Adaptive Coded Modulation for Transmission over Fading Channels - Adaptive Signal Processing in Wireless Communications - page 71

s ( t )

( b i ) Encoder/Modulator Channel Receiver

X

FIgure 3.1 General adaptation framework: The transmitter sets parameters based on ˆ , which
contains information about the channel state, while forming the transmitted signal s ( t ) from the
information bit sequence ( b i ). Note that ˆ can take many forms: path loss/shadowing estimates
[47, 68, 70], number of errors corrected in previous packets [54, 55], explicit multipath fading
estimates [31, 32], etc.

Since all three of these effects change with time or position, they cannot be known at the
time that the system is designed. Thus, if no form of adaptation is performed while the
system is operating, the system must be designed to in some sense deal with the worst
case, which can be very expensive in terms of system resources. For example, if all users
in a cellular system have to assume worst-case path loss and shadowing (e.g., behind a
large building at the very edge of the cell) regardless of their location, they will trans-
mit at maximum power, thus maximizing battery usage and the interference to other
users. Therefore, wireless system adaptation has been a topic of critical interest in recent
years.
Wireless link adaptation can be defined as shown in Figure 3.1: any altering of the
parameters at the transmitter based on information about the current link state. Note that
link state will be defined ver y genera lly; in particular, in addition to wireless channel con-
ditions (path loss, shadowing, multipath fading), user data requirements will be included
in the definition. Methods of adaptation can be classified by the type of adaptation per-
formed (power, code rate, modulation, etc.) and the timescale of that adaptation.
The timescale of the adaptation depends on the link state phenomena for which mea-
surements are provided to the adaptation algorithm. User data rate requirements, path
loss, and shadowing change at a timescale that is long relative to the symbol rate. This
makes measurements of such phenomena relatively robust [26, 64], particularly at high
signal-to-noise ratios (SNRs), and has resulted in widespread penetration into current
and pending systems of slow adaptations, such as power control [45, 66, 68] and data
rate adaptation through variable spreading, code rate, or code aggregation [46]. Thus,
throughout this chapter, it will be largely assumed that the measurements of the path
loss and shadowing are accurate and known at both the transmitter and receiver, thus
yielding a wireless system with a known given average received signal power but experi-
encing variable multipath fading.
Multipath fading is caused by the arrival at the receiver of many signal reflections, the
superposition of which causes the instantaneous received signal power to vary widely
[57, chapter 4] as described in section 3.2. Since significant nulls can occur, this signal
fading is one of the most difficult problems to deal with in wireless communications
systems. In particular, when the received power drops too low, a burst of bit errors can
occur, and such bursts tend to dominate the error probability—even if the occurrence