Interference, Multipath, and Scintillation - Understanding GPS: Principles and Applications - page 294

replicate the desired signal, but rather approximates the derivative of the desired sig-
nal. The resulting correlation between the received signal and the modified reference
has a much sharper correlation peak (along with, for some approaches, small arti-
facts at larger delays) than the original signal, providing better resolution of
multipaths just as P(Y) code signal provides better resolution as shown in Section
6.3.2. These approaches provide little or no benefit for multipaths with very small
excess delay—a few tens of nanoseconds for BPSK-R(1) modulations—but they do
provide enhanced performance for multipaths with larger delays, compensating in
part for the limitations of narrower bandwidth modulations (as long as the
precorrelation bandwidth is wide). As discussed later in this section, however, their
benefits are offset to some degree by poorer performance in noise and interference
compared to use of conventional early-late processing with more capable modula-
tions and the same precorrelation bandwidth.
Most parametric approaches rely on the discrete model of multipath defined in
(6.42) or (6.43). A parametric algorithm either estimates or assumes the number of
multipaths and then estimates nuisance parameters such as MDR, excess delay, and
relative carrier phase of each multipath. Typically, these parametric approaches
employ carrier-coherent processing and use very long coherent integration times
(greater than 1 second), requiring the received multipath characteristics (including
phase relative to the direct path) to be constant over the integration time. One such
approach is the multipath estimating delay lock loop (MEDLL) [24], which applies
maximum likelihood estimation theory to minimize the mean-squared error
between the received signal, modeled as in (6.42), and the locally generated refer-
ence signal. Other approaches have been proposed [25] and shown to minimize
mean-squared error and RMS error for specific multipath models.
Only limited evaluations have been published to describe the effect of noise and
interference on performance of multipath mitigation techniques. The analysis in
[23] shows that modified-reference processing degrades post-correlation SNR by
large amounts. However, this degradation is easily overcome by use of a conven-
tional prompt correlator with a locally generated reference signal matched to the
transmitted signal. The results in [23] also demonstrate, however, that the code
tracking accuracy of modified reference techniques in white noise is degraded rela-
tive to conventional early-late processing by an amount equivalent to reducing the
signal power by 3 dB at higher input signal-to-noise conditions, and perhaps greater
amounts at C / N 0 less than 35 dB-Hz. While the effect of nonwhite interference has
not been evaluated, it can be expected that performance would be degraded more by
interference with power concentrated away from band center, compared to conven-
tional early-late processing. This increased sensitivity to noise and interference can
be offset by use of narrower loop bandwidths, although practical considerations
impose limits on narrowing loop bandwidths.
Multipath mitigation remains an area of active research interest. Designs of new
GNSS signals provide opportunities for new modulation designs, and better perfor-
mance in multipath can be one consideration. There are, however, many other con-
straints and factors that must be considered in GNSS modulation design, including
issues that arise in sharing frequency bands with multiple signals. The increasing
opportunity to process signals at multiple frequency bands opens up new potential
for multipath mitigation processing that takes advantage of multiple carrier fre-