Signal Processing - November 2016 - 65

practice, and incorrect modeling of the feedback path can
lead to instability.

Narrowband feedforward ANC

Ideally, noise cancelation inside a vehicle would require the
total acoustic energy distributed over the entire global region
to be minimized. Since this is an impractical task, the control
region is often sampled using one or more error sensors distributed over the control region. The total acoustic energy to be
minimized is then approximated by the sum of squares of the
sensor output as

Noises generated by mechanical components such as engines,
compressors, motors, and fans are generally narrowband and
periodic. To monitor such noise, it is sufficient to use a nonacoustic sensor (e.g., tachometer), which provides an electrical
M
(3)
J p = / p m2 ,
reference signal that contains the fundamental frequency and
m= 1
all of the harmonics of the primary noise source. This technique
has several advantages compared to a broadband ANC system
where p m is the sound pressure at the mth error sensor posiusing acoustic reference sensors; 1) the effect of feedback is
tion (m = 1g M) . The effects due to the above approximaeliminated, 2) aging and nonlinearities assotion is often comparable for low frequency
The control bandwidth
ciated with acoustic reference sensors are
control but gets increasingly noticeable at
avoided, and 3) the causality condition is
high frequencies. The accuracy of the above
of feedback control
preserved due to periodicity. In narrowband
approximation largely depends on the locasystems are inversely
ANC systems, once an electric reference
tion of the error sensors because, as menproportional to the
signal is available, a corresponding acoustic
tioned in the section "Noise Fields Inside
spacing in between
reference signal is internally generated to
Vehicles," noise field characteristics inside
the error sensor(s)
assist the noise cancelation process.
an enclosure are largely related to the encloand secondary source(s).
sure's active number of acoustic modes and
structural-acoustic coupling. The effects of
MIMO feedforward ANC
structural-acoustic
coupling on ANC inside vehicles have been
When the noise field of interest increases in size and bandthoroughly studied over the last 20 years [1], [14], [15]. In [1],
width, the number of active acoustic modes increases. To
Elliot et al. showed that at low frequencies, a single sensor is
control multiple acoustic modes, it is necessary to use
capable of achieving significant control, however with increasmultiple-channel ANC systems with multiple secondary
ing complexity of the sound field (frequency and geometry),
sources, error signals, and reference signals. A MIMO
multiple acoustic modes become active and, therefore, multiple
feedforward ANC system employs J reference sensors to
sensors are required to successfully couple into all of them. The
observe the primary noise, M error sensors to measure
frequency limit of global control was shown to be directly relatthe residual noise, and K secondary sources to produce
ed to the modal overlap or the number of acoustic modes that
the antinoise. Figure 4 illustrates the block diagram of a
are significantly excited at a given frequency f, which increases
broadband adaptive MIMO feedforward ANC system with
feedback and secondary path transfer functions. The wide
with the cube of f. Therefore, to achieve control over the entire
arrows represent a flow of vectors (multichannel acousglobal region with a size of a car, the number of sensors required
tic or electrical signals). The matrix P represents M # J
are often impractical.
To improve the control bandwidth, an alternative conprimary path transfer functions, matrix S represents
trol
strategy needed to be developed. Such a strategy was
F
K # M secondary path functions, matrix
represents
recently
investigated in [6] and [16], which attempts to conW
K # J feedback path functions, and
represents a matrix
trol the sound field within smaller spatial regions (regional
of K # J adaptive filters each serving an individual feedforward channel.

Application of feedforward
control to car noise cancelation

x (n )

In the application of car-noise cancelation, feedforward
ANC systems are mostly applied for engine noise cancelation. This is because it's easier to obtain a reference signal
directly from the engine resulting in high coherence between
the reference and error signals. Furthermore, as mentioned
previously, engine noise is often periodic. Therefore, ANC
for engine noise is often approached via a cost-effective
narrowband feedforward system using an engine speed
reference sensor, low-cost microphone error sensor(s), and
the vehicle's built-in loudspeaker system as control sources.
Such ANC systems have been commercially implemented
by a number of manufactures as discussed later in the section "Commercial Systems."

d (n )

y (n )

e (n )

x ′(n )
W

Figure 4. A broadband adaptive MIMO feedforward ANC system with
feedback and secondary paths.

IEEE SIgnal ProcESSIng MagazInE

November 2016

Table of Contents for the Digital Edition of Signal Processing - November 2016