diy thermocouple project, connectors,etc for 1:n selection

[linux-works]

I'm thinking of doing a quick project to allow one controller (likely arduino, might be rasp pi) to sample values from many thermocouples.

first problem I encounter is the cost of real type-k female connectors.  lots of fake ones on ebay/amazon and I'm not able to find any on mouser, yet.  the ones I've read reviews on, on amazon, all say that the metal contacts are not the right metals.  maybe for a proof-of-concept build, it might be ok, but I'd prefer to find real connectors.  where can I get the females (blade type) without spending a fortune?

assuming I find real ones, next question is: any time I leave that connector blade and change metals, I'm now making a tc junction, right?  how do the T/M companies do this?  at some point, a tc female will be on a pcb or be wired to one and then you change metals.  if there is solder, there is also a change of metals.  all this affects things, right?

how would I select 1-of-n if I have a bunch of female connectors on wires or on a front panel; how bad would it be to use relays to do the 1:n selection?  and if I step thru using relays, is there a certain settling time needed (after a relay switches to the current tc) before I can get a good reading from it?  if relays are not the right way to go, I'm not sure if a solid state switch of some kind would work.  I'm hoping to switch the 1:n *before* the tc amp module.  I need to support upwards of 30 tc sensors on this diy project and I can't justify 30 tc amp modules running 1:1.  I need to share the tc amps and need some way to switch the tc modules in and out on that single tc amp, under cpu control.

I know the relay solution won't be very high end, but it might be close enough for my needs.  and I may be able to cancel-out the effects of the connectors and relays since it will be common to all the 30+ measurements, and those will be attached to a device-under-test, perhaps even in a temp chamber.


any hints as to how to proceed?

[Kleinstein]

Those yellow (for type K) flat pin connectors are rather good for TC. Usually these should be somewhere is the $3-$5 range.
Sometimes rather expensive round metal ones are used (LEMO), with these I have bad experience, despite there high price.
If price is really critical, one could use direct connection to the cold junction temperature, thus have the connectors in a low temperature gradient area and measure that temperature. So the TC to copper transition would be right at the plugs / screw terminals.

For switching relays are not that good - best avoid them. One can use relatively cheap CMOS switches - even 74HC4051 or similar should work good. Just  make sure there are no significant temperature gradients - which is relatively easy as 74HC... is low power.

For so many channels I would consider more than one amplifier / ADC to reduce the noise requirements. So more time could be spend for each input.

[jpanhalt]

Let's start with the thermocouple    .What do you feel a "real" female type-K  connector should be made of?   What about its male counterpart?

John

[Kleinstein]

A possible source would Omega.com. The connectors are at something like $2-$3.

[Marco]

For so many channels I would consider more than one amplifier / ADC to reduce the noise requirements. So more time could be spend for each input.

I don't know what kind of noise a thermocouple generates, but if it really needs filtering why not simply give each one an RC filter before the switch? A 100 Hz filter with 100K would only add around a uV p2p noise, so you don't need much capacitance.