Wireless Data Could Travel Twice as Fast Using a New Circuit

The high level view of FD is that one removes the transmitted signal from the receive path, allowing reception. This must be done at two levels: 1) a big chunk is removed in an analog way 2) the remainder is removed in the digital domain. The reason is that no A/D converter is good enough to work only in the digital domain. Depending on the team, there is a claim of 85 to 100 dB reduction in the Tx to Rx loop. That's impressive, but still a bit short when in LTE a device can receive at ~ -100 dBm (YMMV) and transmit at +23 dBm. And the analog removal part will have a tricky calibration: not a problem for a controlled demo, but tricky (and potentially costly) in real life.

On top of this, all those techniques remove the internal looping of the Tx signal to the Rx. That's already a challenge, but then there are the external reflection: the Tx signal has a direct loop in the device, but can also bounce on surrounding walls/stuff and get back the the Rx path this way. The issue then is that there is a serie of time lags and distortions, so the UE doesn't really know what needs to be subtracted. En the power level of such reflections is big enough to kill the reception, on its own. This reflection issue doesn't exist for some devices like line of sight micro-wave backhaul, but is very common for all user devices (used in-door or in urban environment: no line of sight, a lot of multi-paths).

The reflection issue makes me pessimistic for consumer devices applications. And for other application with little / no reflections, there may be more cost efficient ways to improve things that do not require the tricky analog calibration.

So the teams working on FD do impressive technology, but the road to a profitable business case is still to be proven. Good luck to them, although I'm skeptical (but would be glad to be proven wrong ;).