Methods and systems are described for sampling a data signal using a data sampler operating in a data signal processing path having a decision threshold associated with a decision feedback equalization (DFE) correction factor, measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler operating outside of the data signal processing path to determine a center-of-eye value for the decision threshold of the spare sampler, initializing the decision threshold of the spare sampler based on the center-of-eye value and the DFE correction factor, generating respective sets of phase-error signals for the spare sampler and the data sampler responsive to a detection of a predetermined data pattern, and updating the decision threshold of the data sampler based on an accumulation of differences in phase-error signals of the respective sets of phase-error signals.

An apparatus includes an FFE circuit, including a clock generator creating multiple sub-rate phases of an input clock, and a multi-phase sampler responsive to a data signal and to the multiple sub-rate phases generated by the clock generator. The sampler is configured to sample the data signal and to generate held sample outputs corresponding to the multiple sub-rate phases. A SC equalization circuit in the FFE circuit has two states and is responsive to inputs from the multi-phase sampler output and the clock generator. The SC equalization circuit is configured to form outputs using the two states. A variable gain output stage in the FFE circuit is responsive to the outputs from the SC equalization circuit and is responsive to gain control signal(s) to provide variable gains to corresponding outputs of the SC equalization circuit to form equalized outputs based on the data signal.

An apparatus includes an FFE circuit, including a clock generator creating multiple sub-rate phases of an input clock, and a multi-phase sampler responsive to a data signal and to the multiple sub-rate phases generated by the clock generator. The sampler is configured to sample the data signal and to generate held sample outputs corresponding to the multiple sub-rate phases. A SC equalization circuit in the FFE circuit has two states and is responsive to inputs from the multi-phase sampler output and the clock generator. The SC equalization circuit is configured to form outputs using the two states. A variable gain output stage in the FFE circuit is responsive to the outputs from the SC equalization circuit and is responsive to gain control signal(s) to provide variable gains to corresponding outputs of the SC equalization circuit to form equalized outputs based on the data signal.

Decision feedback equalization (DFE) is used to help reduce inter-symbol interference (ISI) from a data signal received via a band-limited (or otherwise non-ideal) channel. A first PAM-4 DFE architecture has low latency from the output of the samplers to the application of the first DFE tap feedback to the input signal. This is accomplished by not decoding the sampler outputs in order to generate the feedback signal for the first DFE tap. Rather, weighted versions of the raw sampler outputs are applied directly to the input signal without further analog or digital processing. Additional PAM-4 DFE architectures use the current symbol in addition to previous symbol(s) to determine the DFE feedback signal. Another architecture transmits PAM-4 signaling using non-uniform pre-emphasis. The non-uniform pre-emphasis allows a speculative DFE receiver to resolve the transmitted PAM-4 signals with fewer comparators/samplers.

Methods and systems are described for sampling a data signal using a data sampler operating in a data signal processing path having a decision threshold associated with a decision feedback equalization (DFE) correction factor, measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler operating outside of the data signal processing path to determine a center-of-eye value for the decision threshold of the spare sampler, initializing the decision threshold of the spare sampler based on the center-of-eye value and the DFE correction factor, generating respective sets of phase-error signals for the spare sampler and the data sampler responsive to a detection of a predetermined data pattern, and updating the decision threshold of the data sampler based on an accumulation of differences in phase-error signals of the respective sets of phase-error signals.

Circuits, methods and systems that are to be used for dynamically modulating a high frequency bit duration of data based on a status of one or more previous transmitted bits. One circuit comprises a first data path comprising a first input, a first buffer, and a first output connected to a first multiplexer data input. The circuit further comprises a second data path comprising a second input, a second buffer, a phase interpolator, and a second output coupled to a second multiplexer data input. The circuit further comprises a multiplexer having at least two data inputs, at least one control input, and a common output coupled to a transmitter signal line and wherein the at least one control input is operatively coupled to a generated control signal that is based on the status of one or more previous transmitted bits.

Methods and systems are described for sampling a data signal using a data sampler operating in a data signal processing path having a decision threshold associated with a decision feedback equalization (DFE) correction factor, measuring an eye opening of the data signal by adjusting a decision threshold of a spare sampler operating outside of the data signal processing path to determine a center-of-eye value for the decision threshold of the spare sampler, initializing the decision threshold of the spare sampler based on the center-of-eye value and the DFE correction factor, generating respective sets of phase-error signals for the spare sampler and the data sampler responsive to a detection of a predetermined data pattern, and updating the decision threshold of the data sampler based on an accumulation of differences in phase-error signals of the respective sets of phase-error signals.

An information handling system includes a processing unit that is coupled to a memory device by a communication channel. The processing unit includes a memory controller and is configured to host a basic input output system (BIOS). The memory device, which may include a dual in-line memory module (DIMM), stores serial presence detect (SPD) information. In an embodiment, the BIOS obtains the SPD information and parameters of the communication channel, such as channel impedance and channel length. In this embodiment, the BIOS uses a look-up table to determine an equalization of the communication channel based on the obtained SPD information and the obtained parameters of the communication channel, and utilizes the memory controller to set the equalization of the communication channel, such as by setting or controlling settings of transmission and reception components of the memory controller.

According to an embodiment, a control circuit of an equalizer configured to set a first amount to a linear equalizer, determine a second amount optimizing a non-linear equalizer with respect to a first signal generated by the linear equalizer to which the first amount is set, set the second amount to the non-linear equalizer, update an amount from the first amount to a third amount smaller than the first amount based on a magnitude of the first amount, set the third amount to the linear equalizer, determine a fourth amount optimizing the non-linear equalizer with respect to a second signal generated by the linear equalizer to which the third amount is set, and update an amount from the second amount to the fourth amount.

An equalizer includes a first feed-forward stage that provides a measure of low-frequency ISI and a second feed-forward stage that includes a cascade of stages each making an ISI estimate. The ISI estimate from each stage is further equalized by application of the measures of low-frequency ISI from the first feed-forward stage and fed to the next in the cascade of stages. The ISI estimates from the stages thus become progressively more accurate. The number of stages applied to a given signal can be optimized to achieve a suitably low bit-error rate. Power is saved by disabling stages which are not required to meet that goal.