How about this blast from the past from Bill Frezza, one of the most frequent and vocal Qualcomm bashers from back in the day. Note the last paragraph disclaimer:
Succumbing To Techno-Seduction
My favorite single-panel cartoon shows a proud physicist standing at a blackboard giving a lecture to a group of skeptical dairy farmers. The caption reads-"Assume a spherical cow." Whenever I see it I think of Qualcomm and the digital-cellular standards wars now coming to a head. As potential Personal Communications Services (PCS) operators place their bets on the technologies that could make or break their fortunes and incumbent cellular operators gird for the competitive onslaught, this is no small decision. Not only are billions at stake, but the nationwide compatibility that mobile phone users take for granted hangs in the balance.
Assume a spherical cow. How convenient it is to take the hoary complexities of the real world and reduce them to a few simple formulas. This is the stunning beauty-and possibly the Achilles heel-of Code Division Multiple Access (CDMA), widely touted as a panacea that will usher in an age of unlimited bandwidth.
Here is a brief review of modulation techniques-skip ahead if you already know this. There are several ways to subdivide spectrum. The simplest is Frequency Division Multiple Access (FDMA), in which each user is assigned a specific range of frequencies. Channels are kept separate using either analog or digital bandpass filters, extremely mature and inexpensive technologies. All wireless communications systems use FDMA at the lowest level, often layering additional modulation schemes on top.
Time Division Multiple Access (TDMA) divides a channel into tiny interleaved time slots. Users take turns communicating in rotation. The trick is kee ping the time slots separate, a challenge in wireless environments plagued by variable propagation delays. Fortunately, basic TDMA technology is quite mature, having been used in commercial wireline networks for decades.
CDMA is based on statistical mathematics that not one person in a thousand can comprehend, none of them in the investment banking community. Its roots lie in military systems-the folks that buy $500 toilet seats. CDMA works by modulating each user's data stream with a higher-speed pseudorandom chip sequence, creating a wideband signal that looks like noise. Each user's conversation is then extracted from the noise using a correlator looking for the particular "signature" of that pseudorandom chip code.
Channel separation is a function of the length of the chip code, the extent to which the correlator approaches "ideal" performance, the relative power levels of the signals and the number of users on the system at any given moment. Unlike FDMA and TDMA, the capacity of a CDMA system is not bandwidth-limited but interference-limited based on a raft of complex assumptions, as well as the degree of quality degradation that users are willing to accept each time another caller is added.
Engineering as Performance Art
Five years ago, when I first heard Qualcomm's cofounder Dr. Andrew Viterbi expostulate the wonders of CDMA, I was spellbound. I wanted his autograph. The system he described was the most beautiful thing I had ever seen-and I actually understood most of the math. It promised at least a 20-to-1 capacity improvement over traditional analog techniques, while the first generation TDMA systems offered only 3-to-1. I did have a few nagging doubts-I mean, it sounded too good to be true-but Viterbi handled everyone's questions so smoothly I didn't even notice that he never answered them.
I wasn't the only one impressed. After several cellular carriers signed on for tests and trials, Qualcomm took off on a rags-to-riches story that is the stuff of legends. The headlines were awesome: Bona fide genius founders shake wireless world; Patented scientific breakthrough hailed as end of traditional technologies; Unified industry consensus based on TDMA standard rent asunder to make room for brilliant newcomer; Blue-chip investment bankers take company public in an orgy of speculation. Qualcomm's PR engine began manufacturing an endless stream of news whipping their market capitalization up to $1.5 billion. It didn't hurt that best-selling technology authors touted CDMA in the pages of Forbes magazine as "an end to spectrum scarcity as we know it." Erstwhile competitors scrambled to sign license agreements in fear of being left out. Meanwhile, major vendors focusing on making their TDMA systems work were characterized as a bunch of old fuddy-duddies that just didn't "get it."
Assume a spherical cow. One of the curious things about this approach is that complex mathematics can be exquisitely sensitive to certain assumptions in very unpredictable ways-ways that don't show up until you actually go out and build something.
What, No Beta Test?
Well, it's time to build something. If Qualcomm's technology is going to stumble, as many predict, it will happen at the edges-the edges of the cell, the edges of performance, the edges of capacity. This is where assumptions break down and cleverness can come back to bite you.
Take power control, a critical ingredient in making CDMA work. Get your assumptions wrong and capacity is cut in half. Spread spectrum correlators produce something called processing gain that allows a base station to pluck wanted signals from the noise created by the other signals, as long as they all arrive within a fairly narrow power range. If not, a too-weak signal disappears and a too-powerful signal can wipe out a whole cell.
Coordinating the real-time transmit power of hundreds of roving users in rapidly fading environments with cell footprints that grow and shrink based on load reminds me of the guy on Ed Sullivan who used to spin dozens of plates o n bamboo poles: Turn your back for a moment and you're screwed. In theory, no problem. In practice? It's show time.
CDMA's claim to fame is its promise of enormous capacity. But the key to happy customers has nothing to do with capacity. We'll soon be awash in spectrum, thanks to the FCC auctions. Success will hinge on cost, features and quality. How does a CDMA phone sound when hundreds of users are crowded into one cell? In theory it should sound OK. In practice? It's show time.
Because CDMA phones are based on advanced digital signal-processing technology, they're supposed to be amenable to highly integrated solutions, yielding the lowest-cost, lightest-weight phones imaginable-in theory. In practice? It's show time.
Broadband CDMA systems are supposed to be easier to engineer because less frequency coordination is required. But much of this assumes relatively uniform traffic density. Assume a spherical cow. How do you handle hot spots like major highway intersections near convention centers if you can't build hierarchical cell structures with layer-to-layer hand-off? Come on folks, it's show time.
First-generation CDMA systems are just now being subjected to serious testing, and it's getting difficult to suppress the grumblings of testers who claim it doesn't work as advertised. Not that Qualcomm hasn't tried-including having its lawyers send threatening letters to some Stanford University professors that refute the optimistic claims.
Beta testing with real customers has now been scheduled for the middle of 1995, after three years of schedule slips. Promises are wearing thin. Some form of CDMA may indeed have a bright future but, frankly, it's the bottom of the ninth for Qualcomm-world leader in investor relations-and it better pick up that bat and swing for the fences.
What about fuddy-duddy operators like McCaw and all of Europe that have chosen TDMA? Yes, they struggled to get first-generation systems based on the IS-54 standard up and running-struggles that Qualcomm has yet to face . But GSM deployment in Europe went smoother.
Vendors are now fielding third-generation systems for PCS. Taking a chapter from Qualcomm's book, these systems layer a more practical form of spread-spectrum modulation called frequency hopping on top of the basic TDMA scheme. In dueling white papers, the TDMA advocates claim they have erased CDMA's capacity advantages with a system that is 50 percent less expensive to deploy with better speech quality. But after all is said and done, consumers buy products, not white papers. To date, more than five million phones based on TDMA technology have been sold with more than 50 countries worldwide adopting the technology. CDMA? Well, it's finally show time.
As it gets closer to the time to put up or shut up, a lot of air has come out of Qualcomm's stock price, particularly after the insiders started dumping. I never invest in individual stocks, being about as knowledgeable about Wall Street as bankers are about technology. But this looks like a dicey situation that may get even dicier.
...And a little boy in the back of the room said, "But sir, cows aren't round."
Bill Frezza is the President of Wireless Computing Associates (WCA). In the interests of full disclosure one of the proponents of TDMA, a division of Ericsson, is a client of WCA. Bill can be reached anytime, anywhere at frezza@radiomail.net. |
