[sdiy] Mind blown -- TL074 saturated outputs different

David G Dixon dixon at mail.ubc.ca
Thu Nov 29 01:26:18 CET 2018

Yes, Mark, it is the same idea.  However, it only works well if the opamp
outputs are equal and opposite.  If they are unequal, then the threshold
voltages are too sloppy.
My circuit is a 9-channel morphing scanner.  It is based on a back-and-forth
2164-based crossfader.  For 9 channels, the crossfader goes from B to A to B
to A to B to A to B to A to B.  These correspond to channels 1, 2, 3, 4, 5,
6, 7, 8 and 9.  
There are two main parts of the circuit.  The first part turns a -5V to +5V
CV signal into a series of truncated triangles which go from +5 to 0 to +5
to 0 to +5 to 0 to +5 to 0 to +5.  This is the control signal for the
linearized 2164 VCA in the crossfader.  The actual triangles are -1V to 6V,
but they are clipped at 0V and +5V to give small plateaus where only one
voice is exerted.  This part of the circuit works perfectly.
The second part converts the -5V to +5V CV signal to 8 3-bit codes which
control two DG408 8-channel analog muxes.  The switching is done when the
control triangles are somewhere on their plateaus, and this provides a tiny
bit of leeway for the actual switching voltages.  The DG408s exert pairs of
the input audio signals and send them to the crossfader.  The pattern is:
DG408 A:  1 3 3 5 5 7 7 9
DG408 B:  2 2 4 4 6 6 8 8
In this way, the switching is done in such a way that the same audio signal
is exerted full-strength during the switch, so there is no audible
switching.  It's like this:  1 to 2 (switch) 2 to 3 (switch) 3 to 4, etc.
In this way, a single crossfader can process as many voices as you want.
I built a 5-channel version of this and it worked perfectly (there's a video
on Youtube to prove it).  The AD in that case was only 2-bit and evidently
there was enough slope in the switching voltages such that all four switches
occurred during the control triangle plateaus.  However, the 9-channel
version did not work perfectly using the "rough-and-ready" comparator
method.  There were one or two audible switching points, and some of the
switches were occurring off of the plateaus, when more than one voice was
being exerted.
I don't believe that it would be practical to expand this design to 17
voices, although that would be cool.  33 voices would be even cooler, but
who has 33 separate audio signals to fade?  It would be possible with a more
precise AD converter, like an actual IC.

<http://hem.bredband.net/bersyn/VCO/8p2ch VC switch_s.pdf>  

There's a 3 bit flash converter here, done in the style of John Simonton.
Uses a TL074


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20181128/b482a819/attachment.htm>

More information about the Synth-diy mailing list