Low thermoelectric coefficient solders

Standard lead-tin solder makes a thermocouple with copper of some tens of uV/C. Sorry I don't recall the precise number, it's been a long time since I faced this problem (mid-1980s). My solution at the time was to use Cadmium-Tin solder, which was about an order of magnitude better. It required a special flux and was incompatible with a soldering-iron tip used for standard solders. These days it may not even be available because of its cadmium content.

With a DC gain of a billion, everything is a thermocouple, even for a small delta like 30C. 1uV/C * 30C * 1e9 = 30mV. Do you really need such a high DC gain? Is there any way you can chop? I'd think your signal would be lost in a billion times the thermal noise...

Stephan Goldstein a écrit :

Better make that 30kV!

Tiny thermal gradients, millikelvin air flow variations, silicon chip stress effects, RF pickup, and pure Johnson noise make nanovolt measurements very difficult. And slow. Silicon has huge thermoelectrics against metals, too.

We sometimes glue thin surface-mount pcbs to temperature-controlled aluminum blocks, in insulated/shielded boxes, which helps some.

John

It seems unlikely; after all, a solder joint only bridges a few dozen microns distance between copper and copper... the temperature gradient times the distance is the temperature difference, it cannot be much.

You could eliminate one kind of dissimilar metal by welding, of course. Alas, weldable packages aren't the common ones.

Even commercial copper from different manufacturers can differ at the hundreds of nanovolts per degree level. DC gain of a billion is illusory except for SQUIDs.

Cheers

Phil Hobbs

Micro spot welding? At least you reduce and control different metals involved. Do that high gain circuit on ceramic, like RF modules? Then put module on PCB with other stuff not have the gain issue.

Grant.

Umm, yeah. So much for my touch-typing skills...

DC Gain of over 10^9, Wow! How much temperature drift did you expect? Can you tell us what you are measuring? (Sorry, I can't help with any solder TC effects.)

George H.

Hah! Right, I see I worded the query poorly. There's a (photodiode amp) TIA with a gain of a few hundred thousand at the front end. In this current gain configuration, its offset voltage is only amplified by unity, of course. The DC gain AFTER that is about 3,000. I assumed the drift was coming in at the front end but after you guys tossed some numbers around there, I see TE effects are larger than I thought and, if they ARE the cause, could be coming in after it. So I shall tie some things to 0V, and build a version with lead-free solder, and see where it's really coming from.

Cadmium-tin solder sounds like nasty stuff, but at least I know there's one option if it turns out that lead free solder isn't good enough! Thanks guys.

Oh, TIA with 'gain' of ~100k ohms (volts/amp), then 3k of voltage gain. Is the TIA opamp a FET? The input bias current keeps going up as you raise the temperature. That could change your offset voltage.

George H.

Nemo schrieb:

Hello,

there are a lot more unwanted thermocouples in your circuit than the solder joints. The IC leads are another metall than the traces of the PCB, inside the case there are bonded gold wires, the silicon is covered with aluminium traces. The resistors are also a combination of different metalls.

Bye