We wrote a single throwaway line after the IPTV World Forum, about the respective speeds of the two competing factions in ‘no new wires’ networks, G.hn HomeGrid and P1901Homeplug. It triggered mayhem.
Unwittingly we put HomeGrid’s real world performance claims in a sentence with a huge sign we saw on the Devolo stand ‘ quite simply ‘Why wait for G.hn we do 500 mbps now,’ an unfair comparison because this was clearly the PHY (physical layer maximum theoretical rate) of one system up against the real performance of another. HomeGrid has been claiming a 1Gbps PHY for a while now and Sigma is fond of claiming 3Gbps because it uses all three power cables to deliver it using its ClearPath technology .
Reviewing all major coverage of this debate it seems we have all been brainwashed into believing that just because a technology system is new, it’s better than the one it was designed to replace. It’s usually the case, but it’s not a hard and fast rule. So perhaps instead of asking if G.hn can or will work, we should ask whether or not it is any better than HomePlug, and in particular HomePlug AV2, due out as a standard this quarter.
When the piece ran in a partner publication, it triggered a hail of complaint in particular from Sigma Designs with Michael Weissman, VP of corporate marketing, writing an ‘outraged’ letter, stating ‘facts’ as he saw them.
From the other camp the phone rang a few times as well, and emails went hither and yon, ‘Did we know the difference between PHY rates and real world performance?’ ‘Had we heard of IEE 1901?’ One throwaway line can get you into trouble in this business, but the truth is that we stand by it, if you have a PHY of 500 Mbps, it’s probably enough to deliver HD around a home, so why worry ‘ surely both systems are fine. What everyone wanted to know was did we need help understanding this?
Do we need help? Well perhaps. We certainly felt the need to think about it and come up with some guidance for our readers, because although understanding the technology isn’t all that hard, we also have to understand how the mechanics of the IEEE work and where the power is in this equation. After all large companies have been known to turn up with 50 new members on an IEEE committee, vote out the chairman, and hijack the standard.
So for the record let’s establish what the two sides are? Promoter members for HomeGrid include Sigma, Lantiq, Intel, Gigle and Marvell, as well as VDSL specialist Ikanos, and these are the chipmakers who are driving G.hn, all of them very capable. And yet Intel is very clearly in both camps, and Gigle has been acquired by Broadcom, another chip maker just as clearly pushing the opposing HomePlug camp, and joining as a board director in January.
In the other Homeplug corner are Intel, Broadcom and Cisco and of course Atheros which acquired original homeplug chipmaker Intellon before being acquired by Qualcomm itself. Oh yes and there’s Lantiq again, backing both horses.
The old fight for operator driven TV connectivity over the past four or five years was previously between MoCA and HPNA and that fight was every bit as vitriolic as this one, and, as it turned out, there was room for both systems around the world. Could it turn out that there will be room for two in the coming world? Both sides seem determined not to let this happen this time.
The next fight appeared to be HomePlug versus UPA (Universal Powerline Association). It was that fight which gave a few clues about how this type of thing works. Homeplug 1.0 was built on the wrong technology, using CSMA/CD ‘ effectively Ethernet, while UPA used OFDM, a more modern modulation style which has formed the bedrock of the modern world.
WiMAX, DVB-T, LTE and the later variants of WiFi, all use OFDM ‘ a way of running thousands of signals in parallel in tiny fragments of spectrum, and sending data in parallel over each of them. Other wireless technologies have all fallen by the wayside, primarily because of the ease with which you can handle multiple problems using OFDM. It’s precise timing can take multipath RF signals and combine those which have bounced off something and arrive a little later, and ignore others. Frequency and time interleaving can be used to make it exceptionally robust against interference, signal noise or fading. And the number of signals can be allocated dynamically to create a very simple and yet very sophisticated Quality of Service protocol.
The fact that the WiFi crowd (original WiFi was not OFDM) could blithely abandon its early technology and switch to another, shows what can be done here. This is really a fight of brand names, not technologies, since any rider can change its horse mid-stream. It’s about building momentum, and strong allies behind a brand.
And yet given such a colossal advantage to UPA technically, Homeplug won. How did it do this? It simply copied the UPA idea and used OFDM. Homeplug AV was born with a 200 Mbps PHY and a real world performance of around 80 to 90 Mbps. And what has happened to UPA? Well its biggest fan, chip designer DS2 from Spain went bust and was acquired by Marvell. And is Marvell backing the HomeGrid standard? Well yes, and maybe no. It certainly is a Promoter member of HomeGrid and only joined in January, and it hasn’t turned up on the HomePlug scene anywhere, but equally it hasn’t come out with G.hn chip and yet it probably had the closest to a completed chip of anyone when it acquired DS2, which was the cheerleader for the technology. Why is that?
Todd Antes, vice president, Computing & Consumer Networking, at Atheros talked with us about it this week and he has a theory, ‘Sigma came along with its so called proprietary MIMO technology, which is really just a diversity feature and now they are trying to retrofit it into the standard. The G.hn chip that Lantiq came out with hasn’t got it in and they’ll have to go a make another chip once the Sigma technology is fully in the standard.’
So perhaps this happened at Marvell as well, and it was ready with a chip design and had to go back to the drawing board, but there has been no-reaffirmation of Marvell’s intent since January and it would be mad for a company that large and that ambitious not to think in terms of covering both camps, especially given that the DS2 engineers would have been just as capable of delivering HomePlug, as HomeGrid.
We know that similarly Broadcom bought Gigle, which already had a complete set of HomePlug products and was working on a G.hn chip. Since Broadcom joined the board of HomePlug in January this might equally be seen underlining its backing of HomePlug, or it too might be covering both sides of this battle. If it was, surely Gigle would have shown up on the HomeGrid web site as a promoter by now, but that hasn’t happened, although it is still there as a contributor.
Talking in more detail with Antes at Atheros he brings up the main reason why Atheros, with its Intellon HomePlug chip purchase, just feels it HAS to stay with HomePlug, ‘Companies like France Telecom, and Iliad’s Free in France and Deutsche Telekom in Germany, rely on these chips to deliver IPTV, France Telecom has really high attach rates because it pushes the technology with every set top, and Free gives away Free HomePlug plugs with every order.
‘These operators are not crying out for a 1 Gbps PHY, they need to know they can get 25 Mbps to 30 Mbps at every outlet in the home in order to deliver their service.’ And if that’s right then perhaps they need about double that for 3D, and that can be achieved easily by either standard. The US operators may be a little more bandwidth hungry and perhaps that 1 Gbps claim has more pull with them.
‘Intellon was very involved in the early days of G.hn and we thought it might really be the next generation which created a unified standard, but when they said that it would not be backwards compatible with HomePlug and took the decision NOT to interoperate, that left some of our customers with hundreds of thousands of nodes. What are they going to do, uninstall them or do we have to build an expensive bridge for them?’
We’ve heard both sides of this type of conversation in the past. Intellon wants to protect its existing customers, while newcomers with no customers don’t care too much; but at the same time Intellon wants to get this thing back under its control, in order to give it a head start, so naturally it calls for interoperability. Rivals worry that they will end up paying huge royalties for Intellon patents or that their development efforts will be slowed down and the next thing you know nobody’s talking to one another.
Antes makes it sound like the G.hn jihad were just unreasonable, but it won’t have been that way. His other issue is co-existence ‘ if the devices can’t co-operate why can’t they at least recognize Homeplug is on the network and share the line.
‘This process will be written into HomePlug 1901 over the next year,’ says Antes, ‘but both systems need to see each other and back off and try again, but if one technology backs off and the other doesn’t, then the network just stops working for the one with the co-existence protocol.’ Again he implies this is black and white, and that those pushing G.hn are just being mischievous in not going along with it, but there would have been reasons. Perhaps they are not targeting Europe, and only want to go after networks which haven’t yet taken a position. Perhaps they sense time pressures.
But with this we can see why there are reasons why Atheros and its chip partners are pushing to block G.hn being part of IEE 1905. That group is a new attempt to let MoCA, Ethernet, HomePlug and WiFi interoperate. But if G.hn could interoperate with those other physical network types, then G.hn would get a foothold and soon HomePlug networks everywhere would crash as soon as a G.hn device was plugged in, unless it also adopted a co-existence protocol.
In Europe where there is a huge proportion of Multi-Dwelling Units, that would be a disaster. MDUs often share a single ring main, and so if next door bought a G.hn device, your home’s network might fail. Bloodshed would follow, either literally or in the courtrooms.
Funnily enough when we talked to HomeGrid at IPTV World Forum, this was exactly the point it made ‘ suggesting different versions of HomePlug were regularly being plugged into a network and being unable to work with one another, something Atheros denies emphatically. Everyone can see how ruinous for consumer relations it is to take a device home from a store, plug it into a home power socket and the internet stops working throughout the house, but for both sides to be using this argument seems to be fake. HomePlug either is or is not compatible ‘ there are no G.hn devices out there yet, and we would have thought that past plugfests would have established that even HomePlug 1.0 at least recognizes and will share a network with HomePlug AV.
There is a complete absence of Texas Instruments on either board, and in either platform, and yet a year ago TI was a member of Homegrid, and it has since disappeared. So has Infineon.
The only silicon producers who have publicly committed to G.hn silicon are Ikanos, Lantiq, DS2 (now Marvell), Aware (a DSL test supplier) and Xingtera, along with Sigma, which was showing off silicon at IPTV World forum. Californian start up Xingtera promised devices before the end of 2010, but we have not seen or heard from them on the subject lately and for a start-up, this kind of war is ruinous ‘ imagine trying to raise another round of finance when one VC partner asks if they have backed the wrong horse. Ikanos has made no move in G.hn and neither has DS2 since the merger leaving just Lantiq and Sigma.
There are a few other claims that HomeGrid made to us in March that show just how aggressive this battle is getting ‘ it said that BestBuy would immediately kick out Homeplug from all US BestBuy stores, as soon as G.hn devices are ready and that the technology is destined for installation within AT&T’s U-Verse IPTV service, and AT&T has already put out an RFP for G.hn technology. Well if it has, we are faced with a split market. AT&T can drive enough volumes to keep one technology alive for at least another generation, and it did marvels for HPNA when it was first endorsed by AT&T, and it gave the all clear to other operators and since then many millions of HPNA devices have shipped.
Atheros wouldn’t address the AT&T issue except to say, ‘The AT&T decision is not over yet, just you wait,’ and then Antes clammed up.
Finally we come to the issue of performance. The 1Gbps PHY claim of Sigma has been made many times, but there was a change to the spectrum used by G.hn in 2009 reducing the amount of spectrum it used, after complaints by international carriers, as well as a reduction to its signal power for interference reasons. A paper by DS2 written prior to those changes, but focusing on the spectrum from 30 to 100 MHz, concluded that around 700 Mbps was the theoretical limit, so that anything claimed above that is probably due to the broader spectrum which is no longer used.
The spectrum change was to eliminate 100 MHz to 200 MHz and reduced the baseband operational spectrum from 100 MHz to 80 MHz (to avoid FM radio interference). Atheros makes the point that the HomePlug standard is also authorized to go up to 100 MHz, and that when it produced a chip which adheres to the HomePlug standard and yet uses spectrum up to almost 100 MHz then it achieved a 500 Mbps PHY.
Antes makes it clear that this chip can interoperate with other HomePlug chips, but of course if they don’t use that spectrum they can’t operate at a theoretical limit of 500 Mbps. But then again that’s the same for G.hn, they can only interoperate with other G.hn chips which use the same standard, and the Sigma 3 Gbps claims for its MIMO architecture can only interoperate with Lantiq at the 1 Gbps level (which we have established only operates at around 700 Mbps anyway, with real world results at around 300 Mbps).
So picture a home network that already exists using HomePlug AV. Use the new Atheros chip in two devices and they can speak to one another using a 500 Mbps PHY, but the rest are limited to their respective PHYs. We can’t see much wrong with that. It’s exactly the same with 802.11 a, b, g and n in WiFi. But Antes makes the point that they remain compatible, and by that he means they interoperate and co-exist happily.
Let’s take a longer look at Weissman’s other claims in his letter, which he entitles HomeGrid (G.hn) Outperforms HomePlug. Well does it? HomePlug AV 2 spec is out shortly and will be five time faster than HomePlug AV1. It may well decide to copy Sigma and use all three power cables, and all the spectrum under 100 MHz, but it’s PHY will be largely in line with HomeGrid. Chips will be out later, devices about the same time.
More Weissman facts
‘HomeGrid is 1 Gbps of PHY compared to a PHY of 200 Mbps for HomePlug AV’. Not a fact according to DS2 who invented it, as we said, closer to 700 Mbps.
‘The Atheros 7400 (for HomePlug) has the potential to do a PHY of 500 Mbps, but does so in an incompatible way’ ‘ err No. It can interoperate and co-exist, that’s hardly incompatible.
‘HomePlug AV goes only to 30 MHz of bandwidth ‘ any communication using bands above 30 won’t be heard by HomePlug AV devices.’ It’s true that many early Homeplug AV devices used this limit but they don’t have to, hence the Atheros 7400.
‘The Atheros solution goes to 75 MHz of bandwidth. This means all the performance from 30 to 75 MHz of spectrum will be incompatible with either HomePlug AV and all the performance from 50-75 MHz will be incompatible with P1901 devices.’ ‘Incompatible’ is surely the wrong word (G.hn is incompatible) and the 1901 charter is to 100 MHz, but anyway this will be completely changed with HomePlug AV2 standard, so not that relevant.
Weissman adds, ‘Throughputs are lower than PHY rates (which are theoretical). Throughput is a byproduct of the actual physical channel in the real world multiplied by the inefficiency of the MAC layer. In the best case scenarios, HomePlug AV devices (those that comply with the HomePlug AV spec) run at around 100 Mbps ‘ this is really rare, but possible. In the same setting and conditions, Sigma Design’s G.hn solution was demonstrating 400 Mbps of throughput at the IPTV show in London. This means that we are delivering 4x the performance of HomePlug AV.’
We really question that the MAC on a system designed for power, coax and twisted pair is as efficient as a MAC designed for just one physical layer. And HomeGrid itself cites 300 Mbps, not 400 Mbps.
Weissman goes on, ‘ With ClearPath, Sigma Design’s patented technology, we were demonstrating a 50% improvement over rival Powerline solutions when electrical noise is injected into the network.’
This is interesting. Hear what Antes has to say. ‘First off ClearPath is a proprietary architecture with a patent on, and it’s not yet a part of the G.hn standard.’ The he adds, ‘An AC line is a sine wave, and at certain parts of the cycle certain devices pull their power from the cable, usually in the upswing of the cycle. There are 6 or 7 timeslots in this cycle. The HomePlug system recognizes this and tries not to transmit during that timeslot.’ In other words he accepts that interference slows the system but marginally, and urges us to wait until Homeplug AV2 is out, where no doubt that receives some attention.
Weissman finishes ‘These facts stand on their own. Each is verifiable.’
The facts do anything but stand on their own, but if you add the fact that the biggest posse of WiFi players is backing the other group, and that they have among them the biggest chip makers in the world, and that the only really large chip-makers in Texas Instruments and Infineon that were on the G.hn team have quit that leaves Sigma and possibly Marvell who have volume revenues, and some perfectly excellent smaller players, we can see that the weight of the argument is in the market.
We point you to the resilience of Homeplug when it was up against UPA, a better technology, plus the imminent arrival of Homeplug AV 2, and its installed base inside TV operators and to the fact that G.hn has yet to win over a single operator.
If Sigma and its HomeGrid team wants to change that, it must to two things. Find a way to co-exist with HomePlug devices on the same network and it MUST land its promised deal at AT&T. If it fails to land AT&T then G.hn is likely dead. If it fails to offer some co-existence capability, then it is at the very least marginalized.
We think there is likely room for more than one technology in digital networking, and G.hn needs at least to be onboard with the IEEE1905 interoperability standard (see separate story). But a company the size and ability of Qualcomm failed to sustain its own cellular architecture CDMA (still going strong, but losing out to LTE) in the long run, though it managed two (almost three) generations. And this is looking very similar to that battle ‘ HPNA driven by Coppergate, now owned by Sigma ‘ managed a first generation win at AT&T, and has ended up with millions of installs on the back of it. Now it needs another, similar win.
But in the meantime our vote goes entirely to WiFi eventually winning the job of delivering the last few feet of video whether that is across a room to a HDTV or around the home. And the next significant event is for a company like Atheros and Broadcom, to take on the video upstarts of Quantenna and Celeno, which are trying to claim this space. One of the best steps any of them could do is go and buy (or at least sign a license) with Turkey’s Airties, because this is a company that has managed this with a Broadcom 2 x 2 MIMO standard WiFi chip, simply by writing clever firmware.
But while WiFi remains in the driving seat, both for final few feet to HDTV and to tablets, you have to see which way the WiFi leaders are voting, because they will drive this market and they are Broadcom and Atheros.