Thursday, July 8, 2010

Diversion post

Rediff article: Linky
LTE may score over WiMAX in broadband race

Let me comment on what I know of this business. I admit I know little other than the tech aspects (and that too little here and there), but eh thats enough to understand the business in a big picture (I think).

A Reliance Industries spokesperson declined to comment on the choice of the technology. However, people close to the development say while RIL is exploring both technologies, it is veering in favour of TD-LTE as a long-term technology.

There is a lot of ululation in the article on why WiMAX is an abandoned child. Fact is, I would nt cry if WiMAX kicked the bucket. WiMAX is a 2007-ish standard for a 2015-world. LTE is more like a 2010-ish standard for a 2018-world. Dont take the numbers too much to heart, but you get the picture. WiMAX has been around for ages, and Long-Term Evolution or LTE is the evolution of 3GPP (Third Generation Partnership Project) for higher data rates. There is one other cousin of LTE, LTE-A or LTE-Advanced which is like the steroids version of LTE, which is still in the works. WiMAX, when it came out, was a big boost over HSDPA and much of its progress over HSDPA was based on technological advances in MIMO (using multiple antennas at the transmitter and receiver side) to push higher data rates.

To explain the above, one needs to understand what MIMO is. In layman terms, higher data rate needs higher bandwidth. In contrast to rediff-esque use of the term "bandwidth", bandwidth is a well-established quantity in the communications arena. If you speak into your telephone mike, your voice stream occupies (a little around) 4kHz of bandwidth. Bandwidth is the frequency content of a random signal (and is the natural counterpart of time). In general, a short-burst in time has a huge frequency content and a long-winded slowly decaying signal in time has a small frequency content. Time and frequency are like the yin and yang of a random signal, both are there always and they present two different ways of looking at the same thing. In fact, you can look at things from a time-frequency viewpoint (what frequency content is present at what time) and this makes sense in many contexts, including radar.

So, till like the mid-90s, the only way to push higher data rate was to either make your signal duration long or frequency content large. The best way to utilize these resources was to compartmentalize the resource and allocate one user for every chunk (what is now called the time division multiple access TDMA and frequency division multiple access FDMA). Making your signal duration long (quite obviously!) means that you have to wait long enough (at the receiver) to get your data recovered. This is because almost every error correction coding scheme is built to process the data completely and not partially. With a "data"-stream (as in email, text, etc.), this delay is fine, but with voice, it gets jarring and people complain of terrible quality, sooner than later. If you want to increase your frequency content, folks such as RIL have to buy up bandwidth or spectrum via licensing. That is where FCC in the US and DoT in India make a huge profit. They sell sovereign rights over bandwidth in different localities to the highest bidder and reap a windfall. License is like right-of-way in the air (in the electromagnetic spectrum).

With this stage set in the mid-90s, someone came up with the idea of expanding the pie by bringing multiple antennas. Now there is a reason why this simple idea had to wait till the 90s. Using multiple antennas efficiently means the need for a better signal processing front-end, better analog and digital devices, better antennas, better energy efficiency, etc. While the mathematics of communications could have jumped right away (some optimists say so!), all these had to wait patiently for laggard technologies such as device electronics, RF, and EM folks to catch up (even today, the mathematics is far sophisticated to the point of being called mathematical masturbation, but these comm engineers usually dream things up in the air, the hard part lies in making things work with real devices, not fancy ideas). So in any case, by the 90s, the technologies had shown enough promise that MIMO technology spread like wild-fire. People could show that, with a trade-off of better signal processing (which means more battery power and more cost), higher data rates could be pumped in with multiple antennas. As the devices/ICs matured, we actually saw the economies of scale kick in and things actually became cheaper.

In the meanwhile, folks at Qualcomm (I dont know how true that is especially given that Hedy Lamarr, a Hollywood actress, had ideas about spread spectrum in the 50s, which she had patented) came up with the idea of CDMA. The basic idea behind CDMA is to use a random signature stream for a user to separate them over the same time and frequency. With this innovation, the data rates could be pumped up as more users can be served with some more signal processing. The rise to stardom of CDMA was because it could scale up over TDMA and FDMA, but also because of political and strategic reasons (the US has showed enough indications in communications history to be on its own). However, this isolation has caught up with the US now as other technological innovations such as OFDMA have taken over the sweet spot left behind by CDMA. Most of the LTE and WiMAX business is centered around OFDM and if someone said MIMO-CDMA, they would be laughed out of the place. In fact, in 2008 I heard a talk by Andrew Viterbi where he (finally!) acknowledged that CDMA could (sometimes :)) perform worse than OFDMA. Some say, first confession they heard, but yea nevertheless that is intellectual politics, neither lofty nor fun nor honest.

The other thing going for 3G (1G was voice-centric with your brick cell phones if someone had them in the 80s, 2G was voice- and data-centric with a thick cell phone slab/CDMA standard from the mid-90s, 3G was and is primarily data-centric with sleek little Razrs and what not) was that multiple antennas can be put JUST at one end. The idea is that you have multiple antennas at the base station and you can spread the cost over 100K users (or so), but let each user have only 1 antenna; so his/her device is still ultra-cheap. This boost was good enough for most of the earlier generation systems.

However, with WiMAX, the push was for the cell phone user to get multiple antennas so that innovations in circuitry can be put to use in the real market. This means higher cost for your cheap $0 cell phone, which is of course subsidized by you buying a better cell phone plan over a longer contract period, etc. Remember this is a cut-throat market and cell phone companies actually make little money with your contract as such. The big revenue sources are when you exceed the plan minutes (all those single folks in a Long-Distance Relationship will know ;)). The cost is usually 40c a minute and up, remember that is where the prime rib is.

So in any case, LTE is MIMO with multiple users. In WiMAX, it was pretty much MIMO but with a single user. In short, what that means is that the base station can and will send signals simultaneously to N people and you end up throwing away the rest of the gunk and decoding your piece alone. Needs better signal processing, but feasible with a cheap cell phone as they envision it. Of course, if you are security paranoid MU-MIMO (multi-user MIMO) comes up with some security features, etc. But the level of security is not like what you may get in a public key crypto-system. With some fairly innovative technology, which is what intel agencies have, you can tap and decode data for everyone and figure out strange patterns.

Ok, the "head-fake" (copyright: Randy Pausch) for this lame diversion is this. With all the data rates that cell phone companies are promising, and the cut-throatness of the business as is, we need killer apps. What the hell does a customer do with the data rate? Music/movies/internet on the go is ok, but if my case is any indication, I would nt use internets on the go unless if I necessarily had to. So where is the next innovation? How does one hook the consumers into being cell-phone or wireless addicts? I am sure RIL would be pleased to hear your ideas if you had any. I sure as heck would like to know where the consumer market is headed in the next ten, twenty, thirty years. It looks gloomy you see, sitting where I do. Hunting for killer apps and selling it in a creative way for money is an art that only certain people can do. I am hunting, not for selling, but for knowing which bredator is doing what :)... Please indulge.



Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home