Circuit Operation
The LM832 has two signal paths, a main signal path and a
bandwidth control path. The main path is an audio low pass
filter comprised of a g
m
block with a variable current, and a
unity gain buffer. As seen in Figure 1, DC feedback con-
strains the low frequency gain to A
v
e b
1. Above the cutoff
frequency of the filter, the output decreases at
b
6 dB/oct
due to the action of the 0.022
m
F capacitor.
The purpose of the control path is to generate a bandwidth
control signal which replicates the ear’s sensitivity to noise
in the presence of a tone. A single control path is used for
both channels to keep the stereo image from wandering.
This is done by adding the right and left channels together
in the summing amplifier of Figure 1. The R1, R2 resistor
divider adjusts the incoming noise level to slightly open the
bandwidth of the low pass filter. Control path gain is about
60dB and is set by the gain amplifier and peak detector
gain. This large gain is needed to ensure the low pass filter
bandwidth can be opened by very low noise floors. The ca-
pacitors between the summing amplifier output and the
peak detector input determine the frequency weighting as
shown in the typical performance curves. The 1
m
F capaci-
tor at pin 10, in conjunction with internal resistors, sets the
attack and decay times. The voltage is converted into a
proportional current which is fed into the g
m
blocks. The
bandwidth sensitivity to g
m
current is 70 Hz/
m
A. In FM
stereo applications a 19 kHz pilot filter is inserted between
pin 8 and pin 9 as shown in Figure 16.
Normal methods of evaluating the frequency response of
the LM 832 can be misleading if the input signal is also
applied to the control path. Since the control path includes a
frequency weighting network, a constant amplitude but vary-
ing frequency input signal will change the audio signal path
bandwidth in a non-linear fashion. Measurements of the au-
dio signal path frequency response will therefore be in error
since the bandwidth will be changing during the measure-
ment. SeeFigure 9 for an example of the misleading results
that can be obtained from this measurement approach. Al-
though the frequency response is always flat below a single
high-frequency pole, the lower curves do not resemble sin-
gle pole responses at all.
A more accurate evaluation of the frequency response can
be seen in Figure 8. In this case the main signal path is
frequency swept while, the control path has a constant fre-
quency applied. It can be seen that different control path
frequencies each give a distinctive gain roll-off.
PSYCHOACOUSTIC BASICS
The dynamic noise reduction system is a low pass filter that
has a variable bandwidth of 1 kHz to 30 kHz, dependent on
music spectrum. The DNR system operates on three princi-
ples of psychoacoustics.
1. Music and speech can mask noise. In the absence of
source material, background noise can be very audible.
However, when music or speech is present, the human ear
is less able to distinguish the noiseDthe source material is
said to mask the noise. The degree of masking is depen-
dent on the amplitude and spectral content (frequencies) of
the source material, but in general multiple tones around 1
kHz are capable of providing excellent masking of noise
over a very wide frequency range.
2. The ear cannot detect distortion for less than 1 ms. On a
transient basis, if distortion occurs in less than 1 ms, the ear
acts as an integrator and is unable to detect it. Because of
this, signals of sufficient energy to mask noise open the
bandwidth to 90% of the maximum value in less than 1 ms.
Reducing the bandwidth to within 10% of its minimum value
is done in about 60 ms: long enough to allow the ambience
of the music to pass through, but not so long as to allow the
noise floor to become audible.
3. Reducing the audio bandwidth reduces the audibility of
noise. Audibility of noise is dependent on noise spectrum, or
how the noise energy is distributed with frequency. Depend-
ing on the tape and the recorder equalization, tape noise
spectrum may be slightly rolled off with frequency on a per
octave basis. The ear sensitivity on the other hand greatly
increases between 2 kHz and 10 kHz. Noise in this region is
extremely audible. The DNR system low pass filters this
noise. Low frequency music will not appreciably open the
DNR bandwidth, thus 2 kHz to 20 kHz noise is not heard.
Application Hints
The DNR system should always be placed before tone and
volume controls as shown in Figure 1. This is because any
adjustment of these controls would alter the noise floor
seen by the DNR control path. The sensitivity resistors R1
and R2 may need to be switched with the input selector,
depending on the noise floors of different sources, i.e., tape,
FM, phono. To determine the value of R1 and R2 in a tape
system for instance; apply tape noise (no program material)
and adjust the ratio of R1 and R2 to slightly open the band-
width of the main signal path. This can easily be done by
viewing the capacitor voltage of pin 10 with an oscilloscope,
or by using the circuit ofFigure 12. This circuit gives an LED
display of the voltage on the peak detector capacitor. Adjust
the values of R1 and R2 (their sum is always 1 k
X
) to light
the LEDs of pin 1 and pin 18. The LED bar graph does not
indicate signal level, but rather instantaneous bandwidth of
the two filters; it should not be used as a signal-level indica-
tor. For greater flexibility in setting the bandwidth sensitivity,
R1 and R2 could be replaced by a 1 k
X
potentiometer.
To change the minimum and maximum value of bandwidth,
the integrating capacitors, C3 and C10, can be scaled up or
down. Since the bandwidth is inversely proportional to the
capacitance, changing this 0.022
m
F capacitor to 0.015
m
F
will change the typical bandwidth from 1 kHz–30 kHz to 1.5
kHz–44 kHz. With C3 and C10 set at 0.022
m
F, the maxi-
mum bandwidth is typically 30 kHz. A double pole double
throw switch can be used to completely bypass DNR.
The capacitor on pin 10 in conjunction with internal resistors
sets the attack and decay times. The attack time can be
altered by changing the size of C9. Decay times can be
decreased by paralleling a resistor with C9, and increased
by increasing the value of C9.
When measuring the amount of noise reduction of DNR in a
cassette tape system, the frequency response of the cas-
sette should be flat to 10 kHz. The CCIR weighting network
has substantial gain to 8 kHz and any additional roll-off in
the cassette player will reduce the benefits of DNR noise
reduction. A typical signal-to-noise measurement circuit is
shown in Figure 13. The DNR system should be switched
from maximum bandwidth to nominal bandwidth with tape
noise as a signal source. The reduction in measured noise is
the signal-to-noise ratio improvement.
5