Yet its claim was true. This spread-spectrum introduced a new para- digm for designing multicell voice networks using CDMA [15, 16]. It is based on the fact that the main criterion for efficiency, or capacity, shouldbe this:
FOR A GIVEN TOTAL BANDWIDTH, CAPACITY IS THE MAXIMUM NUMBER OF ACTIVE USERS IN A HIGH-DENSITY
NETWORK OF CELLS.
1 3 1
2
A New Paradigm for Designing Radio Networkshttp://jntu.blog.com 1173
Most earlier works focused on the capacity of a multiple-access system in terms of the number of users per bandwidth or the number of users per iso- lated cell in a star network. The key point is that in a multicellular system illustrated in Figure 3.3, spatial isolation and voice activation are critical for achieving a high overall capacity as defined above. Spatial isolation owing to the typical terrestrial ultra-high-frequency (UHF) propagation attenua- tion with the distance to the fourth power, together with the added protec- tion of spread-spectrum techniques, allows for 100 percent frequency of reuse among all cells in a network [15]. This 100 percent reuse is the reason that, in a well-designed spread-spectrum radio network, the overall capac- ity can be much greater than that with conventional narrowband radio mul- tiple-access techniques.
Note that this is a more global capacity criterion. Capacity does not con- sider the maximum number of users per bandwidth or the maximum num- ber of users per cell, but is more practically based on the total number of users in the total area covered by multiple cells. Normally, evaluating sys- tems on the basis of users per bandwidth, users per cell, and users per area
1174 Commercial Applications
Figure 3.3. High-density network of microcells. For a given bandwidth and size of each microcell, what is the total number of active channels?
20_c03 8/16/01 5:34 PM Page 1174
http://jntu.blog.com
would give the same result.With spread-spectrum radios, however, these are not the same. The key difference lies in the frequency of reuse among cells in the network, which differs for spread-spectrum radios.
The design of networks using spread-spectrum radios can aim to limit mutual interference but cannot necessarily eliminate it completely, as with conventional radio systems. Spread-spectrum radios are designed to toler- ate some level of interference, with their overall capacity limited by how well this mutual inference can be controlled. Conventional radios are limited by the number of non-interfering signals achieved by using complete separa- tion with FDMA or TDMA techniques. Thus, the capacity of spread-spec- trum radio systems is interference-limited, while the capacity of conventional radio systems is limited by the number of non-interfering coordinates. We restate this important difference as follows:
THE CAPACITY OF SPREAD SPECTRUM RADIOS IS INTERFERENCE-LIMITED, WHILE THE CAPACITY OF
NARROWBAND RADIOS IS DIMENSION-LIMITED.
The paradigm for the design of a radio system consisting of a large net- work of cells changes with the use of spread-spectrum radios. When overall capacity is considered, as measured in the total number of active users in an area of many cells (capacity per square kilometer, e.g.), a well-designed spread-spectrum system can achieve a higher overall capacity than one using conventional narrowband radios. In such interference-limited radio net- works, power control is critical to overall capacity. For digital voice networks, the advantage is greater with spread spectrum, because mutual interference can be further reduced (thus increasing capacity) with the use of voice acti- vation. Finally, this type of design puts more pressure on good radio design, in which better performance against interference—which translates to less transmitted power—is critical.Thus, the use of forward error-correction cod- ing can play an important role.
Narrowband signals using FDMA need a frequency guard band to sepa- rate the FDMA channels to maintain a low co-channel interference level.
Similar guard times are needed for narrowband signals to separate assigned time slots in TDMA systems. Up to 20 percent of the total capacity needed for guard bands (FDMA) or guard times (TDMA) is eliminated in the spread-spectrum CDMA systems, which are designed to perform with some level of interference.
In Part 5, Chapter 2, we make a distinction between SSMA and CDMA.
With SSMA, we assume that transmitters are not synchronized in time and that interference from other users is modeled as Gaussian noise. Massey and Mittelholzer [21] have characterized the case in which users are not time-synchronized as “asynchronous code division multiple access” (A- CDMA). Here, one can use the SSMA assumption of Gaussian noise interference or take into account all the random time shifts of the inter- ference signals. Massey and Mittelholzer characterize the case when trans-
A New Paradigm for Designing Radio Networkshttp://jntu.blog.com 1175
mitters are all time-synchronized as “synchronous code division multiple access” (S-CDMA). For S-CDMA systems, specific spread-spectrum spreading codes can enhance performance to a greater degree than in A- CDMA systems.
Although the IS-95 system uses an overall network time reference, the inbound signal (from mobile units to cell site) is treated as an A-CDMA system, and interference can be modeled as Gaussian noise. For large cell sizes with some multipath delays larger than a chip-time interval, mobile unit signals cannot be easily synchronized. In this direction, signals in a cell are not orthogonal. The outbound signal (from cell site to mobile units) is S-CDMA, since all signals originate from one location and can be time-syn- chronized.
In military spread-spectrum radio applications, interference derives from enemy jammers not under any control by the communication system. For a large network of cells in a commercial application, on the other hand, inter- ference canbe tightly controlled among all radios, since all are part of the name system. This tight control of mutual interference among spread-spec- trum radios within a network of cells is the key to achieving high overall capacity for that network.