Just as Broadcom, Marvel and Ralink (MediaTek) all came out with 4 x 4 Multi User MIMO chips for WiFi, Qualcomm and Quantenna this week showed the advantages of coming to market with MU-MIMO a year earlier and showed a handful of devices which are already committed to their MU-MIMO technologies. Tiny Quantenna, a leader in MIMO WiFi chips for broadcast video also has a similar lead, but is sure to be marginalized by these recent efforts from its bigger brothers.
Qualcomm showed off design wins at Amped Wireless, Buffalo, D-Link, NEC, TP-Link, TRENDnet and China’s Xiaomi, who all say they have committed to using the technology.
Back in October Faultline spent about three issues not quite believing the fact that Qualcomm seemed to have a considerable head start over Broadcom, virtually a year since announcement, in Multi-User MIMO. However as wave two 802.11ac chipsets have emerged at CES, Broadcom has finally revealed its hand with a chip which it says has a different philosophy, and which is claimed to be four times faster than a normal SU MIMO chip. Marvell and Mediatek’s Ralink have also put their hat in the ring with chips that offer similar claims and seem very similar in conceptual design.
Qualcomm sold MU-MIMO as being 3 times as fast as single user MIMO, a claim we tended to doubt, but we acknowledged that it was at least, somewhat faster. In November an analysis by MU-MIMO Manny Patel, Director of WLAN Business Development at Broadcom, described that Qualcomm approach as an N minus 1 approach, suggesting that it was configured to best perform simultaneously with three client devices and designed for 1 x 1 WiFi devices. Patel said at the time that when Broadcom comes out with its own chip, it would be better at supporting clusters of 4 devices simultaneously.
The difference appears to be in the mathematical model of how the each of spatial streams can best be used. When a designer assume that the clients are 1 x 1 antenna clients, as he insisted Qualcomm’s design does, it has to use one of the antennas to help the other antennas form beams. However when you are speaking to 2 x 2 clients, which is increasingly the case in smartphones, then it is easier to form beams, since there are twice as many spatial streams being received.
Separately Qualcomm showed off its new 802.11ad 60 GHz technology at CES, which came out of its acquisition of Wilocity last year. Although it only works in line of site, so within the same room,. It was shown transferring a 122 MB movie from a movie rental kiosk to a tablet, taking just 2 seconds.
60 GHz WiFi is clearly a technology lead that buying Wilocity has earned Qualcomm, but we’re more interested in whether its Atheros group has enough of a head start in Wave 2 802.11 ac chips, to dent what has previously been an unassailable WiFi lead at Broadcom for a decade.
These 2nd wave chips have two independently managed radios driving the signals at any given antenna and all of the chips out this week have four antennas, so that means that each antenna can send two streams at the same time, a total of 8 streams. The idea is to null these streams in any direction except the one which points at the receiving device, in order to save transmit power for other streams, and as far as we can work out, the signals from the other antennas have to be calculated to help with both this beamforming and the nulling process to adjusting their transmissions.
Given that many of the receiving devices are handsets and until this year many of those have been single antenna, single radio devices. IN which case the point of having four MIMO in single user mode wastes lots of resource, because such a device can only receive as many streams as it has antennas. So while the transmit power may be divided into 4 streams, only one of them can be received.
When receivers have more antennas, for instance four antennas, then the WiFi Access Point can send one stream to each antenna, and use its full capacity.
The biggest waste is when a Single User Access Point with four antennas sends to a single device which has a single antenna. If the 4 x 4 MIMO 802.11ac is capable of sending 1.7 Mbps, then a single receiving antenna on such a device gets just 433 Mbps as its maximum, wasting 75% of capacity. The AP then moves onto its next job, and if that is also talking to a single antenna device, it continues to waste that amount.
With MU-MIMO then each antenna can engage two client devices simultaneously and this is where Broadcom gets its claim that its chip goes four times faster than previous APs.
The truth of this particular pudding will be in the eating, and if Broadcom genuinely has pursued a different technical philosophy (and we’re not sure Qualcomm would agree that it has) then there are tons of small details which may affect real world performance. Questions such as how much processing power do the chips have, what kind of overhead is there in calculating MU-MIMO beams, and how quickly can they switch to a less congested air channel when they meet interference, and the fact that when a device is speaking back to the AP, it has to do this sequentially since these are not MU-MIMO clients. All of this will affect the performance and we won’t really get an understanding of which of these devices the market will like until we actually have devices.
MIMO market leader Quantenna, which has had its SU-MIMO devices selected by many of the US home gateway operators such as AT&T, launched its own MU-MIMO chip back in May 2013, which to date has only thrown up a design win at Linksys and a demonstration of interoperability with MediaTek MU-MIMO chips.
At CES this week Quantenna was demonstrating what it calls its 10G WiFi platform, an 8×8 Multi-User MIMO chip with 160Mhz channel support which it said even provides a benefit to 1 x 1 clients.
At CES Broadcom was pitching its new MU-MIMO BCM 4366 chips at retail routers, wireless cable/DSL/PON gateways and set-top boxes (STBs) and claimed that it was optimized for the delivery of broadcast quality video around a home.
The chip is coupled with the BCM47094 network processor SoC with a 1.4 GHz dual core ARM processor, which it says will improve video streaming performance. We figure this is the chip that will earn Broadcom some of the home media gateway business at AT&T and other tier 1s, which had to opt for Quantenna in their initial specifications
Broadcom also came out with a 2 x 2 80211 ac chips the BCM53573 and BCM47189 dual-band SoCs for “affordable” and “mid-tier” residential routers, wireless repeaters, power line and MoCA Wi-Fi bridges.
Marvell launched a very similar system to Broadcom with a 4×4 802.11ac Wave-2 WiFi SoC which it calls the AP3200 Reference Platform built around its 88W8964 chip, which it says it aimed at enterprise access points, smart gateways, multi-room carrier-grade video distribution and SMB routers. The Armada-395 SoC also integrates a dual core host CPU running at 1.8GHz.
And finally MediaTek released its MT7615, MU-MIMO chip operating in both 2.4GHz and 5GHz channels, which it says is best for delivering 4K video around a home or office environments. It is designed with MediaTek’s 4x4x4 (4Transmitter, 4Receiver and 4Spatial Streams), and combines with MediaTek’s MT7623 Quad Core ARM-based network processing unit.
Almost all of these chips go to sampling now or in three months and we guess it will be between 6 and 9 months before products which use them hit the market.
Considering that the Wireless Broadband Alliance has been talking about MU-MIMO coming late in 2015 and only plans certification for March 2016, there are an awful lot of chips already on the market, which we think is perhaps because of the huge revenue potential at stake compared, to even prior generations of WiFi..