| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Dear colleagues, After the discussion we had about delay constraints in our Interim meeting of 11 August, I have been thinking how to provide arguments that can make easier to understand the adopted specifications. During the meeting I did a reference to .3ch and .3cy projects, which may be used to compare delay constraints specifications, so I will use them below. Basically, in .3cz, .3ch and .3cy, the xMII to xMII latency will be dominated in a practical implementation by the FEC decoder (I only consider interleaver 1x configuration because .3cz does not include interleaver) and the DSP needed to implement compensation of the impairments like ISI equalization, echo cancelling (not applicable to .3cz), etc. In order to meet timing constraints, real implementations are parallel, e.g. time-interleaved ADCs, parallel structures for FFF filters, parallel structures for FBF filters with pre-computation, parallel structures for FEC decoder, etc. It is important to consider this in the latency upper bound specification. The 3 mentioned PHYs are baed on Reed-Solomon FEC with different n, k, and t parameters. In order to start a Reed-Solomon CW decoding, a full CW needs to be received. Some extra time is needed to implement syndrome computation, locator polynomial computation, error location roots computation and error values computation. Complexity of these computations is proportional to m*t product. Error correction complexity m*t:
A figure of merit (FOM) that I think is illustrative is the ratio between the delay constraint and the RS CW length in bits:
The 27% lower FOM of .3cz can be explained by lower decoding complexity (i.e. 110 vs 170), larger CW length (i.e. 544 vs 360) and lower DSP complexity (i.e. smaller equalizer, no echo canceller, etc). However, the delay constraints are in the same order for all the three PHYs considered, .3cz, .3cy and .3ch, considered normalized by the CW length. Best regards, Rubén Pérez-Aranda To unsubscribe from the STDS-802-3-OMEGA list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-OMEGA&A=1 |