G.hn, a PHY For All Seasons

By Rich Nesin, General Manager and Resident Philosopher, HomePNA 

Back last May we bloged a bit about what makes G.hn different. One caveat, set down at the time by the service providers, required G.hn to start with a clean slate.  They were told to choose the technology based on best of breed rather than interoperability with existing technologies. With evangelical zeal and no existing triple-wire home networking industry specification, it seemed like a good idea at the time.  I think this decision more than anything else helped G.hn stay the course and finish on time

So it wasn't a huge surprise that G.hn didn't pick one of the dominant wired home networking codes used by HomePNA 3.1, HomePlug AV or MOCA. Instead, G.hn used a Low-Density Parity-Check code (LDPC code) which Wikipedia correctly notes is an error correcting code used to transmit messages over a "noisy" transmission channel. Wikipedia goes on to say that that it wasn't practical to implement an LDPC code when it was developed in 1963 however in theory it is the most effective code developed to date.  This doesn't mean the other codes are bad even over noisy channels.  It only means that after evaluating the alternatives, LDPC was the best match to their requirements.

So G.hn selected a code for all seasons, allowing technologists to design one universal wired home networking chip. Selecting the code type was just part of the job (although a fairly contentious part). G.hn standardized a complete PHY specification; in addition to the LDPC coding scheme they selected modulation parameters and other specifications required to actually make it work. The three media types – powerline, coax and phone wires – are not the same. Far from it. G.hn developed different modulation parameters to provide robust operation over each medium.

The result is a PHY standard ready for the chip designers. A PHY standard is not the same as a chip specification so expect different implementations from different vendors. For example, there is no physical interface to the PHY defined; just a logical interface allowing each vendor to design its own optimized product. Should you expect to see stand-alone G.hn PHY chips? I don’t -- any more than I expect to see stand-alone 802.11 WiFi PHY chips. Integrated PHY/MAC products make a lot of sense for a number of reasons (not the least being cost). Should you expect chip vendors to be hard at work putting the finishing touches on their G.hn PHY designs now that the standard is complete? You betcha.

Comments

[Chano Gomez]

Hi Rich,

You are right when you say that the final choice of an FEC (Forward Error Correction) scheme for G.hn was a contentious one. For months, proponents of the two main choices (LDPC and CTC) submitted hundreds of pages of performance simulations and carefully reviewed each other's contributions to find any flaws.

This peer review process was really productive: flaws were indeed found and fixed, which means that the final proposals were even better that the original ones.

I think that the final decision has been a very good one and provides a good FEC scheme that will work very well across all three media targeted by G.hn.

Other "compromise" proposals that were discussed at the last minute, such as choosing one FEC scheme for powerline and a different one for coaxial cable, would have fragmented the market and would have discouraged silicon vendors who wanted to implement support for all three media in a single chip.

Chano Gomez
DS2