Time-Compressing Infrasonic Recordings to Discover New Sounds, by Clark Huckaby
Audio Amplitude Detector: Circuit Description and CharacteristicsThe
audio amplitude detector acquires the envelope of an
audio signal, so that it can be recorded for time-compressed playback
using the Infrasonicon channel. Overview Part 5 has a photo and general discussion of the unit, along with links to example recordings. The schematic is shown below in Figure 1
(figure and component numbers apply to this page only). The unit was
built on a perfboard using soldered point-to-point wiring. It uses
three TL072 FET-input general-purpose low-noise dual op-amp ICs (U1-U3). The
board is framed with 3/4-inch-wide right-angle aluminum stock on which
the power supply
terminals, RCA jacks, and LEDs are mounted. The following text describes how
circuit works and its main characteristics.
Schematic diagram of Audio Amplitude Detector. Overview Part 5
gives a general discussion about this unit and its applications.
Power Supply Interface. As
indicated at the bottom left region of the schematic (Figure 1), the audio
amplitude detector operates on +/-15 VDC power supply rails. This
normally allows adequate audio headroom, which I will discuss further
below. Diodes D1 and D2 help protect the op-amps
by shorting out the power supply if it is accidentally hooked up in
reverse (the power supply should have fuse-protected outputs).
Capacitors C1 and C2 bypass the DC rails at the board's supply
terminals. Red LEDs D3 and D4 indicate power-on status; R1 and R2
are their respective current-limiting resistors. Stacked-film
capacitors C3 and C4 provide power supply bypass where the rails
terminate at U3, which is physically the most distant IC package
from the DC
supply screw terminals.
Audio Inputs and Mixer.
Two parallel-connected pairs of RCA
jacks (J1-J2 and J3-J4) serve as input and "thru" jacks for each of two
audio channels. Either channel may be used for monaural audio. The
"thru" jacks are for monitoring the audio; input impedance is 100K
ohms, so this unit does not significantly load headphone or
"line-level" outputs (common "line-level" unbalanced audio inputs have
10-K impedance). Capacitors C5 and C6 AC-couple the signals to a
inverting summing amplifier (mixer) built around op-amp U1A, in which
input resistors R3 and R4 each equal the value of feedback resistor R5.
Feedback creates a virtual ground at U1A's inverting input (Pin 2), so
the 100-K value of R3 and R4 is the impedance at each input.
output of U1A passes to an active full-wave precision rectifier
(absolute value circuit) that uses op-amps U1B and U2A. This is based
on a “cookbook” design presented by Walter G. Jung (“IC Op-Amp
Cookbook” 3rd Ed., 1986, ISBN: 0-672-22453-4; p.237, Fig. 5-9), where I
refer you for the detailed theory. Basically, U1B along with R7,
R9, D5 and D6 is an inverting half-wave active rectifier whose output
(at the cathode of D6) swings positive in accord with negative input
excursions, but is 0V during positive input swings. Op-amp U2A is an
inverting summing amplifier that mixes the
half-wave rectified signal (via R10) with the original input
waveform (via R8). Importantly, R10 is exactly half the value of R8
(and feedback resistor R13), so when positive current flows through
R10, its magnitude is double that of R8 but has the opposite sign. No
current flows in R10 during positive input excursions. The net result,
with respect to the overall
stage input (at R7/R8 junction), is that U2A outputs the negative
instantaneous absolute value both for positive
and negative input swings.
Next, the rectified audio signal is processed through a second-order
low-pass active filter built around U2B. This stage's design relied on
information from Don Lancaster's "Active Filter Cookbook" (2nd Ed.,
17th Printing, 1995, ISBN: 1-882193-31-8; while the whole book is
great, the following parts are most relevant here: p. 73, Fig. 4-5; p.
128, Fig. 6-8; p. 139, Fig. 6-15). The cutoff (corner) frequency is 20
Hz and the voltage gain in the passband is 1.6-fold; it's the only stage in the unit
with greater-than-unity gain (more about this below). The filter has a
Butterworth characteristic (flattest possible passband response).
Performance was verified in a breadboard version whose cutoff frequency
was scaled up by one decade (i.e., 200 Hz), by substituting 22 nF
capacitors for C7 and C8 while keeping R14-R17 the same.
The preceeding low-pass filter stage is non-inverting, so U2B's output
is a negative voltage, whose magnitude represents the audio amplitude
Since amplitude is an absolute value, the infrasonic waveform it
defines (as audio amplitude changes) never crosses zero--it only
reaches zero when the audio is silent. Recalling the ultimate
goal--converting amplitude envelopes into sound using time
compression--not only must the infrasonic waveform settle at zero during silent intervals, it should average
zero overall (as sound waveforms do). This is the role of the
first-order passive high-pass filter consisting of C9 and R18,
whose corner frequency is 0.10 Hz. Its follower is op-amp U3A,
configured as a unity-gain buffer which puts insignificant load on the
Output Driver Stage.
U3A drives attenuator trim-pot R20, which scales the level of the
infrasonic signal. This is followed by the inverting unity-gain buffer
made up of op-amp U3B and associated components. The infrasonic signal
is inverted here, so an increase in the amplitude envelope magnitute
makes the output voltage change in the positive direction. At U3B's
non-inverting input, trim-pot R23 allows DC-offset adjustment of the
infrasonic output signal, which feeds RCA jack J5. Offset is
necessary to match the operating point of the Data Converter (voltage input),
which is DC coupled and uses single-supply op-amps (more on this below). R24 and C10 help
stabilize the output driver stage in case a long cable is used to connect the Data Converter, although the shortest necessary cable is always preferred.
Overall Bandwidth and Gain; Operating Levels.
The audio amplitude detector's -3 dB bandwidth is 0.10 to 20 Hz. (Playing a
recorded amplitude evelope at the typical 165-fold time compression
scales this to 16.5 to 3300 Hz.) While the 20-Hz low-pass filter
is second-order within the unit described here, I normally set the Data Converter's own
third-order filter for 20 Hz when using the amplitude detector. This
yields an overall fifth-order 20-Hz-cutoff low-pass
characteristic, in which the attenuation slope reaches -30 dB per
octave as frequency climbs above 20 Hz. Importantly, the
full-wave rectifier stage effectively doubles the fundamental frequency
of high-amplitude (zero-crossing) audio signals. This helps the low-pass filter
prevent audio-frequency signals from significantly contaminating
envelope recordings. As a worst-case scenerio, say the peak audio
features are bass notes or drum beats with 40-Hz tones. The
rectifier converts these to 80 Hz, which the filter attenuates by more
than -60 dB. This is close to the Data Converter's
12-bit resolution limit (-72 dBFS [dB referred to full scale]). At 165-fold
time compression, any residual contaminant would be shifted to 13.2 KHz.
of the unit's active stages has unity voltage gain except for the
low-pass filter, where gain is 1.6-fold (4.1 dB) in the passband, as
mentioned above. On +/-15 VDC power supply rails and working into loads
of 10 K or greater, a TL072 op-amp can swing through 24 VPP without clipping (see TL072 datasheet).
To confirm un-clipped signal flow through the unit, I measured the
maximum output of my CD player (an Audio Dynamics CD-2000E) as 6.2 VPP
per channel working into no load. (The test CD track was a 1-KHz sine
waveform recorded at 0 dBFS [the maximum] on both channels.)
Thus, the unit's voltage summing stage (mixer; U1A in Figure 1) may encounter audio peaks approaching 12.4 VPP
at its output, when both channels of a CD are in phase (as they should
be most of the time for parts like kick drums and bass). The
corresponding peak signal at the output of the rectifier stage would be
-6.2 V. A cursory analysis of maximum peak level at the output of the
low pass filter would be 1.6 X -6.2 = -9.9 V, still within the op-amp's
limit which should be about -12 V. But the actual peak voltage's
absolute value can be considerably less, because the filter integrates
the rectified audio-frequency signal. Suffice it to say that CD-level
signals are safe through the gain structure of the audio amplitude
detector. Without attenuation at R20, maximum peak-to-peak amplitudes
of infrasonic (envelope) signals may approach 10 V.
When using this amplitude detector, I normally set the Data Converter's
analog block for its minimum gain, which is about 4 dB. Since it uses
single-supply rail-to-rail op-amps, its output can swing between 0 and
+5 V, which matches the input range of the 12-bit ADC that it feeds.
Thus, the Data Converter's maximum voltage input is about 3 VPP,
and the DC offset needs to be about +1.5 V. As expected, all of the
audio sources I used required some attenuation at R20, which I used to
set the level for audio envelope recordings.
Overview Part 5 has more about the Audio Amplitude Detector, and links to example sounds.
Infrasonicon Home Page
Infrasonicon Overview Sound Example Page
Go to clarkhuckaby.com
Copyright 2013 Clark
Huckaby. All rights reserved. Commercial use of this content (text, images,
sounds, or source code) without first obtaining the expressed
permission of Clark Huckaby is prohibited by
law. Distribution of copies beyond personal use and/or re-posting any or
all of this page is prohibited by law.