Revealing the infrasonic underworld cheaply, Part 2

In Part 1 of this Design Idea (DI), we saw how a standard electret microphone capsule can be persuaded to detect infrasonic signals down to a fraction of a Hertz by adding some fairly simple equalization. In this second and concluding part, we will improve that circuitry and also add an audio output to allow us to hear the otherwise inaudible.

Wow the engineering world with your unique design: Design Ideas Submission Guide

Figure 1 shows the revised schematic. While the equalization is much the same, the circuitry around the mic itself is more elaborate. Originally, the mic was fed straight from the power rail, potentially causing feedback problems. Now, it is powered from a clean reference source and also enclosed in a feedback loop to help stabilize its operating point.

Figure 1 A new input buffer stabilizes the mic’s operation, while other additions improve the circuit’s performance.

R1 and D1 define a nominal +1.24-V reference supply for the mic’s positive terminal. (D1 is not specified. I used an LM385-1.2, but that family is now obsolete—why?? They seemed trouble-free and well-specified, the -ADJ version being especially useful. But LMV431s or LM4041s look good.) In the absence of any offset, A1’s output rests at -1.24 V, needed for the low end of the mic’s load resistor R4. When everything is set up and stable, the junction of R4 and the mic’s negative terminal is close to 0 V (or common). The mic’s signal rides on that, and is amplified by A2, a portion of it being fed back into A1 by R5 to help to stabilize the mic’s operating point.

Calibration and setting up

All microphones behave slightly differently, mainly owing to the spreads on their internal JFETs, so some initial calibration is needed: the load resistor R4 must be trimmed. Sw1a allows R5 to be open-circuited to speed up this initial adjustment, while Sw1b keeps C2 out of circuit (see below) for the same reason. They are shown in the operating position on the schematic.

Once set, operation will be stabilized by feedback, largely compensating for temperature changes. Without the feedback from A2, the tempco of the voltage across the mic measured around 13 mV/°C; with it, it drops to around 1. R18 trims the residual offset, which can otherwise be up to 200 mV at the output of A4.

C1, at 100n, is enough to hold A1 stable during calibration, but it is C2—10 µF—that defines the low 3-dB point of A1’s circuit, once allowance has been made for the reduction in its effective value by the feedback. If C2 is in circuit at start-up, things take takes ages to settle, so that capacitor is pre-charged and only switched seconds after switch-on. (Charge-injection, as shown in Part 1’s Figure 5, scarcely helped here, perhaps because the feedback neutralizes it.)

U1a generates that delay. C3 and R6 control the period, with R7 and C4 adding positive feedback for a snappy action. U1a then switches U1b to control how C2 is connected. During start-up, C2 is connected between A1’s output (-1.24 V) and the common rail, charging it to its expected operating voltage; during operation, it spans R3, defining the circuit’s time-constant.

That hysteresis is necessary because controlling a ’4053 directly with the slowly changing voltage from an RC network leads to it oscillating (at least with the Motorola and RCA devices to hand; those names alone date them). The chosen delay is longer than strictly necessary, given the values of C1 and R3, but it allows time for the rest of the circuit to settle better. This leaves a spare section of the ’4053; using it to repeat the pre-charging trick on C5 or to short out R14+15 during the start-up period made little difference to the overall settling time.

Acoustically isolating the mic and allowing both it and the circuit to settle properly before calibration is necessary. Temporarily enclosing it in two hollow hemispheres of modeling clay, loosely sealed together, works well. During operation, it should be shielded from any air movements. Even a single sheet of fabric suffices, but a block of open-cell plastic foam or a wodge of acoustic fiber should be even better.

 The rest of the circuit

The circuit around A3 is unchanged except for the added offset-trimming pot R18. A4 adds a two-pole Sallen–Key low-pass filter (f_3dB ≈ 12 Hz) to the signal path so that, along with the roll-off from A3, any 50/60 Hz components are attenuated by 35 to 40 dB.

C8 and R14+15 define the overall low-end cut-off which, with the values shown, can vary from about 300 mHz to 1.7 Hz (3 dB points). With R15 maxed out, C1/R3’s time-constant is dominant.

In Part 1, we tried a meter for indication but found it to be rather slow. However, it now becomes a useful add-on, allowing the mic’s load resistance and the offset trim to be set easily. It still indicates the lowest frequencies well.

Figure 2 shows the response to changes in air pressure with R15 set to both its maximum and minimum values. While these are LTspice-derived traces, they closely match the real-world measurements. Compare the top, red trace with Figure 2 in Part 1.

Figure 2 The calculated frequency responses of the circuit with the limiting values of R15.

Actual results are shown in Figure 3. These used the test rig described in Part 1, and were taken with R15 set for maximum bandwidth. Like Part 1’s Figure 5, with which it can be compared, it was scanned manually, so don’t trust the frequency scale to be truly logarithmic. As before, the trace wanders vertically because of flicker or 1/f noise from the JFET, but the overall response and linearity are both clear.

Figure 3 Measured response using a real microphone in the pressure-chamber test rig.

The microphone used was the 10mm-diameter type which was to hand. “Other types are available” but may work differently. Tests using salvaged 5-mm units imply that the sensitivity is roughly proportional to the diaphragm’s area—or the square of its diameter—which seems reasonable. The 5-mm devices were both newer and quieter, presumably like their internal JFETs, so their overall S/N ratios were similar. Use the largest ones you can.

Hearing the infrasound

If we take the infrasonic signals and use them to modulate an audio tone, we can then hear what’s going on, or at least a proxy for it. A recent DI was for a pitch-linear VCO: this DI is of course the project (or gadget) for which that was needed (or wanted).

The oscillator used here is almost identical to one of the variants in that article. It’s shown in Figure 4. We won’t describe its operation here (you can refer back for the details) but there are some changes and additions.

Figure 4 Frequency-modulating an audio tone lets us hear the form of the infrasonics.

As before, the main part of it generates a tone whose frequency is centered at around 500 Hz and which varies by plus or minus an octave—doubling or halving the frequency—for control inputs ranging from plus to minus a volt (roughly), so that it is linear in pitch rather than frequency. That control input is of course the detected infrasonic signal.

Under extreme conditions, that signal can span the power rails—up to ±2.5 V—so it is potted down by R23 and R24. (Something non-linear in place of R24 is tempting but untried. That should allow low-level signals through almost unchanged while compressing the peaks. Perhaps two pairs of back-to-back 1N4148s, with a higher value for R23…)

The audio square wave from U2b feeds R28 and the pair of limiting diodes bridged by C13 and C14 to give a trapezoid of about a volt peak-to-peak. That may be excessive, so C13/14 also pot it down to ~100 mV pk–pk. A8 buffers the signal, R29 providing a ground path while scarcely shunting C14.

For a straightforward ~1 V pk–pk audio output, short out C13, make C14 33 nF, and eliminate R29 and pot R30. Even simpler (and cruder) would be to feed the phones directly from U2 through a 5k pot acting as a volume control. Extra filtering caps across the phones could then be added to taste.

Mixing and matching

We may want that lower signal level because it’s then comparable with the output from A2, which is the mic’s amplified wideband audio signal. R30 lets us cross-fade between that and our tones, should we want to. (And we may, a little later.) The output can now be fed to a power amp (I used a TDA7052A—not shown) and speaker or earphones (with a series resistor). With the left and right ’phones in parallel, the sound is roughly in the middle of your head; connecting them in series (out of phase) gives an “out there” effect, which can be less distracting if you also want to hear the complex soundscape of the planes, trains, and automobiles causing the infrasonics.

While it might be nice to include a sound file here so that you can hear the results, we will have to make do with something visual: a typical trace, showing a passing high-speed train a few hundred meters distant while a couple of planes much further away approach Heathrow Airport.

Figure 5 The effects of a nearby train, a plane or two, and some local traffic can be seen in this trace. Note the x-axis time scale.

Air is not the only element

We mentioned cross-fading earlier, but why might we want it? Two things that I have yet to try will be under the ground and underwater. Sealing a mic in a suitable and appropriately-weighted drinks bottle should make an interesting hydrophone, and the local canal will soon be awash with propeller noises. But will these mainly be directly audible or much lower in frequency? Pressure variations should couple through the bottle walls to the air within and thence to the mic, though probably with low efficiently. Filling the bottle with oil might improve that, at least until it seeps into the mic’s innards.

Another bottle, buried in and grouted into the ground (the local bedrock is chalk, exposed in places) may work as a geophone, not that there should be much audio in that underworld. Seismically quiet it may be round here, but we still get the odd rumble at magnitude 2 or 3.

Final flights of fancy

Wouldn’t it be nice to have a pair of mics, suitably spaced to make a stereo pair and each with its own equalization, perhaps with their summed outputs controlling the tone’s frequency and their individual ones adjusting the relative left–right amplitudes? Control of phase might also be needed, that being the main source of directional information at low frequencies, which could involve some rather complicated voltage-controlled all-pass filtering. Or something. Or a DSP.

If you want to see what electrets can do if customized by serious funding, search for “NASA infrasound” which will produce a slew of fascinating results, spanning environments from the ocean depths to the edge of space. While this DI cannot compete with NASA’s sub-millihertz detection capability, it should be just as much fun and is certainly rather cheaper.

Editor’s Note:

 Part 1 of this DI uses an electret mic to detect infrasound. It starts with a basic equalization circuit validated with a DIY test fixture and simulations, and ends with a deeper analysis of the circuit’s real response.

 Part 2 includes refinements to make the circuit more usable while extending its detectable spectrum with an additional technique that allows us to hear the infrasonic signals.

—Nick Cornford built his first crystal set at 10, and since then has designed professional audio equipment, many datacomm products, and technical security kit. He has at last retired. Mostly. Sort of.

 Related Content

• Revealing the infrasonic underworld cheaply, Part 1
• Squashed triangles: sines, but with teeth?
• A pitch-linear VCO, part 1: Getting it going
• Earplugs ready? Let’s make some noise!
• Supersized log-scale audio meter

The post Revealing the infrasonic underworld cheaply, Part 2 appeared first on EDN.