Multiplexing thermocouples, what to look for in multiplexers?

Question

I would like to use a MAX31855K IC for amplifying K type thermocouple response. This alone seems rather straightforward, but I'm also considering connecting 2 or 3 thermocouples to one IC. I don't need high frequency (a measurement every 0.5s is enough) and I also don't need it to be super accurate. I'll be measuring high temperatures, in ranges of 300-600C (570-1100F) and I'm fine with accuracy of 10-15C in that range.

I found a project that uses ADG407 analogue multiplexer for that, which was described as a successful try.

Before I try this on my own, I would like to ask on what should I consider when looking for multiplexers for this project? Or even, is a multiplexer the right choice? There're multiple ways of routing signals. I could also probably use an analogue switch IC, or maybe even some kind of transistors?

Initially I was thinking that a really low Rdson (like a few ohms) would be a key, to not alter a signal, but on the second thought the current there is probably close to 0, so ADG407's 80 ohms sounds fine. Is there anything else to consider? Is there any Rdson level that would not be acceptable for my requirements?

Answer 1

I use DG409 multiplexers (100 ohm max) and switch both thermocouple wires in order to multiplex up to 16 devices. Ditto the cold junction RTDs and I also have spare multiplexers to give me a "zero" reference and an actual positive voltage reference of about 75mV. Done it this way for years.

So, I believe the trick is to not worry too much about on resistance but keep an eye on relative on-resistance specifications between the internal multiplexing switches. Run the mulitplexers from both positive and negative supplies because you cannot be sure about the polarity of the thermocouple signal AND you want to have decent headroom to each supply to avoid strange on-resistance effects. I use +/- 5V.

Because you are going at fairly low speed I'd hang a 10nF on the common output of the multiplexers and also consider 10nFs on each input wire to 0V.

Answer 2

Bias current is 100nA (between inputs) and type K is about 41uV/°C so 100 ohms per line would represent an error of less than 0.5°C.

Make sure your mux supplies are both plus and minus and at least 5V (more voltage gives you lower resistance too).

It's not clear to me how capacitors on the inputs will interact with the error detection test, which appears to dynamically change the inputs, so you might want to test that thoroughly to ensure it does not cause errors.

A small cap across the mux output (like 10nF) should be okay.