3.6 SPREAD-SPECTRUM CDMA FOR PCS AND PCN
3.6.2 S-CDMA Equivalent to Bit-Level TDMA
The idea of using a matched filter in direct-sequence spread-spectrum receivers is based on the fact that such filters are known to be optimum for signal detection in white Gaussian noise. In our S-CDMA system, however, we are limited not by noise but rather by interference caused by other
Spread-Spectrum CDMA for PCS and PCNhttp://jntu.blog.com 1183
spread-spectrum signals. Here, we consider an alternative to the matched fil- ters based on Ruprecht [26].
Rather than assign a unique orthogonal codeword to each mobile unit in a microcell, an alternative S-CDMA scheme is to assign to each mobile unit a common codeword with a unique time slot. Suppose we have three mobile units, each using a common spreading codeword. As illustrated in Figure 3.5, if each mobile radio uses a different time offset from the basic reference derived from the base station’s broadcast signals, then the common inphase and quadrature codeword-matched filters will have at their outputs differ- ing peaks, owing to the different offset transmissions of each mobile unit.
As shown in Figure 3.5 with the in-phase matched filter, signals from the different mobile unit signals can then be separated at the output of the com- mon matched filter. This method for achieving a set of orthogonal signals is referred to as the time-slotted S-CDMA technique.Recall that the RAKE processor in Qualcomm’s system is designed to separate the different mul- tipath signals from a mobile unit if the multipath delays are longer than the chip time. Such a RAKE processor combines the outputs to take advantage of this path diversity.
Note that this time-slotted S-CDMA technique also requires time syn- chronization and that orthogonality is provided by time-slot assignment sim- ilar to that in TDMA. The output of the matched filter is essentially a TDMA signal at the bit level. Thus, in a microcell, there are as many S-CDMA orthogonal signals as TDMA signals. The important difference is that in a high-density application with many microcells in a network, S-CDMA can
1184 Commercial Applications
Figure 3.5. Time-slotted S-CDMA with matched filters.
20_c03 8/16/01 5:34 PM Page 1184
http://jntu.blog.com
tolerate more interference from other microcells and thus allow greater fre- quency of reuse, which results in an overall higher capacity when it is mea- sured in the number of mobile units per high-density area.
This slotted S-CDMA system has the advantage that only one set of inphase and quadrature matched filters is needed for all the mobile units assigned to the base station. The problem, however, is that with matched fil- ters, signals are not completely orthogonal. In general, each signal’s matched filter output has some non-zero terms that contribute to interference at other time slots. This is illustrated in Figure 3.5.
Ruprecht [25] noted that the matched filter is optimum for detecting sig- nals in white Gaussian noise but not necessarily for other situations. For example, he considered an application in which spread-spectrum signals function as sounders (or radar signals) used to determine propagation chan- nel characteristics (or target characteristics). In particular, if such spread- spectrum signals are used for estimating multipaths in a channel, it is important that pulses from different paths not interfere in the measure- ments.
In the interest of minimizing mutual interference between pulses from dif- ferent multipath returns, Ruprecht determined that the maximum-likelihood estimation of multipath channels would use filters that are inverse to the transmitted spread-spectrum pulse signal rather than matched filters.
Suppose we regard a codeword as an impulse response of a filter which, of course, is of finite duration. The ideal inverse filter is a real-valued wave- form of infinite duration. Naturally, by definition, when a codeword enters the inverse filter, the output is ideally a delta function.A similar discrete time version can also be shown to give only a single non-zero output sample for the discrete time-inverse filter.
If such an ideal inverse filter is used in place of the matched filter, no inter- ference of the time-shifted signals is seen at the filter output. Thus, in the- ory, inverse filters in a time slotted S-CDMA system produce truly orthogonal signals at the output of the inverse filter.
As discussed earlier, this time slotted S-CDMA technique can be con- sidered as a bit-level TDMA scheme in which the CDMA signals are con- verted to TDMA signals. The inverse filter for the common codeword treated as an impulse response, however, has infinite duration. Ruprecht considered various levels of time truncation for the inverse filter and found that in most cases truncation to 3 times the codeword duration resulted in
“slight” interference among signals at the filter output samples. He also con- ducted an exhaustive search for short codewords and their truncated inverse filters and found the best ones when white Gaussian noise is in the channel.
Although the use of a truncated inverse filter results in a single filter rather than a bank of matched filters for separating all mobile units’ signals at the base station, implementation of such filters is complicated because, unlike codewords, they are not binary levels. Many ways exist for developing eas-
Spread-Spectrum CDMA for PCS and PCNhttp://jntu.blog.com 1185
ier-to-implement approximations to the inverse filters at the cost of some mutual interference.