Building a vocoder?

Started by StephenGiles, October 08, 2007, 03:43:29 PM

Previous topic - Next topic

StephenGiles

"I want my meat burned, like St Joan. Bring me pickles and vicious mustards to pierce the tongue like Cardigan's Lancers.".

StephenGiles

"I want my meat burned, like St Joan. Bring me pickles and vicious mustards to pierce the tongue like Cardigan's Lancers.".

Processaurus

Many thanks for the link to the MS20, Stephen.  Now I can fix the family synth if it ever goes south.

Mark Hammer

We've had this discussion before but I'll bring it up here again.

Vocoder designs are predicated on the assumption that it can work with ANY voice, without knowing anything about that speaker's vocie in advance.  As such, much like a 31-band graphic EQ that holds no assumptions about what frequency bands are critical to you as a user, the "better" vocoder will have more bands (12 is considered noticeably better than 8).  Does it really need all those bands?  I don't know for sure one way or the other, but there is the theoretical possibility that if the bandpass filters are tuned to the critical vocal information of the user, you can get away with fewer bands.  The hurdle is figuring out what the formant properties of the user actually are.

StephenGiles

What drew me to post these links was the clarity of the circuit diagrams.
"I want my meat burned, like St Joan. Bring me pickles and vicious mustards to pierce the tongue like Cardigan's Lancers.".