Adaptive Coded Modulation for Transmission over Fading Channels - Adaptive Signal Processing in Wireless Communications - page 76

Such adaptations will be called “slow” adaptations throughout this work, and they will
be characterized by schemes that adapt the transmitter at an interval on the order of (at
least) many (hundreds of) symbols. A good tutorial on slow adaptations, particularly in
current standards, is provided in [48].
First, consider wireless system adaptations that adapt depending on user needs. In
particular, one of the key features of third-generation cellular systems is supporting
users with high data rates. This is often done by simply allocating more of the time/
bandwidth/code space to the users. For example, in Enhanced Data Rates for GSM Evo-
lution (EDGE) systems, which are built on a time-division multiple-access (TDMA)
framework, users with high-data-rate needs are allocated more time slots. In the code-
division multiple-access (CDMA)-based IS-95 Revision B, high-data-rate users are allo-
cated multiple spreading codes, which is termed code aggregation [48].
Next, consider adaptations based on the current channel conditions for a given user.
In fixed-rate systems, such as first- and second-generation cellular telephone systems,
where the rate of the vocoder is generally fixed, the key is to adapt the system such that
acceptable performance is maintained at this fixed rate. The transmission technology
and channel assumptions fix a minimum average received SNR γ 0 required for accept-
able operation—the goal of adaptation is to maintain γ 0 , which can be done by adapting
the transmitted power in response to measurements of the path loss and shadowing.
Methods of performing such adaptation include channel inversion, where the trans-
mitted power is set proportional to the channel loss, and truncated channel inversion
[e.g., 17, 66]. Truncated channel inversion is defined by a threshold L 0 , which breaks the
policy into two cases:

1. L ( t 0 ) ≥ L 0 : The transmitted power is set to γ 0 / L ( t 0 ), which results in an average
received SNR of γ 0 .

2. L ( t 0 ) < L 0 : The transmitted power is set to zero, which results in an outage.

Using this policy, the required average received SNR (and no more) is obtained when-
ever possible, but excessive power is not wasted by inverting the channel when there are
large losses on the channel.
Outside of current standards, the setting of the rate (coding, modulation, spreading
factor) of the system to the current average received SNR has clearly emerged as a criti-
cal topic. In particular, turbo codes [4] and low-density parity-check (LDPC) codes [23,
46] are approaching channel capacities on a variety of channels. Thus, assuming enough
receiver complexity for the decoding of such codes and enough latency to allow perfect
interleaving of the coded symbols, the rate of nearly error-free systems should approach
the Shannon capacity [59] of the independent and identically distributed (IID) discrete-
time Rayleigh fading channel, defined by:

YXn
i

= + α ,

i i i

where Y i is the received sequence, α i is the IID sequence of Rayleigh channel fading val-
ues, X i is the transmitted sequence, and n i is the noise sequence. The Shannon capacity for