Infrasound sensor - what makes a narrowband 2.7 Hz signal?

[JBeale]

> http://www.openmusiclabs.com/learning/sensors/electret-microphones/index.html
Thanks for that link, it's a very good explanation of the electret capsule. I didn't realize they were that simple inside.  The model I was using, I think it was a "Primo EM182" is about 10mm diameter and claims higher SNR than most other capsules; maybe due to a larger diaphragm and and/or better JFET (?)

The experiment could be fun! I can almost guarantee you can detect a door opening from quite some distance away, more than you would expect just by listening with your ear anyway.  That's on a day with calm weather. If it's windy outside (and/or if you have HVAC running) mostly you just pick up pressure fluctuations from that.  Otherwise I'd expect this type of sensor would be available in burglar alarms, because when conditions are quiet you can really pick up motion from all over the house.

[CatalinaWOW]

A fun extension of your project would be to set up two or more of your sensors as a phased array.  You could then get direction information which would help in identifying the source.

[StillTrying]

A fun extension of your project would be to set up two or more of your sensors as a phased array.  You could then get direction information which would help in identifying the source.

At these VLFs you'd need quite a bit of space.

"Arrays are also used to estimate the direction of arrival of a wave and its propagation velocity. Typical sizes, i.e., apertures, are in the order of 100 to 1000 meters."
https://www.knmi.nl/kennis-en-datacentrum/achtergrond/infrasound-and-seismology-in-the-low-frequency-array-lofar

[JBeale]

I definitely have thought about multiple sensors for a direction-finding array. The spacing required is exactly the problem. Doable if you've got a big farm, etc. but for us suburban folks, it means you need to get some friends living some distance away also involved. Maybe I could interest someone at the kids' school for a science project.

There is actually a kickstarter running now with exactly what you would need, a networked infrasound sensor. If interested check out "Raspberry Boom". It is from a group "OSOP" and I have their previous kickstarter item, a home seismograph which does locate earthquakes using a network of sensors. How feasible that is with infrasound, I do not know.

[CopperCone]

m.youtube.com/watch?v=IWR3dc7kkAA

[CatalinaWOW]

Space requirement depends on accuracy required.  Think of this in the time domain.  Speed of sound is around 330 meters/sec.  If your sensors are 10 meters apart (should be possible in many urban situations), the distance difference is approximately sin(theta)*10 where theta is the angle between the source of the sound and the perpendicular to your sensor baseline.  Or 1.7 meters for a 10 degree offset.  So if you can resolve 1.7/330 second (or 5 millisecond) time of arrival differences you can tell angle to roughly ten degrees.  For nice sinusoids of the type you are hunting measuring the phase will probably be easier than measuring time of arrival, but the basic concept is the same.

There are issues, but there are advantages to a short baseline.  Such as no ambiguity due to full cycle phase shifts from large angular offsets.  There are lots of other things to consider.  Multi-path being probably the worst.

[CopperCone]

Do you have any examples of such a low frequency phased array system?

I am interested in building one.

[CatalinaWOW]

I don't though I am sure that the nation's who fear submarines have built them.  The people playing these kind of games don't publish much.

[texaspyro]

google:  infrasound smu

[StillTrying]

A couple of rough experiments...

Even in doors in very quiet conditions the air always seems very noisy ~0.5 - 20Hz, so noisy that I wired up a bare electret Yellow, to compare with the 0.2Hz<80Hz X 50 amp one Purple. There's audio, - TV news on the Yellow one, you can see how large the LF noise is compared with the audio.

The .wav is 2 X 1 sec. of 8-bit samples of the randomness, collected off the scopes .csvs and converted to a mono 8 bit .wav, playing at 8 X speed. Experiments continue...

[David Hess]

Usually when I see something like this, it is caused by either aliasing, mixing of two much higher frequency signals, or AM detection.  EMI protection should be used at every susceptible analog input and output.  The various cellular communication and wireless networking standards which use TDMA are particularly troublesome because the detected AM shows up as DC offset shifts.

What was the sample rate and what kind of filtering was used?

[JBeale]

Thanks for the scope plot; interesting to see. If the yellow trace is simply a bare electret going direct to the scope, with no low-pass filters, I think it really is air pressure because I would expect RFI to look more bursty. If you directly pickup RF from wifi, cell phones etc. you should see very short vertical spikes during the brief transmissions.  If there is even the slightest stir of wind outside, it can cause low-frequency pressure changes indoors like this. In my area the winds die down after sunset and there is very little wind noise from midnight - 6 am.

You could prove it by enclosing the electret in some kind of a sealed, airtight jar.

[Kalvin]

If you want to make signal audible in real-time, you could try AM modulating for example 1000 Hz carrier with the signal. Alternatively, you could FM modulate the carrier. FM modulation can be achieved quite easily with a MCU by creating a sine lookup table (for example 1000 values) and make the signal-to-be -detected adjust the index step size. Of course you can use any other more sophisticated spectrum shifting algorithms.

[StillTrying]

"What was the sample rate and what kind of filtering was used?"

I've just got an op amp on a bread board, it's definitely not very sophisticated!

"You could prove it by enclosing the electret in some kind of a sealed, airtight jar."

It took me 4 tries to get the electret completely air tight sealed into a small glass salt cellar with a ball of blutack in the neck, purple trace.
The second purple trace is with a very small air leak in the blutack - the electret still at the bottom of the glass.
After seeing the quiet trace in the sealed salt cellar, I think a lot of the mains hum is actually audio through the air, from the TV/PC/Lights.

"If you want to make signal audible in real-time"
I'll try adding some sine to the wav files, but the sample rates are usually quite low 100 - 5000 Sa/s.

[JBeale]

Nice work. Sure enough, from that experiment it seems pretty clear you are mostly picking up actual sound (or infrasound) with the bare electret.  That was part of my interest in fact- what kinds of sounds can I pick up, which we don't normally notice?

[StillTrying]

Longish post got 502'd 

"what kinds of sounds can I pick up, which we don't normally notice?"

Any interest in trying to make the infra sound real-time hear-able?

I tried just adding/mixing some 400-600 Hz at amplitudes above and below that of the infra sound, that did make some of the VLF changes hear-able.

I tried fully modulating the amplitude of 400-600 Hz with the VLF, that was better at making the VLF changes hear-able. I can't figure how to simply do FM.
These first 2 were very tiring to listen to for the VLF changes, even for just a minute or two.

I tried just squaring up the VLF comparator-like, that seems to be the best so far in making the VLF changes audible.
The squared up version sounds very much like a buzzing bad connection, I think this version is worth a few more lines of C, - a 50 sample running average to accurately follow the the mid-level, and at least some variation to the square's height/volume based on the actual current p-p height of the VLF. The running average is easy enough, I'm still thinking on the reached peak detection. I'll post an audio sample when if I get anywhere.

[CopperCone]

post the raw data im sure some people know exactly what to do

[JBeale]

Interesting ideas with the FM modulation. All I had done before with audio was simply a direct speed-up, for example with playback at 100x speed I can identify some infrasound signals.  Other than that I was just using spectrograms, which make really narrowband features easier to pick out, but for any more complex things I'd agree your ear is likely better than your eye.

[Gregg]

It could be paving equipment.  Caltrans does a lot of work at night; they usually try to finish by 5 AM, but will go to 6 if it has to be done. 

[StillTrying]

Here's a few seconds of the squared infra sound, it might sound awful but it certainly makes the silent LF very audible at X1 speed. I'd still like to get the squared amplitude to increase when a 'loud' infra sound comes along.

Here's the lines that make some adjustment to the center 127/128 position of the input wave 15 samples ahead of the squaring comparator, ...there's probably a better way.

Code: [Select]

const char    Hyst=3;       // hysteresis
float   Center = 127.5;
for (i = 1; i<NumSamples; i++)
    {
    if  ( ADCSample[i+15] > Center ) Center += 0.08;
    if  ( ADCSample[i+15] < Center ) Center -= 0.08;

    if ( ADCSample[i] > (Center+Hyst) ) ADCSample[i] = 127 + 85;         // set o/p high
        else if ( ADCSample[i] < (Center-Hyst))  ADCSample[i] = 127 - 85; // set o/p low
        else    ADCSample[i] = ADCSample[i-1];                // no change
    }
Edit: font=courier goes italic for some reason.

[StillTrying]

Anyone still interested in infra sound experiments. 

Using the VLF to FM modulate a 600Hz sine wave certainly makes the VLF wobbles hear-able at X1 speed.

[JBeale]

FWIW, I've most recently been using the enclosed circuit (PDF), which goes from low-audio right down to DC (no series caps in the signal path).  This shows the sustained increase in pressure in the room when the HVAC blower comes on, for example.  Interestingly it does not go back to normal right away when the fan stops; maybe thermally driven air currents? I don't think it's purely a sensor temperature dependence, as the room temperature variation has a still longer time constant.

[thermistor-guy]

It could be paving equipment.  Caltrans does a lot of work at night; they usually try to finish by 5 AM, but will go to 6 if it has to be done.

Another possibility: a Kreepy Krauly or similar pool cleaner. The flow valve, in the main body housing, can oscillate at a few Hz, depending on water flow and pump pressure. If you stand near the pool (concrete), you can feel vibrations through your feet as the cleaner passes close by.

Vibrations can carry through clay-based soils (and the pool's water of course). The concrete pool walls and floor might couple vibrations efficiently into the surrounding ground, especially if the Kreepy Krauly frequency is near a resonance frequency of the pool's structure.