JOST A MON

In the good old days when digital wireless telephony was in its infancy, there were several contending technologies slugging it out for supremacy in the US. There were no standards, and the policy of the regulators was to allow the market-dominant technology, when one such emerged, to become the de facto standard.

This took several years and a lot of money was spent, contributing both to my ever-fattening pocket and the stock market, and then the big bust and the associated flattening of my wallet.

Meanwhile, the Europeans, in their infinite wisdom, had established the GSM system by fiat. Rigorously enforced by publicly available standards, it galloped its way to dominance across the world.

Countries such as India and China followed the US in allowing multiple competing technologies. But after all the dust settled, two techniques remained standing: CDMA and GSM.

It was all very exciting in those early days. These were the 1990s and the mobile world was dominated by Motorola. The Nokias and Ericssons and Siemenses of the world were only coming into their own in network infrastructure; the Japanese and the Koreans with their nifty and miniature designs of cell-phones were not too visible in the global market; there were multiple behemoths in the US alone (Hughes, Nortel, Lucent) fighting to corner the market.

At the time, a little example was used to explain the competing technologies. The old analog method was, in effect, a two-way FM radio system. Each mobile - in conversation - would be assigned a frequency pair to transmit and receive on. One could liken this to a room of people each speaking at a different pitch. You would be able to follow any conversation by just tuning your ear to a particular frequency. Humans can do this rather well, especially at higher pitches. Try it in a room of men and women and kids talking. You should have no problem zoning into the various speakers just by virtue of their pitch of delivery.

Of course, this was immensely wasteful. For one thing, there are large pauses in conversation when one person listens and the other speaks. During this time, the frequency allocated to the listener's speech is inactive, and could be well used for another conversation. In fact, there are small pauses in the midst of active speech as well, such as when one catches one's breath, or hems and haws. While these are brief moments in a human's perception, for computers doing the allocation of channels, these are sufficiently long that they can switch the use of the frequency to another speaker. Unfortunately, signalling the activation and deactivation of channels is not very efficient in an analog domain, and so this brief enhancement was not developed further.

As the capacities of the analog networks began to fill up, regulators demanded a more efficient use of the spectrum allotted to the mobile providers. The only way forward was to go digital. The first attempt was called TDMA (Time-Division Multiple Access), which can be illustrated thus: in the same room with multiple speakers, assign to each speaker a frequency pair (as in the analog case), but allow them to speak only for a fixed period in time, pause for a fixed period, speak again, pause, and so on. Now, a listener only needed to know the 'active' periods, the time-slots during which to listen in. At other times, he would tune out. In this fashion, TDMA allowed three times as many speakers on the same bandwidth as the old analog system (by creating six time-slots on each frequency, and assigning one pair for each speaker, fixed for the duration of the phone call).

Hughes went one step ahead with their Digital Speech Interpolation technique: it sensed the silent periods and released the allotted time-slot, which could then be assigned to another conversation that entered the active phase. I was working at the company at the time, and their great idea was to develop a large unit from which 96 telephone lines could be drawn; the unit - placed on top of a block of flats, and communicating wirelessly with the phone network - would then provide a rapid deployment of a telephony service in countries where these were not so readily available. As an alternative to digging up roads to lay copper cable and wiring up huge sections of a town, this was superb. (And that's how I got involved with a large-scale installation of the technology in Jakarta in 1995. But that's another story.)

The problem with the method was that the slot release-and-assignment cycle took a bit of time, and resulted in delays in the voice path. Behavioural studies have shown that a human conversation can handle loss or degradation in the sound (the ear-to-brain pathway is very robust) but it can't tolerate delays in the speech. You may recall from the old days of satellite-based trunk calls (if you were around at the time) how so much was time was spent repeating yourself while your voice travelled into space and back. Exactly the same flaw dished the Hughes system, and after the Jakarta installation, I don't think there was much takeup of the idea.

An alternative to TDMA was Code Division Multiple Access, possibly the most elegant technique in communications that I have ever come across. CDMA worked as follows: imagine that, in that same room of people, the conversations are happening simultaneously (unlike TDMA, where there would be stops and starts as people awaited their allotted time-slot to converse), but in different languages. Now all a listener needed to do was to tune into a particular language, and as long as nobody was speaking too loudly, it would be easy to isolate a particular conversation amidst all the background noise. In CDMA, also known as spread-spectrum, there was little to distinguish an active call from the background chatter; only an observer in exact coherence with the code governing the call would be able to distinguish it from white noise. The sensitive power-control required to prevent mobiles from swamping each other by 'shouting' was one of the trickiest aspects of CDMA to master. As they moved around, the mobiles might lose or regain line-of-sight to the base station, resulting in large variations of received power.

Now, spread-spectrum can be achieved either as above with codes, or with high-speed frequency hopping. Once the receiver and transmitter are synchronised and the sequence of frequencies to hop are known, they can remain in sync throughout the conversation. The constant jumping across the spectrum results in a wash of low-power radio almost indistinguishable from background noise, just as in the case of codes. The full implementation of this method had to wait for the development of fast signal processing chips and powerful processors. But the original idea was developed in 1937, and by the unlikeliest figures imaginable.

This is surely one of the most romantic tales in the sciences. Fleeing from the Nazis and her Fascist arms dealer of a husband, Hedy Lamarr emigrated from Austria to the US. She was already well known as a stunningly beautiful and talented actress; what was less well known was her sharp mind, and the fact that she had listened very carefully to her ex-husband's colleagues as they discussed armament and technology. At the time, researchers had been struggling to develop a method to prevent remote-controlled torpedoes from being jammed by enemy radio signals. Lamarr realised that by sending control signals to the torpedo by hopping across a pre-defined set of frequencies - unknown to any interceptor, but with the transmitter and receiver in perfect sync - the torpedo could be effectively immunised against hostile jamming. Together with the pianist George Antheil, she developed a hopping technique based on piano rolls, and applied for a patent, which was granted in 1942.

As it was a military secret, its details were not published for years. 17 years later, it expired, but it entered the public domain only in the 1980s. Antheil died in 1959, and Lamarr in 2000; they never received any monies for the industrialisation of their patent. But in 1997, the Electronic Frontier Foundation honoured Hedy Lamarr for her pioneering work, and a sin of omission was finally atoned for.