| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Ryan
I have copied Mr. Roy Cideciyan on this email as well as provided link to his analysis of MTTFPA work originally presented in the
IEEE802.3bj May 2012 meeting
Roy considers two scenarios one with a=0.5 and one with a=0.1, where a is the probability of next bit being an error bit. Potentially we may
need to look at even smaller value of a like 0.01 or 0.001 and see what happens to Roy's model. Looking at Roy's work, it does appear that for
any reasonable implementation of 20 dB link like OIF MR to avoid MTTFPA issue you will need FEC.
A possible way to get around of MTTFPA is by operating with small DFE taps like max of 0.125 and combined with link BER of ~1E-25.
But this approach does not make sense as it will require likely a receiver with KR4 DFE capability, since the tap are constrained.
The whole purpose of 20 dB chip to chip was to have lower power dissipation and the receiver will have 2-3 tap DFE.
Another option would to borrow KR4 with all its capabilities and just define a 20 dB channel. Are we then fragmenting the market for
too little to be gained?
Thanks,
Ali
On Oct 3, 2012, at 9:48 AM, Ryan Latchman wrote:
Colleagues, |