Inexpensively switching/multiplexing a large number of sources?

[timb]

Hey guys, so I'm working on a project where I'll need to be able to multiplex a large number of sources together. The thing is, these sources can be anywhere from -15 to +15V and up to 500mA. Some of the sources might just be a voltage, some might be digital (I2C, UART, SPI, etc.) or even an analog signal up to a few MHz.



Here's a bit of an example I drew up. Let's say I've got 8 rails (A-H), I need each rail to be able to make a connection to any of the other rails. So, rail A could connect to D and G and rail H could connect to F, B and C. In the example above I've only drawn out connections for A and H, but B through G would be identical.

I need to be able to control the switching from a microcontroller.

I've looked into dedicated analog MUX chips, but it would be sort of expensive and I'm not sure if it can meet the above specifications. I've also thought about BJTs or JFETs but I'm not sure how reliable that would be. There's always relays, I can always get small ones but that's still a lot of space and the cost would add up (just for the 8 sources in the example that's 56 relays).

Does anyone have some other options I might be overlooking?

[Marco]

If you can ignore the current requirements (just buffer the outputs instead) then this would work.

http://www.digikey.nl/product-detail/en/CD22M3494MQZ/CD22M3494MQZ-ND/1034975

[timb]

Innnnteresting. That might just work. So I'd connect A-H to Y0-7 *and* X0-7 which should let me make any connection I want. If I wanted 16 connections I could use two and connect X8-15 from each IC together.

[timb]

I've been sort of swishing other options around in my head the last couple of days and had a thought. If I was only talking about digital signals, I should be able to use an FPGA for this, right?

[alm]

If the signal has a +/- 15 V voltage swing, then you're going to need some signal conditioning to keep the FPGA from releasing magic smoke. It's also unlikely to be the cheapest or simplest option.

[David_AVD]

Instead of a full matrix, could you switch the required input and output onto a separate line?  Somewhat like a telephone exchange.  It would depend on how many simultaneous (but separate) connections you need.

[dfmischler]

Can I ask a dumb question?  Why would you want to multiplex digital signals together with analog signals?  What would you do with them at the other end?  What about an SPI signal that needs clocking and other signals with it?

[timb]

If the signal has a +/- 15 V voltage swing, then you're going to need some signal conditioning to keep the FPGA from releasing magic smoke. It's also unlikely to be the cheapest or simplest option.

Right, I should have added that I'm only talking about 5V max digital signals in this case. I was just thinking of going a different direction in which case it would just be the digital signals.

Instead of a full matrix, could you switch the required input and output onto a separate line?  Somewhat like a telephone exchange.  It would depend on how many simultaneous (but separate) connections you need.

That's one of the first things I thought of actually! Sadly for this use it wouldn't quite fit the bill.

Can I ask a dumb question?  Why would you want to multiplex digital signals together with analog signals?  What would you do with them at the other end?  What about an SPI signal that needs clocking and other signals with it?

Well, I wouldn't want to mux digital and analog signals together. But I would potentially want to mux digital with digital or analog with analog. Or maybe just route analog signals or digital signals independently.

I guess a little clarification is in order. I was just investigating the feasibility of a programmable breadboard, for lack of a better term. It would allow any one column to connect to any other column on the board. In addition, you'd be able to route power and ground plus GPIO, UART, I2C and SPI lines from a uC to any of the columns.

The idea being faster, less error prone experimentation with ICs. It's obviously not feasible for analog signals (not cheaply, anyway), but I've found a few devices that  make it very doable for digital signals, so I still might pursue the idea, just the latter half (signals from a uC to any pin).

[Skimask]

Well, I wouldn't want to mux digital and analog signals together. But I would potentially want to mux digital with digital or analog with analog. Or maybe just route analog signals or digital signals independently.

I guess a little clarification is in order. I was just investigating the feasibility of a programmable breadboard, for lack of a better term. It would allow any one column to connect to any other column on the board. In addition, you'd be able to route power and ground plus GPIO, UART, I2C and SPI lines from a uC to any of the columns.

The idea being faster, less error prone experimentation with ICs. It's obviously not feasible for analog signals (not cheaply, anyway), but I've found a few devices that  make it very doable for digital signals, so I still might pursue the idea, just the latter half (signals from a uC to any pin).

ADG1406 / ADG1407
They'll handle +/-12v, but won't handle much current...~20mA max.
And at ~$12 each?  meh...probably not...  Likely would be cheaper to implement a crapload of DIP switches.