-
+
Noise
source
y(n)
Signal
source
LMS
Primary
sensor
Reference
sensor
d(n)
x(n)
W(z)
e(n)
z
-1
Figure 8.8 Basic concept of adaptive noise canceling
x(n)
y(n)
d(n)
e(n)
+
+
+
-
Digital
filter
Adaptive
algorithm
s(n)
x
(n)
P(z)
Figure 8.9 Block diagram of adaptive noise canceler
estimate of noise x
H
n, is then subtracted from the primarychannel signal d(n), produ-
cing e(n) as the desired signal plus reduced noise.
To minimize the residual error e(n), the adaptive filter W(z) will generate an output
y(n) that is an approximation of x
H
n. Therefore the adaptive filter W(z) will converge
to the unknown plant P(z). This is the adaptive system identification scheme discussed
in Section 8.5.1. To applythe ANC effectively, the reference noise picked up bythe
reference sensor must be highlycorrelated with the noise components in the primary
signal. This condition requires a close spacing between the primaryand reference
sensors. Unfortunately, it is also critical to avoid the signal components from the signal
source being picked up bythe reference sensor. This `crosstalk' effect will degrade the
performance of ANC because the presence of the signal components in reference signal
will cause the ANC to cancel the desired signal along with the undesired noise. The
performance degradation of ANC with crosstalk includes less noise reduction, slower
convergence, and reverberant distortion in the desired signal.
Crosstalk problems maybe eliminated byplacing the primarysensor far awayfrom
the reference sensor. Unfortunately, this arrangement requires a large-order filter in
order to obtain adequate noise reduction. For example, a separation of a few meters
between the two sensors requires a filter with 1500 taps to achieve 20 dB noise reduction.
The long filter increases excess mean-square error and decreases the tracking abilityof
376
ADAPTIVE FILTERING