Mixed Signals
Any sufficiently advanced information is indistinguishable from noise
Monday 22 October 2007, 6:39 PM
60 GHz networking, the story so far
Some faffery is going on over IBM's announcement of a deal with MediaTek to produce 60 GHz wireless chips that can squeeze out data at 2.5 Gbs sometime in the next two to three years.
But that's all it is - a co-development deal. There's no new technology mentioned here over and above that which IBM talked about more than eighteen months ago, and nothing substantial on power requirements, range or modulation techniques.
60 GHz is extremely short range: it's never going to be a direct competitor for any of the 802.11 wi-fi standards - a fact reflected by the IEEE's standardisation group devoted to the new standard, 802.15.3c. That's in the same family as Bluetooth, Zigbee and ultrawideband... and it's funny that nobody's mentioning UWB, which is far further advanced and has pretty well identical speed and range potential. It should, moreover, be lower power.
It's noticable that IBM is referencing Wi-Fi in its announcement, when that isn't a competitor, and not UWB, which is. That's usually a sign that a company is hoping to keep the debate going on its terms - which, so far, seems to be happening.
Nevertheless, there's a lot of work going on in 60 GHz - and unsurprisingly so. There's 7GHz of bandwidth available between 57 and 64 GHz for unlicenced devices, and companies such as SiBeam are gearing up to get there first. SiBeam's schtick is that it can make its chips out of ordinary silicon using reasonably normal CMOS techniques (guess who's hoping to be snapped up by Intel), while IBM is working with silicon-germanium. There are other companies doing far whackier things.
But even with the mainstream engineering, expect the unexpected. One radio engineer i talked to pointed out that, at these extremely high frequencies, you can counter the much greater loss of signal by making very small yet very high gain antennas. One an inch across could easily add 25dB of oomph, and beam the resultant very tightly focussed signal in just the right direction to hit the distant receiver -- or bounce it off a wall to get past an obstacle.
Which sounds like good news, until you do the sums. If you put a mere 500 milliwatts into a 25dB gain antenna, you get an effective radiated power of nearly 160 watts. That's quite a lot of microwave to have zooming around your front room, chasing the cat.
PreviousComments on this post
Greetings from the world of “wackier” wireless technologies!
Thanks for your post. It’s always good to see some attention paid to the emerging world of 60 GHz.
A couple of things to clarify with regard to your posting.
While it is true that small, high gain antennas are indeed a benefit of working in the mm-wave regime, the FCC (in the States) has stepped in to mandate that power at the transmit antenna for 60 GHz be limited to 10mW. (No cats will be cooked in the name of wireless HDMI!)
I concur with you notion that antennas are a big deal for 60 GHz radios. One of the areas where we engineers at Phiar (pronounced “fire”) differ from many others is our contention that these radios must be as simple as possible to come in at consumer-friendly prices. We advocate “wasting” some bandwidth by choosing battery-friendly OOK or ASK modulation schemes rather than power-hungry OFDM – this saves money, complexity and power.
I agree that IBM’s choice to compare 60 GHz to WiFi rather than UWB is a little off. What you missed, however, is the fact that mm-wave radios will blow past UWB in terms of bit-rate. (Even if the comparison is less breathtaking than 60 vs. WiFi.) Given Phiar’s “inefficient” choice of 1 bit / Hz modulation, a theoretical 7 Gbps of throughput will result. I am not aware of any in the UWB camp who can make such claims.
Also, think about UWB’s spectrum: Just about everything in the existing wireless world operates between DC and 10 GHz, and they have to operate their radios damn near the noise floor to avoid interfering with incumbents. At 60 GHz, on the other hand, there are far fewer interferers. 60 is also an oxygen absorption peak in the atmosphere. This allows a 60 GHz radio in the living room to operate without even detecting the system in the office – the spectral reuse is that good.
Finally, let’s talk about whacky. Metal-insulator electronics are increasingly seen as a viable alternative to semiconductors for mm-wave analog applications. These devices have better performance at 60 GHz than GaAs Schottky diodes, but that cost less to manufacture than 2 GHz CMOS.
If it’s whacky to want higher performance for less money in an analog technology, then I graciously accept that moniker!
--- Adam Rentschler
Aha! Thanks for replying...
I stand by whacky, for three reasons.
First, it's 60GHz. Not so long ago, peasants would take their pitchforks and flaming brands, and drown as a witch anyone who claimed to make things work at 60 GHz. Rightly so.
Second: quantum tunnelling. This is deeply whacky all by itself. I know we've had Esaki diodes for a while now, and that it is perfectly respectable in certain circles to entertain the idea that in this libertarian age a particle can manifest anywhere within its probability function it jolly well chooses, but I think it's impolite at best just to appear on the wrong side of a potential barrier without so much as a note through the door first. Deeply suspect.
Third: “In fact, Rentschler maintains that Phiar devices can be fabricated on almost any substrate. “Tree bark is probably a really bad choice,” he says, “but we can pretty much deal with anything.”. You're the man talking about making treebark resonators at 60GHz using tunnelling – m'sieur, the epithet is justly deserved.
Now, onto some meatier issues. UWB may or may not be able to aspire to the full theoretical speed of 60 GHz, but let's be careful about our definitions. UWB is a modulation method, 60GHz a band. If we're talking about the practical side of things, then the question is whether when 60GHz turns up (and it it won't be sporting the full 7Gbps in issue one, I bet), will UWB be competitive in power, speed and price? I think the chances are, it will. And what's to stop UWB overlaying that 7GHz of spectrum (or, indeed, the full DC to daylight monty) too?
In the end, bandwidth is bandwidth, and I think we're getting good enough at modulation methods to expect to make the theoretical maximal use of any bit of spectrum we can get. At that point, devices that can operate cheaply and efficiently at 60GHz will have whatever part to play that this sort of spectrum is best equipped to offer – and, as you say, absorption and other issues mean it should be exceptionally well suited to very high bandwidth very high re-use purposes.
As for antennas – 10mW EIRP is certainly going let let Tiddles rest easy. That's only, what (watt?), 50 microwatts from the final amplifier?
I don't doubt that metal insulator devices are capable of great things. The question is whether they're so much better than the alternatives that they can overcome the considerable inertia and commercial advantage that more mainstream techniques have. History doesn't offer much comfort to ideas that are only quite a lot better than the incumbents they strive to displace.
