Identification of Meteor Infrasound Signatures

The best way to identify meteor signatures is to visually or photographically observe the meteor, and then relate the timed observation to likely infrasound signatures. However, not all meteors are visible, particularly during daylight, or cloudy conditions. As an amateur astronomer, one of the things that attracted me to get a Shake and Boom was for detection of meteors.

An excellent reference for meteor infrasound signatures is the following paper: “Optical Observations of meteors generating Infrasound - I: Acoustic Signal Identification and Phenomenology”. It is available here:

I have been applying the criteria described in this paper for potential identification of meteor signatures in my RBOOM infrasound recordings. I use an Excel spreadsheet to calculate the Dominant Signal Period (measured from the first cycle) and Dominant Signal Frequency and from that to classify the speed of the meteor. i.e. Slow is DSP<0.4s, Fast is DSP>0.61s and Medium in between (obviously).

The signature to look for and the methodology for classification of the meteors is according to this diagram:

And these are the notes relating to classification:

For now, I am ignoring any potential meteor signatures that coincide with a seismic signal. I realise that large meteors can and will excite the seismometer in the Shake and Boom, but without a confirmed sighting, I am not yet confident that a seismic signal could not produce a linked infrasound signal similar to what a meteor would produce. So for now, I am possible ignoring some larger meteors or bolides. As a result, the potential meteor signals I am finding that I have reasonable confidence in are relatively small signals - they are not the strongest signals on the helicorder (though I do still look at them just in case there’s a big one!)

The other things I look for in the signal is that the first cycle amplitude has to be the biggest. Any signal with an increasing amplitude at the from of the signal (rather than a step change) is not likely to be a meteor.

The Dominant Signal Period also needs to be in the range of 0.1s to 1.6s. Outside this range and the signal is not likely to be a meteor.

The Frequency Spectrogram will also be strong below 10Hz and relatively weak above. A 10Hz low pass filter may be useful to eliminate noise. If the expected meteor signature disappears when the filter is applied, it’s probably not a meteor.

Sporadic meteors (ones that happen all the time - not associated with meteor showers) occur at the rates of 3 to 4 up to 8 or 10 meteors per hour. That is a for visual detection. I don’t know how that rate relates to infrasound detection yet. Do some visual meteors not register on infrasound? Does infrasound detect some meteors that can’t be seen? I don’t know and haven’t yet found a reference for that.

I hope this will help people to identify meteor signals on their RBooms and encourage people to look for them.


This is all extremely fascinating sheeny, thank you so much for sharing your elaborations and finding with us!

As a fellow amateur astronomer and also as someone that is interested in meteor monitoring (I have a radio echoes rx station installed at home) this can be a step forward in what we can see with our Boom sensors.

In particular, the following passage

Do some visual meteors not register on infrasound? Does infrasound detect some meteors that can’t be seen?

is what should push us to further analyze the data from our Boom network, in particular during the major meteor storms (so that there is a higher chance of capturing good signals). Perseids cannot come any faster now.

1 Like

I agree. Radio echoes sounds like a good detector. Is it automated to log times automatically 24 / 7? That could be really handy to point out times to check for Infrasound signals! Timing is always the issue with visual and photographic methods.

I had an interesting signature yesterday. I’ve just re-tweeted it as I forgot to add @raspishake to the tweet.

It was another potential meteor a class Ia- but with a bit of a messy tail to it. I noticed it had 2 peaks in the power v frequency plot, so I processed the signal with a low pass filter (<3Hz) and a band pass filter (3 to 10Hz) found two separate Class IIa meteor signals.

While the paper I’ve based my methods on have related the Dominant Signal Period to observed speed, I suspect the DSP is also affected by the size of the particle. The DSP is determined by the time the air takes to collapse back into the vacuum behind the supersonic particle, so to my my mind there’s 2 ways to punch a big hole in the air - speed, and size.

So I’m suggesting the meteor was actually two particles of different size moving together.


I received notification of a new paper on infrasound signatures from bolides today:

It would seem that the Dominant Signal period has no relationship with meteor speed.



Yes, my system is automated with a combination of software, SDR# to connect with my AirSpy2 and Spectrum Lab to log all entries above a certain tresholds in both .csv and image format.

Unfortunately, the radio source that I use to capture echoes’ pings is in France, so I cannot correlate what I see on the radar with particular waveforms at my Boom which is distant. However, calculations could be made for infrasound sensors in France (or, at least, inside the general detection area) combining their data with visual and radar traces.

I will have to take a look at the new paper as soon as I have a bit of time, it seems definitely interesting.

Meteor detections have been rare for me lately, but I got a good one this morning. It was a complex medium sized Class III c or d +.

There was a small seismic signal in the 18 to 22 Hz range, but as this started at the end of the first half cycle of the infrasound signal, I interpreted this as the infrasound exciting the seismometer rather than the other way around.

I was able to split this into two potential meteors: a medium class III + and a small class II + by judicious filtering.

It appears from the accurate measurement of the start time of the first cycle that the small meteor was trailing the larger one by about 1/20 of a second at the time of impact with the atmosphere. The signal duration of the smaller meteor is also less than the larger one indicating less time to ablate and/or slow below the speed of sound, confirming smaller mass/size. I didn’t measure that, but the duration of the signal is easily measured.

Notice that I’ve interpreted the Dominant Signal Period (DSP) to indicate particle size. Originally this was interpreted as an indicator of speed, but intuitively is also affected by particle size, as the DSP is determined by the time for the vacuum behind the particle to collapse.



This is very detailed and well presented sheeny, I quite like the calculation sheet that shows all the steps done to classify the meteor signal.

The usual start-of-the-year ‘dry period’ is about to end, the Lyrids are just a couple of weeks away, so more chances to acquire some good meteor data from your infrasound sensor!

The calc sheet is just based on the sheets I’ve used for years for engineering calcs. No point reinventing circular things! ;o)

mmm eta Aquarids in a couple of weeks! 50 per hour… how will I keep up? lol! Luckily(?) I don’t think I detect as many meteors by infrasound as are visible by eye, otherwise I should see about 3 to 6 per hour just from sporadics. The advantage with infrasound is it doesn’t matter if it’s cloudy or daylight or even if I stay up late (as long as I’m not visually verifying).

I suspect I’m only catching those that are close or a bit on the larger side.


1 Like

Same here when I have to re-use older projects part for new assignments. It’s always faster to take pieces here and there and re-assemble them.

That’s what I love about radio meteor detection too, snow, sun, clouds, day, night, nothing matters and I am always getting echoes, especially during the major swarms. Now that you have given me a base I will try to see if I can get some infrasound echoes from my location.

My RS&B detected another meteor this morning - a large Class I -.

As the first phase is negative, this suggests that at the time of impact with the atmosphere the meteor was heading away from the RS&B. That’s the only way the negative pressure wave from the vacuum behind the particle can lead the signal.

The other thing to note is that the second half of the N wave is lower in amplitude and wider in period, so there are some remnants of the near field pressure wave still evident. So the meteor must have been relatively close for the signal not to attenuate to a balanced N wave.



Thanks for the information. I’m planning a trip out to Anza Borrego park in Southern California for the Lyrids April 20. I’m planning to do some wide-field timelapse for the night, but I just stumbled on this article and am thinking that bringing along my Shake and Boom might an interesting thing to do.

My campsite happens to be in an east-facing canyon in Anza Borrego park in Southern California:,-116.3112682 and I’m wondering if that might create an amphitheater-like effect help to amplify any infrasound waves from shower. I would be interested in any ideas about how the infrasound might interact with the acoustics of the canyon. The park (and nearby Joshua Tree National Park) are full of lots of rocks and canyons, so it might be a fun quest to seek out optimal conditions.

I would also be interested in any radio frequency recording I might do with another Pi and an SDR setup, if anyone has any suggestions.

Thanks for any advice.


1 Like

I can’t comment with any authority on the acoustic of the canyon, but I doubt it would be the most significant issue. Assuming you have good weather (no rain) the biggest issue is likely to be from wind. Rain makes a lot of infrasound which will just swamp any meteor signals. Likewise wind and wind gusts. If you do a search online for infrasound detectors you should find some pictures of the “star wheels” that they make to eliminate the effects of wind. They have, say, 12 pipes radiating out from a central hub which is where you connect a tube from the RBoom to. This averages the pressure to all the 12 pipes at the hub before it is read by the RBoom so it’s supposed to eliminate the effect of wind.

Hopefully you will have good conditions and be able to capture some signatures.

One final comment - the signatures are typically not the biggest signals on the helicorder (at least not with my setup!). I normally look for small single spikes on the helicorder especially in areas where the seismometer is also quiet, but if you can actually time when you see a meteor you can then specifically check times to see if you have a signal.


1 Like

Latest detection from this afternoon, another two particle (that I can resolve) meteor. Both particles are small: a class II+ followed by a class IV+ 0.12 seconds later.


Another complex meteor detection from a couple of days ago.


And a simple one from today. I still ran it through a band pass filter though, and the N wave cleaned up nice and smooth and balanced without the noise.


I wish there was some way to automate the process of finding the N wave signals. It’s pretty time consuming doing it manually. After I find one I usually stop looking for the day - other things to do! ;o)

You could try to implement an RMS process that analyzes the signal in the timeframe you have selected and then locates the spikes, presenting to you a list of times that you can then manually investigate for actual N wave signals.

It could be a bit tricky to do, but maybe with Python processing or similar languages it could work. There is reference on the net regarding such an implementation, so there could be useful experience from other people who are using this RMS identification.

I’m pretty low on that learning curve. I’ll have to learn Python first. ;o)

1 Like

I’m working on a quicker way to detect meteors, so I’ve just done a trial on last night’s data. We had rain last night so conditions are not ideal. Rain makes a lot of infrasound, so for last night I’m trying to detect meteors through more than normal nightly noise.

First step is to tile the EHZ and HDF channels side by side in SWARM so I can quickly check and eliminate any seismic sources.

I then setup a 0.5 Hz to 10 Hz bandpass filter for the HDF channel to cut as much noise as possible that’s not related to meteors.

I then set the zoom window to about 15 minutes wide, so I can scan each 30 minute line of the helicorder in two goes, and look for very sharp little spikes equally balanced above and below zero.

It’ll look something like this:

I note the time of the spike to the nearest minute - I’ll come back to later. Scan as much of the helicorder as you want to recording the times.

Then come back to each of these times and zoom in to confirm if it’s a likely meteor candidate: i.e. N wave form (even if noisy), first cycle largest, etc and no seismic trigger. This stage will look like this:

Note in this example there is a tiny seismic signal that coincides with the potential meteor signal, but as there are other seismic signals of similar size and frequency that have not triggered an infrasound response, there’s no reason to suspect this one is triggering this significant infrasound signal.

Doing this for last night, in not ideal conditions, I was able to detect 24 possible meteors occurring in 9 hours, and eliminated 5 of these as definitely not meteors, 14 likely meteors and 5 marginally possible meteors that are very weak signals and difficult to separate from the background noise.

This gives me an hourly detection rate of 1.5 meteors per hour. Sporadics occur at 5 to 10 meteors per hour, so about a 15 to 30% detection rate compared to visual detection in good to ideal conditions.

This still needs more testing, but I think it’s a step in the right direction. The analyses I’ve been doing on potential meteor signals to date is interesting, but time consuming.

I’m hoping this means my RS&B isn’t as deaf to meteors as I first thought. I think the key to detection is to filter as much unwanted noise out first (obviously).


I got an interesting signal tonight that doesn’t quite fit the classic N wave model at first glance, but resolves into a complex meteor of at least 2 particles. After processing meteor signals over the last few weeks, I’ve slowly been coming to grips with the idea of a meteor of indeterminate phase which Silbe and Brown mention in their paper but give no further information on or example. Analysing this signal has helped me to get to grips with this. Rather than repeat myself please read the notes in the screen shots.

1 Like