal combination corresponding to the specific sub-channel. Such detection may be implemented, for example, by a multiple-input comparator (MIC) that performs a weighted summation operation on two or more line signals associated with orthogonal modes of the respective sub-channels and directly generates one bit of received data. In the [ Shokrollahi ] example, fully decoding five data bits requires the use of a set of MICs corresponding to a combination of two to six line signals. Since the codeword and the subchannel processing model are fully equivalent to each other, it can be said that interoperability can be achieved between the two regardless of the specific encoding and decoding models used. For example, in the case of using the same orthogonal vector signaling code, a codeword transmitter may be used in conjunction with a MIC-based subchannel receiver.

As described above, the line signals in an orthogonal vector signaling code system may take on a number of different values, while the subchannel detection results (e.g., the weighted sum results at the MIC outputs) are typically binary results. In this way, the reception processing functions such as ISI reduction and time offset correction can be performed on a simpler sub-channel signal rather than a more complex line signal, and are therefore more efficient.

In the practice of conventional high-speed integrated circuit receivers, the end point of each received signal channel, sub-channel, or line signal is a sampling device. The measurements performed by the sampling device impose constraints in both the time and amplitude dimensions. In an exemplary embodiment, the sampling device may be comprised of a sample and hold circuit that imposes constraints on the time interval to be measured and a threshold detector or digital comparator that determines downstream whether the signal within the interval is greater than or less than a reference value (or, in some embodiments, whether the signal is within upper and lower limits set by the reference value). Alternatively, it may be constituted by a digital comparator that determines the amplitude of the signal and a downstream clocked digital trigger that collects the determination at a selected point in time. In other embodiments, a combined time/amplitude sampling circuit that samples the amplitude state of the input signal in response to the occurrence of a clock transition may also be employed. In some embodiments, the acquisition time of the sample values may be adjustable within some or all of the samplers of the receiver. In some embodiments, the threshold level that is the subject of the sample value comparison may be adjustable within some or all of the samplers of the receiver.

Since the receiving means for obtaining the input measurement values impose constraints on the measurement of both time and amplitude, for the convenience of description, the receiving means is referred to herein as "sampling means" or, more simply, "sampler" without the use of the term "slicer" which is equally referred to in the art but is lacking in descriptive terms. In some embodiments, the acquisition time of the sample values may be adjustable within some or all of the samplers of the receiver. In some embodiments, the threshold level that is the subject of the sample value comparison may be adjustable within some or all of the samplers of the receiver. For example, the well-known receiver "eye pattern" is typically obtained by iteratively adjusting such parameters, with the result being a plot of signal amplitude versus time.

The embodiments described in the present application can also be applied to a permutation and combination ordering method in the prior art that is not covered by the vector processing method in Shokrollahi. More generally, these embodiments are applicable to any communication or storage method, particularly such methods that require a coordinated cooperation between multiple channels or channel elements to produce a coherent overall result.

Clock data recovery

Clock Data Recovery (CDR) or Clock Data Alignment (CDA) circuits described in Hormati et al extract time information from the data lines themselves or from a dedicated clock signal input and use the extracted information to generate a clock signal to control the time interval used by the received signal sampling means. The actual clock extraction operation may be implemented by well-known circuits such as Phase Locked Loops (PLLs) or Delay Locked Loops (DLLs). To support operation of the receiver, such circuitry may also generate higher frequency internal clocks, multiple clock phases, etc. during its operation. It has been found in the art that the practical differences between the reported CDR and CDA embodiments are not relevant herein, and therefore the application will be referred to hereinafter collectively as "CDA" but this is not meant to be limiting.

In a common CDA implementation, a first sampling time is used to optimally obtain data sample values, and a second sampling time is used to optimally determine whether the phase of the internal clock remains aligned with the input signal transitions, which may differ in time from the optimal data sampling time by up to 1/2 received signal Unit Intervals (UIs). Since in such embodiments two samples are taken per receive unit interval, a dual baud rate CDA scheme may be used. Such systems are extremely common in low speed communication systems or where the received signal has extremely sharp transitions (i.e., where a large time difference is observed between the time point of the signal transition and the optimal data sampling time point).

Equalization of received signals

At high data rates, even relatively short, high quality communication channels have considerable frequency dependent signal loss, and thus data receivers often have the capability to equalize the received signal. Continuous Time Linear Equalization (CTLE) is generally used to increase the high frequency gain of a received signal path to compensate for the phenomenon of increased attenuation of the channel at high frequencies.

In addition, compensating for signal propagation anomalies in the transmission medium by Decision Feedback Equalization (DFE) has also become a common practice for data communication receivers. The DFE system performs nonlinear time domain equalization on the received signal by: maintaining, by the receiver, a history of previously received data values; the historical data values are then processed by the transmission line model to infer the expected effect of each historical data value on the currently received signal. The transmission line model may be pre-calculated, obtained by measurement, generated by trial and error, and may cover the effects of one or more previous data intervals.

In a typical receiver design, a correction signal that more accurately represents the received data values is generated by subtracting the DFE compensation calculation from the current received signal input. Those skilled in the art will appreciate that the DFE compensation value generated as described above can only be calculated after the completion of the detection of the data value in the previous unit interval. Thus, as the data rate increases, the information used to generate the DFE compensation values always becomes at some point inapplicable to the samples of the next unit interval in time. In fact, at the highest data rates currently in practical use, this condition may occur for several unit intervals before, since the detection time of a single data value may correspond to the length of the unit interval, so that the receiver has to pipeline or parallelize its detection operations. Thus, in various embodiments, this "closed loop" DFE approach is often discarded for one or more recent unit intervals, and instead one or more elements that generate the DFE compensation values in an "open loop" ("loop unrolling") manner are employed.

To speed up such DFE operations, some embodiments speculatively generate a corresponding DFE compensation value for each possible detected data value for a given unit interval. One embodiment includes a plurality of data detection samplers, each provided with a uniquely associated DFE compensation value for one or more previous unit intervals of possible detected data values. The result of each sampler is not saved until the corresponding previous data value becomes known, and at the same time the corresponding saved result is selected for data detection.

A set of DFE compensation values, speculatively generated representing a series of possible detected data results for one or more previous transmission unit intervals, represents a set of measured levels spanning some portion of the received signal amplitude range. For example, when "0" signals are successively transmitted in the past, the predicted threshold level set for the subsequent receiver data measurement compensated with the speculative DFE may be made low; whereas when the "1" signals were previously transmitted successively, the predicted threshold level set for the same data measurement may be made high. Thus, for any data measurement used to detect an actual data value, the multi-sampler receiver may perform the measurement operation in a manner such that the threshold value may be too high or too low for the actual signal for the corresponding interval.

Combined use of CDA and DFE

In high speed communication systems where the channel has a large frequency dependent attenuation effect, the received signal tends to have significant ramp up and ramp down times even after equalization. Thus, a signal sampler that is set in time to trigger at the "eye center" may still cross a signal transition from one data value to a subsequent data value in some cases (especially when the corresponding received signal is more heavily affected by ISI). An example of this is shown in FIG. 2B, where the sample points marked "+" intersect the signal trace labeled [0, 1, 1 ]. For this case, the receiver employing the DFE adjusts its effective data sampling threshold further away from the signal trajectory, thereby compensating for this phenomenon, while a fixed threshold sampler (i.e., a sampler without the beneficial effects of the DFE) may obtain a false detection value.

In the above case, the data value and the clock phase may be simultaneously determined by using a single sampling time within each reception unit interval. This single baud rate CDA implementation is based on the following observations: in some cases, the interaction between the received ISI and the detection sampling threshold results in data sampling characteristics that are less desirable, i.e., they have a greater probability of crossing the input signal that rises and falls more slowly and is undergoing such changes. Thus, by limiting the receive equalization to impose constraints on transition speed and then limiting the clock timing concerns to only the sampling thresholds and received data patterns that exhibit the crossover described above (associated with a particular ISI level), simultaneous sampling of clock and data at a single sampling time can be achieved.

At least one embodiment in Hormati utilizes the above to perform the measurement operation with a plurality of samplers or comparators that employ speculative DFE operation. In such embodiments, the stored speculative result is not used to determine the received data value (i.e., a measurement that has a signal skew up or down relative to the corresponding DFE correction value for the corresponding interval but has a measurement time that is the same as the valid data sample value), but rather may provide information related to clock recovery.

Operational considerations

It is envisaged that careful configuration of the parameters of the receiving system and constant monitoring and adjustment during operation may be required to enable the received signal quality, degree of equalisation, sampling level and clock timing to interact with each other in the advantageous manner described above. Furthermore, since each such parameter has an impact on the received signal quality, adjustments in the receiver operation process must be made in a manner that does not affect normal signal reception operation, both in terms of information measurement and configuration or variations in operating parameters.

For simplicity of description, the term "interference" will be used hereinafter to adj ust system measurements, adjustments or configuration changes that are made during normal operation of the receiver that would or may not interfere with that normal operation or that would or may not cause incorrect detection of data values from a given signal source. Conversely, the term "non-interfering" is intended to describe system measurements, adjustments, or configuration changes that can be made during normal operation that do not interfere with the normal reception of correct data values from the signal source.

Obtaining accurate or real-time data of critical parameters, such as received signal level, may require a significant penalty in terms of both circuit complexity and power consumption. In a practical implementation, the signal sampler does not record the actual amplitude of its input signal, but only information indicative of whether the input value was above or below its threshold level at its trigger. Thus, some embodiments may use a backup sampler operating outside of the data signal processing path to measure the "upper edge" or "lower edge" of the signal eye as follows: the decision threshold is incrementally adjusted up or down until the data produced at its output becomes invalid compared to the data received by the data sampler operating in the data signal processing path. Similarly, the backup sampler may also measure the "eye-open width" of the eye pattern by: incrementally advancing and downshifting a variable phase measurement clock (e.g., of a phase interpolator) and simultaneously comparing the sampling results with data decision results produced by the data sampler.

In addition, another factor that complicates the above operational considerations is that the large number of components within the system implementation, particularly the sampling device itself, mostly employ analog processing functions that present stability issues. Since device parameters such as transistor gain and threshold voltage vary with temperature and supply voltage, the threshold level of the sampling device and its measured signal amplitude and baseline voltage may drift, requiring periodic calibration and adjustment. Furthermore, process variations within an integrated circuit device may cause different samplers to have different characteristics, in particular different threshold voltages, and thus different sampling thresholds. Thus, some solutions employ redundant analog processing elements, wherein one set of elements is switched out of the data signal processing path for calibration or adjustment, while another set of elements is switched into the data signal processing path for active processing of the received data. The use of a large number of identical high power consuming circuits such as analog amplifiers, comparators, and samplers is a non-interfering practice, and while this may increase system power consumption to an unacceptable level, replacing the samplers and redirecting the DFE feedback may be a real-world problem. Thus, the above solution enables to avoid replacing the sampler.

Illustrative embodiments

For non-limiting illustration purposes, the serial data receiver shown in fig. 1 is taken as an example below. The exemplary receiver includes: at least one speculative DFE processing stage 150 supported by two comparators 120 that simultaneously perform time-sampling operations at two different magnitude thresholds; and a receiver clock system 170 for generating a sampling clock, the phase of which can be adjusted by CDR correction to achieve data sampling time optimization. As shown, the data sampler includes two comparators 120 that generate comparator outputs that slice the data signals received from the CTLE 110 and sample the comparator outputs according to the sampling clock. Prior to sampling, the CTLE 110 may amplify and/or equalize the received data signal.

In some embodiments, an apparatus includes two comparators 120 for generating two comparator outputs for comparing a received data signal with a first threshold by comparator 121 and a second threshold by comparator 122 based on a sampling clock, the first and second thresholds being determined by an inter-symbol interference estimate for a multi-wire bus. In some embodiments, the first and second thresholds correspond to correlation Decision Feedback Equalization (DFE) correction values of a previous unit interval. The apparatus may further comprise a data decision selection circuit 130 for selecting one of the two comparator outputs as a data decision result based on at least one previous data decision result, which may be stored in a data value history memory 140. Multiplexer 130 may select the output of comparator 121 as the data decision result if the previously received bit stored in data history memory 140 is a logic "1". If the previously received bit is a logic "0," the output of comparator 122 is selected as the data decision result. The apparatus further includes a phase error indication selection circuit 160 that selects the other of the two comparator outputs as the phase error signal in response to receiving the CDR select signal from the pattern detection circuit 155, the pattern detection circuit 155 being operable to identify a preset data decision pattern in the data value history memory 140.

In some embodiments, one or more backup samplers 180 provide additional measurement functionality. Although fig. 1 includes two spare samplers 180, it should be noted that such embodiments are not intended to be limiting, as the number of spare samplers 180 may be either single or more than two. In one such embodiment, one or more backup samplers 180 may be used to acquire eye range data using amplitude thresholds and/or variable phase adjustments. Such a variable phase adjustment clock may be set to move left and right around the sampling interval or may correspond to the sampling clock provided to the comparator 120 corresponding to the eye center sampling instant. Wherein a clock selection circuit (not shown) may be utilized to select between these two clocks. In another mode, the measurement controller 190 may set one or more of the back-up samplers 180 to perform non-interfering measurements of received signal amplitude or transition time.

In some embodiments, the apparatus further comprises a receiver clock system 170 for receiving the phase error signal and responsively phase adjusting the sampling clock. In some embodiments, the phase error indication is a too early/too late logic decision made on a transition of the received signal. In some embodiments, data decision selection circuit 130 and phase error indication circuit 160 select different comparator outputs. In some embodiments, the apparatus further includes a Decision Feedback Equalization (DFE) circuit 150 for generating the first and second thresholds.

In some embodiments, the apparatus further includes a subchannel detection multiple-input comparator (MIC, not shown) that operates on signals received via the plurality of lines, the subchannel detection MIC being used to generate the received data input signal. In such embodiments, the signals received over the plurality of lines correspond to codeword symbols of a vector signaling code, the codeword corresponding to a weighted summation of a plurality of sub-channel vectors, each sub-channel vector being mutually orthogonal. In one such embodiment, the inter-symbol interference is specific to a particular subchannel, and the subchannel-specific ISI corresponds to modulation of the components of the corresponding subchannel vector associated with the received signal. In some embodiments, the associated subchannel-specific ISI of each subchannel vector are mutually orthogonal to each other. In some embodiments, the apparatus may further comprise a filter for filtering the received signal before the comparator output is generated.

In some embodiments, at least the data sampler 120 of fig. 1 is a plurality 180, where one set of data samplers measure and operate in an interfering manner on a received signal, while another set of data samplers measure and operate in a non-interfering manner on the same received signal. In other embodiments, at least one of a comparator threshold, a sampling time, a CTLE gain, and a CTLE equalization are adjusted for at least one set of data samplers during the measuring and/or adjusting.

Fig. 2A shows an example of a typical received signal "eye diagram", which is an effect of a plurality of unit intervals of a received signal superimposed on a time axis. Thus, each individual curve making up the eye diagram represents a received signal "trace" transitioning from one unit interval to a subsequent unit interval.

In fig. 2B, the partial signal traces that make up the eye diagram of fig. 2A are labeled as triplets of [ previous data value, current data value, next data value ]. It is envisioned that the [1, 1, 1] traces converge at the top of the figure, the [1, 1, 0] traces transition from top to bottom at the right of the figure, the [0, 0, 1] traces transition from bottom to top at the right of the figure, and so on. The corresponding trace for the previous received data value of "1" (best suited for detection at the top DFE sample position) is shown by the thick black line, while the corresponding trace for the previous received data value of "0" (best suited for detection at the bottom DFE sample position) is shown by the light line.

If the previous data value is "1," the DFE system selects the top DFE sample position for detection of the current data value, which is marked with the black symbol "+" and marked with "+ VH 1". It may be noted that this sample position is located exactly in the center of the upper half of the data eye, but directly above the [0, 1, 1] received signal trace (the current data value of this trace is detected by the bottom sample position since the previous data value was "0"). Thus, a sampler with a decision threshold set to "-VH 1" (indicated by a white "+" sign) effectively corresponds to an edge sample that can be used by the CDR system to determine whether the sampler time is too early or too late with respect to a signal transition. Since the CDR will adjust the phase of the sampling clock to a ratio of early/late indications selected from the data sampler in response to the transition data pattern of approximately 1:1, by using the sampler output as the early/late indication, the sampling clock will be made to have a lock point associated with the DFE correction factor ± VH 1.

In some embodiments, the signals shown in fig. 2A and 2B correspond to the output of a subchannel detection multiple-input comparator (MIC) for receiving a plurality of symbols of a codeword and for detecting a modulation scheme of a corresponding subchannel vector among a plurality of subchannel vectors, wherein each subchannel vector is orthogonal to each other. The codeword corresponds to a summation of the plurality of sub-channel vectors, each sub-channel vector being separately modulatable by a respective information bit received by an encoder that generated the codeword. In such embodiments, the ISI shown in fig. 2A and 2B is specific to a particular subchannel, and thus ISI in the corresponding subchannel vector modulation is seen only at the corresponding MIC. Therefore, in a system with at least two sub-channel detection MICs, ISI due to the modulation of a first sub-channel vector is not seen in a second MIC that detects a second sub-channel vector, because the first and second sub-channel vectors are orthogonal to each other.

Adjustment of decision threshold for dynamic data sampler

As described above, reliable error-free detection of a received data signal may involve accurately adjusting the data sampler threshold to a predetermined time and a predetermined amplitude position in the received signal "eye pattern". Drift in the relationship between the pre-set sampler and the signal over time, temperature or supply voltage may lead to increased reception error rates and may eventually lead to detection failures.

In one known solution, the spare sampler is calibrated and adjusted offline (i.e., non-interfering), and then the preconfigured sampler is swapped with the currently active data sampler to take the latter down for calibration and adjustment. In such systems, switching circuitry must be provided for all signals entering, controlling, and output by the sampler to enable these signals to be directed to the data path or calibration function as required.

To avoid using the switching circuit described above, one embodiment takes measurements with a backup sampler operating outside of the data signal processing path and then uses information obtained from such measurements to adjust the operation of the data sampler operating within the data signal processing path in a non-interfering manner. Since the spare sampler is not part of the data signal processing path in the active state, the threshold level and the clock time can be adjusted without affecting the received data, thereby enabling identification of the two extreme points of normal operation (i.e., the boundaries of the open eye region of the receive "eye pattern"). In some embodiments, the spare sampler may be referred to as an eye sampler in the following of this application, but this is not meant to be limiting, as such adjustments to the spare sampler are similar to adjustments made to obtain the statistics required in drawing an eye.

Since both the DFE compensation and the data sampling threshold depend on the total eye height, the above measurement process may include: the vertical eye opening of the data signal is determined by the normal data sample clock timing and an initial eye center threshold value close to the eye center value. In one specific example, the determination threshold of the spare sampler is gradually increased until the difference between the determination result of the spare sampler and the data determination result reported by the data sampler reaches a certain preset ratio to determine the top edge of the open-eye region, and then the determination threshold of the spare sampler is gradually decreased until the difference between the determination result of the spare sampler and the data determination result reported by the data sampler reaches the preset ratio again to determine the bottom edge of the open-eye region. For example, the decision threshold of the spare sampler may be stepped up until the spare sampler reports one "1" output for every eight "0" outputs, resulting in a ratio of 1: 8. This ratio can be selected as a degree of accuracy. The eye center value corresponds to a midpoint between the determined eye top edge value and the eye bottom edge value. In some embodiments where ISI is significant (i.e., a large amount of DFE compensation is required to achieve proper data reception), the comparison between the eye data results and the data sampler results is performed only for specific received data patterns such as [1, 1, 1] and [0, 0, 0] by logic gate control when determining the top and bottom edges of the eye. In some embodiments, a preset threshold in the form of a ratio of positive and negative data comparisons (i.e., a ratiometric limit) is used in determining whether the results match.

While some embodiments may perform the collection and analysis of results in real time, other embodiments may simply record the sampler results and then perform the results comparison, pattern matching, etc. on the historical or stored information. In embodiments of the present application, such operations may be performed by boolean logic, a finite state machine, or software executing within an embedded processor or management system.

It should be noted that this measurement method does not produce calibration results such as absolute values of eye opening in millivolts. Rather, the units of the resulting results are typically associated with internal threshold adjustment means, such as values input to a digital-to-analog converter (DAC) used to generate the analog sampler threshold levels (in arbitrary internal units). Since the DAC is not calibrated on an external reference and the sampler knows that there is an offset (which is variable and unmeasured), the resulting eye opening value is a relative value and not an absolute value. That is, since the measurement results do not contain sampler offset, the set values obtained from these results will produce consistently consistent results when used to set the threshold values for the same eye sampler.

After determining the vertical eye opening, the determination of the appropriate sampling threshold level for data detection may continue. In a receiver system employing one DFE speculation processing stage as shown in the example of fig. 1, the threshold level will contain the H1 term of the DFE compensation amount corresponding to the associated signal level of the measured eye opening.

After initializing the vertical decision threshold level of the spare sampler based on the previously determined eye center value and the associated DFE correction factor of the selected data sampler for calibration purposes, an optimal improvement in data reception is indicated if the resulting decision for the vertical decision threshold level is different from the decision for the data sampler. In at least one embodiment, since the vertical decision threshold is calibrated based on the true eye center value, the decision result of the spare sampler is considered as the correct result, and the decision threshold currently used by the data sampler is incrementally modified based on the decision result obtained from the spare sampler, e.g., the decision threshold of the data sampler is incrementally increased or decreased by a unit increment. Due to the above-mentioned circuit characteristic differences (e.g., intra-sampler offsets), the parameters determined for the spare sampler are generally not directly applicable to the data sampler. Thus, when the difference is significant, an intermediate conversion step may be implemented.

In one embodiment, a second measurement process is also carried out, in which the decision threshold of the spare sampler is incrementally adjusted away from the initial decision threshold until the decision result generated under the spare sampler characteristic matches the decision result of the data sampler. In one embodiment, the process analyzes the decision corresponding to the phase error signal determined by the mode detection circuit. The amount of progressive adjustment used in adjusting the decision threshold of the spare sampler from the initialized decision threshold to a decision threshold that is close to the data sampler characteristic corresponds to the amount of offset correction representing the amount of progressive adjustment that can produce better data detection results by being applied back to the decision threshold currently set by the data sampler. The application of the adjustment amount may be accomplished in a variety of ways including: applying the entire differential adjustment; applying a portion of the differential adjustment; progressively applying portions of the adjustment amount at different time points; after a single application of incremental adjustment, the entire measurement procedure is repeated.

As described above, the determination of the sampler characteristic difference may comprise a more sensitive process than simply comparing the sample value outputs. Referring again to fig. 2B, it will be readily appreciated that a large change in time and/or threshold may occur near the sample point at the horizontal position of the "+" symbol before a significant change occurs in the detection of any data bit "1", which is basically a prerequisite for the selection of the "center of eye" sample point, and which maximizes the probability of obtaining correct data detection results under a variety of different conditions. However, for edge sampling (e.g., the vertical position of the sample point at black "+" on the [0, 1, 1] signal trace in fig. 2B), even a very small change in the vertical decision threshold will produce very different results.

In one embodiment, a comparison is monitored between edge sample values obtained by a spare sampler and edge sample values obtained by a data sampler at the same unit sample time, wherein the edge sample values are logically gated or selected by pattern matching sequentially received data values, and wherein the edge sample value results are obtained by a loop-expanded DFE sampler not used to provide the data values during the corresponding unit interval, as described above. In one embodiment, the difference (e.g., determined by an exclusive-or operation) of the edge sample values represents a different sampler characteristic, wherein the positive and negative of the edge sample values and the value of the logic gate control data pattern represent the direction of the difference. Fig. 3 is a block diagram of one such embodiment. Fig. 3 includes a data sampler 301 and a spare sampler 302, both operating on a data signal Vin. In some embodiments, the data signal may be a test mode, while alternative embodiments may operate on an informational data signal. Data sampler 301 operates according to a vertical decision threshold + Vh1, which may correspond to the application of DFE correction factors associated with information received during a previous signaling interval. Spare sampler 302 operates according to the same DFE correction factor + Vh1 and is shown in fig. 3 as operating according to a control signal received from accumulator 304. As described above, the decision threshold of backup sampler 302 may be initialized based on the determined eye center value and the DFE correction factor associated with the selected data sampler to be calibrated, which in the embodiment shown in fig. 3 is the + H1 data sampler. Control logic 303, which may be part of measurement controller 190, for example, receives Edge sample values Edge _ DS and Edge _ SS from Data Sampler (DS) and Spare Sampler (SS), which may correspond to phase error signals determined for mode detection. As shown in fig. 2B, the decision threshold, indicated by the black "+" symbol marked for + VH1, is located approximately on the 0, 1, 1 track and therefore can be used as a phase error signal. The pattern detection may be performed by data bits Dn-1, Dn +1 received from the data history memory 306, and the three bits may correspond to the data determination result of the unit interval before the edge determination, the data determination result of the same unit interval as the edge determination, and the data determination result of the unit interval after the edge determination, respectively. After the data pattern is detected, the edge samples may be compared to each other and used to adjust a decision threshold for one of the data sampler and the backup sampler. An accumulator 304, which may be part of measurement controller 190, for example, may accumulate the differences in the phase error signals and ultimately be used to update the data decision thresholds of the data samplers operating within the data signal processing path.

Fig. 4A, 4B, and 4C illustrate further details of control logic 303 according to some embodiments. As shown, fig. 4A includes a + H1 data sampler 401 and a spare sampler 402. For the sake of brevity, the-H1 data sampler is not shown in FIG. 4A. The + H1 data sampler 401 generates a phase error signal Edge _ DS that the data signal compares to the + VH1 DFE correction factor in the [0, 1, 1] data pattern, while the spare sampler 402 generates a phase error signal Edge _ SS that the data signal compares to a spare sampler decision threshold Vt initialized with the previously determined vertical eye center value and the + VH1 DFE correction factor. FIGS. 4B and 4C illustrate logic for obtaining control signals for registers associated with the + H1 data sampler and the-H1 data sampler, respectively. As shown, [ previous data, current data, next data ] data model [0, 1, 1] is used to identify edges suitable for evaluating the + H1 sampler alignment condition determined by logic gate 404 in FIG. 4B, and data model [1, 0, 0] is used to identify edges suitable for evaluating the-H1 sampler alignment condition determined by logic gate 403 in FIG. 4B.

As described above, the logical AND gate 404 in FIG. 4B is used to identify the data model [0, 1, 1] from [ Dn-1, Dn, Dn +1 ]. The logical exclusive or gate 405 is used to verify that the phase error signals Edge _ DS and Edge _ SS provided by the data sampler and the spare sampler, respectively, are different from each other. The logic and gate 406 generates a count enable signal c _ en _ + H1 for the register 501 associated with the + H1 data sampler, and the logic and gate 408 converts the phase error signal Edge _ DS of the data sampler into the early/late indication signal + H1CU/CD according to the truth table given in table I below.

The logic gates 403, 415, 416, 407 function similarly to the logic gates 404, 405, 406, 408, respectively, described above to generate the count enable signal c _ en _ H1 and the early/late indication signal-H1 CU/CD for the-H1 data sampler. Table I below is a truth table of the relationship between the detected data model and the data sampler decision and the early/late difference:

TABLE I

Fig. 5 shows one embodiment of accumulator 304. As shown, registers 501 and 511 accumulate differential indications of the + VH1 and-VH 1 data samplers, respectively, with their high order bits controlling the DACs contributing to the spare sampler 504 thresholds. The + H1 comparator threshold for generating a phase error signal by comparison between the decision result of the data signal and the + H1 DFE correction factor in the [0, 1, 1] data model is controlled by counter 501, DAC 502, and summing node 503, while the-H1 comparator threshold for generating a phase error signal by comparison between the decision result of the data signal and the-H1 DFE correction factor in the [1, 0, 0] data model is controlled by counter 511, DAC 512, and summing node 513. The control signal H1CU/CD and the count enable signal C _ en _ H1 for controlling the count up/down direction are provided by the logic of FIGS. 4B and 4C, for example. After performing the edge sample value comparison under logic gate control for a sufficient number of times and completing the update of the Least Significant Bit (LSB), the set of high order bits corresponding to the counter Most Significant Bit (MSB) will cause the spare sampler threshold to change through the DAC until the spare sampler and data sampler characteristics indicated by the CDA edge decision result become identical. The number of bits in the counter downstream of the set of MSBs used to control the DAC determines the amount of processing to average or smooth the multiple sampling events between the progressive threshold changes applied to the spare sampler, or in some embodiments directly to the data sampler. Subsequently, a logic circuit, finite state machine or processor executing a pre-stored program may review the DAC settings of the standby sampler by comparison with the DAC settings found to enable detection optimization and apply a corresponding amount of difference, in part or in whole, to the DAC settings of the current data sampler. As further shown in fig. 5, the apparatus may include some steering logic to select between the calibration of the + H1 data sampler and the calibration of the-H1 data sampler, e.g., via multiplexer 515, upon receipt of a select signal.

Although the fig. 5 embodiment is shown with the MSB portion of the register controlling the decision threshold of the spare sampler, in an alternative embodiment, the decision threshold of the data sampler is directly controlled during the difference accumulation process with the equivalent accumulator based on a logic sum counter as shown in fig. 3, 4A-4C and 5, rather than the decision threshold of the spare sampler.

Although the above embodiments describe updating the decision threshold of the spare sampler and/or the data sampler after obtaining the comparison of the phase error signal, it should be noted that in at least one alternative embodiment, the sampling instants of the spare sampler are updated, for example, with a sampling clock provided by the phase interpolator. Such embodiments may enable horizontal adjustment of the spare sampler based on the comparison of the phase error signals. As shown in FIG. 2B, the + VH1 decision threshold, denoted by the "+" sign, of the data sampler is on the trace of the [0, 1, 1] data pattern under lock-in conditions. Two possible scenarios where the spare sampler decision threshold and the data sampler decision threshold differ from each other are indicated by the "x" symbol in fig. 2B. When the + H1 data sampler is calibrated, if the decision threshold of the spare sampler is smaller than the decision threshold of the + H1 data sampler, the sampling time of the spare sampler can be shifted forward until the phase error signal starts to become matched; conversely, if the decision threshold of the spare sampler is greater than the decision threshold of the + H1 data sampler, the sample time of the spare sampler is moved in the opposite direction to that described above. In this manner, the direction of the sample time stealing can be used to determine the direction of adjustment of the + H1 data sampler decision threshold. the-H1 data sampler can be adjusted by similar techniques.

In some embodiments, as described above, the spare sampler 504 is used to determine the amount of offset correction to apply to the + VH1 and-VH 1 data samplers, respectively, before updating the vertical decision thresholds of the data samplers. After the offset correction amounts of the two data samplers are determined, the vertical decision threshold values of the two data samplers are updated at the same time. In some embodiments, when the vertical decision threshold values of both the data samplers are updated, the entire update may be performed in a direction determined only by the offset correction amount. In an alternative embodiment, the incrementing of the data sampler vertical decision threshold may also be done incrementally in accordance with the direction and magnitude of the offset correction amount determination.

Fig. 6 is a flow diagram of a method 600 according to some embodiments. As shown, the method 600 includes: the data signal is sampled 602 with a data sampler operating within a data signal processing path having a decision threshold associated with a DFE correction factor. The method 600 further comprises: the eye opening of the data signal is measured 604 by adjusting a decision threshold of a backup sampler operating outside the data signal processing path to determine an eye center value of the decision threshold of the backup sampler. After initializing 606 the decision threshold of the spare sampler in accordance with the eye center value and the DFE correction factor, a corresponding set of phase error signals for the spare sampler and the data sampler is generated 608 in response to the detected preset data pattern. The decision threshold of the data sampler is then updated 610 according to the difference accumulation of the phase error signals in the respective set of phase error signals.

Some embodiments may include: updating the decision threshold of the spare sampler in response to a difference accumulation of the phase error signals in the corresponding set of phase error signals before updating the data sampler. In some such embodiments, the decision threshold of the spare sampler is updated until the difference accumulation results of the phase error signals in the corresponding set of phase error signals are within a threshold, e.g., the number of differences in the phase error signals is less than a preset percentage threshold for a given number of phase error signals. In some embodiments, the method 600 further comprises: an offset correction value between an updated decision threshold of the spare sampler and an initialized decision threshold of the spare sampler is determined. Subsequently, the decision threshold of the data sampler is updated with the determined offset correction value. In some embodiments, the data sampler decision threshold is updated only as a whole in the direction determined by the offset correction amount. In an alternative embodiment, the updating of the data sampler vertical decision threshold may also be done according to the direction and magnitude determined by the offset correction amount.

In some embodiments, the method 600 includes: determining a difference accumulation result between the respective sets of phase error signals by comparing the respective sets of phase error signals. Such comparison may be accomplished, for example, by a logical exclusive or gate.

The differences between the respective sets of phase error signals may be accumulated by a register. In some such embodiments, the LSBs of the register are updated each time a difference between a pair of phase error signals, each phase error signal of the pair of phase error signals being selected from one of the respective sets of phase error signals, is determined. The update direction of the LSB may be determined by one phase error signal among the pair of phase error signals. In some embodiments, the decision threshold of the spare sampler is controlled by a set of high order MSBs of the register such that a change in the set of MSBs caused by an imbalance between two parties acting in opposite directions in the LSBs causes the decision threshold of the spare sampler to change over time. In the register, a number of demarcation bits between the set of MSBs and LSBs may be used as a low pass filter, such that the MSBs gradually change as the LSBs are updated more and more. Although some embodiments control the decision threshold of the spare sampler by the set of MSBs, it is noted that in alternative embodiments, the decision threshold of the data sampler may also be directly controlled by the set of Most Significant Bits (MSBs) of the register.

In some embodiments, each phase error signal of each set of phase error signals among the respective set of phase error signals is determined according to (i) a sampled value of the data signal and (ii) the preset data model.

In some embodiments, a method includes correcting an eye range offset globally subversively by: firstly, measuring the top edge and the bottom edge of a data signal eye diagram by using a standby sampler; the decision threshold of the spare sampler is then centered vertically. The spare sampler decision threshold for each subchannel may then be set equal to the + H1 value currently selected for use by that subchannel. The decision threshold of the spare sampler is aligned with the corresponding data sampler at each phase by balancing the early/late indication votes of the two samplers for each sub-channel. The early/late indication votes may be balanced by an accumulation circuit, such as a count register. Once an equilibrium is reached between the early/late indication votes, the decision threshold of the data sampler is equal to the decision threshold of the back-up sampler, enabling the decision threshold of the data sampler to be calculated by subtracting the + H1 value from the offset of the back-up sampler.

The decision threshold of each sub-channel spare sampler is then set equal to the-H1 value currently selected for that sub-channel, and the decision threshold of the spare sampler is aligned with the corresponding data sampler at each phase by again balancing the early/late indication votes of the two samplers for each sub-channel to determine a corrected offset of the spare sampler with respect to the data sampler. After the correction offsets for both the + H1 and-H1 data samplers are known, the decision thresholds for both data samplers are updated at the same time. In some embodiments, the update may be done only in the direction determined by the correction offset, but alternative embodiments may include an amplitude component in the update.