TCS Daily


Told You So...

By Steven Den Beste - October 21, 2002 12:00 AM

I used to work for Qualcomm designing cell phones. Qualcomm is the company that invented the CDMA cell phone and made it practical and made it into a market success. It is now the leading system in the American market.

And right now I'm basking in the evil glow of a major case of schadenfreude.

But allow me to explain why. And bear with me: this article has a lot of technical jargon. But it's worth it if you plow through to understand how flexibility and ingenuity in a competitive system will trump planning and central control any time.

Cell Hell

The original cell phones were analog, using fairly straightforward FM channels for voice communication (technically, Frequency Division Multiple Access or FDMA). When your phone was in a call, it was granted an FM channel by the cell and used it exclusively until the call ended. FM encoding used spectrum extremely inefficiently. Spectrum (RF bandwidth licensed from a governmental authority) was scarce and expensive, and it rapidly became clear that FM wasn't able to handle the traffic that was necessary to make cellular telephony a profitable business through economy of scale. Though the system as a whole was known as the Advanced Mobile Phone System (AMPS), it became obvious that it wasn't very advanced..

One obvious approach to this problem was to digitally encode the voice traffic and compress it. This approach led to the first Time Division Multiple Access (TDMA) digital systems, which take a single channel and timeshare it among several phones. Each phone digitizes and compresses its voice traffic and transceives it during its assigned timeslice. In the first TDMA-based cell phone system deployed in the US (which was controlled by TIA/EIA standard IS-136), a 30 KHz (times 2) channel which had carried only one voice call with AMPS could now carry three digitized calls, making it a nice upgrade.

GSM Gloaters

Meanwhile, a consortium of European companies designed the Global System for Mobile communications (GSM) which abandoned the old channel size entirely. They allocated 200 KHz channels (times 2) and divided them into 8 slices, giving each phone somewhat less than 25 KHz effective bandwidth (each direction).

"FDMA" and "TDMA" are examples of air interfaces. "AMPS", "IS-136" and "GSM" are full cellular protocol stacks, which include specifications for the air interfaces they use but contain much more besides, most of which is not directly related to the air interface. If GSM is a Corvette, TDMA would be the tires mounted on it. The tires carry the car, and ultimately do set limits on performance, but the tires don't have anything to do with the stereo in the car, or how comfortable the bucket seats are.

GSM was clearly superior to both IS-136 and IDEN (a Motorola design which Moto refused to license to anyone else, killing its commercial viability). As one of its first significant actions, the European Commission decided that the Europeans had designed the ultimate digital cellular system, and they passed laws making it illegal to deploy anything except GSM, whose primary supporters/suppliers were Nokia, Ericsson, Siemens and Alcatel.

Meanwhile, the FCC decided that it would not mandate any industry standard. It granted licenses for spectrum but permitted the licensee to choose whatever equipment and standard it wanted. (Within limits: there were certain certification standards required by the FCC to guarantee safety and to avoid interference between neighboring systems.)

And all through the 1990's, everyone in the U.S. cell phone industry put up with constant ragging from Europeans about the evident virtues of GSM and the equally evident virtues of a government mandated standard. While in the U.S. you had what seemed at the time to be utter chaos, with a huge number of small companies using a bewildering array of different standards which were mutually incompatible, in Europe anyone could carry their phone almost anywhere in the continent, and if they couldn't use it they could move their SIMM into a local phone and use that.

U.S. Emerges

Of course, that apparent chaos in the U.S. was only a temporary phenomenon, and I think maybe the FCC and the rest of the government knew it would be. There's always shakeout and consolidation in a developing market environment. But this government policy of keeping its hands off meant that industry players were given broad ability to experiment. And within that environment, early in the 1990's, the founders of my former employer Qualcomm began to work on a radically different kind of cell phone air interface called Code Division Multiple Access, or CDMA.

CDMA is a form of Direct Sequence Spread Spectrum (DSS) and is ultimately based on an idea patented in the 1940's by Hedy Lamarr. (Who says beauty and brains don't fit in the same package?) CDMA is radical in many ways. By far the most obvious radical shift is that all the phones in the system and all the cells in the system operate simultaneously on the same carrier frequency. They don't "take turns" because they don't need to. Qualcomm and other companies created a cellular phone specification based on CDMA, which was eventually published as IS-95 under the auspices of the EIA, TIA and JEDEC.

In fact, Qualcomm's CDMA proposal was so revolutionary that when it was first discussed, many thought it couldn't be made to work. At least one European company deeply involved with GSM, Ericsson, went through the three classic stages of Not Invented Here syndrome:

1. It's impossible.
2. It's infeasible.
3. Actually, we thought of it first.

When I worked for Qualcomm, I had to soft pedal how I felt about this. Now I'm no longer associated with the company, and I can indulge in a rant.

At first, the most vocal top brains at Ericsson tried to claim that CDMA violated Information Theory, Claude Shannon's vitally important contribution to nearly everything in electronics. In the IS-95 implementation of CDMA, a single carrier frequency has a bandwidth of 1.2288 MHz, each direction, and up to 40 cell phones in a given sector can all be transmitting chips at that rate on the same carrier frequency, which seemed on first examination to be a case of trying to fit fifty million bits into a one-and-a-quarter MHz sack, which would indeed violate Shannon. The mistake they made was that chips aren't "information" based on Shannon's definition. Each chip contains only a small fraction of a bit of true information (which is why they're called "chips") and though those phones were sending chips at an immense rate, the actual bit-rate per second was either 9600 or 14,400 (depending on the voice compression codec being used).

Unfortunately for Ericsson's claims of physical impossibility, Qualcomm did a field test in New York City where several prototype phones mounted in vans were able to operate at once on the same frequency while communicating with multiple cells all of which also operated on the same frequency. (And they proved that soft handoff worked, too, by driving around during calls.)

OK, Ericsson's voices said, maybe it does work, but it's going to be too expensive. Everyone knew that the electronics required to make CDMA work was a lot more complicated than what TDMA needed, and Ericsson's loud voices claimed that it could never be reduced in price enough to make it competitive.

Only problem was that Qualcomm proved that wrong, too, by beginning to produce both infrastructure and phones at competitive prices. (Qualcomm did this to bootstrap the industry. It's no longer in either business.) Both were more expensive than competing systems, but not enough to make it commercially uncompetitive, and with rising volume and increasing experience, as well as continued advances in the IC industry, the prices are now pretty close to the same.

At which point Ericsson suddenly decided that it had applicable patents and took Qualcomm to court. Over the long drawn out process of litigation, every important preliminary court judgment went in favor of Qualcomm, and after most of Ericsson's infringement claims were summarily dismissed it became obvious that Ericsson didn't have a case, and equally obvious that Qualcomm wasn't going to be intimidated. Ultimately, the entire case was settled in a massive omnibus agreement where Ericsson became the last of the large companies in the industry to license Qualcomm's patents (on the same royalty terms as everyone else) while taking a large money-losing division off Qualcomm's hands and assuming all the liabilities associated with it, and granting Qualcomm a full license for GSM technology. The industry consensus was that this was a fullscale surrender by Ericsson.

Nokia wasn't as foolish and had licensed several years before. (Just in passing, the people at Ericsson who have my contempt are in the front office. Their engineers are as good as anyone else's.)

Order Out of Chaos

In the years of apparent chaos in the U.S., when loud voices in Europe proclaimed the clear advantage of a single continental standard, order began to appear. Small companies using compatible standards set up roaming agreements, and then started merging into larger companies, which merged into yet larger ones. One company (Sprint) started from scratch to build nationwide coverage using the 1900 MHz version of IS-95 (ANSI J-STD-008, later merged into IS-95). Bell Atlantic Mobile acquired GTE Mobile (who had been a joint partner in PrimeCo), and eventually merged with Airtouch (itself the result of several mergers) to form Verizon, almost all of which was based on IS-95. The last major nationwide system to form was Cingular, after the various GSM carriers in the U.S. realized they were in big trouble competing against Verizon and Sprint and AT&T (which uses IS-136). Cingular doesn't actually have a single standard; some parts use GSM, some use other things.

Once the existence and commercial feasibility of the CDMA air interface were established beyond doubt by its use in IS-95, its advantages became clear. The CDMA air interface was obviously drastically superior to any kind of TDMA. For one thing, in any cellular system that had three or more cells, CDMA could carry far more traffic within a given allocation of spectrum than any form of TDMA. (Depending on the physical circumstances, it's somewhere between three and five times as much. It's difficult to quantify because it depends on terrain, and also because CDMA has "soft capacity" .) Another advantage was that IS-95's implementation of CDMA was designed from the very beginning to dynamically allocate spectrum.

In TDMA, a given phone in a given voice call is allocated a certain fixed amount of bandwidth whether it needs it or not. In IS-136 that's a bit less than 10 KHz each direction, and in GSM it's somewhat less than 25 KHz. But human conversation doesn't use bandwidth that way; when you're talking, I'm mostly listening (maybe with an occasional "uh-huh"). So your 25 KHz channel to me is carrying your voice, and my 25 KHz channel to you is carrying the sound of me listening to you silently.

In any protocol that uses a CDMA air interface, the bandwidth used by a given phone in a call changes as needed. In IS-95, for instance, it changes 50 times per second, and can vary by a factor of 8:1, and the phone uses as much or as little as it needs for each 20 millisecond frame. When I'm silent, I only use 1/8th of the peak bandwidth I use when I'm talking. While that's very useful for voice, it's essential for data, which tends to be extremely bursty.

Moreover, IS-95 had the ability for different phones to be given different overall allocations of bandwidth, because the initial standard included both 8K and 13K codecs (which respectively use 9600 bps and 14,400 bps). The CDMA air interface does not have any hard bandwidth cap for any given channel (up to the total bandwidth of the carrier). So when higher data rates were desired, it was possible to upgrade the hardware in IS-95 cells and create new cell phones which could transmit 64 kilobits per second (and later 153 kilobits per second), while operating along side existing handsets which had not been upgraded, sharing the same carrier frequencies. So the same spectrum can be used for both voice and data, with the allocation changing between them up to 50 times per second.

The TDMA air interface used by GSM had a hard physical bandwidth limit per channel of somewhat less than 25 kilobits per second. (Some capacity is lost to protocol overhead and to time guard bands in the timeslicing.) GSM permitted data to be sent on voice channels, but increasing the bandwidth beyond that fundamental allocation wasn't really possible. It turned out that the only reasonable way to do it was to graft on an entirely new system to support data, which they called GPRS. GPRS does not share spectrum with the existing voice system in GSM; it has to be allocated its own portion of the licensed spectrum, which meant that operating companies that deployed it had to decrease the spectrum allocated for voice, decreasing the number of subscribers they could support. This turned out to be very unpopular with the operating companies.

3G

With the push to greater and greater data rates, everyone recognized that a new Third Generation of cellular equipment would be needed, the legendary "3G." And for the reasons given above, and several others, it was equally clear that it had to use a CDMA air interface (though not the precise version used by IS-95). GSM was the very best propeller-driven fighter aircraft that money could buy, and its implementation of TDMA was the Merlin engine. But CDMA was a jet engine, and ultimately TDMA could not compete. The fundamental weakness of TDMA at the RF layer could not be compensated for at any layer higher than that, no matter how well designed it was, and any 3G system which used a CDMA air interface was going to have an overwhelming commercial advantage over any TDMA-based system. TDMA was a dead end, and to create a 3G version of GSM, Europe's electronics companies were going to have to swallow their pride and admit that Qualcomm had been right all along about the virtues of the CDMA air interface. (And not incidentally, to license Qualcomm's intellectual property, and pay royalties on it.) The new system being designed to replace GSM is called Universal Mobile Telecommunications System (UMTS), and its air interface is referred to as W-CDMA. Meanwhile, the industry body which supported IS-95 was beginning work on its own 3G system, which ended up being named CDMA 2000 (CDMA2K).

This article in the Economist says that the development process for UMTS is not going well because of technical problems in making W-CDMA work. Among other problems, handsets from one company often don't work with infrastructure from other companies, and all the handsets tend to have short battery life.

When Qualcomm and its industry partners designed CDMA2K, they were able to make it backward compatible with IS-95 (in part because the CDMA2K chiprate is an exact multiple of the IS-95 chiprate). Because of that, a CDMA2K handset will work with IS-95 infrastructure, and an IS-95 handset will work with CDMA2K infrastructure, and CDMA2K cells can sit next to IS-95 cells and use the same spectrum. This is a tremendous commercial advantage, because existing operating companies using IS-95 can upgrade by incrementally replacing individual cells as budget allows and selling new handsets without having to wholesale replace all existing ones. They can take an existing IS-95 system using an existing spectrum license, and phase it over without acquiring any new spectrum, and do so over a period of years if need be.

None of that is true for the switch from GSM to UMTS. CDMA and TDMA are fundamentally incompatible air interfaces and there's no way to make UMTS support existing GSM (TDMA) handsets using the same spectrum as UMTS (W-CDMA) handsets. So GSM operators who want to upgrade to UMTS either have to acquire new spectrum (at huge expense), divide their already precious spectrum up into parts for legacy GSM and new UMTS, or cut the entire system over all at once. Each choice is lousy, in different ways. In actual practice, UMTS isn't an upgrade for GSM. It's a replacement.

If it happens at all. For the other thing they're discovering in Europe is that making CDMA work well is a lot harder than they thought it was. Interoperability problems have been particularly serious. This article talks about the experience that DoCoMo had in Japan when it deployed the first UMTS system in the world. The article doesn't mention that DoCoMo has had to recall and replace thousands of handsets at its own expense when it was discovered that the handsets had fatal interoperability problems with DoCoMo's infrastructure which could not be fixed. In fact, DoCoMo had to do this twice. Both times were fantastically expensive, and both times represented really bad public relations fiascos. DoCoMo's name is mud in Japan now; they may never fully recover.

CDMA2K, on the other hand, is real and it works now. Commercial shipments of infrastructure and handsets began a long time ago. Both Sprint and Verizon began upgrading more than a year ago, and CDMA2K has been deployed elsewhere in the world, including by DoCoMo's rival KDDI. And what everyone is discovering is that it works. The transition is clean. There haven't been any unfortunate surprises. And it's making possible all kinds of new cool features. In Japan, half the CDMA2K handsets which have been sold have cameras built in and their users send each other pictures.

On the other hand, in Europe the GSM service providers are in deep trouble. They spent truly vast amounts of money, tens of billions of euros, on licenses for new spectrum which they can't actually use yet (so they are making no revenue off the investment). The licenses specify that they can only be used for UMTS, and none of the equipment suppliers is actually ready for full scale deployment. Some of the operating companies are talking about giving the licenses back just to get out from under the payments. And others are beginning to ask if they can have permission to deploy CDMA2K, but the bureaucrats in the EU aren't having any of it. (Yet.)

Schadenfreude

I confess to a deep feeling of satisfaction about this on a personal level, primarily because of all the harassment I put up with from GSM fans over the years when they talked about how superior the European approach was.

If the U.S. had followed the same policy as Europe and mandated a standard, the CDMA air interface would never have been given the chance to prove itself. We in the U.S. now have just as good of nationwide systems and just as much roaming ability as the Europeans do, only our best systems are fundamentally better on a technical level than the best European systems, and are upgrading sooner with less pain. I can use my Verizon handset nearly anywhere in the U.S. or Canada that has cellular coverage at all (which is an area much larger than the EU), either directly on Verizon or roaming on a compatible system (such as Canada's Bell Mobility).

So I'm sitting here basking in the warm glow of schadenfreude. Regardless of the merits of GSM versus IS-95, CDMA has soundly defeated TDMA, and until something better comes along, all future proposals for cell system protocols will use some form of CDMA air interface.

Morality Tale

This turns out to be a morality tale which more broadly shows the difference in approaches to most things between the Europeans and the Americans, and it demonstrates quite clearly why the American philosophy is more successful.

Though the adoption of a continent-wide standard for Europe in the 1990's did have certain substantial benefits, it also had some hidden prices. It gave them compatibility, but it was also protectionism, and as is always the case with industries shielded by protectionism, the European cell phone manufacturers became arrogant and complacent, and as a result they fell badly behind. Now they're trying to catch up, and it isn't turning out to be easy.

Like all protected industries, the GSM companies didn't make the investment they should have early enough. Part of why they're way behind is that they started late, and much of that was because of ego, because they didn't want to admit that Qualcomm had been right (or to pay Qualcomm royalties). So they lost two full years in lawsuits and negotiations with Qualcomm before the real design process could begin. And then they discovered that the problem was harder than it looked. As it now stands, it's going to be interesting to see whether they can ever get it to really work well (with low power usage in the handsets and interoperability problems solved), and more importantly, whether they'll take so long doing it as to miss the market window. I think they will eventually make it work, but I think it will be too late. Much of GSM's dominance worldwide over IS-95 was due to its mindshare victory, but as UMTS continues to have problems and CDMA2K continues to work well, that is extremely likely to erode.

Here are some key lessons:

First, Europe pulled this decision up to as high a level as it could. The legal mandate to use GSM was passed by the European Commission. In the U.S., that decision was pushed down as far as possible, and the superiority of CDMA over TDMA was decided by individual operating companies and millions of cell phone users who voted with their wallets.

Second, Europe tried to stop the clock. It decided that it had the final answer with GSM and that no further experimentation was necessary because no further improvement was possible. In the US, the government kept its hands off and still does, and if a newer air interface technology comes along which is superior to CDMA, it will have the same opportunity commercially that IS-95 had. (Not quite; the market has evolved and we're into the "consolidation, standardization and shakeout" phase now, which is much more difficult for upstarts. But there won't be any government mandate preventing deployment of some future technology, because the FCC still doesn't mandate any standard.)

Europe emphasized cooperation over competition, centralized wisdom and control over distributed creativity, bureaucracy and regulation over freedom of choice. It was viewed as important that there be compatibility over the whole continent, and to achieve that they outlawed competition and created a monopoly. In the U.S., we valued competition, and ironically we not only ended up with compatibility over the whole continent but got that compatibility with a superior system which emerged out of competition.

And the final irony is that the GSM upgrade to UMTS is using a version of the CDMA air interface, which would never have had the chance to prove itself if the U.S. had adopted the European approach. Europe's cellular technology will be improved (if they can get it to work) because the FCC ignored Europe's advice.

In the mean time, the long-term fallout of the EC/EU's policy is that a lot of the cellular operating companies in Europe are in deep financial trouble, not to mention facing legal deadlines for deployment of UMTS which cannot possibly be met. MobilCom in Germany is near death, for example, and just announced that it would lay off 40% of its staff. Apparently it would already be dead were it not for a €400 million loan from the German government, which has angered the EU. And because the telecom companies in Europe are all so heavily cross invested, this is potentially a cascading problem. Part of why MobilCom is in trouble is because France Telecom SA is in trouble and had to renege on an investment commitment. You're eventually going to see a chain-reaction sequence of commercial failures as the money runs out and each failing company pulls another over the edge of the cliff. (Actually, what you'll see is huge government subsidies to prop them up, creatively packaged to evade EU regulations.)

Another of the ironies in this is that "cooperative" Europe has turned out to not cooperate as well as "competitive America." The UMTS interoperability problems clearly show that they're having difficulty cooperating on a detailed technical level. On the other hand, the companies involved in the CDMA2K process are cooperating closely because it's in their own narrow self interest to do so. (For instance, there's an independent testing laboratory for compatibility established by the industry which is used by all manufacturers including Qualcomm. The cellular operating companies are part of this industry body, and they insisted on it.) The companies in the CDMA2K process are cooperating closely because they know they'll be killed if they don't, not to mention the fact that they smell GSM's blood and know that they've stolen a march on UMTS and have an opportunity to cause a major shift in the balance of power in the world cellular industry (all the more so because UMTS, as a replacement for GSM rather than an upgrade, isn't really much more attractive for existing GSM service providers than CDMA2K would be).

This kind of thing has played out much the same way hundreds of times before between Europe and the U.S., and nearly always it's had the same result. And as Europe increasingly centralizes and "harmonizes" and moves more and more authority to Brussels, it's going to keep happening. Decisions will be made from the center, and a lot of the time they'll be made wrongly because the "center" is not the infinite repository of all knowledge and wisdom. The "center" chose GSM (and thus TDMA) to be the winner; America decided to let the market pick the winner, and it didn't turn out to be TDMA. And now Europe is switching to the superior CDMA air interface which could never have been developed in Europe because of government regulation.

European centralization turned out to be a competitive advantage - for the U.S. And that's going to keep happening. If I were vicious and wanted to wish commercial failure and misery on Europe, I could think of nothing better to inflict on it than the process going on now whereby more and more authority will move to Brussels to be used by unelected bureaucrats who answer to no one, and will make binding technological decisions based on politics and ideology.

Steven Den Beste is a retired software engineer who now spends his time writing for his web site, USS Clueless.
Categories:
|

TCS Daily Archives