B. P. Lathi, Zhi Ding - Modern Digital and Analog Communication Systems-Oxford University Press (2009)

12.10 Blind Equalization and Identification 71 1
bandwidth per station and provides a higher data rate. With OFDM, the FM !BOC subchannel
bandwidth equals 363.4 Hz, and the maximum number of subcarriers is 1093. Each subcarrier
uses QPSK modulation. On the other hand, the AM !BOC subchannel bandwidth is 181. 7
Hz (half as wide), and as many as 104 subcarriers may be used. Each subcarrier can apply
16-point QAM (secondary subcarriers) or 64 point QAM (primary subcarriers). Further details
on IBOC can be found in the book by Maxson. 1 6
12.10 BLIND EQUALIZATION AND IDENTIFICATION
Standard channel equalization and identification at receivers typically require a known (training)
signal transmitted by the transmitter to assist in system identification. Alternatively,
the training sequence can be used directly to determine the necessary channel equalizer.
Figure 12.18 illustrates how a training signal can be used in the initial setup phase of the receiver.
During the training phase, a known sequence is transmitted by the transmitter such that
the equalizer output can be compared with the desired input to form an error. The equalizer
parameters can be adjusted to minimize the mean square symbol error. At the end of the training
phase, the equalizer parameters should be near enough to their optimum values that much of
the intersymbol interference (ISi) is removed. Now that the channel input can be correctly
recovered from the equalizer output through a memoryless decision device (slicer), real data
transmission can begin. The decision output s[k - u] can be used as the correct channel input to
form the symbol error for continued equalizer adjustment or to track slow channel variations.
The adaptive equalizer then obtains its reference signal from the decision output when the
equalization system is switched to the decision-directed mode (Fig. 12. 1 8). It is evident that this
training mechanism can be applied regardless of the equalizer in use, be it TSE, FSE, or DFE.
In many communications, signals are transmitted over time-varying channels. As a result,
a periodic training signal is necessary to identify or equalize the time-varying channel response.
The drawback of this approach is evident in many communication systems where the use of
training sequence can represent significant overhead costs or may even be impractical. For
instance, no training signal is available to receivers attempting to intercept enemy communications.
In a multicast or a broadcast system, it is highly undesirable for the transmitter to start
a training session for each new receiver by temporarily suspending its normal transmission
to all existing users. As a result, there is a strong and practical need for a special kind of
channel equalizer, known as blind equalizers, that do not require the transmission of a training
sequence. Digital cable TV and cable modems are excellent examples of such systems that can
benefit from blind equalization.
There are a number of different approaches to the problem of blind equalization. In general,
blind equalization methods can be classified into direct and indirect approaches. In the direct
Figure 12. 1 8
Channel equalization
based on
a training phase
before switching
to decision
feedback mode.
Channel input
Sn
LT!
channel
H (z)
Channel
noise
w[n ]
d[n]
Sn - u
Equalizer t-----,---.i
QAM -
--
F (z)
Equalizer
design
dee(.)
Training
storage