I think a shallow dome shaped plate would be best. Piezo film is what I thought of for the transducer itself, kynar (?), but I can’t recall where I saw it available in small pieces. It was edmund scientific or somewhere similar.
For anyone interested, I’ve put 3 wavs up on one of my sites, recorded last March with the setup detailed above.
They are 10-12 megs, and about a minute long.
http://www.tbo.net/rainwavs/lightrain.wav
http://www.tbo.net/rainwavs/modrain2.wav
http://www.tbo.net/rainwavs/hevrain5.wav
Found this radio astronomy strip chart software:
http://www.radiosky.com/skypipeishere.html
It works with a sound card, so I think I’ll see what rain looks like with it.
I thought you could have one for the light rain drizzel, and one for the heavy rain ect.
Obviously going down the wrong road with this #-o
You may have a good point, although SLOChris said earlier that he thought the sound card would have enough dynamic range (difference between the softest and loudest sound it can measure) I was wondering this AM (driving through the rain) if that is really true. Seems to me there could be a huge difference between fine/soft/misty rain and a good heavy downpour, so maybe it needs a large plate to detect the fine stuff, but a small plate for the heavy stuff because the signal from the large plate would be too big/overloaded when the rain is heavy?
The captured WAV files are amazing! I wasn’t expecting anything like that kind of noise, although thinking about it afterwards I shouldn’t have been surprised because it it in effect a big drum.
Light rain has less high frequency component than moderate/heavy. Moderate and heavy seemed to have similar audio characteristics (by eye/ear) and I guess the main difference was the number of impacts/second rather than a different frequency response.
This next bit is me thinking out loud…
Maybe there are two ways to ‘measure’ the rain. Firstly by frequency response of the audio signal and secondly by trying to record the number of impacts/second.
To process the audio, it needs digitising and FFT processing to extract the various frequencies involved. The tricky bit is then looking at the frequencies and decide what they mean. It’s possible that after FFT processing the digitised audio you might find a low frequency component which gives an indication of the rate of drop impact on the plate. It’s also possible that the other frequencies might give an idea of the droplet size? If/when I get over the initial teething problems with the dsPIC development board/compiler and learn a bit more about how to program the beast, it will be interesting to feed these signals into the board’s ADC and then see if I can DSP the signals to extract anything useful.
I also wonder whether a smaller plate might be better and maybe with thinner material, e.g. niko’s kynar suggestion. With a large plate you’ll get a lot of drops at the same time in moderate/heavy rain and that might make it more difficult to differentiate between the individual drops and also reduce overloading due to the sheer number of drops hitting almost simultaneously. Also with a thinner material you would get better response to smaller drops in light rain.
If heavy rain overloads the piezo transducer/sound card then maybe you could fix a second transducer to the plate but via some light foam to act as a kind of damper. Use the damped transducer in heavy rain and the undamped one for light rain (this is like Clanger’s suggestion of multiple plates, but all on one plate).
IIRC the plate of the vaisala is around 2-3 inches dia. I was afraid FFT was going to rear its ugly head sooner or later 
Heh. We have aluminum shingles on the roof, and the sound in the house in a rainstorm is incredibly cool. I keep thinking about miking the attic with a couple of mikes and recording it in stereo. Then I could play it back through speakers up there and get the same effect.
…or maybe not, this paper makes it sound easy
That paper is more worrying than FFTs. It includes some partial differential calculus 8O
Now I know that FFTs and partial differentials are closely related, but having had to derive FFTs by hand using differentials during my time at university, I don’t want to go back to basics and try deriving them all over again. The ability to do solve FFTs in the digital domain using software is one of the best things that has been invented since sliced bread!
I still have nightmares about page after page of partial differentials trying to derive various things. One simple mistake at the beginning spirals out of control 20 pages later! Mind you, my derivations sometimes gave interesting outcomes. In an exam once I managed to simultaneously derive the fourth and fifth laws of thermodynamics…which was a shame because firstly I was trying to derive the third law and more worryingly the fourth and fifth laws don’t exist (or maybe the do and no-one but me has ever derived them) #-o
Did you lose your deriving license?
Fortunately not. Luckily there’s more to Chemical Engineering that proving partial differential calculus equations!
One way to do calibration would be brute force and awkwardness. Just note the output for various types of precip, and create a lookup table. Then interpolate for intermediate readings.
I hadn’t considered the plate as a tympanic membrane before. Like a drum, it resonates at essentially the same frequency for any impact on it, but presumably at higher amplitude for bigger drops. Hail would probably cause some more high freq components because of the more elastic collision. Hmmm. Piezo film, eh?
Or maybe damp the plate more, so the impact sound is heard, rather than the resonance of the plate.
Chris,
I believe that you are on the right track. I have a little experience with processing digital audio, so I ran the rain clips through a spectrum analysis. It confirmed what I suspected (and you and sloweather suspected). The sensor acts like a drum and has a prominent frequency no matter what hits it. In this case, it’s 86 Hz or F2 on the piano. That’s the “lower” tone you hear in the recordings. So, I believe you would have to use an impacts/sec type measurement or possibly set a noise gate of some kind and count how many times the audio pops above it? I’m just thinking out loud here. Interesting project. It’s got my wheels turning a bit.
Johnny
Read the paper I referenced a few posts back. They damped their sensor plate.
Niko, which page? I didn’t see where they damped it. It seems to me that dampening it would kill their ability to detect the hail resonance.
Here’s a nice donationware PC sound card scope and spectrum analyzer - Visual Analyser8.
Sorry, missed your post. They wrote “The material and dimensions of the detector cover are selected such that the resonant vibration excited by the impacting raindrop is attenuated rapidly.” To me “vibration … attenuated rapidly” is equivalent to saying it’s damped, although I agree that my “They damped their sensor plate” overstates it.
Thanks! I’ll check that out. I just wish it would rain here again so I could play with it some more.
There are a variety of PC signal processing tools available for download here:
http://www.web-ee.com/Downloads/Sound%20Cards/sound_card.htm
I don’t know for sure about the others, but Wave Tools is free and includes a nice oscilloscope, spectrum analyzer and set of meters.
Steve