Is there a standard way to build a configurable circuit topology mesh using mechanical or solid state relays (or/and other components)?

Question

Preamble:

1. The motivation is to build a hybrid analog/digital computer this is reconfigurable (the topology of the opapms, resistors, capacitors, adders are all wired automatically -- without human hand)
2. For that I need to be able to route/connect analog signals (and loads)
3. There is no need for high speed switching
4. Though there is a need for low noise, low cross-talk and low resistance, impedance, capacitance across the connections.
5. Although I do not know too much about them -- I reckon I could extend how cross-bar switches in the old telephone exchanges worked
6. Cost is an important factor
7. Solid state would be preferred over mechanical
8. Digital FPGAs are not candidates for this soliton as they only deal with digital signal
9. FPGAs have digital cousins (see this question) -- but they seem to have a very low analog component count -- 2 OpAms, 2 comparators, and seem to suffer from noise and some of the challenges I cited above.

I want to build a configurable circuit topology mesh using mechanical or solid state relays (or/and other components).

1. What I envision is that I have a set of electrical pins: 1 through N; The pins could be left open; connected to a pin of a component like a capacitor, inductor, OpAmp, Adder, Multiplier, etc.
2. I have a digital (bus?) that specifies the topology of the circuit by a set of pin short circuit specifications:

A set of pin short circute specifications looks like:

• Short pins: 1, 11, 13
• Short pins: 22, 12, 43
• Short pins: 44, 66, 89

Note the short circuits are intended to create circuit nodes (as in Kirchoff's laws nodes)

I've tried to illustrate this using the diagram below.

Is there a standard way of constructing such a routing mesh?

Answer 1

A system for mechanically making electrical connections between pairs of points? That sounds like ... a (pre-digital) telephone exchange!

Provided that you're not using most of the connections most of the time, you could build it out of Strowger Switches.

Edit: plan B, the robot-pegboard: take a large sheet of "hardboard pegboard" (large scale) or veroboard (small scale). Run wires horizontally/vertically on each side, arranged so that a metal peg pushed through makes contacts on both sides. Having thought about it, reliability of this might be a nightmare, you'd probably have to mass-produce sprung inserts.

If you insist on all points to all other points potentially all at once (which the telephone network can't do!), then there's no shortcut. You will need 100 x 100 switches, arranged in a grid. Each of those analog switches will let you handle a 16x16 section of the grid. You'll need 40 of them. Note that each switch has a 200 ohm on resistance, which might be a problem for small signal applications. Designing the routing so that crosstalk doesn't destroy everything is also going to be interesting - a ground shield for every signal?

(Basically there's a reason nobody builds huge analog computers, and the noise problem is a big part of it. It'll work at audio frequencies - see modular analog synths - but at tens of MHz it'll just disintegrate)

Answer 2

As jms and pjc50 previously mentioned, there are ways to wire up (pairs of) FETs to act as bidirectional analog switches for analog signals in some reasonable range of voltages and frequencies.

Such circuits are used internally in digitally controlled analog switches such as the ADG1311YRZ, analog multiplexers such as the ADG731BSUZ, analog crossbar switches such as the AD75019, etc.

One standard way to build an automatically-reconfigurable circuit topology mesh for analog signals is with a field-programmable analog array (FPAA) such as the ones in a Cypress PSoC. ( " Are there any Analog FPGAs? " ).