VCOs, VCAs, and vocoders...

[gstopp at fibermux.com]

     If you need a sawtooth VCO to have precise tracking over a wide range, 
     I haven't seen anything much simpler than the 3140/LM311/FET VCO that 
     forms the basis of the ASM-1 VCO (that's why I chose it).
     
     That VCO can be reduced even further in complexity if you only have 
     one modulation source - you can get rid of the input summer and drive 
     the transistor pair directly, for example. An initial frequency pot 
     can be resistor-summed with the input CV.
     
     I've seen an even simpler design using a 555 plus transistors, but 
     I've never tried it.
     
     The VCA in the ASM is real simple too - you can get rid of the op-amps 
     and just use the 3080's, and if they all go to one audio output you 
     can just tie all the 3080 outputs together into a single op-amp. You 
     may need an op-amp to control pin 5 of each 3080 depending on the 
     voltage range of the modulating waveform. The 3080 is great for VCA's 
     as long as you don't need super-high-end performance.
     
     The more bands a vocoder has, the more understandable the speech will 
     be from the synthesized output. The more bands a vocoder has, the 
     tighter the filters must be - they must have steeper cutoff slopes. 
     I've never built a vocoder before so I can't really provide a lot of 
     details - there are others on this list who know more about them - but 
     I do think that 12 or 16 bands will be better, with 1/3 octave spacing 
     and 18 dB/octave slopes on the bandpass filters. On top of this you 
     can add voiced/unvoiced logic to substitute noise for the excitation 
     source to simulate fricative sounds like "ssss" and "ffff". You can 
     get very complicated if you want.
     
     Looking at the datasheet for the LM394, since there are examples of 
     log converters, I'd say that it would probably work fine to replace a 
     MAT-02 for a VCO exponential converter. However I've found that 
     randomly picked 2N3904's work fine too.
     
     - Gene


______________________________ Reply Separator _________________________________
Subject: VCOs, VCAs, and vocoders...
Author:  klund at MIT.EDU at ccrelayout
Date:    1/3/97 11:45 AM
     
I have built the Paia (clever design!) Shepard Function Generator, 
http://www.hyperreal.com/machines/manufacturers/Paia/images/ and now I 
want to produce the "barber pole" tones, so I need 8 VCOs and 8 VCAs. 
What topology would give me a simple exponential VCO for a bank of them? 
Should I just use a stripped-down version of Gene's ASM-1 VCO?  Also, 
can anyone recommend a simple VCA topology?  I have a bunch of 3080's, 
so I'd like to use them, if possible.
     
While on the subject of banks of VCA's, I've been thinking about 
building a vocoder.  I was reading the archive, and I came across some 
posts saying that an 8 channel vocoder isn't the best...  How many 
channels should a good vocoder have?  How many octaves should it cover, 
with how many bands per octave?  If anybody would be willing to send me 
an example band, I'd really appreciate it.
     
Also, for the ASM-1, can the LM394 be substituted for the MAT-02?