SPEECH, MUSIC, AND NOISE MM
White and Pink Noise
References to white noise and pink noise are common and sometimes
confusing, What is the difference? White noise is analogous to white
light in that the energy of both is distributed uniformly throughout the
spectrum. In other words, white noise energy exhibits a flat distribu-
tion of energy with trequency (Fig. 5-18).
While light sont through a prism is broken down into a range of
colors. The red color is associated with the longer wavelengths of
light, that is, light in the lower frequency region. Pink noise is noise
Lieviny ligher egy ia the luw fequeucies, Lu fact, pik uvise Las
come to be identified specifically as noise exhibiting high energy in
the low frequency region with a apocific downward slope of 3 dB por
octave (Fig. 5-18C). There is a practical reason for this specific slope.
These two colorful terms arose because there are two types of
spectrum analyzers in common use. One is the constant bandwidth
Pink noise filter characteristic
10 Fiz 20 Ha
Frequency - Hz
CTE
Random noise has constant energy per Hz. If the spectrum of random noise is measured
(white) with a wave analyzer of fixed bandwidth, the resulting spectrum will be flat
with frequency as in A. If measured with an analyzer whose passband width is a given
percentage of the frequency to which it is tuned, the spectrum will slope upward at 3
4B per octave, as in B. By processing the white noise spectrum of A with a filter that
slopes downward at 3 dB per octave, such as in C, a flat response results when constant
percentage bandwidth filters are used such as octave or one-third octave filters. In mea
suring a system, pink noise is applied to the input and, if the system is flat, the read
out response will be flat if one-third octave filters, for example, are used.112 CHAPTER Five
analyzer, which has a passband of fixed width as it is tuned through-
out the spectrum. One well-known analyzer of this type has a band-
width of 5 Hz. If white noise with its flat spectrum were measured
wilh & Coustanl-bandwidth analyzer, auoler flat specirum would.
result because the fixed bandwidth would measure a constant energy
throughout the band shown in Fig. 5-18 A.
Another very populer and convenient spectrum analyzer is the
constant percentage bandwidth analyzer. In this instrument the band-
width changes with frequency. An example of this is the one-third-
octave analyzer, commonly used hecanse its handwidth follaws
reasonably well with the critical bandwidth of the human ear through-
out the audible frequency range. At 100 Hz the bandwidth of the one-
third-octave analyzer is only 23 Hz but at 10 kilz the bandwidth is
2,300 Hz. Obviously, it intercepts much greater noise energy in a one-
third octave band centered at 10 kllz than one centered at 100 Hz.
Measuring white noise with a constant-percentage analyzer would
give an upward-sloping result with a slope of 3 dB/octave, as shown in
Fig. 5-18B.
In audio-frequency measurements, the desired characteristic of many
instruments, rooms, etc. is a flat response throughout the frequency
range. Assume that the system to be measured has a characteristic almost
flat with frequency. If this system is excited with white noise and mea-
sured with the very cuuve
1 cuustan-perceutaye analyzer, Une wesull
would have an upward slope of 3 dB/octave. It would be far more desir-
able if tho measured result would be close to flat so that deviations from
flatness would be very apparent. This can be accomplished by using a
noise with a downward slope of 3 dB/octave. By passing white noise
through a filter, such as that of Fig, 5-19, such a downward sloping exci-
lation noise can be oblained. Such a noise, sloping downward at 3
dB/octave, is called pink noise. A close-to-flat system (amplifier, room)
excited with this pink noise would yield a close-to-flat response, which
would make deviations from flatness very obvious. For such reasons
pink noise is here to stay.
Signal Distorti
Our discussion of the various signals encountered in audio is
incomplete without at least an acknowledgment of what can happen‘SPEECH, MUSIC, AND NOISE 113
COTE
A simple filter for changing white noise to pink noise. It changes random noise of con-
stant energy per H2 to pink noise of constant energy per octave. Pink noise is useful in
acoustical measurements utitizing analyzers having passbands with bandwidth of a con-
stant percentage of the center frequency. General Radio Company!
to the precious signal in passing through transducers, amplifiers,
and various forms of signal processing gear. Here is an incomplete
Tist:820
Bandwidth limitation Lf the passband of an amplifier cuts lows or
highs, the signal output is different from the input. If the scratch filter
reduces record surface noise, the overall effect can be improved, but
basically the signal itself is the poorer for it
Nonuniform response Peaks aud valleys williu the passband lsu
alter the signal waveshape.
Distortions in time If tape travels across the head at any other
hifted up or
down in frequency. If there are slow or fast fluctuations in that speed,
wow and flutter are introduced and the signal is degraded.
Ph jortion Any phase shifts introduced upset the time rela-
tionship between signal components
Dynamic distortion A compressor or expander changing the
original dynamic range of a signal is a form of distortion.
Crossover distortion In class-B amplifiers, in which the output
devices conduct for only half of the cycle, any discontinuities near
zero output result in what is called crossover distortion
Nonlinear distortion If an amplifier is truly linear, there is a
one-to-one relationship between input and output. Feedback helps
to control nonlinear tendencies. The human ear is not linear. When
than the recording speed, the frequency components are