Blind adaptation of a decision feedback equalizer for use in a 10Gbps serial link

The Decision Feedback Equalizer (DFE) is commonly used in recovering data at the receive end of a multi-gigabit per second rate serial backplane channel. This form of equalization scheme typically relies on a training sequence. However this training sequence may not be shared by all implementations of the DFE, thus causing significant compatibility issues. A model of a DFE and adaptation algorithm was developed in MATLAB in combination with measurement data of a typical backplane channel. A method was developed to process the incoming signals from the channel involving a known adaptation algorithm in order to determine appropriate tap coefficients. These coefficients are shown to be effective for a data rate of 10 Gbps. The design presented would yield itself to standard-cell implementation except for the analog DFE structure and several stages of the up/down counters. Furthermore the design was evaluated using representative error sources and across several design parameters.