AW: Re: home-made vca

[Haible Juergen]

 ----------
> Von: gstopp at fibermux.com
> An: synth-diy at horus.sara.nl
> Betreff: Re: home-made vca
> Datum: Mittwoch, 19. Februar 1997 19:02
>
>
>      Okay this is good feedback from all you guys about VCAs...
>
>      I *have* seen the data sheet circuits, usually in the
>      matched-transistor sections like the LM394. I have more reading to do
>      - this is the direction I want. However if I pick a design that uses
>      something like a MAT-02 or such, I might as well try an SSM VCA as
>      well. I'd really like to avoid using dedicated chips.
>
>      I still have this desire to make something out of parts that you can
>      buy at the local electronics store. I haven't seen any Gilbert
>      mulitpliers, or else I would have tried that already - maybe I need to
>
>      look a little harder. I should probably get out my NTE cross-reference
>
>      and do some reading... if matched transistor pairs like the LM394 are
>      in the NTE catalog then that may be something to investigate.
>
>      How about the VCA from the Electronotes Preferred Circuits Collection
>      that uses a 3046 transistor array plus a couple of op-amps? The
>      original Dave Rossum design uses 741's, but what if a TL072 is used
>      instead? Is the 3046 quiet enough, or matched well enough, or whatever
>
>      is needed spec-wise? I know that the same 100K/220 divider scheme is
>      used as in the 3080 designs, so noise may be an issue.

Level, Distortion, Noise:

If you have no linearisation diodes, the input level versus distortion
would
be pretty the same for any type of transistor pair. So the only way to
improove
SNR would be using low noise transistors. I *guess* that the cheap low
noise
types I used (BC550C, DM 0.15) have a better noise figure than a 3080
ota
or a 3086 array, but I never cared to actually calculate it.

Linearisation:

I think the OB-X (not Xa) uses a 3080 and an extra diode array to do
some
linearisation on the main VCA (for all voices together). Sure we would
use
a 13700 or 3280 nowadays, but this old design may give some hint that
linearisation would somewhat work even if the diodes are not on the same
chip as the differential pair. Remaining questions: How good did it work
in
the OB-X ? Did Oberheim match the diode array to the 3080 or not?

Matching, CV rejection:

For a good CV rejection, the transistors have to be matched.
The good news: You can match single transistors much better than a 3086
!
Don put me right about this long ago, when he pointed out that the
original
Minimoog filter has much better matched transistors than later Moogs
which use cheap arrays. A 3086 has 5mV specs (if memory serves), but
you can match single transistors to 1mV easily. All you need is an
hour glass and a little patience. And make sure that no one opens the
door to the room where you do your measurments. I gave up the thought
that I could pick a transistor with my fingers (even with gloves)
without
changing its temperature. But if you grap it quickly and plug it into
the test socket, you can watch the vbe fading in with an exponential
slope, as temperature stabilizes. I found that after 3 minutes
(hourglass
for cooking tea ... I love tea ... especially Pai Mutan ... a green
Chinese
sort that keeps me awake at night to do my electronic projects ...
sorry, I disgress)
Ok, after 3 minutes the vbe hardly changes more than 1mV any more.
Takes some time, but it's quite relaxing (especially with Klaus Schulze
on the turntable ... I disgress again ..)
Now the really good thing is that these matched transistors have a much
better
CV rejection than a 3080, 13700 or the like. You still need an offset
trimpot,
but on a 3080, you still have a remaining "thump", as the offset
compensation
only works for one specific current. I don't know why, but the discrete
transistors are far better here. I can really make the "thump" inaudible
by carefully adjusting the trimpot. I don't know what's the reason for
this.
Maybe it's due to the different location of the trimpot (I change the Rc
balance, like in the Minimoog VCA, not the Vbe (which is the only method
that works on a 3080). Don't know why (anybody knows?) - but it works.

JH.