More about switching

[Haible Juergen]

	> That aside it's a damn good idea. And one worth investigating.
Though I
	> would still need at least 8 OTAs per channel. That's 40 13600
packages.

If you have so many channels, I would really consider using a
"discrete ota" approach. You only need the back end of the ota
(current mirror) and a buffer once per output channel, and only
8 pairs of transistors for 8 front ends. Just connect all of the
collectors in parallel ( 2x8, of course) and do the mixing with 
currents.
I built something like this to mix the 4 filter outputs of my JH-4
synthesizer module. Yes, *mix*, not just switch. So I have no
feedback, but I had to match the transistors in pairs to avoid
offset problems. But if you only want to *switch*, and thus
have feedback, you don't even have to select the transistors.

You could probably use cheap pnp pairs for the inputs, and
a norton amplifier (LM3900) as common back end.

> If I were to use OTAs I'd be real tempted to arrange it to fade in and
> out.
> Integrate the control signal over a cap and slow it right down. That
> would
> be so silky smooth. No thumps and very sexy. But I'm not sure that it
> would
> work within a feedback network. I know I've had OTAs with feedback for
> some
> reason but I can't recall what or why. 
> 
Otas with feedback are mostly used in VCFs, slew limiters,
S&H circuits and so on. Works great in most applications.

Slow transitions are surely possible. During the transistion,
the signal will also be soft clipped, as the feedback looses its
influence when the ota gain is small. But you can surely live with
a signal that is slightly distorted for a few milliseconds during
switching on and off. I guess it's better than a glitch from hard
switching.

	>In fact the chip I'd be real interested in investigating is the
SSM2163.
	>This little puppy is a complete, digitally controlled, 2 bus, 8
channel
	>mixer. 2 of 'em together makes a stereo 8 channel mixer with 2
stereo
	>busses. I don't know how much they are. Nor where to get 'em
but they look
	>pretty interesting on paper. But then again it's not quite
what's required
	>in this app.

Sounds interesting indeed! Will probably make small mixing
desks even cheaper than today. (The next revolution after the 
introduction of the SSM2017 ??)

	>There's a FAB called quality-semiconductor who make a series of
multi
	>channel change and change-over FET bases switchers and muxes.
Once again >not
	>quite what I'm looking for but interesting none the less. And
it's a shame
	>about the Harris ones being only 5 volt. Otherwise you could
build a
	>digitally controlled pin-block like the VCS3 has. Only without
the pins of
	>course. Would make an interesting patch bay.

What I never quite understood about switch arrays for VCS3-like
patching: You need one source going to several destinations,
and one destination being fed by several sources, both at the
same time. So:

1) Your array must be able to make more than one connection
     per column and per row at the same time.

2) To avoid shorting, each connection must have a reasonable
     *Resistance*. The VCS3 chose the elegant way to place
     these resistors inside the pins.
     You could do the same thing with discrete switches and one
     discrete resistor for each switch - but how would you do this
     in an integrated mux or even cross point matrix ?? 

>But anyway OTAs would most certainly overcome any crosstalk ambiguity. 

That's what I hope. But I am not completely sure how the feedback
would affect crosstalk. Someone should try it ...

JH.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 3593 bytes
Desc: not available
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/19970814/c1a3bcdd/attachment.bin>