Chatting this week with the Marvel/DS2 employee who relatively recently was put in charge of the HomeGrid Forum’s G.hn promotional effort, and we find we are up against someone with a freshly aggressively state of mind, after the G.hn successes in China, announced at Computex.
John Egan fires an opening volley at rival powerline communication group HomePlug, by highlighting a clear and present weakness in the roadmap, a segment of technology where he feels HomeGrid got it right and which is responsible for more or less all of its success in China (see last week’s Faultline).
‘China Telecom did try out the rival technology HomePlug before it decided to go with G.hn, and it tried it in a single Multi Dwelling Unit, and said it did not meet its performance criteria,’ said Egan. It was only last week that he was given public permission by China Telecom executives to publicize the choice of G.hn as its official system for powerline. ‘That doesn’t mean that there aren’t some HomePlug installations out there. China Telecom operates in 7 regions, and each operating group is allowed to pick any technology it likes as an interim technology, so there may be the odd HomePlug installation out there, but that won’t continue after this decision.’
The fatal technical flaw that Egan attributes the G.hn win to is Neighbor Network Interference (NNI) whereby more than one home can be on the same electrical circuit in an MDU. About 50% of Chinese people live in MDUs, so this is important. The existing approach that HomePlug uses to protect neighbors from this interference in their broadband is to share the network. If there are 3 homes, then you officially get a third of the bandwidth using a CSMA approach like Ethernet. But with noise considerations that are common on powerline networks, Egan says that this sometimes comes down to 10% of the speed of HomePlug if you were not sharing it. ‘In China some circuits are shared by up to 16 residences,’ he added, implying that it could be catastrophic in performance terms.
The MDU issue is more complex than it first seems. You could introduce filters between each property, but this ends up being expensive and the connections are hard to get at. And even if you manage this, the cabling often runs parallel to other parts of the circuit, so the interference comes back through inductive crosstalk after you’ve filtered it out.
Egan brought out the same old chestnut that has been used to beat up both HomePlug and MoCA as well, the fact that both of them are, in essence, systems which rely on single sources of silicon, or in MoCA’s case used to rely on a single source for quite a long period, while it was gaining ascendency. ‘Intellon designed all the chips, so it said what went on them, there was no standards process, no interoperability and it decided that this was a good enough approach for MDUs.’
Egan goes on to compare key dates in the development of the two standards, HomePlug AV and G.hn, mapping dates for the completion of the specification, availability of first silicon, first certified silicon, 2nd certified silicon, when the first multi-vendor plugfest was held, the development of the MIMO specification, first MIMO silicon and first certified MIMO silicon. The graph does speak for itself and although there may have been drop outs and uncertainties along the way, it shows that this all took 9 years within HomePlug, while G.hn claims to have got through it in 3.
HomePlug is now saying that it will have a more effective Neighbor Network Interference strategy within its silicon inside a year. The G.hn NDIM approach centers around sending a pre-amble to identify each message and setting the broadband lines from different homes orthogonally out of phase using this pre-amble. The receiver in that home then also treats everything else as interference which is not its own message bandwidth and tunes it out.
Egan claims that this system uses patents held by Marvel and says that it will happily license that technology inside G.hn, but not to other technologies, like HomePlug. We advised him against this route because patents can be used for anything people want to use them for once they are licensed, and the idea behind one patents it never to prevent an idea being used, but to get a fair payment for the use of your invention. We suspect that HomePlug will go ahead and use a similar method and let the legal actions fly, if need be.
The truth seems to be that had HomePlug (Intellon is now controlled by Qualcomm Atheros) taken these concerns more seriously and helped China Telecom over the problem say a year ago, G.hn may not have had anywhere to go. But it didn’t. Egan insists that the reason that it didn’t is endemic in the way Intellon was the only chip vendor in HomePLug for so long. Another part of the reason for this has to be because by then, HomePlug was already a retail product and retail products don’t get re-written because a major operator wants an improvement, they instead live or die by the consumer’s interest.
Egan claims that this system was invented by Marvell’s Spanish acquisition DS2, and it creates no more than a 10% fall-off in performance when different homes on the same electrical circuit use G.hn.
Egan gave more details about the China Telecom acceleration of the technology saying that testing has been held at China Telecom for G.hn silicon and systems, that it wants customer trials later this month (July) which will go into at least 3 provinces, with 100 homes in each, with 2 or more devices per home. If these are fine China Telecom wants to go to deployments by end of year.
But what we missed last week was that Chunghwa Telecom in Taiwan also said it would they deploy as well, as soon as it gets its hands on HomeGrid certified systems. The Marvell chip had already been certified, and Sigma has just announced certification, with Metanoia and Xingtera not far behind and one device maker in Comtrend taking part in interoperability tests.
When Ericsson announced last year that it had replaced the IPTV systems inside
Taiwan’s Chunghwa Telecom, the operator had over 900,000 subscribers at that point, so while not as big as China Telecom, it is an older and more sophisticated IPTV environment, where there is more likely to be need of multi-room connections, using powerline as the domestic backbone.
As Egan’s parting shot against his old enemy, he bring the conversation around to the IEEE P1905.1 technology, whereby rivals HomePlug and MoCA can interoperate with WiFi and Ethernet. G.hn was denied involvement in the standard on the pretext (or reason, depends on how you view it) that at the time G.hn was not a completed and deployed standard.
What the HomeGrid Forum has done is go to the ITU and within its auspices it has defined its own connection standard for P1905 which shows all the G.hn (and HomePNA) players how to successfully implement the same interoperability with WiFi and Ethernet and its rival networks, simply by adding some key management messages to the basic G.hn stack.
Egan says that HomeGrid is now ready to come out of its purely Chinese support environment and begin to round up the HPNA vendors that it landed last time. HPNA was a rival standard to MoCA, which allowed signaling over both twisted pair and coaxial cable, and which was installed by many Telcos around Europe, as well as AT&T in the US, in its U-Verse IPTV system. It is true that G.hn has had its back to the wall for some time, but that now that it has a solid relationship in China, it will not go away, and as it solves perhaps the last few technical interoperability questions such as this P1905 connection, it is preparing to push out into HomePlug and dare we say it, MoCA territory, with some confidence.