Digital Signal Processing Reference
In-Depth Information
that the power spectral density (psd) of noise is stable over time and has a
uniform characteristic over the entire frequency band. In most circumstances
this is a fairly good approximation of reality, since many narrowband appli-
cations are designed to operate in their own, dedicated frequency band, safely
separated from the interference caused by other users. This 'separation' can
manifest itself under several forms. Apart from the physical separation such as
in wired modem applications, communication channels can also be separated
from each other using frequency selectivity (fdma), time division multiple ac-
cess (tdma) or code division multiple access (cdma). The common denomi-
nator to all these narrowband systems is that a high signal power is set against a
low (in-band) thermal noise density. This high signal-to-noise ratio allows the
use of a multilevel modulation scheme which is exploited by clever error cod-
ing algorithms such as Trellis coded modulation to achieve their near-Shannon
throughput performance.
Wideband radio systems face a very different set of environmental circum-
stances, ensuing from the frequency-dependent characteristic of a wideband
channel. First of all, there is the problem of frequency-selective fading in a
multipath environment. From a frequency domain point of view, the receiver
sees a signal-to-noise ratio which is heavily dependent of frequency. Some sec-
tions of the frequency band are completely unusable since the received signal
power drops way below the noise floor. From a time domain point of view,
a multipath channel suffers from intersymbol interference (isi) if the period
of the transmitted symbols becomes too short. Unfortunately, it turns out that
high symbol rates are inherent to all broadband systems. Some modulation
techniques, such as ofdm (802
.
11a/g), solve this problem by using multiple
carriers. Each subcarrier is modulated at a reduced symbol rate, so that the
resulting symbol period is below the delay spread of the multipath channel.
Ofdm also offers the possibility that the modulation depth of each subcarrier
is controlled independently and is adapted to the current snr as seen by the re-
ceiver in each of the subbands. This method can be regarded as a system-level
noise whitening
technique: by adapting the data throughput in each individual
subband, it can be avoided that channels with a poor snr determine the overall
error rate of the system.
The main reason why adaptive loading is still not supported by the 802
.
11a/g
standard is that the use of adaptive loading requires a feedback channel through
which the receiver must inform the transmitter of the actual signal conditions
in each subband. The major flaw of this method is that the transmitter must
always work with outdated channel information. This is because of the delay
between the actual channel estimation done by the receiver and the moment
when the transmitter is informed of these new measurements and puts them to
use [Tho00].
