Cellular Voice and Data
By far the largest category of PCS system is the group of standards used for digital cellphone networks. Although these were all designed primarily for voice, they can also support data transmission at varying rates. The data rates shown in Table 4.1 are those that use the voice circuit, usually by connecting a phone to a laptop computer. Many also include far slower data services that can operate at the same time as the user is talking, usually used to display simple text messages on the phone itself.
Most PCS systems are based on TDMA (Time Division Multiple Access). FDMA (Frequency Division Multiple Access) is too wasteful of bandwidth, while CDMA (Code Division Multiple Access) had not yet been invented at the time they were standardized. The exception is cdmaOne, which most scientists agree is technologically the most advanced system. However, this doesn't mean it is necessarily superior to the others; for example, GSM offers the best coverage, while D-AMPS is compatible with older networks still in use across North America.
The arguments between proponents of different systems are often heated, even referred to as "religious wars." Most of the debate centers around spectral efficiency, which means how much capacity a system can squeeze out of its allocated frequencies. Because operators each have a relatively small slice of spectrum, it is in their interest to use this efficiently. It is also in the public interest, as the total spectrum is finite and use that is more efficient should keep costs down. Up to a limit, any system can be made more efficient by installing more lower-power base stations, which was part of the rationale behind spectrum auctions—if operators have to pay for spectrum, they are more likely to invest in equipment to use it more efficiently.
Unfortunately, spectral efficiency depends on so many variables that it is not easy to calculate accurately. Table 4.2 tries to do this for the four main systems, with the important results in the final column. Note that these are very approximate, though they do show that GSM is generally less efficient than the others. This is in part because it provides better quality and a more reliable connection than the other TDMA systems, though cdmaOne offers similar benefits without the extra spectrum requirements. This is because it was standardized more than a decade after GSM, by which time technology had moved on.
| System | Channel Bandwidth | Calls per Channel | Cells per Channel | kHz per Call | kbps per Call | Hz per bps |
|---|---|---|---|---|---|---|
| GSM | 200 kHz | 8 | about 4 | 100 | 14.4 | 7 |
| D-AMPS | 30 kHz | 3 | about 7 | 70 | 9.6 | 7 |
| PDC | 25 kHz | 3 | about 7 | 58 | 9.6 | 6 |
| cdmaOne | 1250 kHz | about 15 | 1 | 83 | 16 | 5 |
normal: How is Spectral Efficiency Calculated?Every system divides its frequencies into relatively wide FDMA channels, which are shared between a number of calls. The amount of spectrum used by each call should therefore be
This is true for a single transmitter, but most real cellular systems don't allow every frequency to be used in every cell. Except in CDMA, neighboring cells cannot use the same frequencies, so the equation becomes
This is the traditional measure of spectral efficiency, shown in Table 4.2 under the kHz per Call heading. It applies only to voice calls and does not take into account the type of codec used. GSM and cdmaOne use higher bit-rate codecs than the other systems, resulting in better quality speech, which in part explains why they require more kHz. In the age of mobile data, it may be more appropriate to divide by the capacity available if the voice calls are replaced by data. This measures the numbers of wave cycles needed to transmit each bit.
The results of this calculation are in the final column of Table 4.2. Again, this is not the full story. GSM and cdmaOne both include built-in error correction not found in D-AMPS and PDC. |
GSM
Now known as the Global System for Mobile Communications, GSM originally stood for Groupe Spéciale Mobile, the name of the committee that designed it. The technology is used by more than half of all mobile phones, and by 1999 was expanding at a rate of 1 million users per week. Its popularity is due partly to relatively high quality voice, but mostly to early standardization by European governments. Another advantage is that it is easily upgradeable to higher data rates, a deliberate decision on the part of its inventors, but not something appreciated by its early adopters.
The system now operates on four different frequency bands, shown in Table 4.3. It was originally designed for frequencies around 900 MHz, to reuse the spectrum intended for Europe's analog TACS networks. It was later adapted to bands around 1800 MHz, licensed in Europe specifically for GSM, and then to the 1900 MHz band used in America for several different digital networks. These higher frequency variants are sometimes called DCN (Digital Communications Network), but they're really just GSM. The latest variant uses the much lower 450 MHz frequencies, so that it can replace aging analog networks based on the Scandinavian NMT (Nordic Mobile Telephony) system.
| GSM Type | Uplink Frequency | Downlink Frequency | Cell Size |
|---|---|---|---|
| GSM 450 | 450.4–457.6 MHz or 478–486 MHz | 460.4–467.6 MHz or 488.8–496 MHz | Biggest |
| GSM 900 | 880–915 MHz | 925–960 MHz | Big |
| DCN 1800 | 1710–1785 MHz | 1805–1880 MHz | Small |
| PCS 1900 | 1850–1910 MHz | 1930–1990 MHz | Smallest |
Like other digital cellular technologies, GSM encodes data into waves using a form of phase modulation, a system which uses the different parts of a waveform to represent information. The precise type is known as GMSK (Gaussian minimum shift keying), which achieves a symbol rate and data rate of 270.8 kbps in each of its 200 kHz channels. This transmission is only one-way, so GSM uses separate paired channels to send and receive.
As anyone who has tried to send data over GSM knows, the achievable rate is a tiny fraction of the maximum 270 kbps. The main reason is that, like other TDMA systems, GSM divides the channels into a number of time slots, in this case eight. Each phone only transmits and receives for one eighth of the time, reducing the data rate by the same factor to about 33.9 kbps.
Some of this capacity is effectively wasted. It's left empty to counteract propagation delay, the time taken for a signal to travel from a base station to a mobile phone. At the speed of light, radio signals take just over 3.3 microseconds to travel a kilometer, which adds up to delay of around 100 microseconds for a phone just 30 km (20 miles) from the base station. Each of the time slots lasts only 577 microseconds, so this delay is enough to make around 27 percent of the slot unusable.
Each 577 microseconds slot has room for exactly 156.25 bits, arranged in the structure illustrated at the bottom of Figure 4.1. Each bit can be allocated to four different uses, whose proportions are shown in Table 4.4.
The header and footer are empty space at the beginning and end of the slot. They serve to separate a slot from its neighbors, negating propagation delay at any distance up to 35 km from the base station. Beyond this limit, GSM phones cannot be used, even if a clear signal is available.
The training sequence is a fixed pattern in the middle of the slot. It is used to help a receiver lock on to the slot, but does not carry any useful information. It also acts as a header and footer if the slot is further divided, as happens when a half-rate codec is used.
The two stealing bits identify whether the slot is used to carry actual data or control information. The name comes from the perception that the slot is being "stolen" from a call for use by the GSM system itself.
The traffic bits are available for other uses, which include control information and error correction, as well as actual speech and data. In a typical GSM system, this payload is 24.7 kbps.
| Data Type | Bits in Slot | Equivalent Data Rate | Proportion of Total |
|---|---|---|---|
| Head and Tail | 14.25 | 3.1 kbps | 9.1% |
| Training | 26 | 5.6 kbps | 16.6% |
| Stealing | 2 | 0.5 kbps | 1.3% |
| Traffic | 114 | 24.7 kbps | 73% |
Figure 4.1. GSM slot, frame and multiframe structure
A complete cycle of eight time slots is known as a frame, illustrated in the center of Figure 4.1. Not all frames are open to the user; for every 24 that carry voice or data, one needs to be "stolen" for signaling and another reserved for other types of traffic, such as short text messages or CLI (Caller Line Identification, the service that displays a caller's number on the phone's screen). This means another unit of structure is needed, known as a multiframe. Each contains 26 frames, labeled A to Z in Figure 4.1. The two "stolen" frames are E and R in the diagram, but each of these move on by one every multiframe to help with timing.
Signaling reduces the total capacity per user to 22.4 kbps, but even that is more than the user will ever see. FEC (Forward Error Correction) and encryption reduce the data rate by at least another third, though the precise amount depends on how the phone and network encode voice and data. GSM originally used a voice codec, which required 13 kbps, producing what was then, in 1982, considered high quality voice. Since then, advances in microelectronics have allowed phones to be developed with a half-rate voice codec. This halves the bandwidth required per call and so doubles the capacity, with a slight decrease in quality. Data cannot take advantage of half-rate codecs, which is one reason why some operators charge more for data than for voice calls.
There is also an enhanced full-rate codec, which uses the same bit rate as the regular full-rate but provides better quality. Most phones and networks now support all three codecs, though because of the lower quality of half-rate they usually don't advertise it. Some operators use the enhanced codec as a standard, but drop down to half-rate if the network becomes busy. Others offer the enhanced rate to specific customers, usually those paying higher tariffs.
The earliest schemes for sending data over GSM actually had to route it through the full-rate codec, at a rate of only 9.6 kbps. An enhancement, not supported by all phones or networks, bypasses the codec entirely and pushes it to 14.4 kbps. This rate is higher than the output of the voice codec because data can miss out some of the error correction; it's actually more tolerant of errors than highly compressed voice because it doesn't matter in which order packets arrive. The full range of codes is shown in Table 4.5.
FEC anticipates that errors will occur and tries to preempt them by sending extra correction code. Most data protocols, including those of the Internet, take a different approach: they wait to see if errors actually happen, then ask for garbled information to be sent again. Users of these protocols can omit FEC entirely, for a raw data rate of 21.4 kbps. Salespeople sometimes quote this as a realistic capacity, which is misleading—some errors will still creep in, requiring retransmission that pushes down the effective speed.
| Voice | Data | |||||
|---|---|---|---|---|---|---|
| Full-Rate | Half-Rate | Enhanced | Regular | Enhanced | Raw | |
| Codec | 13 kbps | 6.5 kbps | 13 kbps | 9.6 kbps | 14.4 kbps | 21.4 kbps |
| FEC | 8 kbps | 4.0 kbps | 8 kbps | 11.1 kbps | 6.6 kbps | 0 kbps |
| Encryption | 1.4 kbps | 0.7 kbps | 1.4 kbps | 1.4 kbps | 1.4 kbps | 1.4 kbps |
HSCSD
High Speed Circuit Switched Data (HSCSD) is a very simple upgrade to GSM that gives each user more than one time slot in the multiplex. Standardized by ETSI in 1997 and first released commercially in 2000, it is the equivalent of tying two or more phone lines together and aggregating their capacity.
All HSCSD-capable networks or terminals use the enhanced data codec, so that each channel allows rates of 14.4 kbps. The standard allows up to four of these to be tied together, for a maximum of 57.6 kbps. Intermediate steps of 28.8 kbps and 43.2 kbps are also possible, and actually more common. It's also possible to have asymmetric data rates; for example, three slots from the base station to the mobile and only two the other way.
The limit of four channels was not just picked at random. It would be possible to aggregate all eight channels in a time slot together, but this would have made handsets more difficult to design, and HSCSD was supposed to be a simple upgrade. With four channels, a phone never has to both transmit and receive at once; each time slot in a GSM uplink frame is paired with one "opposite" it in the corresponding downlink, as shown in Figure 4.2. Other problems with using additional channels include power consumption and capacity.
Many GSM networks are already near the limit of their capacity, struggling to build new base stations to keep up with their growth in customers. Full-rate HSCSD means that each customer requires four times as much bandwidth as for a regular call, and eight times as much as a call using the half-rate codec. Most people are unwilling to pay four or eight times the cost of a regular call, particularly with the perception that data should be free. For this reason, many GSM operators have decided not to deploy the system at all, instead waiting for other systems that use bandwidth more efficiently.
HSCSD terminal design is also proving difficult. Four time slots needs to transmit for four times as long as a regular GSM phone, with corresponding increases in battery drain and radiation emission. Aside from possible health risks, there is a more immediate problem: early full-rate terminals are reported to have burst into flames.
Because of the overheating problem, most HSCSD devices are asymmetric, allowing greater download than upload speeds. Happily, Web browsing is a highly asymmetric application, with pages of text and graphics traveling one way and only a few mouse clicks the other. Unhappily, the paired spectrum system means that there is no way to reallocate the unused upstream bandwidth to downstream, meaning that some slots have to be left empty.
The only symmetric application touted so far is two-way video, long a dream of science fiction, but lacking popularity in the real world. The British operator Orange has even developed a videophone with a built-in camera and 28.8 kbps HSCSD terminal. It was supposed to be launched in early 2000, but was postponed for at least a year when the company was successively taken over by Mannesmann AG, Vodafone Airtouch PLC, and France Telecom SA.
Figure 4.2. Four-slot HSCSD
GPRS
Of all the wireless Internet schemes, GPRS (General Packet Radio Service) is the one most popular among operators. Designed for data, it promises to give every user a permanent and high-capacity connection to the Internet.
This promise may be fulfilled, but not yet and not for everyone. GPRS was envisaged as an upgrade to any TDMA-based system, but in fact works only with GSM. The first generation of terminals supports data rates less than that of the more primitive HSCSD technology, rather than the 115.2 kbps seen in promotional literature. And it arrived late, with products announced in 1999 that had still not shipped a year later.
Nevertheless, GPRS represents a major step forward in mobile networks. Its key improvement is packet switching, which for most data applications is more efficient than circuit switching. Under GSM or HSCSD, each user has to keep open a full circuit of 9.6 kbps or more for the duration of their time online, even though surfers spend more time reading Web pages than actually transferring information. This is wasteful for both the customer and the operator—one is paying call charges for an idle connection, while the other is committing spectrum that could perhaps be more profitably deployed elsewhere.
Packet switching uses bandwidth only when needed, freeing up gaps in the data stream for other users. Under GPRS, a single 14.4 kbps time slot can be shared by hundreds of people, provided that they don't all try to use it at the same time. Each has a continuous connection at a very low data rate of 0.1 kbps or less, which bursts to higher speeds when they receive email or click on a hyperlink.
Because of this greater efficiency, the full specification calls for terminals capable of using all eight time slots at once. This would effectively eliminate the rigid TDMA structure, allowing each user up to 115.2 kbps. However, such speeds have so far proved elusive—GPRS suffers from the same overheating problems as HSCSD, and until they are overcome, it will have the same speed limit.
Like HSCSD, most GPRS systems are asymmetric. This is partly because of terminal design issues and user demand, but also because of the way packet switching is implemented. A base station can monitor all downlink traffic and arrange it in the most efficient way possible, but a phone cannot do the same for the uplink because it does not have access to other transmissions. The result is that there is actually more downstream than upstream capacity available.
The plan is that GPRS networks will eventually carry voice over packets, using a variable-rate codec so that data can be transmitted during gaps in conversation. In the meantime, GPRS has to coexist with GSM and perhaps HSCSD, meaning that phones have to support both packet and circuit switching. Three different grades of GPRS handsets have been defined, corresponding to how these are supported.
Grade C terminals can operate in either packet-switched or circuit-switched modes, but not both at once. When a voice call is made, the otherwise permanent data connection has to be dropped.
Grade B terminals can maintain a packet-switched connection while a circuit-switched call is in progress, but cannot send any data over it. A customer can be notified of, for example, incoming email while a call is in progress, but has to hang up in order to receive it.
Grade A terminals are capable of simultaneous circuit-switched and packet-switched connections. A customer can talk on the phone and surf the Web at the same time.
To make GPRS even more complicated, terminals are also divided into 29 classes, depending on their particular combination of slots. A full list is given in Table 4.6, which shows the maximum number of slots and their corresponding data rates. The slot gap is the time that the terminal takes to find a channel and frame in which it can transmit; a shorter gap will mean a lower latency.
Not all classes are capable of full duplex operation, which means utilizing both their full uplink and downlink capacities at once. For example, a Class 3 terminal can use up to two slots in each direction, but cannot exceed a total of three. This means that if it transmits using two, it can only receive using one, and vice versa.
The most ambitious classes are numbers 15 through 18, as these require a terminal that can transmit and receive on two time slots simultaneously. Only a Class 18 terminal would fulfill GPRS's promised data rate of 115 kbps in both directions.
In addition to new terminals, GPRS requires extensive new investment by operators; they need to build an entire new backbone network, replacing the existing telephone system with one more similar to the Internet. However, this can be reused for 3G technologies, most of which are based entirely on packet switching.
| Slot Class | Maximum Time Slots | Slot Gap | Capacity (kbps) | Full Duplex | Networks | |||
|---|---|---|---|---|---|---|---|---|
| Up | Down | Total | Up | Down | ||||
| 1 | 1 | 1 | 2 | 3 | 14.4 | 14.4 | Yes | GSM, GPRS |
| 2 | 2 | 1 | 3 | 3 | 28.8 | 14.4 | Yes | HSCSD, GPRS |
| 3 | 2 | 2 | 3 | 3 | 28.8 | 28.8 | No | HSCSD, GPRS |
| 4 | 3 | 1 | 4 | 3 | 43.2 | 14.4 | Yes | HSCSD, GPRS |
| 5 | 2 | 2 | 4 | 3 | 28.8 | 28.8 | Yes | HSCSD, GPRS |
| 6 | 3 | 2 | 4 | 3 | 43.2 | 28.8 | No | HSCSD, GPRS |
| 7 | 3 | 3 | 4 | 3 | 43.2 | 43.2 | No | HSCSD, GPRS |
| 8 | 4 | 1 | 5 | 3 | 57.6 | 14.4 | Yes | HSCSD, GPRS |
| 9 | 3 | 2 | 5 | 3 | 43.2 | 28.8 | Yes | HSCSD, GPRS |
| 10 | 4 | 2 | 5 | 3 | 57.6 | 28.8 | No | HSCSD, GPRS |
| 11 | 4 | 3 | 5 | 3 | 57.6 | 43.2 | No | HSCSD, GPRS |
| 12 | 4 | 4 | 5 | 2 | 57.6 | 57.6 | No | HSCSD, GPRS |
| 13 | 3 | 3 | 6 | 1 | 43.2 | 43.2 | Yes | HSCSD, GPRS |
| 14 | 4 | 4 | 8 | 1 | 57.6 | 57.6 | Yes | HSCSD, GPRS |
| 15 | 5 | 5 | 10 | 1 | 72 | 72 | Yes | GPRS |
| 16 | 6 | 6 | 12 | 1 | 86.4 | 86.4 | Yes | GPRS |
| 17 | 7 | 7 | 14 | 1 | 100.8 | 100.8 | Yes | GPRS |
| 18 | 8 | 8 | 16 | 0 | 115.2 | 115.2 | Yes | GPRS |
| 19 | 6 | 2 | 8 | 3 | 86.4 | 28.8 | Yes | GPRS |
| 20 | 6 | 3 | 8 | 3 | 86.4 | 43.2 | No | GPRS |
| 21 | 6 | 4 | 8 | 3 | 86.4 | 57.6 | No | GPRS |
| 22 | 6 | 4 | 8 | 2 | 86.4 | 57.6 | No | GPRS |
| 23 | 6 | 6 | 8 | 2 | 86.4 | 86.4 | No | GPRS |
| 24 | 8 | 2 | 8 | 3 | 115.2 | 28.8 | No | GPRS |
| 25 | 8 | 3 | 8 | 3 | 115.2 | 43.2 | No | GPRS |
| 26 | 8 | 4 | 8 | 3 | 115.2 | 57.6 | No | GPRS |
| 27 | 8 | 4 | 8 | 2 | 115.2 | 57.6 | No | GPRS |
| 28 | 8 | 6 | 8 | 2 | 115.2 | 86.4 | No | GPRS |
| 29 | 8 | 8 | 8 | 2 | 115.2 | 115.2 | No | GPRS |
normal: GPRS or HSCSD?There's a lot of misinformation about GPRS and HSCSD, and not all is due to misleading sales pitches from vendors. Many press reports have focused on the idea of a conflict between the two systems, implying that they are incompatible or even that vendors and users will have to choose one or the other. In reality, the systems are very similar and entirely compatible with each other. Every GSM operator plans at some point to upgrade to GPRS, and the only choice they have to make is whether to bother with HSCSD in the meantime. HSCSD is very easy to deploy, so this is really just a question of whether they have enough capacity to give every user several dedicated time slots. GSM, GPRS, and HSCSD can all coexist in a single network, and the terminal will also be backward-compatible for roaming purposes. For example, if Class 18 (eight-slot) GPRS terminals are ever produced, users will be able to drop down to four-slot operation when roaming on HSCSD and one-slot while on GSM network. Some GPRS operators that claim not to support HSCSD may in fact do so for roamers, but not for their own subscribers. This is because operators usually subsidize the cost of handsets to their subscribers, so have some control over what technology they use. Similar policies are already in force covering half-rate codecs in regular GSM; operators have put their own customers on the more efficient system, but still allow connections at full-rate when necessary. |
D-AMPS
As the name implies, D-AMPS (Digital Advanced Mobile Phone System) is designed to be compatible with the older analog AMPS technology, which is widely deployed in the U.S. It uses the same 30 MHz frequency channels as AMPS, but divides each one up into three TDMA slots, so is often known simply as TDMA. Some people also refer to the same system as ADC (American Digital Cellular), NADC (North American Digital Cellular), USDC (U.S. Digital Cellular), or by the name of one of the standards that govern it, IS-54 and ANSI-136.
D-AMPS uses the same paired spectrum and structure as regular AMPS, described fully in Chapter 3, "Cellular Networks," Table 3.1. The difference is that instead of sending a single FM radio transmission over a 30 MHz channel, it allows each one to be used by three simultaneous conversions. Operators of AMPS networks can selectively allocate channels to either digital or analog, allowing the two systems to coexist.
The modulation scheme is called DQPSK (differential quadrature phase shift keying) and is even more complicated than GSM's. It also requires more power than GMSK, resulting in shorter battery life and higher radiation emissions, but is about 20 percent more spectrally efficient. The resultant raw data rate is 48.6 kbps, shared between three users. Each occupies the channel for one third of the time, giving each one 16.2 kbps.
Time slots last 6.67 ms, more than ten times longer than GSM's. These bigger slots have room for signaling information, allowing the original IS-54 (Telecommunications Industry Association Interim Standard) specification to mandate no separate signaling frames. However, the later ANSI-136 (American National Standards Institute) standard added a hyperframe structure similar to GSM's, allowing advanced services such as text messages to be transmitted in one or more frames.
The complete D-AMPS hyperframe, frame, and slot structure is illustrated in Figure 4.3. It looks more complicated than GSM, because the uplink and downlink slots have a slightly different internal arrangement of bits. However, the data rates in each are the same. As Table 4.7 shows, more than 80 percent of each slot can be used for traffic, which compares well to GSM's 73 percent. The greater efficiency is achieved by using longer slots, which waste proportionately fewer bits on training and tail sequences.
| Data Type | Bits in Slot | Equivalent Data Rate | Proportion of Total |
|---|---|---|---|
| Tail | 12 | 0.6 kbps | 3.7% |
| Training | 28 | 1.4 kbps | 8.7% |
| Control | 24 | 1.2 kbps | 7.4% |
| Traffic | 260 | 13 kbps | 80.2% |
The uplink slot may seem unnecessarily complicated, with traffic carried in three separate sequences. The reason for this is that when the half-rate codec is used, extra bits are needed for signaling. Under the half-rate scheme, the small 16-bit sequence is used instead for control data, while the larger 122-bit sequences each go to a user. To allow for the extra signaling, the half-rate codec actually codes at less than half the full rate.
D-AMPS's transmission speeds are listed in Table 4.8. As with some versions of GSM, data has access to higher capacity than the output of the voice codec because it requires less error correction. However, the total capacity is less than GSM, meaning that voice has noticeably poorer quality.
| Full-Rate Voice | Half-Rate Voice | Data | |
|---|---|---|---|
| Codec | 7.95 kbps | 3.73 kbps | 9.6 kbps |
| FEC | 5.05 kbps | 2.37 kbps | 3.4 kbps |
Figure 4.3. D-AMPS (ANSI-136) slot, frame and multiframe structure
PDC/JDC
Japan has deployed a system based on D-AMPS, but designed for backward compatibility with its own J-TACS analog system. This is known as PDC (Personal Digital Cellular), or sometimes JDC (Japanese Digital Cellular). Despite being used in only one country, it is the world's second most popular mobile standard (behind GSM), thanks in part to the wireless Internet service i-mode.
J-TACS used channels of only 25 MHz, which means that some changes were needed to the D-AMPS system designed for 30 MHz. The same size time slots and modulation are used, resulting in a lower overall bit rate. However, the D-AMPS voice codecs and data rate of 9.6 kbps can still be achieved, by missing out some of the error correction and protocol overhead. This makes the system less reliable, a problem overcome by keeping cells small so that every user has a relatively clear link to the base station. Table 4.9 shows how PDC differs from D-AMPS and GSM.
Japan is much more crowded than the U.S., so small cells are not a problem—operators want to make cells as small as possible to provide the highest density coverage. However, interference and handover problems put a minimum practical limit on the cell size of any system, which meant that PDC had already run out of capacity by 2000. To counter this, operators introduced packet switching, to use the existing capacity more efficiently.
The main reason for PDC's success is i-mode, a mobile Internet system developed by the largest operator, NTT DoCoMo. Using the regular 9.6 kbps or slower data connection, it provides access to thousands of Web sites specially adapted to fit onto the phone's small screen.
| System | Channel | Slots | Slot Length | Bits/Slot | Modulation | Raw Bit Rate | Data Capacity |
|---|---|---|---|---|---|---|---|
| GSM | 200 MHz | 8 | 0.577 ms | 156.25 | GMSK | 33.9 kbps | 14.4 kbps |
| D-AMPS | 30 MHz | 3 | 6.67 ms | 324 | DQPSK | 16.2 kbps | 9.6 kbps |
| PDC | 25 MHz | 3 | 6.67 ms | 290 | DQPSK | 14 kbps | 9.6 kbps |
D-AMPS+
When ETSI designed GPRS, they hoped that it could be applied to every TDMA-based system, including D-AMPS and PDC. The principle is simple: use more slots in the multiplex, increasing capacity until a single user has the entire channel all of the time.
GSM has eight time slots, so theoretically allows data capacity to be multiplied by a factor of eight. Unfortunately, D-AMPS has only three, and each of these already offers a lower data rate of 9.6 kbps compared to GSM's 14.4 kbps. The result is that the maximum capacity, even if all slots are used, is only 28.8 kbps (3 x 9.6), a long way from the 115.2 kbps hyped by GPRS enthusiasts.
Yet another problem is that to achieve even this rate, terminals need to be able to transmit and receive at the same time. A standard D-AMPS phone transmits for one-third of the time, receives for another third (on a different frequency), and is idle for the remaining third. Without simultaneous transmit and receive, there is only one spare time slot (compared to six in GSM), limiting speeds to only 19.2 kbps in one direction and 9.6 kbps in the other.
Nevertheless, this scaled down version of GPRS is still being implemented by some operators in the U.S., under the name D-AMPS+. Japanese operators have chosen not to apply it to PDC because they already have so many users that there is no spare capacity. Instead, Japan has accelerated its plans to deploy third-generation systems based on CDMA.
cdmaOne
The only twentieth-century system to use CDMA is cdmaOne, developed by the company Qualcomm but now supervised by an independent organization called the CDG (CDMA Development Group). It has been standardized by the TIA (Telecommunications Industry Association) as IS-95a.
CDMA systems seem superficially simpler than those based on TDMA. They involve no slot or frame structure; every phone just transmits and receives all the time, sending many duplicates of the same information to ensure that at least one gets through.
The number of duplicates is known as the gain and depends on the channel bandwidth and the length of the code used. cdmaOne uses the Walsh codes, a set of 64 numbers, each 64-bits long, that have been calculated to cancel each other out. Each user has a different Walsh code, by which they multiply every bit of data before it is sent. Thus, every bit is effectively sent 64 times: the raw speed per user on the downlink is 19.2 kbps, but the phone is listening to 1228.8 kbps (64 x 19.2).
The very high transmission rate is achieved by using two different phase modulation techniques, QPSK (Quadrature Phase Shift Keying) on the downlink and OQPSK (Offset Quadrature Phase Shift Keying) on the uplink. The second type requires more forward error correction, necessary because individual phones cannot coordinate their transmissions in the same way that base stations can. The extra correction means that the raw uplink data rate per user is higher, 28.8 kbps.
The uses of the 19.2 kbps are shown in Table 4.10. The codec is similar to GSM's, producing the same high quality (for a cellphone) voice. Note that although the system may appear to be asymmetric in favor of the uplink, it isn't; all of this extra capacity is needed for error correction.
| Voice | Data | |
|---|---|---|
| Codec | 13 kbps | 14.4 kbps |
| FEC (Uplink) | 15.8 kbps | 14.4 kbps |
| FEC (Downlink) | 6.2 kbps | 4.8 kbps |
Sending every bit 64 times may seem incredibly wasteful. However, the same channel can be used by several different phones and by every base station. This means that adjacent cells can use the same frequencies, unlike FDMA and TDMA systems. This alone makes CDMA very efficient in terms of spectrum, as an operator's entire allocation can be used by each cell.
As there are 64 codes, up to 64 users could theoretically share each channel. Unfortunately, this doesn't yet work in practice, and real cdmaOne systems typically fit between 10 and 20 users on a channel. Even this is slightly more efficient than other systems, and the prospect of increased efficiency in the future is real.
One big disadvantage of CDMA systems is their power consumption. By transmitting everything 64 times, a cdmaOne phone would seem to drain its battery 64 times faster than necessary, and also bathe its user with 64 times as much microwave radiation. It gets around this problem by carefully controlling the transmission power. Every phone has to be as quiet as possible, with the allowed power increasing steadily for phones moving further away from the base station. The aim is to ensure that the signal strength at the base station is the same for every user, ensuring that all can be heard equally.
Base stations also have to synchronize their transmissions accurately to prevent too much interference—unlike TDMA, every cell is using every frequency. They do this using signals from GPS (Global Positioning System) satellites, which can pinpoint any location on Earth to within 5 m and measure time more accurately than the Earth's own rotation. Receivers for the GPS signal are built into every base station, and the location of each phone is calculated by measuring the time it takes for a radio signal to travel there and back from various base stations. Known as triangulation, this is similar to the process used by radar.
There are technically two different versions of the cdmaOne standard, listed in Table 4.11. The only difference is the frequency that they are designed for, and thus how large the cells can be. The two different versions are needed because of the precise timing requirements, which make CDMA very sensitive to differences in cell size.
| Version Name | Uplink Frequency | Downlink Frequency | Cell Size | Competes With |
|---|---|---|---|---|
| TIA/EIA/IS-95A | 824–849 MHz | 869–894 MHz | Bigger | (D-)AMPS |
| ANSI J-STD-008 | 1850–1910 MHz | 1930–1990 MHz | Smaller | GSM |
cdmaTwo
Phones based on cdmaOne already transmit and receive simultaneously and at very high data rates. Most of this data is redundant, but it doesn't have to be. A phone could use more than one of the available Walsh codes, multiplying its 16 kbps capacity by any factor up to the number of calls per channel. The theoretical maximum is 64, pushing speeds up to more than 1 Mbps. This may not be possible in reality, but the practical maximum of about 15 codes still gives a respectable data rate of 240 kbps.
Unfortunately, a lack of operator capacity and problems with electronics in phones have so far made these high rates unattainable, but cdmaOne does have another upgrade that provides speeds similar to those of GPRS and HSCSD. Known as IS-95b, or occasionally cdmaTwo, it uses four Walsh codes to produce a total bit rate of 64 kbps. Operators can deploy it with a simple software upgrade, though their customers have to buy new phones.
IS-95b doesn't mark the end of cdmaOne evolution. The prospect of 15 or more codes is tantalizing, and vendors have touted several rival schemes to achieve megabit speeds. These are discussed further in Chapter 5, "Third Generation Standards," along with upgrades that add CDMA technology to GSM.