Qualcomm this week has taken the Wave 2 802.11ac chips it came out with precisely a year ago, and changed them radically, to offer a new “tri-radio” architecture, two 5.0 GHz alongside a single 2.4 GHz band. To us it looks like a curious attempt to lengthen the life of what we shall call the “big” chip approach to WiFi. We had expected that tri-radio would mean 2.4 GHz, 5 GHz and 60 GHz, in honor of its Wilocity acquisition, but not in this case, instead it means a three ring circus, where no-one can be clear of its real roadmap.
Most Home Gateways today have a single SoC in them and they are supposed to offer all the WiFi capability for an entire home. And yet none of them, not even the 8 way MU-MIMO Quantenna devices, reach every crevice at every home, and the result is that 1 Gbps devices run at under 100 Mbps sometimes in some parts of a home, while others still fail to get a household wide signal of any description. More recently operators have gone for a dual mode strategy 5.0 GHz and 2.4 GHz together, which was initially triggered by 802.11n, and has just begun to happen in 802.11ac. No-one has put three radios in like this design.
Instead all RFPs from tier 1 players right now are going in a different direction, multiple Access Points (APs) meshed around the home, and responsibilities shared between them. So why is Qualcomm not going down that route on its roadmap?
What Qualcomm is revealing certainly looks similar to that, except that while it mentions a “mesh” architecture it doesn’t describe one, and it has no mention of a multiple AP strategy. We have emailed queries, but not received any response by the time we went to press – curious in its own right given Qualcomm of old.
Reading between the lines this is a chip built for a specific bid, for one of the larger operators, and this chip may or may not become a mainstream Qualcomm offering. The new Qualcomm Atheros chips work with two options around the IPQ40x9, with either a 2 stream MU-MIMO radio or one with 4 streams.
Qualcomm offers little explanation of why it has two radios for the 5 GHz space, but we can only assume that it plans to “prioritize” the use of one for heavy lifting, either backhaul or stationary devices like 4K TVs which require high volumes in real time, and WiFi backhaul and extenders. The press announcement talks about WiFi Mesh in the sense that it says it can support one, but although there are many proprietary mesh technologies for WiFi the official WiFi mesh remains largely unused and we have not come across Atheros using one at all, although OEMs may add one using their chips. It also mentions WiFi extenders, and using this as a backhaul for those.
Sometime during 2014 Qualcomm came out with an ingredient essential for this type of arrangement to work, which it calls Band Steering. What it does essentially is to decide which AP a particular device should be attached to, and in order to get it to reach out to it, it weakens the signal that it is currently receiving. The radio gives up on the weaker signal and jumps on the signal that will give best system performance. So far so good, and this is something that it is becoming clearer is absolutely essential for healthy WiFi.
So where devices only have a 2.4 GHz radio, they automatically attach to the single 2.4 GHz radio. But where they have both 5.0 GHz and 2.4 GHz, the system can automatically steer the client to the 5.0 GHz radio which has the least amount of traffic. We first came across this at AirTies, and now it has become commonplace and Arris mentioned this to us at TV Connect.
So a 4K TV can get a stable connection in one 5 GHz channel, and anything that detracts from this can be steered away to work on either the other 5.0 GHz channel or the 2.4 GHz legacy channel.
Now MU-MIMO was invented so that an AP could cope with 10 or 15 devices, sharing them across 4 separate streams. But by allocating two 5.0 GHz channels, especially if they are 160 MHz wide, means that your neighbor’s Home Gateway will have even less clear channels to work with and this will almost certainly create more interference. And since we know of few homes with 30 devices to cope with, it seems overkill to offer 2 radio channels which also offer multi user MIMO. Again no guidance from Qualcomm on this.
Other topologies have used one 5.0 GHz channel and a fall back to powerline or MoCA as a backhaul technology. By using a second WiFi channel, there is the potential to create far more interference.
The system can allow all three radios to appear as the same SSID, so that there is seamless attachment between each radio, and it also uses what Qualcomm calls WiFi-SON (Self Organizing Networks), which of course is nothing like cellular SON, but instead puts in a layer of security, and allows smart phone management, plug and play installation, content prioritization, extender or node discovery, it reports on intrusion attempts and it provides the intelligence for the band steering. Basically it’s a software layer on chip. It even offers mesh re-routing for slowed down packets, but as we said, we were unaware that Qualcomm-Atheros has a mesh unless it’s the CSR Bluetooth one.
Qualcomm says that Wi-Fi SON now supports HomePlug AV2 powerline solutions, so quite clearly it is working on offering a powerline backhaul too, but hasn’t got it yet, or at least not in this product. Which again confused the roadmap.
Qualcomm rolled out the usual suspects in Amped Wireless, AsusTek, Linksys and TP-Link to praise this chip, which we seem to remember were the same companies from a year ago. So has Qualcomm not added to its design wins since then? Not that we’ve heard.
While Qualcomm also brought out another WiFi device to support IoT, we’ll leave our IoT publication RIoT to explain that. Essentially it has created a low power WiFi in the QCA401x and offered support for HomeKit, Google Weave and AllJoyn.
Its third offering this week, and these are all being lined up to show at Computex in Taiwan, is called GigaDSL, which appears to be the brand name at Qualcomm for a full G.fast chip. We asked Qualcomm if this was the same chip Ikanos had out before Qualcomm acquired it, based on VDSL 35b, using a full 35 MHz of spectrum on the twisted pair, instead of the more common 17 MHz used in VDSL. No answer as yet, but deep in the release it mentioned it as synonymous with G.fast, which uses frequencies go up to 106 MHz in G.fast in phase 1 and then will rise to 212 MHz, for some vendors during 2016.
Qualcomm claims 1 Gbps, but does not cite operational distances except to says it is aimed at Japan and Korea, by which we assume that it is built as an upgrade for MDUs, from VDSL. Most first cut G.fast chips did not hit 1 Gbps so this may also not do so over anything more than 50 meters and it may need two twisted pair lines to reach that speed. Qualcomm bought Ikanos in August last year and it we’d be very surprised if this is really a full G.fast chipset, ready to ship.
Qualcomm says that OEM designs are already on the way from Sumitomo Electronic, NEC and Wave Electronics for that region. Additionally, the company said that it plans to move to components for 10G-EPON in Japan and it says that transition from VDSL to G.Fast ‘GigaDSL’ is an important segment of this upgrade. It is said to work in multi operator environments and it can interoperate with deployed VDSL CPEs. Sumitomo has been a customer of Ikanos VDSL chips since 2006.