[sdiy] Anyone built the ELEKTOR VOCODER?

[Scott Bernardi]

Your graphic of the filter response curve with the three staggered fc's
and the middle one (which is the middle frequency band for the channel)
running at 1/2 the Q of the two side bands is how you get a "maximally
flat" response. I understand that, but I'm having a bit of trouble
visualizing where the bands fall. Could you label your schematic with Q
and fc values for each filter stage? Maybe even a table (not schematic)
of your whole vocoder?

Also, can someone address this question for me: 30 channels at 1/3
octave sounds very impressive, but is it necessary? I would think you
could have closer spaced channels (1/3 octave) in the frequency range of
the human voice plus some harmonic multiples, and have them 1/2 octave
spaced outside of that.

Anybody come across any papers analyzing the optimum (and practical)
configuration for the frequency bands? (Actually I asked this question
months ago and someone directed me to >Roger M. Golden "Vocoder Filter
Design: Practical Considerations" >Journal of the Acoustical Society of
America vol 43 #4 1968 p803-810, but you have to be a JASA member to get
access to the articles). 

 
> Although the above is quite true, I found a way to reduce the number of
> op-amps, and still have 30 bands with 1/3 octave filters.
> I actually found the trick in another Elektor article (spectrum analyzer),
> and it consist in detuning slightly the central freq of each filter, so that
> Fc of filter 1 is the central freq of the band, and the central freq of the
> 2 other filters respectively the 2 cut freqs (high & low) of the band.
> Carefully adjusting gain & Q allows to obtain a rather linear curve (all
> filters
> have Q slightly > 4, "central" filters have a gain of 1, and filters for Fh &
> Fl
> have a gain of 1.4).
> Furthermore, 2 adjacent bands share the same filter.
> 
> Schemos can be found here :
> http://perso.wanadoo.fr/jbv.silences/DIY/Vocoder1.jpg
> http://perso.wanadoo.fr/jbv.silences/DIY/Vocoder2.jpg
> 
> The total amount of op-amps for the analyzer part is reduced from 90
> to 75. The same is true for the synthesizer part which is built around
> the same structure, which means a total of 150 instead of 180.
> 
> JB

-- 
Scott Bernardi
sbernardi at home.net