Handbook for Sound Engineers

Consoles 843

25.7.12 Band Limiting

One of the first great superficially appealing results of
using the enormous feedback inherent to op-amps at the
relatively low gain requirements of the audio world was
a close approach to dc-to-light frequency response. The
author remembers well the hysterical peals of laughter
as the response of a new mixer was measured as still
0 dB right to the end of the testing ranges of the oscil-
lator and the badly disguised puzzled looks and worried
glances when we put real audio through and actually
listened to it.

Many audio signals, especially live ones from micro-
phones, analog tape-machine returns with a high vesti-
gial bias content, keyboards, and a range of other
sources, have a fair amount of ultrasonics present. If an
analog remix of a digital recording takes place (it
happens) many digital-to-analog (D/A) converters have
an embarrassment of out-of-band noise that is of no
program relevance whatsoever. A good microphone is
going to hear all manner of stuff in a space:
TV/computer scan whistles, motion alarms,
switch-mode power supply or light-dimmer inductor
screechings just for a start, none of which can be
pretended to be musical. Depending on how good a
following A/D convertor implementation may be, some
of these may well get aliased down into the audible
frequencies to less-than-subtle effect.

There is a proverb: the wider the window, the more
muck flies in. Returning focus here just to analog signal
processing, it would be perfectly all right if the
following circuitry were capable of dealing with signals
much higher than the audio band; sadly at the time (and
to a lesser degree even now) that is not so. The root of
the difficulty is the worsening open loop gain of the
individual op-amps; as it drops off at 6 dB/octave with
increasing frequency, there remains less closed loop
feedback available to maintain the op-amp’s linearity. In
other words, the circuitry becomes less and less linear
as the frequency increases and the feedback dwindles.

Fig. 25-21 is representative of the open loop (no
feedback) input-output transfer characteristic of an
op-amp—i.e., what comes out in relation to what goes
in. Not at all linear. In fact, rather nasty. (Incidentally,
most big power amps have similar curves.) The good
in-band linearity and low distortion of op-amps come
from the application of monstrous amounts of negative
feedback. Take the case of a noninverting 741-type amp
with 40 dB of gain around it, Fig. 25-22. At 100 Hz
there can be 60 dB of feedback, which is great nonlin-
earities are being corrected by roughly the tune of
1000:1! However, the open loop gain plummets above

this frequency, leaving a still respectable 40 dB of feed-

back at 1 kHz (100:1). (This figure of 40 dB is widely

regarded as the lowest amount of feedback for good

performance from an op-amp.) At 10 kHz it’s down to

20 dB; it is 14 dB at 20 kHz; and at an ultrasonic

40 kHz, there is a bare 8 dB! There is still gain, though,

and the amplifier is quite capable of supporting and

amplifying a signal up at those frequencies; it’s just not

very good at it.

Harmonic distortion of ultrasonics that would be

generated by passing through a transfer function like

Fig. 25-21 is unimportant; the frequencies would be

even more ultrasonic. The problem lies in the intermod-

ulation of two or more signals, products of which more

often than not fall into the audible band; even reciprocal

mixing with noise results in in-band noise products. A

whole slew of intermodulation products are produced. It

is no wonder that early op-amps sounded bad.

Figure 25-21. Operational amplifier open-loop gain curve

typical of a bipolar device.

Figure 25-22. A 741 with 40 dB gain.

Vin

Vo

A. Input-output curve.

B. Test circuit.

Vo

Vin

Vin

741 V 0

100 k 7

1 k (^7) Approximately 40 dB gain