Cellular telephone technology
Below is the text of a mail message I sent to some friends on 19 May 1998. It reflects the industry (and my understanding of it) at that time.
Since then, Sprint Spectrum, which had pretty solid service, has been dismantled. Its customers were switched over to Sprint PCS, which basically sucks. It has continual problems with poor call quality (lots of digital artifacts and complete obliteration of portions of a conversation) and dropped calls. In fact, perhaps more than half of my phone conversations over Sprint PCS involve roughly four to six individual calls, all but the first beginning with the words "Sprint PCS sucks!"
I’m currently looking into VoiceStream, who appears to have bought the old Sprint Spectrum network. I’ll include more detail after I look into their service plans and coverage.
TDMA, CDMA, spread spectrum, and other concepts
From: Geoff Adams
Subject: Cellular phone technology confusion
Date: Tue, 19 May 1998 18:46:38 -0400
The current cellular phone industry is a bit confusing. Here’s what I’ve figured out so far. First, some background. (This may sound technical, but I’m trying to keep it conceptual, with the exception of the occasional concrete example. So don’t despair until you try to follow it.) :)
Spread spectrum is a technology which reduces the impact of interfering signals on the reception of the desired signal. The effect is to spread the signal over a wide band of the frequency spectrum, resulting in a low power per unit bandwidth (watts per Hz) in the transmission. In fact, the power per unit bandwidth usually puts the signal below the noise floor.
Now, what happens at the other end reminds me a little of the conversion between time and frequency domains, or perhaps of convolution. The receiver despreads the signal spectrum it receives to reconstitute the original signal. In the process, it spreads the noise. So, imagine a bit of interference caused by someone transmitting a powerful signal in your spectrum space that interferes with part of your signal. That powerful interference will now be spread out to a much lower power across your entire signal, and, assuming you’re using digital modulation techniques, it’ll likely be well below the signal strength required to cause a bit error. Neat, huh?
So, there are a couple ways of doing that. In Frequency Hopping Spread Spectrum (FHSS), a pseudo-random bit stream is used to select discrete frequencies. The other technique, Direct Sequence Spread Spectrum (DSSS), apparently achieves better interference immunity, although I don’t yet understand how it works, but it also requires a pseudo-random bit stream. The idea is that any two pseudo-random bit streams will have very low correlation, so the affect of two signals being transmitted simultaneously, but using different pseudo-random bit streams, on each other will be minimal. Many users can be piled up in the same spectral space without interfering.
That’s basic spread spectrum theory. Now, some background into current cellular technologies. GSM uses a technique called Time Division Multiple Access (TDMA). TDMA uses a single frequency band (channel), and communicating mobile units (phones) are allocated a time slot in that channel during which to communicate with the base station (cell). There is a clear inherent limit on the number of users per channel using this technology. Also, adjacent cells must use different freuquencies, or the users in adjacent cells (and indeed, the base stations themselves) would interfere. Also, note that TDMA is not a spread spectrum technology.
In contrast, Code Division Multiple Access (CDMA) systems are far cooler. CDMA, as far as I can tell, simply refers to spread spectrum technology as applied to mobile communications. All users in a cell operate on the same frequency band, and all phones receive all signals. However, each phone is assigned a unique (orthogonal) code, which is its pseudo-random sequence. Each phone then uses this code to demodulate (despread) the information in the band to recover the desired signal. This also means that adjacent cells do not interfere, and so, each cell is able to use the entire bandwidth for communication.
(I’m looking at a book which claims that TDMA should allow simultaneous use of a 12.5MHz band by 165 users [three users each in 30-kHz channels, each user occupying 8.5 kHz], while CDMA should allow 367 [more complex derivation, based on statistical analysis of S/N ratios]. Interesting numbers.)
Obviously, CDMA is beneficial to the network provider, since it allows a greater number of subscribers per cell. This means providing the network is cheaper per customer, they can have more customers, etc. Other than the potential for that monetary efficiency to be passed to the end user, I don’t know yet whether there is a difference in quality of service for the user. Perhaps it simply means that as a cell approaches saturation, TDMA users will simply be denied service, since there are no available time slots, while CDMA users in the cell will notice a uniform increase in noise.
Also, there are technological limitations that I believe currently prevent reaching the theoretical maximum number of users per cell, at least in the CDMA case, but these will be worked on as time progresses, I’m sure. One of the limitation is how quickly the transmitting units can change frequency without introducing errors (and the more quickly you move to more widely- spread frequencies, the better your processing gain, a measure of the extent to which noise is spread into nothing as the signal is reconstituted).
It may be important to note that CDMA is a relatively recently-looked-into technology. It’s only just now (within the past two or three years) gaining any significant adoption in the telecommunications industry.
One more bit of background: GSM, as adopted for use in in the United States, is known as PCS-1900. This is basically (as far as I can tell) GSM, or a subset of GSM, transposed onto the 1900 Mhz spectrum allocated by the FCC for use by Personal Communication Services. GSM, of course, is the current global standard for mobile telecommunication. It provides a number of unique service advantages to the user, including swappable SIM cards and transparent cell- switching behaviour.
Now, for the confusing bits. The new (at least in this area) Sprint PCS service is based on CDMA technology in the 1900 MHz band. This means, simultaneously, that it’s spread spectrum, and that it’s not PCS-1900, as its name might seem to imply. I must assume that it takes its name from the more general sense of the term, which refers to the FCC naming of the 1900 MHz band. As I understand it.
On the other hand, Sprint Spectrum is a PCS-1900 technology. This means it uses the GSM "air interface," which is TDMA. (Yeah, it’s the older, less-cool technology, but this is an older, well-established -- but still very cool -- system.) Sprint Spectrum service is therefore *not* spread spectrum, as the name would seem to imply, but it *is* PCS.
Shouldn’t the names simply be switched?
In contrast, it appears that the new AT&T’s Digital PCS service is TDMA on 1900 MHz, but it does not appear to borrow anything from GSM. And please note at this time that I haven’t even gone anywhere near the 800 MHz band, either the traditional Analog Mobile Phone System (AMPS) or the efforts to do digital communication on that band.
And one final note: Here’s what happens when you want to combine the cool features of GSM with the cool spectrum usage of CDMA:
<http://www.qualcomm.com/news/pr980217.html>
I hope you’ve enjoyed reading my little essay,
- Geoff