[sdiy] digitally switching cv and audio lines

[Magnus Danielson]

From: "jhaible" <jhaible at debitel.net>
Subject: Re: [sdiy] digitally switching cv and audio lines
Date: Sun, 28 Sep 2003 21:32:39 +0200

> I just looked at the data sheet for this.
> 
> I don't understand how this will work as a signal and CD routing
> matrix for a synth.
> 
> I'm familiar with the EMS system, where pins with built-in resistors
> will
> (1) mix signals from seperate sources, and
> (2) isolate seperate sources which share one, but not all, destinations.
> 
> Does this really work with that chip? The 200R on-resitance
> looks awfully low for that purpose. (For comparison: EMS uses
> several kilo-ohms.)
> 
> Are you going to scale down your CVs and signals so much that you
> can mix with 200R, or what is the trick ?
> 
> One "critical" example would be:
> (with xi = sources, and yi = destinations)
> 
> You want to make these 4 connections:
> x1 -> y1
> x2 -> y1
> x1 -> y2
> x2 -> y3
> 
> How will you avoid to also get the (unwanted) connections
> x1 -> y3  and
> x2 -> y2
> 
> ???

You need to run things "buffered". Essentially you want the outputs to be
current-summations inputs, i.e. I/V-converters. The input (source) end has
resistors tied to them. Better still is to use a current-output. With
current-outputs and current-inputs it actually works. However, a current-output
will produce a different signal if shared between outputs.

Now if you run a analog-switch setup this is not really very nice since you
can't in a good fashion control modulation (this is done in the EMS with the
use of different pins). The best plan for this is to use an OTA as the input
driver.

Now, this again has problems, since you can't use that for multiple inputs.

So, what I propose to do is this:

First use an analog switch to make point-to-point routing (i.e. no mix-routing)
and then route signals to modulation controlers (i.e. OTAs) and fixed gain
(i.e. op-amp setups) and those instead drivers another analog switch for the
mix-routing or if suitable directly drives into the mixer (without the second
routing).

The first switch-stage is really a global one, which can take on almost any
topology. The second switch-stage is really a number of local ones and has much
less complexity. It actually is all very similar to PAL-PLD style of designs.

For the first-stage I would recommend a Benes network (similar to Clos network)
and the assignment algorithm for this is chockingly simple and proved to handle
the situation. A small micro-controller should have no problem of implementing
the algorithm. 

For the second-stage we are really down to some form of MUX-setup and standard
4000-series CMOS-switches might work equally well as matrix-switches.

OUPS! I think I just told my secret plan on a virtual digital synth...

Cheers,
Magnus