Martin Czech martin.czech at
Tue Nov 2 09:24:29 CET 1999

:::Deconvolution is a very common technique in DSP world. It make wonders if properly
:::implemented. To avoid the singularities of the inverted signal spectrum I would 
:::recommend to look at the Wiener filtering algorithm. His trick is to bias the whole 
:::spectrum by a very small amount and thus eliminate the division by zero problem.

I hope that if the impulse looks right (not much ringing, Gaussian),
there will be no singularities. Encouraging to hear that all this is 
already done...

:::Quite another approach is to calibrate your speaker+microphone using sine waves first,
:::which should be performed in possibly neutral and echoless environment (some kind of
:::desert ;-)) and then generate the impulse using an artificial signal reproduced by
:::your speaker.

Thought about this, too. The best would be a little speaker, and a electrolytic
cap to discharge into (... popppp).
I hope this will be loud enough. There is certainly a dynamic problem, the ..pop..
will be much louder than the room response, forcing me to turn down input volume.
The best thing would be a very short, reproducible loudspeaker pulse, because
I could use the very same signal to gate the input.


More information about the Synth-diy mailing list