SPREAD-SPECTRUM CDMA IN HIGH-DENSITY NETWORKS For practical commercial spread-spectrum radios, the potential capacity dis- cussed in the previous section is that achievable with direct-sequence spread- ing. In principle, frequency-hopping spreading can achieve similar results but requires fast hopping implementation, which is less practical in low-cost commercial applications. Thus, this discussion is limited to direct-sequence spread-spectrum radio systems.
To understand the apparent paradox of achieving higher overall capacity with spread-spectrum radios than with conventional radios, consider a com- parison of TDMA versus CDMA for a single-star network as shown in Figure 3.4. Assuming the same fixed total bandwidth available for both TDMA and CDMA, Figure 3.4 shows the number of channels on the hori- zontal axis and the intolerable level of interference on the vertical axis.
Initially, consider the performance of a single radio or multiple radios located at the cell site that receive TDMA or CDMA signals from mobile units surrounding the cell.
TDMA is a time-dimension-limited system in which there can be no addi- tional users when all slots have been assigned. As time slots are filled by increasing numbers of TDMA users, there is no interference caused by one mobile radio to the reception of another mobile radio at the cell site. The number of TDMA users can increase until the number of dimensions (in this case, time slots) is exhausted. It is not possible to increase the number of
1176 Commercial Applications
20_c03 8/16/01 5:34 PM Page 1176
http://jntu.blog.com
Direct-Sequence Spread-Spectrum CDMA in High-Density Networks 1177
Figure 3.4.TDMA versus CDMA.
http://jntu.blog.com
users beyond the dimension limit without causing an intolerable amount of interference to reception of a mobile unit at the cell-site receiver. Thus, as shown in Figure 3.4, the TDMA system has a “brick wall” limit on capacity.
Spread-spectrum radios can tolerate some interference, as shown in Figure 3.4, so the introduction of each additional active radio raises the overall level of interference to the base station receivers receiving CDMA signals from mobile transmitters. Each mobile radio can introduce a unique level of inter- ference, reflecting its cell-site power level, timing synchronization relative to other signals at the cell site, and specific code cross-correlation with other CDMA signals.
The number of CDMA channels allowed in the star network depends on the level of total interference that can be tolerated. Unlike the time- dimension-limited TDMA system, the CDMA system is limited by inter- ference, and consequently the quality of the radio design plays a key role in its overall capacity. Certainly, a well-designed radio will have a required bit error probability with a higher level of interference than a poorly designed radio. Forward error-correction coding techniques can also increase the threshold of tolerable interference and, thus, increase overall CDMA capacity. Spread-spectrum CDMA systems, therefore, place a greater premium on good overall mobile radio design and overall system design than conventional narrowband FDMA and TDMA multiple-access radio systems.
Returning to Figure 3.4, suppose that at the cell site the signal level of all mobile users is the same and interference between radios is modeled as Gaussian noise. This is the case with A-CDMA, in which we assume that each radio has a required bit error probability that defines a required energy-per-bit-to-noise ratio given by Eb/N0. This Eb/N0defines the thresh- old shown in Figure 3.4. Given these assumptions, the relationship between the number of users M, the processing gain PG, and the required Eb/N0is given by
(3.1) which is a special case of (5.2) in Part 5, Chapter 2.
For a given bit error probability, the required Eb/N0depends on how well the radio is designed and how much error-correction coding is used.The ideal Shannon limit in white Gaussian noise shows that error-free communication is possible [22] for
(3.2) For this Shannon limit, we have
(3.3) M1.44PG.
Eb>N0ln 20.69 1.59 dB.
M PG Eb>N0
1178 Commercial Applications
20_c03 8/16/01 5:34 PM Page 1178
http://jntu.blog.com
This theoretical Shannon limit shows that spread-spectrum A-CDMA sys- tems can have more users per cell than traditional systems, which are lim- ited by the number of dimensions, such as non-overlapping frequency bands (FDMA) or time slots (TDMA). Of course, with practical coded A-CDMA radios in a cell, it may be difficult to accommodate this many users in a sin- gle cell. FDMA and TDMA systems, however, use up to 20 percent of this theoretical capacity to ensure the practical isolation between non-interfer- ing channels.
3.5.1 Data versus Voice Applications
For data applications, the bit error probabilities are typically less than 105, and consequently the threshold of tolerable interference shown in Figure 3.4 may be lower than that in digital voice applications, resulting in fewer CDMA channels. Another way to increase CDMA capacity for voice appli- cations is to use voice activation circuits that cut off radio transmit power when there is no voice activity. Since the average voice activity factor is 35 percent, using such circuits in a CDMA system dramatically lowers the noise floor or allows proportionally more active voice channels.
Owing to the higher tolerance to error and the option of voice activation, the spread-spectrum CDMA’s capacity advantage is greater with digital voice systems than with data networks.
3.5.2 Power Control
For an interference-limited CDMA system, each active mobile transmitter produces some level of interference in receivers of other mobile radio sig- nals at the cell site. Therefore, mobile radios should not transmit any more power than is necessary for satisfactory performance. Power control of mobile radios would ensure that for a fixed number of active radios, the interference caused by each radio is minimal and approximately equal.
Another benefit of power control is that the life of usable battery time is extended for mobile spread-spectrum radios.
Generally, for spread-spectrum CDMA systems, cell-site radios provide
“master” control of “slave” mobile radios. Although power control and voice activation circuits may be complex, they are typically implemented in digi- tal processing chips for which the cost keeps dropping as technology advances.
3.5.3 Time Synchronization and Orthogonal Codes
Another factor affecting the level of interference experienced by base sta- tion receivers is the time synchronization of all mobile radios. If all mobile radio signals arriving at the cell site are synchronized to within a fraction of a chip-time interval, then it is possible to reduce the level of mutual inter- ference dramatically. For such synchronized CDMA (S-CDMA) star net-
Direct-Sequence Spread-Spectrum CDMA in High-Density Networkshttp://jntu.blog.com 1179
works, the use of orthogonal codewords effectively reduces interference to zero, given sufficient orthogonal codes for distribution among mobile radios. Even with some timing errors, interference can be significantly reduced with a careful selection of codewords and time synchronization of all radios in a cell.
Codes (collections of codewords) used for such S-CDMA applications have been studied extensively and are discussed in Part 1, Chapter 5, Section 5.7. With S-CDMA, the number of non-interfering spread-spectrum mobile radios can be roughly the same as with TDMA. This is so because for the same time and bandwidth, the number of orthogonal dimensions is also the same, regardless of whether the dimensions use orthogonal frequencies (FDMA), time slots (TDMA), or spreading codes (S-CDMA). In practice, differences reflect implementation. (As we shall see in Section 3.7, S-CDMA systems can exceed the dimension limit by using more complex optimal mul- tiple-access receivers.) With S-CDMA, it is theoretically possible to have as many non-interfering mobile users as with FDMA or TDMA. If this is truly the case, why does spread-spectrum CDMA offer more capacity? As illus- trated by Qualcomm, the answer is that the 100 percent frequency of reuse allowed with spread-spectrum systems can be applied to a network of many cells in a high-density area.
3.5.4 The Outbound Channel
Up to now, the discussion of spread-spectrum CDMA focused on the recep- tion of mobile radio transmitters at the cell site. Normally, for large cells with large multipath delays, it may be difficult to time-synchronize mobile trans- mitters so that their signals arrive at the cell site within a fraction of a chip time. Such synchronization in the inbound direction becomes even more dif- ficult as the data rate increases, and, in data networks, the data rate is usu- ally very high. Thus, the inbound channel may be characterized as asynchronous CDMA (A-CDMA), with mutual interference characterized by random delays and larger delay multipath components relative to the chip-time interval. Such A-CDMA channels are usually analyzed by mak- ing an equivalent noise model of the mutual interference. With S-CDMA channels with small, random time offsets, performance analysis is usually per- formed for specific codes.
The outbound channel (cell site to mobile radios) performs very differ- ently from the inbound channel. Since transmitters for the outbound chan- nel are all located at the cell site, all radio signals to mobile radios can be easily synchronized. Each mobile radio then receives from the cell site the radio signals synchronized in time, or S-CDMA. Since the outbound chan- nel can be synchronized, orthogonal codewords are typically used for trans- missions from the cell site to mobile radios. Of course, multipath can result in interference in the mobile receivers.
For large cells such as in the digital cellular IS-95 standard radios, the inbound channel is A-CDMA and the outbound channel is S-CDMA.
1180 Commercial Applications
20_c03 8/16/01 5:34 PM Page 1180
http://jntu.blog.com
Consider the S-CDMA outbound channel in wireless LAN applications in which a wired LAN can be accessed by wireless laptop computers. On average, the amount of data outbound (LAN server to mobile laptop) is about 10 times as much as the data inbound. Hence, for inbound channels, typically only one mobile laptop can transmit at any given time, using car- rier-sensed multiple access (CSMA) protocols. With S-CDMA on the more critical outbound channel, a much higher overall capacity can be achieved. The S-CDMA channel is limited by the multipath conditions in the propagation paths from the cell site on the wired LAN to the mobile laptop units.
3.5.5 Frequency of Reuse and Antenna Sectorization
The primary advantage of direct-sequence spread-spectrum radios is that they allow reuse of the same frequency in adjacent cells in a multicellular network. Because narrowband radios are not designed to withstand much interference, narrowband radios in adjacent cells cannot share the same fre- quency band. If, as discussed above, the number of mobile users per cell is roughly the same for the same bandwidth per cell, a direct-sequence spread- spectrum radio system can employ frequency of reuse to attain a higher over- all network capacity.
Figure 3.3 illustrates an idealized array of cells in a high-density network of many cells. In this configuration, each cell has six neighboring cells. For narrowband radios, each cell would have available only one-seventh of the total available bandwidth, while the direct-sequence spread-spectrum radios with 100 percent frequency of reuse would have available the entire bandwidth available in each cell. If a spread-spectrum system has as many mobile radios per bandwidth in each cell as a narrowband radio system, its overall capacity would be 7 times greater than that of the narrowband system.
The actual achievable capacity depends on the tradeoff between many parameters in a network of direct-sequence radios. These parameters include the type of direct-sequence spread-spectrum modulation, processing gain, coding gain, level of network synchronization, data rates, use of power con- trol, specific code selection, and voice activation. Environmental issues include cell size and multipath/propagation conditions. In Section 3.6, we present an example of a radio system for an idealized microcellular network for voice personal communication service applications in which S-CDMA is used in both inbound and outbound directions.
3.5.6 Narrowbeam and Delay-Line Antennas
For a large network of microcells, one of the key advantages of having complete frequency of reuse is the elimination of the complicated fre- quency-management planning required when conventional narrowband radios are used. Also, spread-spectrum radios afford far greater flexibil-
Direct-Sequence Spread-Spectrum CDMA in High-Density Networkshttp://jntu.blog.com 1181
ity with antennas. Because these radios are less sensitive to interference, sectorized antennas, which can increase a cell’s capacity, can be used at a cell site with fewer constraints on overlap in sector antenna beams. Fixed and adaptive multibeam antennas can also be more easily employed with spread-spectrum radios. Moreover, owing to their inherent ability to over- come multipath delays greater than a chip-time interval, distributed antennas on a long cable facilitate covering all areas inside a building [16].
Again, these advantages are inherentin direct-sequence spread-spectrum radios.
Since slow frequency-hopping spread-spectrum radios are merely nar- rowband radios with time-varying center frequencies, they do not have the same frequency of reuse capability of direct-sequence spread-spectrum radios. Therefore, the higher overall capacity possible with direct-sequence radios in a multicellular network is not achievable with slow frequency-hop- ping radios. Fast frequency-hopping spread-spectrum radios with coding and interleaving can attain the same kind of capacity advantage as direct- sequence radios. However, the low cost requirements of commercial radios make such fast frequency-hopping systems less practical.