Hello,
On the SMF ad-hoc call yesterday, there was some discussion on the impact of the width of the measurement window used when the TDECQ equalizer is optimized. Specifically, the question was brought up whether
the MMSE tap optimization should be performed over the central 0.14 UI of the eye, instead of the center 0.1 UI. The logic was that the central 0.14 UI corresponds to the region of data that will be included in the final TDECQ measurement histograms.
We did an experiment where we kept the input signal constant and adjusted that optimization width, and then measured the resulting TDECQ. We used a 5 tap, T/2 spaced equalizer (as called out) and ran standard
MMSE (as called out in draft 2.2) as well as our proposed adjusted MMSE approach that pre-loads the waveform with noise prior to optimization. The input signal was band-limited to get a TDECQ value around 1.9 dB.
As you can see from the following chart, if you pre-load noise prior to optimization the results are very insensitive to the width of the optimization window, up to about 0.4 UI:
Regards,
Greg LeCheminant
Keysight Technologies