Page 1 of 3

Hearing Radiation

Posted: Sat Nov 17, 2012 5:50 am
by Frank Sanns
Here is an interesting link on hearing radiation: http://youtu.be/Eas4p8Sob9E You might want to skip out to 9 or 10 minutes to cut to the chase.

More accurately, it is converting a pulse height over to an audible frequency. After that they round the frequency of the note up or down to so that it is a discrete note.

Interesting sounds that accompany various radioactive materials. I think the technique could be refined better to give a good way to listen to various sources of radiation. A pure source should give a single tone for each of its peak energies. Neutrons would scream out of the audio with their high energies. It would certainly give a researcher far more information than just hearing audio clicks and rate.

Frank Sanns

Re: Hearing Radiation

Posted: Sat Nov 17, 2012 6:23 am
by Carl Willis
That's very cute, in part because it sounds surprisingly good.

I'd love to take this idea to the Explora Museum here in Albuquerque where I run some little hands-on activities relating to physics on occasion.

I think I can modify my LabVIEW MCA program to do this almost trivially, but the question really comes down to making it sound good--tone persistence, timbre, and temperament, and how pulse height is converted to an audible frequency (linear, logarithmic, etc.) Thanks for posting this very endearing idea.

-Carl

Re: Hearing Radiation

Posted: Sat Nov 17, 2012 7:10 am
by Steven Sesselmann
Frank,

Very cool, thanks for posting, that brings a new meaning to sound card spectrometry.

Steven

Re: Hearing Radiation

Posted: Sat Nov 17, 2012 11:17 am
by Chris Bradley
Good effort, worth evolving from there, I think: I'd throw my own suggestions at this. I'd point out that, frankly, one sound sounds much as another. OK, sure there are a few more high notes in response to one source than the other, but similar.

So what could be done is instead of fixing the note frequencies according to a given key, the computer could do a 'best fit' of the incoming signals that most closely match a musical key. Then, the key would, I suspect, likely change with different sources and that would really bring out some 'colour' between different sources.

The next thing would then be for the computer to be 'told' (programming-wise) to recognise maybe 3 or 4 different 'keys' within the signals, and then to give a proportionate amount of time to each, effectively filtering different 'harmonic sets' in the signal energies. So, for a given time ('bars') it'd play those that come within some correlation to one key, then change to another. That would then give a change of tempo too, as the computer selected between different harmonic sets within the data because the populations of signals matching each key will no doubt be different. Bringing different isotopes together (that stimulate different keys) would then have the audible effect of changing key and tempo every so often as the computer plays the different combinations.

Also, as well as discretising the frequency space into different notes of a key, the computer could also provide some temporal discretisation and match the note outputs to particular timing positions. This would then increase the probability of chords being struck. Another way would be to 'compile' a certain number of notes, randomly between 1 and 10 in number (according to a randomising algorithm), then play those notes altogether once the due number of notes have been received. This will give a mix of single notes and simple, or more complex, chords per music normally played.

Rather than single notes to single signals, there could also be a time-delay so that when the same note is struck repeatedly within a time-frame it is played as a continuous note.

I like the 'bumbling' bass effect of the low energy background that adds a colour and atmosphere to the lower end of the register. Maybe it doesn't need to be temporally discretised below some frequency.

... just a few thoughts for development, if anyone has the programming skills (that I don't!!) ....

Re: Hearing Radiation

Posted: Sat Nov 17, 2012 11:47 am
by bpaddock
Might get some tips on sounding good and sounding bad from here:

"Why Dissonant Music Sounds 'Wrong'"
http://entertainment.slashdot.org/story ... unds-wrong

Re: Hearing Radiation

Posted: Sat Nov 17, 2012 5:28 pm
by Rich Feldman
Very nice.

The OP reminded me of a different photon-to-acoustic spectral converter, seen via the Internet a year or two ago. But more than an hour of obsessive searching failed to turn it up last night.

It made audible tones out of chemical elements's -optical- spectra, from their -electron- energy transitions. Maybe someone else here has seen it, or one like it.

For each user-selected element, we got a nicely rendered picture showing the discrete energy levels. Then after simulated excitation events (random or user-directed, I don't recall), we would hear beeps or bell tones as electrons dropped back to lower energy states. The acoustic frequency went up with the energy step size, as would the optical frequency in real life.

I guess one could bend the pitches to fit a musical key (whose well-tempered scale closely approximates true harmony). Or let the Lyman and Balmer series, etc., form their own key!
Isaac Newton, after discovering that white light is a "mixture of rays with different refrangibility", deliberately chose as many color names as there are musical pitches in an octave.

Again: does anyone here know where to find a tool to "audible-ize" atomic emission spectra?
Thanks
Rich

Re: Hearing Radiation

Posted: Sun Nov 18, 2012 5:18 am
by Frank Sanns
I would imagine that a lookup table would be the easiest solution. When a pulse height is between a given range, a particular tone is called out of an array and that tone sent to the sound card for a predetermined duration. Don't know if you are programing in Visual Basic or C+++ but back in the day I did this for inputs from a solar cell and from random sources quite trivially using the above method. Actual notes are the best and it does not take many to get the effect. Ideally, I could imagine something like Audacity being used to slew tones to what you want or to even put different sounds to different energies like a gong sound when the energy is above the top thresholds. An alternative to this is to make up short mp3 files of very short duration and of the tone that you would like then call them up out of an array. There are short sound clips for free of just about anything you want or use one of the free music programs to make them like Garage Band even. A bit of work up front but I can imagine the final product being much better than the chaps in the video that I posted. Let me know if I can help in any way.

Frank Sanns

Re: Hearing Radiation

Posted: Sun Nov 18, 2012 7:45 pm
by aka47
Wow that's cool, well spotted frank.

I was musing the other day how an electric or acoustic guitar would sound played by Alpha particles if it was set up as a spark detector after looking at some of Carl's work.

http://www.youtube.com/watch?v=HO8VDf_rCXo

Re: Hearing Radiation

Posted: Sun Nov 18, 2012 8:04 pm
by Edward Miller
Not as complicated but Jeff from Mighty Ohm setup something like this for the Maker Faire with his geiger detector kit.
http://mightyohm.com/blog/2012/05/pics- ... ker-faire/

Re: Hearing Radiation

Posted: Wed Nov 21, 2012 10:48 pm
by Carl Willis
I wrote a little LabVIEW VI that emulates this demonstration with a 2x2" NaI detector.

I'm not a very good programmer, even for labVIEW; It was more of a challenge than I thought, and it still needs a LOT of work if it's going to be any good, but I will share a little video later today. I'll also send the VI to anyone who wants it, but it is specific to the Canberra 556 MCA and requires EPICS running under Cygwin or Linux.

My program's basic philosophy is outlined under the dashed line below. I fiddled with several approaches, and this one is the winner so far, running on my little Windows 98 netbook. At first, I thought I could collect spectra with a single count in them and search the spectrum array for the first instance of a 1 to get channel number and directly turn that into audio output, but even in a shielded 2x2" the background counts come in too fast to make that practical. So I collect short spectra and then pick out the channel with maximum counts to derive an audio signal. Cs-137 sources give a nice monotonous train of beeps, most at the same pitch, and Co-60 gives a warbling, lively high-pitched sound. The big problem seems to be audio buffer underflow that causes some outputs to echo and click at times, and the irony is that it is occasioned by the long time it takes to write to the buffer in LabVIEW's function for that purpose. The time needed to complete the main loop in the program is dominated not by spectrum acquisition, communication with the MCA, or math operations, but by the sound output write operation. I'm trying to better understand that and fix it.

Some future directions:
-Playing sound samples rather than waveform synthesis may be much faster and open the door to more complicated tonalities
-Gathering timed spectra rather than total-counts-based spectra would allow V(0) in step 9 below to be scaled by the sum of the spectrum ROI, so that high count rates would correspond to louder tones.
-Adjustment of timbre with pitch to accentuate high-energy events and help distinguish them from Compton continua. Above certain pitches, second, third, fourth and fifth harmonics can be mixed to brighten the tone and make a reedier, more penetrating sound.
-It would be nice to re-bin the spectra such that the FWHM resolution of the detector as a function of energy would correspond to about one or two semitones regardless of energy. Of course, to make it sound good, the frequency difference between semitones may have to be adjusted away from 12-tone equal temperament.

Any other ideas?

-Carl

----------------------

(1) Configure MCA to accumulate X counts in the ROI between channels L and H, then stop. (A good value of X for a 2x2" NaI crystal and 512 channels of memory / 3 MeV seems to be about 15 counts; L = 20, H = 511).

(2) Configure and start a continuous audio task and pre-fill the sound output buffer with an all-zero array.

(3) Start MCA acquiring.

(4) Poll MCA for status. If done acquiring, go to (5). Otherwise, repeat.

(5) Get MCA spectrum.

(6) Search spectrum array for max value and return the index (channel number).

(7) Convert index to a musical note relative to a reference note. E.g., semitone number S corresponds to index N by S = round[(R/A)*N], where A is the channel number selected to play at reference semitone R. On a 512-channel / 3 MeV spectrum, A = 150 and R = 49 (key A4) sounds decent. I have not played with a non-linear relationship between note and channel number.

(8) Convert note to frequency on a 12-tone equal-tempered scale, i.e. f = F*2^((S-R)/12) where F is the reference pitch for key R. Concert pitch F = 440 Hz @ R = 49.

(9) Calculate an audio waveform with a damped envelope. V(t) = V(0)*exp(-q*t)*sin(2*pi*f*t).

(10) Sample V(t) and write sample array to the sound output buffer to be played.

(11) Return to (3).