TURBO EQUALIZATION - Adaptive Signal Processing: Next Generation Solutions - page 168

Figure 3.11 Bit error rate for the iterative decoder and interference cancellation
schemes, using a (5,7) encoder and a Proakis B channel.
This channel imparts a greater amount of intersymbol interference than the pre-
vious channel. Figure 11 shows the bit error rate versus the rate-adjusted SNR
again for the forward-backward channel decoder and for the linear decision feed-
back equalizer. The gains from the turbo effect are more prominent in view of the
single iteration curve that shows the performance in the absence of the turbo effect.
We observe also that poorer SNRs provoke a greater performance degradation in
the linear equalizer scheme.
3.7 CAPACITY ANALYSIS
The turbo decoding algorithm attracted significant initial attention due to its ability to
approach Shannon capacity on additive noise channels [1]. As the turbo equalizer
setup is simply a serially concatenated coding scheme, one would naturally ask if
some capacity-approaching property is inherited. In particular, by interpreting the
convolutional channel as part of a concatenated code, can the capacity be increased
beyond that attainable in a channel without intersymbol interference? We answer
here in the negative: Subject to a constraint on the SNR at the channel output, the pres-
ence of intersymbol interference can only decrease channel capacity. As such, any
gains brought by diversity from multiple transmission paths occur solely due to an
increase in the SNR: Multiple transmission paths result in greater signal power at
the receiver whenever the multiple paths combine constructively.
We first review the notion of channel capacity in terms of the number of messages
that may be reliably distinguished at the channel output [44], and then examine a
log-determinant form [45, 46] for channel capacity that lends itself readily to the
context at hand.