Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[10GBT] Why not do PAM4?


Several task force members have asked me privately, "If PAM8 is better than
PAM12 because of the increased separation of levels, then why not do PAM4?"
This is a legitimate question and I would like to address it and put it to
rest once and for all.

Because of THP and 12dB coset partitioning, the only PAM levels of interest
to 10GBASE-T are multiples of 4. For any 4N levels, the symbol rate of the
PAM system in a straightforward implementation is

Fs = 10000/(4*log2(4N) -2) MHz

For N=1 (PAM4), Fs works out to 1666.6667MHz or the Nyquist frequency, Fs/2,
works out to 833.33MHz. At this symbol rate, even if the cabling
characterisitcs can be extrapolated to such an extreme, the optimum DFE SNR
would be less than the required 13.9dB for 1E-12 BER. Therefore, there is no
margin and PAM4 is out of the question.

Sailesh Rao.

FREE pop-up blocking with the new MSN Toolbar  get it now!