| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
TP2'rs -
During review of my presentation on penalties vs. EQ configurations with
measured TP2 waveforms (see http://grouper.ieee.org/groups/802/3/aq/public/upload/TP2measuredwaveformpenalties_finiteEQsTP3D1.1stressors.pdf),
a question came up about non-monotonic behaviors in some of the plots - where in
some cases EQs with fewer taps worked better than cases with more taps.
Here is the explanation. It is not due to timing - timing has been
optimized. It is due to the fact the EDC is based on MMSE, where it is
effectively minimizing an overall rms value at the slicer input. With ISI, the
pdf of the ISI/noise combination at that point may not be Gaussian. The TWDP
simulator uses a bit by bit BER-based method, where outliers in the pdf can
dominate the BER - and therefore - the penalty. If we could look inside, we
would see that the MMSE's rms value indeed reduces monotonically with tap
count (I have plots that in fact indicate that). If a EDC attempted to
optimize BER, the behaviors would swap.
MMSE has been assumed for all of LRM's work to date, so we
believe the BER/penalty behaviors observed in the plots are realistic.
Tom Lindsay
ClariPhy Communications tom.lindsay@clariphy.com phone: (425) 608-0209 or (949) 480-9210 cell: (206) 790-3240 fax: (425) 608-0232
|