Full Duplex Wireless Tech Could Double Bandwidth

CWmike writes "Rice University researchers announced on Tuesday that they have successfully demonstrated full-duplex wireless tech that would allow a doubling of network traffic without the need for more cell towers. Professor Ahutosh Sabharwal said the innovative technology requires a minimal amount of new hardware for both mobile devices and networks. However, it does require new standards, meaning it might not be available for several years as carriers move to 5G networks, he added. By allowing a cell phone or other wireless device to transmit data and receive data on the same frequency, unlike with today's tech, the new standard could double a network's capacity. Rice has created a Wireless Open-Access Research Platform (WARP) with open source software that provides a space for researches from other organizations to innovate freely and examine full-duplex innovations."

Comments

More than double?

[Geoff-with-a-G]

In wired Ethernet topologies, going full duplex yields significantly more than double the throughput, since you no longer have collisions, back-offs, and re-sends. The article doesn't elaborate whether their full-duplex wireless would still be multi-access (think WiFi, with many clients on the same AP and same channels) or if each frequency would be carved out for one client and the base-station (in which case you'd see the same gains you did on wired Ethernet).

M point is that while they cite "allow a doubling of network traffic", the reality is even better than that. Full duplex gets you more than double throughput, as well as improved jitter/latency since you no longer have to randomly re-transmit frames (or randomly wait to transmit, as with WiFi collision avoidance).

more on why this is difficult

[Trepidity]

The idea, as they mention, has been around for a while, in fact since at least the early 1970s, with some information-theoretic work putting bounds on ideal full-duplex operation. The main idea is that you can cancel your own transmitted signal locally because you know what you're transmitting. The difficulty is that the transmitted signal is much stronger locally than the received signal, so there is little margin for error for imperfect cancellation; even if you cancel out 99.9% of the signal, there might still be too much noise left to decode the incoming signal. Errors can come from nearly anything; slightly imperfect knowledge of the characteristics of your device, changes due to weather or motion, interference from surrounding objects, etc.

Also note that terminology here is a bit confusing. In some uses (esp. radio), "full-duplex" just means any system that is capable of having people speak in both directions simultaneously, even if it's done by using separate frequencies for each direction, or by using a multiplexing scheme. In contrast, this usage of full-duplex means that both directions are transmitting simultaneously on the same channel, without segmenting or multiplexing it.

I don't actually know how they solved the problem, though, and the article is light on details.

Re:more on why this is difficult

[vlm]

Errors can come from nearly anything; slightly imperfect knowledge of the characteristics of your device

Non-linear effect anywhere in the RX or TX chain, or intermod from surrounding objects is a big problem.

I've done quite a bit of RF design work, microwave ham radio stuff, etc. The big problem is historically low noise stuff which makes a great receiver tends to blow up when subjected to power, and high power gear tends to have horrific weak signal noise characteristics.

A great low noise fractional dB noise figure preamp is off the shelf and cheap, and it'll be vaporized by say 20 dBmW.
A great 30 dBmW MMIC 1 GHz amp is off the shelf and cheap, I have used the watkins johnson devices (yes I know they have a new marketing name which I've temporarily forgotten), and its weak signal noise performance ... is not good.

In contrast, this usage of full-duplex means that both directions are transmitting simultaneously on the same channel, without segmenting or multiplexing it.

I don't actually know how they solved the problem, though, and the article is light on details.

If I had to do it, I'd do traditional 70s era spread spectrum code division multiple access CDMA. Imagine a psuedorandom voltage generator feeding the RX VCO attached to the RX mixer. Then imagine a different psuedorandom voltage gen, or at least the same generating polynomial at a different offset, feeding the TX VCO attached to the TX mixer. Two completely separate RF paths, maybe up to the antenna. Synchronizing two separate psuedorandom voltage gens is merely twice as fun as just one, kinda, I guess.

The other way was to use an old fashioned yet highly effective RF circulator. They are large, and heavy, and frankly kinda hard to make. Think like a hockey puck of ferrite with a big ole magnet. RF only flows clockwise. This is old, old stuff. Larger and heavier than a "brick" cellphone from the 90s, although they worked perfectly fine at the base station.

There's another way to do it using PLLs and the two transmitters in quadrature, but that's getting bizarre (like, have I been drinking this morning already?) and synchronization is gonna be an absolute bear. The hard part isn't static stability, but dynamic as it switches in and out of sync, or multipath interferes with it.

Small-scale DIDO?

[Namarrgon]

MIMO uses multiple antennas and the Rice team was able to send two signals in a way that they cancel each other out, allowing a clear signal to go through over the single frequency.

Doesn't this sound an awful lot like the DIDO approach (pdf) that Steve Perlman was talking up recently?

This isn't very helpful

[The O Rly Factor]

When AT&T, et. al. are in a position where they are the DeBeers of wireless bandwidth. I think instead of actually spending money to upgrade infrastructure, they would rather just continue to artificially limit the amount of available bandwidth so they can keep it grossly overvalued. Gotta keep those profits rolling in for the shareholders somehow.