Chapter 13 - Decision feedback channels

Abstract

The main idea of decision feedback is to cancel inter-symbol interference (ISI) using a feedback circuit. This approach is different from partial response channels, where the input readback pulse is equalized to a target waveform. As the channel density exceeds 1.5, the ISI becomes a critical factor―the readback voltage pulses from adjacent transitions overlap and the transitions can no longer be detected reliably with peak detection channels. The partial-response maximum likelihood (PRML) channels cancel ISI by equalizing the input readback to a target waveform so that higher channel densities become possible. A special class of partial-response channel, Class IV partial response (PR4) system, is the first widely used PR channel. The PR4ML channel density is approximately 2.0, and the extended PRML channels are achieving channel densities of around 2.5. However, the extended PRML channels are somewhat complicated, and require sophisticated ML detector, equalizers, and accurate timing and gain recovery circuits. To achieve even higher channel densities in PRML channels, it may require either excessive boosting of noise in the equalizer, or unacceptable hardware complexity and power consumption. Therefore, there are considerable interests in alternative data detection channels. Decision feedback equalization (DFE) channels are designed for channel densities above 2.5 and allow simpler channel architecture than PRML channels. The DFE channels are relatively new, so very few hardware implementations are available and most estimates of DFE performance are obtained by modeling.