AW: [Re: thermal tips re expo converters:]

[jhaible]

> Thanks for the further clarification and especially for the simulation
> results!
> 
> I think some things could be done to increase the cooling rate.  One --
> as I saw mentioned elsewhere -- would be to use fat heavy traces and
> pads off of all the pins. Another might be to remove some of the plastic
> from the package. Third, a chimney over the chip could provide a laminar
> flow cooling draft while at the same time reducing fluctuations due to
> air turbulence. 
> 
> What so you think?
> 
>   Ian

Sounds interesting, but I can't say much about it. I have a collegue at
Siemens who
is doing thermo analysis with finite element stuff full time, but most of
it is the
proverbial "black magic" to me.

BTW, I have just suffered a setback with my new VCO. The waveshaping stuff
is fine, but
(you guessed it) the thermal stuff is not. This was a quad pnp array
(EP1015), with one
pair for expo conversion and the second pair for temperature compensation.
I wouldn't
say it's useless, but almost any other expo converter I've built before was
working better.
It's good enough for rough and animated sounds a la VCS3 over 4 octaves,
but transposing
to 4' or even 2' and still being in tune is not.

Recently I was thinking much about the VCO circuit of the MS-20. I used to
wonder
why they would have a V/Hz keyboard output voltage, and an expo converter
nevertheless.
No I start to understand the brilliant concept behind this. It looks to me
like the best
compromise between Yamaha's linear VCOs and a typical expo VCO.
They have an expo converter, but they only use the expo CV input either for
small CV's
(Tuning, Detuning), or for rather uncritical larger CVs such as ENV
modulation. Therefore
there is no need for a tempco resistor. A dual transistor for 1st order
compensation is enough.
For the critical stuff, they change the expo converter's reference current
(Where you would
normally find a linear FM input on expo VCOs). In order to minimize offset
voltage problems,
they only use the limited range of the keyboard output voltage (without
dividing it down,
that is), and the footage of the oscillator is chosen by switching
resistors. The expense
for these resistors is a little higher than for a V/Oct VCO, because you
need R, 2R, 4R
and so on. But that's still much more pleasant than all this tempco stuff.

Maybe I see all this in the light of my recent frustration with the EP1015
array ... but
I think the MS-20 scheme really has some points to it. The MS-20 still
needed an
offset trimmer for its Hz/V parts, but I think I can use an OP-07 here and
probably forget
the offset voltage ...

So maybe I come back closer to what I originally intended to build; that
was some 
mixture of MiniKorg and MS-20. I don't know if I have mentioned it before,
but I recently
found that portamento on a V/Hz system hs a different shape than on V/Oct.
(This may
be obvious to most of you, but I used to think that one could emulate this
by switching from
liner to exponential response - or lag vs. slew limiting - on a V/Oct.
system. But that
prooved to be wrong.)
So I guess it's finally time now for some Hz/V experiments !

JH.