[sdiy] Evolution of the vocoder?

[Richard Wentk]

On 15 Dec 2008, at 11:45, Ken Elhardt wrote:

> Somebody on the Northern Sounds forum (Old Bob) did a good emulation
> of Roland's technique by taking sampled male and female "AH" choir
> sounds and running then through a high band count vocoder (he didn't
> specify what vocoder he used).  I was told by a VP-550 owner that's
> pretty much what Roland is doing.  I was skeptical at first but then
> later thought, yes, they'd kind of have to do it that way so that
> people with really odd or awful voices would still come out sounding
> relatively decent. He pulled down his demo, but I uploaded it because
> it sounds pretty damn good.  It's here:
>
> http://home.att.net/~elhardt2/Sampled_Choir_Thru_Vocoder.mp3

Convolution is cheap enough now that doing it in real time is  
trivial, so some of the vocoder plug-ins, and possibly some of the  
hardware units, do Fourier-based cross-modulation with the usual  
power-of-2 'band' count, rather than trying to use bandpass filters.

The results sound very different - with more than a hundred bands the  
sound is much cleaner and less processed-sounding.

This isn't always a good thing - frequency multiplication doesn't  
have the charm of analogue vocoding. But it can produce some unique  
effects which are much closer to the ideal of adding vocal  
articulation to a sound in a way that the analogue approach can't  
emulate.

Richard