Your browser is not supported. Please update it.

20 October 2011

Qualcomm Atheros argues once again for HomePlug ‘ we argue back

We have said in the past that the home networking wars are so vehement that it is tempting to abdicate all responsibility for picking a winner. The problem with all the adherents in the Home Networking space is that they expect analysts to come down on their side and operate the principle of ‘if you’re not with us, you’re against us.’

So it was with no surprise that Purva Rajkotia, product management director at Qualcomm Atheros, was onto us the moment we covered a piece about Marvell’s chip entry into G.hn, saying that the G.hn eco-system was building momentum.

Rajkotia began trying to set the scene, but the scene he subscribes to is perhaps slightly one eyed – in his world he sees HomePlug as a winner, and bundles it in with MoCA and talks about them jointly shipping 120 million chip sets compared to G.hn’s none. And breaking these out he sees 70 to 75 million HomePlug devices being the sole owner of the world’s use of broadband modulation over electrical power home cabling. He sees HPNA as being a dead end, because it has no follow on technology and DSL as a declining market.

But surely DSL is still growing, HPNA operators will be looked after by G.hn players putting both protocols into a single chip (as Sigma has already said it’s done) and HomePlug means different things to different people.

At each stage we interrupted and argued with him. If you start on the wrong foot you can agree that black is white and wherever you go from there ends up being nowhere. HomePlug comes in a variety of generations and they don’t even use the same modulation, so they are different. The IEEE P1901 standard uses two PHYs, one based on OFDM and the other based on wavelet modulation, which we understand is only sold in Japan and parts of Asia Pacific, but still makes it confusing. This was done entirely to carve up the world between them and without such a settlement the Japanese suppliers would have fought the standard.

IEEE 1901 also allows for a broader bandwidth than the original HomePlug and HomePlug AV, so that Broadcom has just come out with a chip that works up to 30 MHz only, the therefore maxes out at a PHY of 200 Mbps, and likely real world performance levels substantially below that, depending on interference etc.. but likely to lead to something like 25 Mbps at every power outlet in a European home, no more, but with some devices performing up at the 65 Mbps to 90 Mbps.

Qualcomm Atheros by comparison has just come out with a new AR7420 chip which it says adheres to the 1901 standard, but should be seen as a superset of HomePlug AV, addressing as it does spectrum above 30 MHz, all the way to 68 MHz, and which Qualcomm Atheros says will generate a PHY of 500 Mbps, but we cannot hazard a guess at real world performance, partly because rivals claim there are some technical issues with the higher spectrum here, and that 1 MHz above 30 MHz delivers less payload than below it, and that there are interference issues in some countries. They may be right, or not, but regardless you have to understand why Atheros is doing this.

This is because, like WiFi, Atheros is seeing HomePlug as the answer now and also the answer to home networking in the future. It wants us to believe that simply by opening up the same spectrum on powerline that G.hn uses, 80 MHz of it, it can produce the same outcome. Logically that makes sense.

The IEEE 1901 standard specifies an InterSystem Protocol (ISp) so that all of these very different powerline systems (with the same HomePlug name) can interoperate, but nothing had been produced to test it on, and so as yet none of the supposedly IEEE P1901 chips can really say they are compliant. Already companies are saying that HomePlug AV2.0 will also be compliant with P1901, which of course is very easy when you have 200 companies supporting HomePlug and only 60 or so supporting G.hn and they can vote a new standard into compliancy.

So the answer today is a cheap chip that does its job under 30 MHz and offers 20-25 Mbps reliably around the home, the answer in the future is a chip that has two to three times the bandwidth, a little more efficiency, in a chip cut in a smaller chip geometry. So of course Rajkotia’s right, because that’s just what G.hn is proposing with the additional of MIMO characteristics, and now Qualcomm Atheros, after the fact, and all of 1901.1 is also proposing something extremely similar.

People like Rajkotia can only see the differences, but from outside this community, we can only see the similarities. The question in fact becomes, do you want a chipset that continues to work entirely with powerline, and pays no attention whatsoever to coaxial cable and twisted pair cabling or to WiFi, and leads to topologies around the home which need some form of translation to move from powerline to fixed line cabling, and again when moving into WiFi? That world would mean a simpler, different kind of chip at every termination point in your home, in fact at least two, one to take it out of say MoCA or HomePlug and another to put it into another format.

The other world is where one (slightly more) complicated chip does all of the above. Because make no mistake, in a G.hn network the signal has to be altered when it goes from Powerline to coaxial cable, they modulate at different speeds, the guard intervals are all different, even the FEC (forward error correction) is calculated differently for each medium. So the signal is taken off one medium by one chip and another version of the same chip puts it onto another medium. In the end this will lead to cheaper chips for G.hn simply because there will be more chips that are made the same ‘ so through scale. Rajkotia argues the opposite. But does he really. Later there will be hybrid SoC chips with MoCA, WiFi and HomePlug all onboard, and in what was is that different?

There are advantages and disadvantages to both of these approaches. If you never move from one medium to another, for instance use powerline for everything, then why not keep the silicon simple. That’s what the existing HomePlug chips, such as Broadcom’s, are choosing to do. But if you need a network that has traffic shared across two mediums, because neither have sufficient bandwidth or because a home has very little coax, then G.hn comes into its own.

Rajkotia makes the comment that the G.hn chips are expensive, and even more so for businesses which already have an investment in another kind of home connectivity ‘ he makes the point that G.hn chips and the companies making them, have made no commitment to coexistence on a network which has existing HomePlug chips on it. He almost convinces himself that the G.hn crew are demonic in wanting ALL of any given home and unreasonably hate the HomePlug incumbents, as if they were doing it out of malice. Clearly they are not. G.hn has adopted the same ISp, but it is not mandatory, it is optional. ISp is a way of allocating a timeslot to each of the technologies in a home, so that they switch off when the other technology is being used. This is the only way that the two networks in the P1901 standard could interoperate, since they use different modulation techniques entirely.

So while we have no knowledge of any of the existing G.hn chips going the extra mile and putting in ISp, we also know of no testing framework to test ISp in P1901 chips, although it should be there. Our feeling is that if it cannot actually be retrofitted, it can be easily added in a later version of the chip. Right now G.hn chip makers are going after Green fields regions, such as China, and existing HPNA customers, which have no existing powerline reliance. Of course we can imagine the support call at say a Telefonica client in Spain, a company that already has RFPs out for G.hn equipment. ‘You say it’s not working, do you have any other powerline plugs, for instance HomePlugs in your home. Please unplug them. Is it working now? Yes? You have to replace the HomePlugs with something compatible, we can sell you some.’ Which is perhaps why very few HomePlug devices have sold in Spain, the origin of DS2, the company that pioneered UPA powerline, which Marvel bought, and which helped draft G.hn. Marvel G.hn chips can of course co-exist with the old DS2 UPA chips already.

Rajkotia has an answer to this. ‘Telefonica RFPs are for very low customer numbers, around 100,000, whereas HomePlug has been used in France at France Telecom in around 1 million homes.’

He goes further and says that the only reason that Telefonica has standardized on G.hn is that it is being pushed by DS2, a Spanish company. Of course DS2 is now a US company, since it is owned by Marvell. And anyway perhaps we could argue that the only reason that US operators are considering using HomePlug Av is because it is dominated and was invented by a US company, so he must be careful of where his logic takes him.

It must of course be remembered that the IEEE P1905 (which has many of the same members of the P1901) voted not to allow G.hn (HomeGrid) and HPNA to be included within its scope. P1905 is an attempt to develop a bridging layer that will allow multiple home networking technologies to work together in a single network and is set to bridge between WiFi, MoCa, HomePlug AV 1.0 and Ethernet. The exclusion of G.hn was on the basis of it not really being finished as a standard when 1905 was started, and the basis of HPNA being put outside its scope was never explained, despite it having 20 million chips out there.

P1905 is intended to allow multiple and differing network technologies to appear to upper-layer applications as a single connection type. It would allow products like IPTV STBs, gateways, PCs, smart appliances and other networked devices to use different technologies and still pass along data without the device having to be aware of the medium used or the settings each has.

So while Rajkotia is convinced that G.hn is not committed to coexistence and interoperability, it was not the G.hn group who made this happen. In fact it mustered a complaint to the IEEE over being left out of P1905. Rajkotia insists that the committee was an entity based group, meaning that one company has one vote, to it can’t be rigged, but then again the 200 or so companies that make up the HomePlug and MoCA eco-systems make it tough to get such a vote past, when HomeGrid has perhaps 60 members, or less. So suggesting that making ISp optional is hardly a group that is committed to isolation, especially when it chased so hard at an interoperability layer.

However the announcement by HomeGrid last week that it was planning not one but three more plugfest style meetings, and that it now has five separate chip makers planning to bring chips to them, made us think that although HomeGrid has no massive chip champion like Broadcom or Atheros involved, it is starting to build momentum.

Rajkotia adds, ‘Broadcom bought Gigle which was planning a G.hn device, but it has made it clear that it has dropped HomeGrid membership and is only interested in HomePlug.’ That may be true, although there is no such announcement from Broadcom, more a simple fact that its name is no longer on the HomeGrid web site. That also leaves it open so that if G.hn ever starts to have the types of volumes that MoCA and HomePlug have, then Broadcom can easily drop back into the G.hn debate, on the back of Gigle’s research. This is not something that Atheros can do, since it bought Intellon, the original HomePlug chip leader, and has no prior research along the lines of G.hn (actually it does have some, but pulled out early).

In fact we could easily make a case that if G.hn is a success it would disadvantage Atheros more than any other major powerline player, which may account for the company’s level of aggression, approaching belligerence, towards its G.hn rivals.

But one point that Rajkotia makes is true, that the choice of which home networking wins is down to the power of the brands which support it, although he risks the argument sounding like, ‘We will win because we are bigger,’ and it’s as well to remember that history is rife with Giants who we downed by smaller opponents.

But in the Battle of the brands, HomePlug has Broadcom and Qualcomm Atheros leading the charge and enjoys the backing of Cisco, Renesas, France Telecom, Huawei, Motorola, Ralink and Sony as well as support from G.hn players Marvell and Sigma. The roster at G.hn is considerably smaller but it does have AT&T, Intel and Telefonica on it.

We pointed out to Rajkotia that even 120 million chips, with at least two chips per installation, between HomePlug and MoCA, means that just 60 million pay TV homes could be wired up using these technologies. But there are 690 million pay TV homes around the world, so there is barely 10% of the market established with video class home networks so far, and plenty of business for many competing technologies. And when you take into account that many HomePlug chips have nothing to do with pay TV, it means even more market is unserved.

‘Over 66% of HomePlug chips have been installed at pay TV operators,’ Rajkotia insisted, ‘and it may only be 10% or 20% of the pay TV market, but HomePlug and MoCA between them take 70% of the part of the market which has bought a home network.’ That makes sense, giving HPNA the other 30% perhaps.

But he has to understand that WHDI, WiGig and other 60 GHz efforts, and video capable WiFi from Celeno and Quantenna, as well as 802.11ac, are all going to make a hole in the remainder of the market. And the fanaticism with which each of these players pursue the role of being the ONLY capable technology in the mix, seems hugely unrealistic.

There is not only room for all of these technologies, but for more, such as HDbase-T from Israel’s Valens Semiconductor and the almost defunct WirelessHD.

But if we had to say that one network was on the right track, it would be HomeGrid, quite simply because the eventual winner in Home Networks will be a technology that just lays itself on all the wires that are already there, and uses wireless where it can, and so far G.hn’s ambition comes the closest, and it is this vision that HPNA based operators may well buy into, which will continue this schism well into the future.