Cooperative Diversity: Capacity Bounds and Code Designs - Adaptive Signal Processing in Wireless Communications - page 361

vations of the bounds in the BPSK case are not included here due to space limitations
and interested readers are referred to [47] for details). The transmission rate is fixed at
0.5 bit per channel use, and the average relay power is P r = 70 dB.
It is assumed that the relay is located along a straight line from the source to the
destination, which are 10 m apart. The channel coefficient of the link from sender i to
receiver j (sender i could be the source or relay, and receiver j could be the relay or des-
tination) is c i 2 = K o d ij - n [28], where d ij is the distance from sender i to receiver j , n is the
path loss coefficient, K o = ( c /4π d o f c ) 2 , c is the light speed, d o is the free-space reference
distance, and f c is the transmission frequency. The experimental setup is fixed with f c =
2.4 GHz carrier frequency, path loss coefficient n = 3, and free-space reference distance
d o = 1 m.
Figure 12.10 shows the average source power P s as a function of the distance d between
the source and relay. The additive white Gaussian noises over the transmitting channels
are all set to be with unit variance. It is seen that, when d > 8 m, CF outperforms DF
theoretically, and the practical CF code of [47] is preferable to the practical DF code
[29]. Note that when 7.5 m < d < 8 m, where DF is superior in theory, the CF scheme still
performs better than the practical DF code.
The above practical WZC-based design can be extended to the case of the two-
receiver cooperative channel by following the CF coding strategy with forward individ-
ual decoding described in section 12.3.4. Each transmitter is equipped with one LDPC
channel encoder, and each receiver performs one distributed joint source-channel
encoding step and two channel decoding steps.
In contrast to the scarcity of WZC-based CF designs for receiver cooperation, there
have been more code designs for transmitter cooperation. For example, since the pub-
lication of the work on user cooperation [1], several research groups [32, 52-55] have
developed practical designs based on AF and DF for the wireless two-transmitter coop-
erative channel. A common characteristic of these designs is the avoidance of receive
collision by transmitting signals over orthogonal channels, which simplifies the code
design. Specifically, the two transmitters (or cooperative partners) send encoded mes-
sages over orthogonal channels during the first fraction of a time slot. Each transmitter
decodes the signal it has received from its partner and, in the case of successful decoding,
either re-encodes the recovered message using the partner's codebook (repetition-based
DF) or generates additional parity symbols out of a rate-compatible code (DF based on
incremental redundancy). The resulting codewords are then forwarded over orthogonal
channels to the receiver during the second fraction of a time slot. If a transmitter cannot
successfully decode its partner's message, it switches to either the noncooperative mode
or AF. Orthogonal signaling is achieved by using time-, frequency-, or code-division or
space-time coding.
A DF design based on incremental redundancy for a flat-fading transmitter coopera-
tive channel, dubbed coded cooperation, is proposed in [52, 53]. Two users, partners,
first exchange their messages; then each user sends only additional redundancy bits for
its partner message. That is, after successful decoding, the user transmits only additional
protection bits of a rate-compatible code for the partner's message, so that its overall
code rate is decreased. In addition, whenever the partner's message cannot be decoded,