Beamforming Do-OverBy CWNP On 01/20/2009 - 70 Comments
Cisco has officially entered the Transmit Beamforming (TxBF) game. Let's take a look at their thankfully-updated (inside joke for those of you who missed the first version) whitepaper.
Cisco's TxBF implementation, called ClientLink, is the closest that I've seen in the market thus far to standards-based TxBF, though it's not really all that close to the 802.11n draft implementation. ClientLink doesn't use explicit or implicit feedback mechanisms (because those are 802.11n draft features), but instead uses the 802.11a/g client's uplink traffic as the feedback mechanism - even though their whitepaper explicitly states, "without requiring feedback." Using uplink traffic as the feedback mechanism has the obvious problem of sub-optimal beamforming if there isn't a continuous stream of traffic between the AP and client, but there isn't any other way of doing it really. One thing to point out here is that the smaller the beamformed area, the more continuous uplink feedback is needed.
One of the particular statements in this whitepaper to pay attention to is, "Cisco has added advanced signal processing into the Wi-Fi chipset." This is likely a bit of marketing spin considering that Marvell makes the chips, but looking past that, they are certainly going high-tech on us. Beamforming at the individual subcarrier level in an indoor, non-LOS, reflective RF environment (due to some innovative technology likely acquired from Navini), Cisco purports to exceed the direct line-of-site maximum gain of 3dB for a two-antenna transmission system. The Cisco whitepaper references new technology found in a new book from Cambridge Press, which I'm currently reading. Interesting spin...I'll have to address the technical details of this after I've done a little more math homework.
It's important to note that ClientLink is TxBF for 802.11a/g, but NOT for 802.11n. Spatial Multiplexing (SM) is used rather than TxBF for 802.11n clients. Doing TxBF at the chip level rather than in a separate controller leaves the implementer with an "or" choice: TxBF or SM.
I'm glad to see Cisco enter the TxBF game. It validates the technology in a big way. Now there's two completely different implementations of TxBF technology in the market: Cisco's ClientLink and Ruckus's BeamFlex. I wonder who will be next? Anyone? Anyone? Beuller?