Low thermal EMF scanner and ordinary non-latching relays

[manganin]

The question of selecting relays for a low thermal EMF scanner pops up now and then. And often turns into a competition who knows the most exotic (and expensive) relay type, contact material or contruction technique. But has anyone actually tested if those things are important in real life?

When designing my new 120 input scanner, the decision was easy because I got 150 Panasonic TQ2-5V relays for free. It is a small footprint non-latching general purpose relay. The datasheet values may not be the best in the market, but according to my experience in other applications, it performs very well in low level switching.

A latching relay driven by short pulses would of course make things easier because it produces the minimum amount of heat. But ordinary single side stable (non-latching) relays are widely available, much cheaper and also easier to drive. May be meaningless in an eight channel scanner, but for example in my 120 input system both the component price and the drive circuit complexity become very important.

Luckily the temperature rise itself is not crucial because the circuit is balanced. The most important thing is the mechanical symmetry of the relay i.e. how evenly the heat distributes. Based on the tests that I made many years ago there are expensive but poorly behaving special relays, as well as cheap general purpose relays which perform very well. Try before you buy! The contact material had very little to do with the actual performance.

And don't forget the physical orientation of the relay because there is air convection inside and outside. Mounting the cards vertically (like Dataproof) is not a good idea.


Test circuit.




Test jig.




And the results.




Some recovery time is needed to achieve the thermal balance again. If the swithing period is shorter, you will see a cumulative error component. But in a multi-channel system the ON time will be only a fraction of the OFF time anyway.

Note that the point of interest is the energized state (green). What happens when not energized doesn't matter because the relay is not in the signal path.

[Andreas]

Hello,

thanks for the measurement.
I think the energizing time is too short to come to equilibrum.
The diagram for the coto relais shows around 22 minutes warm up and cool down time.

For me also a single switch measurement would be interesting.
(one pole the common and the other pole the 2 positions).

with best regards

Andreas

[EmmanuelFaure]

Amazing! What measurement equipment have you used?

[plesa]

Interesting, thanks. I was also in my plan to measure it with nanovoltmeter. Bookmarked:)

[Kleinstein]

The time to equilibrium can be longer and depends on the environment / mounting. So just the relay in a foam block may behave different from one on a board.
For a scanner 1 min at one port is already quite long.

[manganin]

I think the energizing time is too short to come to equilibrum.

The goal is to produce as little heat as possible. Why would we want to achieve thermal equilibrium by extending the energized period?

The 1 minute ON and 5 minutes OFF periods were chosen to match the real life as well as possible. 1 minute is pretty much the maximum for an 8.5 digit measurement. In most situations the energized time will be shorter than that (and so will be the recovery time required).

What measurement equipment have you used?

Good old Keithley 148 and a HP 34401A for logging.

The time to equilibrium can be longer and depends on the environment / mounting. So just the relay in a foam block may behave different from one on a board.

Yes, it depends on the rate of the heat exchange between the relay and the environment. In my scanner there is an aluminium equalizing plate close to the relays. The "heat sink" reduces the time constants as well as interaction between the adjacent relays.

[Echo88]

Interesting measurements! 

Maybe a Alucore-PCB is in this case useful, since it will act as a very good isothermal environment by spreading the produced heat of the relays.
https://www.multi-circuit-boards.eu/en/products/printed-circuit-boards/metal-core-pcb.html

[Andreas]

The goal is to produce as little heat as possible. Why would we want to achieve thermal equilibrium by extending the energized period?

Hello,

my intention was to see the "worst case"
and perhaps to compare with the coto datasheet.
But of course you are right this would be no practical measurement condition for a non latching relay.

With best regards

Andreas

[Andreas]

Maybe a Alucore-PCB is in this case useful, since it will act as a very good isothermal environment by spreading the produced heat of the relays.

My opinion is: we rather have a good isothermal between the relay contacts.
So a good thermal isolation of the pins might improve the EMF.

With best regards

Andreas

[manganin]

For me also a single switch measurement would be interesting.

Please see the raw data attached. Could you please make a graph?

1 min OFF --- 1 min ON --- 5 min OFF --- 10 min ON --- 10 min OFF

Scaling: 1.0 = 3 uV

[Andreas]

Thanks.

Here it is:
I hope the scaling *3 in uV is ok:
Obviously a single contact is much worse than compensating two identical contacts.

with best regards

Andreas

[Echo88]

Wouldnt have thought that it would be that much error for a single contact. But hey, if it would be otherwise then nobody would buy COTO-relays.