We have to bring your attention to yet another anti-LTE-U blog or series of blogs from Joey Padden, the lead architect at CableLabs. You can see it here http://www.cablelabs.com/wi-fi-vs-eu-lbt-houston-we-have-a-problem/
He in particular attacks Licensed Assisted Access, the new idea for a service which should jump between LTE in licensed and LTE-U in unlicensed spectrum and he focusses on what we have said from the very start – that it is unlikely that LTE in 5 GHz will leave well alone when it comes to existing WiFi installations.
At Faultline we have gone on record suggesting that LTE-U is nothing more than deliberate interference for WiFi and coming at it with a) Perhaps some pro-cable bias and b) in a far more technical analysis, Padden concludes that we are right.
His argument is based on just how these new chips are planning to avoid interference with WiFi, as there are a number of methods and different regulators in different parts of the world insist on different levels of avoidance.
There are three broad coexistence methods in contributions at 3GPP so far; channel selection, some form of duty cycling the LTE signal, and Listen-Before-Talk. Selecting and sitting in a channel is hardly going to prevent interference, but he says that Listen Before Talk (LBT) is widely thought to be the best way of ensuring avoidance, except that there are many ways of achieving this.
WiFi has its own versions of this in Distributed Coordination Function (DCF), or Enhanced Distributed Channel Access (EDCA) depending upon which WiFi you are using.
Recent 3GPP contributions show that companies are mostly pushing a system called Load Base Equipment or LBE, which uses a different back off process – the back off is the period of time an access point waits after it senses a collision. The system the 3GPP wants to use requires multiple back off points 20 microseconds part. WiFi uses a 9 microsecond back off period and a system called exponential back off, where more time is left between successive collisions.
So Padden created a simulation of what might happen if these two systems were made to work together, and produced some quite alarming results which suggest that WiFi would lose 77% of its capacity in this scenario where both networks had 15 devices on them, and when it was operated with 20 devices per network, it would show an 88% decrease in transmission success. When between 1 and 4 devices were used both networks were fine.
Padden suggests that if you don’t tune the system right with the right back off periods, then performance could be even worse.
Now if cellular goes ahead and tries to continue with the back off process that it appears to be ready to adopt, it will operate fine when the spectrum is in cellular and WiFi will almost stop operating. And given that most environments do not have 15 devices attached now, but will have in the future – if cellular goes ahead with this unsophisticated back off process instead of an exponential variant, the problem will only be found out after these LTE-U services are installed – by which time it will be too late. We expect this debate to run and run.