From: gstopp@fibermux.com
Date: Wed, 08 Nov 95 09:17:56 PDT
Encoding: 85 Text
Subject: Re: Has anyone used ICL8038 chip?
X-Pmflags: 35651712

     
     Using the 8038 as an electronic music VCO was discussed in 
     Electronotes quite a long time ago (early '70s) and I'm turned to that 
     page as I write this. Apparently the problem with starting out with a 
     linear VCO and trying to "exponentialize" it is that there are too 
     many circuit sections required that are subject to error. These are:
     
     1. Temperature in the exponential junction
     2. Typical exponential errors in the converter semiconductors
     3. Errors in the current-to-voltage converter
     4. Errors in the positive supply reference difference amplifier (the
        8038 CV is referenced to V+)
     5. Errors in the 8038's linearity
     6. Errors in the power supply that affect item #4 above
     
     These are the type of errors that exist in any design; however, in 
     this case there are too many in a row and the cumulative effect is 
     that the resulting VCO will not be accurate or reliable enough to have 
     "professional" performance specifications. In other words, if you 
     reached out to play the keyboard at any given moment you couldn't be 
     certain that exactly the right note would play. But it is mentioned 
     that for non-critical applications it would probably work fine, like 
     as an LFO or for special effect sounds.
     
     In typical Electronotes fashion there follows a "cheater" circuit for 
     the 8038 which bypasses the current-to-voltage problem altogether and 
     injects exponential current directly into what would normally be pins 
     for timing resistors, making the VCO exponential with a 30,000:1 
     voltage/frequency sweep. The outputs are Sine, Triangle, Sawtooth, 
     Square, and Pulse (with PWM). This circuit uses an NPN matched 
     transistor pair, followed by a PNP matched transistor pair, along with 
     5 op-amps and a FET.
     
     While this sounds attractive, if the performance of this VCO is 
     expected to be as good as possible, then the two matched transistor 
     pairs will add up to about twenty bucks. This combined with the cost 
     of the 8038 itself makes this design less attractive than say one of 
     the later Electronotes designs which use only one matched pair or even 
     a 3046 transistor array, plus a few op-amps, with performance specs as 
     good as anything ever available commercially (and better than most).
     
     I ordered some 8038 samples recently to play with, so that will be my 
     first experience with these. I suspect they will end up as cheap and 
     easy VCLFO's.
     
     BTW maybe we should continue this thread on the synth-diy list.... but 
     for a first-poster I think you picked the right place!
     
     - Gene
     gstopp@fibermux.com
     


Subject: Has anyone used ICL8038 chip?
Author:  K.Dunnicliffe@uk22p.bull.co.uk at ccrelayout
Date:    11/7/95 12:54 PM


Hi all -- first posting!
     
I'm currently scratch-building a VCO module, and messing about with a 
number of different VCO types - mainly using discrete op-amps and 
transistors.
My *current* prototype is based on an ICL8038 chip (Harris). 
Data is available at 
http://www.semi.harris.com/datasheets/lin-tel/icl8038/
It has linear sweep voltage with a 1000-1 range and *seems* to generate 
good quality square/pulse, triangle (and pseuodo-sine) waves. Looks like 
PWM should be straightforward too. I'll probably put a moog-style log CV 
generator with tempco on the front end.
Has anyone else tried DIY with this chip?  Any hints, tips or warnings? 
Like to share schematics ?? (I'm currently just following the data sheet 
examples).
Does anyone know of any commercial synth using this chip?
     
My rate of building is *very* slow so I won't get to hear this thing 
through a VCF/VCA for a while yet.
     
Happy burbling,
Kevin.