A Coefficient-error Robust Feed Forward Equalizer

Abstract

This thesis presents a new 4-tap feed forward equalization transmitter which is robust to coefficient errors while nominally behaving identically to the conventional one. The proposed transmitter improves eye variation due to coefficient errors without calibration per I/O by utilizing channel loss at negligible hardware cost. A simple digital transition detection filter inside the proposed transmitter modulates signal perturbation by coefficient errors and reduces its low frequency portion while the low-pass-filter channel attenuates the error signal’s high frequency portion. In contrast, the conventional transmitter, which does not have transition detection filter, suffers from the low frequency portion of the signal perturbation. According to our simulation, the proposed transmitter can improve the eye variation by 230% and by 560% for 25-dB and 40-dB, respectively. To verify the proposed concept, both the conventional and proposed transmitters were fabricated in 65-nm CMOS technology and tested at 8Gb/s compensating 25dB channel loss. The experiment shows that the eye variation is improved by 230% without calibration, which implies that 69% of I/O area for calibration overhead can be potentially saved.