Original URL: https://www.theregister.com/2014/07/23/5g_between_hypegasm_and_hohum_the_research_reality/

All those new '5G standards'? Here's the science they rely on

Radio professor tells us how wireless will get faster in the real world

By Richard Chirgwin

Posted in Science, 23rd July 2014 07:02 GMT

The 5G arms race has commenced, but beneath the duelling “my 5G is faster than your 5G” demos, there's serious work going on – and whatever the future of 5G, that work will change the future of mobility one way or the other.

With that in mind, The Register spoke to Professor Eryk Dutkiewicz of Macquarie University. In May, Macquarie (with professor Dutkiewicz at the helm) was tapped as one of number of universities, all but three of them in America, to form the backbone of Intel's 5G development efforts.

Macquarie's brief – to work on how cognitive radio plays into the embryonic 5G future – puts the research group near the centre of many of the big questions surrounding the future of mobile comms: from the spectrum crunch to antenna technology, from the silicon to the batteries.

Spectrum re-use

The accepted wisdom the world over is that there's a two-pronged attack on the spectrum available for mobile communications: the number of users is exploding, and new technologies devote more radio spectrum to each user to increase throughput.

There are two dominant ways to re-use a slice of radio spectrum, both in the spatial domain, and both of interest to Dutkiewicz's group: one is to make cells smaller (so that a given transmission frequency of, say, 2.3 GHz can be re-deployed nearby without interference); the other is to use MIMO (multiple in, multiple out) antenna technologies.

However, as Dutkiewicz explained, neither of these are as simple as they seem.

It's obvious that if your base station has a coverage footprint of 30km or so (common in the days of analogue mobile technologies and still common on highways), you need a significant amount of separation between masts before a frequency can be reused by another base station without interference.

If your base station's range is only in the hundreds of metres – or tens of metres in a world of nanocells and “hetnets” – a given transmission frequency could easily find itself in use in dozens or hundreds of base stations in relatively near proximity.

If that can be done without the next mobile standard requiring a new air interface, carriers will be very happy indeed: having gone through the cycle of AMPS-to-digital, then GSM to CDMA, then CDMA to OFDM, operators would be very happy if “they don't have to throw out what they have right now”.

While operators might feel constrained not to say so publicly, the idea of deploying their 4G networks today and throwing them away tomorrow is, Dutkiewicz says, something they're afraid of.

However, the proliferation of base stations creates “very complex management issues and interference issues that need to be solved”, he said.

MIMO, on the other hand, creates its own – and completely different – set of issues.

The value of MIMO is twofold. The earliest research focussed on pure spatial multiplexing – the same signal following different paths (therefore having a different travel time), meaning different data can be encoded onto each spatial path.

In this use-case the amount of multiplexing is limited mostly by the number of antennas at the receiver, since the receiver is a more constrained physical environment than the transmitter.

The other “trick” MIMO offers is beam-forming – using the way that signals cancel each other or reinforce each other to shape different beams towards different clients of the same base station.

Both MIMO applications, Dutkiewicz explains, bump into the same basic problem: you can only optimise spectrum you have the right to use.

“The problem is that now we're looking at the spectrum map, and we're running out. Spectrum's a limited resource and it is not possible to just say 'give me some more', because there isn't.

“And of course governments and military people have grabbed chunks for themselves, and it's very expensive. Some of it is very precious because the signal propagation is much nicer for some frequencies, like around 800 to the gigahertz range.”

Spectrum sharing

Which leads to the discussion of spectrum sharing, something The Register has discussed in the past.

Dutkiewicz's interest in spectrum sharing is to research the cognitive radio techniques that will get the world past the slow and relatively inflexible arrangements that exist today.

Right now, he said, sharing is done on a long timescale: the incumbent leases out the spectrum for “six months, or maybe twelve months, so that the operator has the certainty of having the spectrum for a reasonable length of time.”

That makes spectrum sharing pretty static: it's based on database lookups.

“That's the beginning. What we'd like to be able to do is shorten the timescale as much as possible, so the secondary user can get access … on a very short timescale of seconds, or even milliseconds. That's the ultimate goal of what's called cognitive radio.“

That means the equipment – the handsets and the base stations – need to be able to tell what's going on at any given moment, and that has to be represented in standards.

His group is working on “how to do resource management at the algorithmic level”, Dutkiewicz said. “For that, we need to produce RF maps very accurately and with very little overhead, because before you decide what spectrum to use where, you have to know what's going on.”

That's more complex than it seems at first sight. For example, Dutkiewicz explained, a mobile carrier's base station might be too distant to a primary user's transmitter or receiver to detect it, and might grab the spectrum and cause interference to the primary user.

Hence the reason that handsets might have to be involved in the cognitive radio environment – the handset might see the primary user's interference and report that back to the base station. The base station would then negotiate a different frequency with the handset.

“So we are also looking at putting appropriate hardware interfaces on the mobile handset – how to make the RF front-end more dynamic in terms of looking for spectrum, and responding to changes in spectrum.”

The opposite is also true: the cognitive radio system has to avoid reporting false positives. Otherwise, vacant spectrum might be considered “in use” and unnecessarily left unused.

And all of this needs reasonably high resolution in multiple dimensions: the radio spectrum is one dimension, the dimensions of the spectrum mapping, and the time dimension.

“We need to know a reasonable spatial resolution of the channel state. At the moment we don't have that. LTE is doing rudimentary channels state measurements, but if you move 10 metres to the right, it changes.

“Some people in the literature suggest putting an infrastructure in, consisting of a sensor network, continuously take the information and send it to the base station. Make the base station the central entity that knows what's going on.”

Dutkiewicz suspects that the end goal of a fully automatic, highly responsive cognitive radio system will probably “not be quite met in time for 5G.” The year 2020 is too close: at best, he said, components of this will be ready in time.

Antenna technologies

Antennas are also big in the research. After all, take a look at what's coming: massive MIMO, millimetre-wave technologies, and handsets that might have to deal with base stations communicating at a bunch of different wavelengths.

If your handset looks way too small for more than a couple of antennas, that's probably because you're thinking of handsets operating at (relatively) low frequencies. At 800 MHz, the wavelength is 37 centimetres, and a 1/4-wave antenna is still 9.25 cm long.

At 20 GHz, a quarter wave is a much more tolerable 3.75 mm, and at the 60 GHz favoured by WiGig and other high-frequency standards, a quarter-wave is in the vicinity of 1 mm.

That's physically feasible in quite a small form factor, even if the short propagation range at that frequency means it would be restricted to handoffs to micro-cells.

Interestingly, even at such very close separation between antennas, Dutkiewicz said MIMO techniques still work – with Australian research house CSIRO's Ngara system operating on-campus, there was a chance to test this.

“One of our undergraduate students had a project to measure the impact of antenna separation of MIMO performance at millimetre wavelengths. She found that the impact was not very significant."

“If you're unlucky and make the separation one wavelength, you will be destructively interfering.”

But as long as antennas are separated enough to avoid that, MIMO works even with closely-spaced antennas.

The other problem that 5G poses is that antenna topology is, of course, different according to wavelength. While it's early days, Dutkiewicz told The Register that this is another research priority in the lead-up to 5G: wideband, intelligent, self-configuring antennas.

There's no guarantee that the spectrum your mooted 5G handset needs to use is contiguous (one 20 MHz chunk might be at in the 800 MHz band and the other at the 2.3 GHz band, for an extreme example).

“We would like antennas that can reconfigured themselves fast enough to respond to changes in the primary spectrum – if the channel disappears, the system has to find another channel quickly enough that the user doesn't know it's happening. The antenna and RF front-end have to make that seamless”.

Power consumption

By now, alert readers will have seen the other croc in the duckpond: power consumption. Take another look at the checklist:

“You don't want to have your handset burning in your pocket,” Dutkiewicz said – not to mention that nobody wants smartphones with a battery life of ten minutes.

So there will be lots and lots of patents to come out of researchers working out how to do all of this – without needing to attach phones to brick-sized batteries – before 5G becomes the all-gigabit all-the-time reality that engineers are dreaming about. ®

Bootnote: The author thanks Professor Dutkiewicz for his extensive and detailed briefing. Any errors are The Register's own. ®