So far the running for LTE-U (using LTE in unlicensed spectrum) has been made by Qualcomm, but this week both Huawei and NTT DoCoMo joined that battle in earnest, offering a single trial in the DoCoMo Beijing labs using Huawei technology.
First and foremost they claimed their system had greater throughput than WiFi, and we have found this tough to believe. Inherently WiFi is MORE efficient in that the latest current version uses a maximum Modulation and Coding Scheme (MCS) that goes up to 256 QAM, and uses very similar OFDM to organize separate carriers and provide multipath protection to LTE. LTE encodes using up to 64QAM, so where 256 QAM is being used directly against 64QAM, it goes 4 times the speed.
But to begin with Huawei is actually measuring previous generation 802.11N technology against LTE-U, which won’t be with us for at least 2 to 3 years. By then the world will have 802.11ac and most likely in MIMO versions of 4 x 4 antennas or even 8 antennas, while LTE will use just one. It will also be likely that it will use MU-MIMO, a multi-user version, which will also improve throughput.
Additionally WiFi already uses those extra antennas for diversity, and for beamsteering and for beamforming – technologies currently beyond LTE and there is considerable improvement even if the Access Point has 4 antennas and the device has one or two. When we asked, it turned out that 802.11N with 2 antennas was in fact tested against LTE-U, and Huawei said that it managed to deliver 1.6 times the speed in a 20 MHz channel in LTE.
Now that is possible, in that in LTE there is centralized intelligent selection of which devices to speak to, at what speed, using which modulating scheme, when the network can see ALL such devices. The trial was carried out with 4 devices, but there are no details as yet of how far they were away from the base station and what barriers where in the way. In WiFi, it is far more Self Organized based on a strict order of requests – so less sophisticated in its planning. However there are some specialist Firmware upgrades which can be added to WiFi, including Mesh architectures, that make its behavior far more predictable, offering higher throughput.
But even if we accept that LTE-U managed to get 1.6 times the data out of unlicensed spectrum than WiFi N, that makes it less than half the speed of AC which is already shipping in volume. There are other issues, but a three year run at this problem may allow LTE to patch up its obvious weaknesses. For instance no LTE chips have Dynamic Frequency Selection to avoid interference in the way WiFi does and that is mandated in the spectrum standards, and will need to be added. This can hardly be beyond Qualcomm (or Huawei) because it runs a WiFi chip company in Atheros.
Another key element is that paid for spectrum has very different power limits (the amount of radiated power that each transmitter is allowed to have) from WiFi. This is WHY WiFi has developed the helping hands of beamforming etc… because it is confined to power constrained spectrum. And no LTE base station built so far has ever contemplated being power constrained. Huawei admits that so far the chips do not avoid interfering with WiFi, but did not mention what power level they worked at.
This is interesting because if it was higher power than WiFi then the comparison is not a fair one, and it could have been because the spectrum it used it said was around 5.735 GHz. Now this is a 20 MHz slice made up of 10 MHz inside and 10 MHz outside the legal WiFi channels in China, so there should be no WiFi chip optimized for this spectrum, which makes the trial look a bit dodgy.
But even if the test was fair, such a system still has to be built, and it still has to be shipped and then selected to go into devices – which will take all of three years to happen and cost a lot of money – and where is this money going to come from? It will cost a lot more than equivalent WiFi chips, which are already promised vast shipment numbers and which take with that the economies of scale of chips produced in their billions. Also WiFi does not need a system of centralized control and billing.
What Huawei is talking about doing for DoCoMo it is calling Licensed Assisted Access (LAA). This means that where LTE is running out of steam in an area of cellular congestion, some of the communication can be shipped off dynamically to Unlicensed spectrum, relieving the congestion. When the cellular space is no longer congested, it can leave the unlicensed spectrum alone and let WiFi have it back.
There are a lot of issues here. Most congested areas will already have 80 MHz or 160 MHz AC WiFi operating in them already when the cellular operator wants to “borrow” this Unlicensed spectrum. And when it tries a DFS call to see if there is a spare channel, it will find there is none. What will it do then? Probably tell a device which is hogging 160 MHz to drop 20 MHz so that it can have some. It depends how its detect and avoid mechanism is designed so that’s all in the future, and also how the WiFi backs off.
But even if an LTE-U base station manages to grab a piece of spectrum in the congested region, it will need a separate and extra base station built and backhauled and powered, in order to use this spectrum. It is unlikely that it can do this using existing kit and Network Functions Virtualization (NFV) to emulate WiFi, because it is already using all the capacity that it has out there. And if a cellular operator does grab 20 MHz once, then it will end up “keeping it” because busy sectors only get busier, so this will eventually deny WiFi capacity, possibly it would halve capacity wherever it was instaled.
But because most 4G is in 700MHz or 2.5GHz bands (definitely so in Japan) the size of the cells there with better propagation, and more radiated power, will be huge – tens of kilometers for 700 and perhaps 2 or 3 KM for 2.5 GHz. But for 5 GHz, even outdoor (which allows a little more power and has less interference), it is still a lot less than LTE, and 5GHz propagates far less. The cell size would be the size of a WiFi cell – about 300 meters radius at best and not go through more than one wall.
So in order to end congestion in the cell, you would need lots and lots of these new base stations. And it would be MUCH cheaper in LTE to create what are called Capacity cells, that re-use the same spectrum you have already paid for, but with their power turned down a notch to create “optimally” tuned local pico cells. Also there would be no need for an extra radio in the phone.
That would be cheaper and better. And at the same time it would not interfere with WiFi or reduce its potential capacity.
Another really key point is that the Unlicensed LTE cell may only be used for a single cellular network. If it is, then it deprives all the customers of other networks. It is possible that there would be RAN sharing of these Unlicensed nodes in the network, but that adds more complication and expense in the network build and potentially adds roaming charges.
So while two LTE players may decide they can both attach to such an Unlicensed WiFi cell, what about a tablet customer that only has WiFi – he can only use that capacity that is reserved solely for WiFi, if there is any left at this point.
The truth is that usage would be confined only to customers who pay cellular operators, creating the situation where consumers “could have had WiFi” often free WiFi, and instead used up their cellular data caps at a far more expensive cost for the data. If consumers could choose they would “ban” cellular operators from using LTE-Unlicensed or LAA, especially as this is all possible to do in WiFi, and Hetnet designs exist that support that strategy for cellular operators, more cheaply adding to the vast pool of WiFi resource there is around the planet, and simply offloading to it, with much the same result, but far less cost to both the operator and the consumer.
It is time that the cellular operators admitted that they are going down this expensive route purely to disadvantage WiFi, so that they can exact more and more cash from consumers, and fight the effect of a consumer population that has begun to question how expensive (and given that over 80% of all cellular device usage is on WiFi, useful) cellular alternatives can continue to be.
We think that eventually cellular operators will be forced by weight of economics to embrace WiFi Offload and forget about LTE-U or LAA.