[sdiy] Percussion MIDI controller - again

[Magnus Danielson]

From: Ingo Debus <debus at cityweb.de>
Subject: Re: [sdiy] Percussion MIDI controller - again
Date: Mon, 28 Mar 2005 15:06:05 +0200
Message-ID: <24F52D7C-9F8A-11D9-980F-000A9571C136 at cityweb.de>

> 
> Am Sonntag, 27.03.05 um 21:35 Uhr schrieb Ian Fritz:
> 
> > Let's see ... the main peaks at one transducer should be separated by 
> > the time it takes a sound wave to travel to the other end of the tube 
> > and back, no?  Sound in solids travels about 5mm/us, so a 2m tube 
> > should ring at about 0.8ms. Sound roughly like what you are seeing?
> 
> I don't have a storage scope here. Now I recorded the signals into my 
> computer at 96 kHz sample rate. This reveils more than I could see on 
> the analog scope. First, the waveform does *not* look like a dampened 
> oscillation. The peaks don't have the same distance in one impulse. It 
> looks more like a bass drum sample (remotely), with different time 
> scale of course.

I would really recommend you to experiment with this with a storage scope at
hand. My experience is that it saved us a whole lot of time using a good
storage scope. It took only some time to adjust things and then we where doing
measurements. There are photos of this around somewhere, maybe even on the
web.

> Anyway, the peaks are more like 0.4ms apart. I don't think this ringing 
> is caused by the refections at the ends of the tube. There are also 
> visible echoes, but these are about 10ms...20ms apart. I think these 
> are caused by bouncing. I used a small screwdriver as a beater.
> I suspend the tube very loosely from two rubber bands hanging off two 
> microphone stands (already fell down several times ;-)). Perhaps that's 
> enough to dampen the reflections?

The sound propagates by a multi-path from the place you are hitting the tube.
There is one path which goes on the same side as your hit directly into the
piezo-element. This is the initial slope and this is what you should be looking
for. Then, the waveform actually expands like a ring out, and that will hit the
piezo-element as a long succsessive wavefront. The trouble is that the hitting
isn't a clear dirac impulse, so the responce from the object will muddle this
picture even further. The multipath and the hitting objects impulse responce
together with the piezo-elements resonance behaviour will create the full
responce. The thing to do is to look for the initial impulse, which is about as
clean as you can get it, the rest is multipath responce and other means of
dispersion which will just muddle the view and not be helpfull to you.

If you are looking at the initial rise on both elements you are home safe, no
signal can come before that anyway, unless it is noise from the surrounding,
which you can cancel to some degree if you need to.

> > Your best bet for detection might be to simply look for the first 
> > arrival of the excitation signal: amplify the signal by a large 
> > amount, set a threshold somewhat above the noise and detect the first 
> > crossing of the threshold using a comparitor followed by a pulse 
> > stretcher (to keep the final output from further switching during the 
> > time the excitation rings down).
> >
> 
> Yes, I think I'll get a real comparitor like a LM393, this should work 
> better. The 74HCT14 has a very large hysteresis.

Toss the 'HCT14s as the initial trigger. You can use the 'HCT14 as digital
cleanup after the real comparators if you need to.

> The signal from the piezo at the far end builds up slowly. So setting 
> the threshold will become tricky.

Look at it with a high BW storage scope. 100 MHz or something is a good BW to
use. I don't have pictures of those fronts stored, but I could make some new
measurements and make available.

Cheers,
Magnus