|
Diversity Techniques for Broadband Communication Systems
This section presents a novel
selection diversity combining technique, Partitioned General
Selection Combining (PGSC(M, L, D)), that adaptively combines M
strongest paths from each of D partitions where each
partition consists of L diversity branches as shown in Fig 1.
PGSC receiver
structure provides a trade-off between the receiver circuit
complexity and required level of performance. Bit error rate
performance of various digital modulation schemes in conjunction
with coherent and non-coherent PGSC receivers are shown below for
all common multipath fading channel models (Rayleigh, Rician and
Nakagami-m) with the assumption of independent and identical fading.
It can be observed that for large number of received signal
paths, PGSC offers comparable performance to that of GSC but at
reduced selection circuit complexity. Also, coherent and
non-coherent PGSC receivers offer comparable performance with the
former having slight diversity gain over the latter.
Comparison of
mean-SNR
Figure-1 compares
the normalized mean output SNR versus number of diversity branches
combined (M×D) for a PGSC(M, L, D) receiver operating in a
mixed fading channel and exponentially decaying MIP (δ
= 10-4) environment. The fading parameters of the
resolvable multipaths are assumed to be distributed as [K1
= 2.5, m2 = 1, m3 = 0.75, K4 = 1, m5
= 1.5, K6 = 2] with the pattern repeating itself in the
given order.
It is apparent as
the ratio M/L increases; the advantage offered by GSC becomes
minimal. Also the rate at which the mean combined SNR increases for
a given receiver configuration, declines as more number of paths are
combined.
Fig.1. Comparison of normalized mean output SNR versus number of
diversity branches combined
Comparison of SC Combiner, MRC, GSC and PGSC
Figure 2 provides
comparison between various selection combining schemes of fixed
diversity order. All the performance curves are upper and lower
bounded by PGSC(1, 12, 1) and PGSC(12, 12, 1), which
correspond to selection diversity (SD) and the maximal-ratio
combining (MRC) schemes. At very low values of mean SNR/bit/branch,
all PGSC schemes with equal number of total combined paths offer
similar ABER performances. It may be noted that as
δ
is small,
the transmitted energy is symmetrically distributed across all the
groups; hence resulting in almost identical performances of PGSC(M,
L, D) provided equal number of paths are combined.
Fig.2. ABEP
performance of BPSK with a coherent PGSC receiver in a mixed fading
channel and exponentially decaying MIP (δ
= 0.1) for fixed diversity order (L×D). The fading parameters of the
resolvable multipaths are assumed to be distributed as [K1
= 2.5, m2 = 1, m3 = 0.75, K4 = 1, m5
= 1.5, K6 = 2] with the pattern repeating itself in the
given order.
Effect of number of
partitions and MIP index
In Fig.3, the effect of number of
partitions used on the performance of PGSC receiver is considered
for different values of
δ.
It is apparent that for large values of
δ,
transmitted energy is contained in first partition, and increasing
the number of groups doesn’t have any significant impact on the ABER
performance. However, when the energy is distributed less
asymmetrically across the group, i.e. δ is close to zero, an
increase in the number of partitions used results in significant
diversity gain.
Fig.3. ABEP
performance of BPSK with a coherent PGSC(1, 3, D) receiver in a
mixed fading channel for different values of D. The fading
parameters of the resolvable multipaths are assumed to be
distributed as [K1 = 2.5, m2 = 1, m3
= 0.75, K4 = 1, m5 = 1.5, K6 = 2]
with the pattern repeating itself in the given order.
Coherent and non-coherent PGSC
Fig.4 shows the
relative performance of coherent and noncoherent PGSC(M, 6, 2)
receiver structures for π/4-DQPSK modulation scheme. The coherent
PGSC performs only slightly better in comparison to the noncoherent
PGSC receiver. Further the improvement offered by coherent PGSC
diminishes at higher values of average SNR/bit.
Fig.4. Average bit
error rate performance of π/4-DQPSK with coherent and noncoherent
PGSC(M, 6, 2) receiver structures in a mixed fading channel and
exponentially decaying MIP (δ
= 0.1) for varying M. The fading parameters of the resolvable
multipaths are assumed to be distributed as [K1 = 2.5, m2
= 1, m3 = 0.75, K4 = 1, m5 = 1.5, K6
= 2] with the pattern repeating itself in the given order. | 