[volt-nuts] Some questions to zeners (thermoelectric effects)

John Beale beale at bealecorner.com
Sun Jan 27 15:31:40 EST 2013

On 1/27/2013 9:36 AM, Ed Breya wrote:
> I think your expectations are not realistic - even if you could make such a
> reference, you could not transport its voltage to the ADC without
> thermoelectric effects causing error that would swamp the performance. To
> keep everything below the 1 ppm/deg C range you would have to put the
> entire circuit in controlled temperature - the reference, the ADC, and the
> signal connection to the outside world.

I don't have the practical experience or measurements to back this up, but 
I understand Seebeck thermoelectric effects are a function of the 
temperature difference between dissimilar-metal junctions, and not absolute 
temperature. So if you have perfectly balanced both the thermal mass and 
the thermal conductivity to ambient of every bimetallic junction in your 
circuit, there should be zero tempco of the system due to thermocouples, 
regardless of both absolute ambient temperature and the rate of temperature 
change with time.

So in theory, if you use a symmetrical circuit layout with balanced thermal 
mass* and then surround your battery-operated device with a large enough 
block of metal (to minimize both thermal gradient, and rate of change with 
time), you can get d(temperature)/d(time) of the circuit and the associated 
internal temperature differentials to be arbitrarily small. How practical 
this "large metal block" would be to meet a <1 ppm/C tempco spec, I do not 
know.  Assuming you have avoided the copper oxide problem (Cu-CuO: 1000 
uV/C) the worst thermocouple will be Cu-Kovar at 40 uV/C so layout at and 
around the IC packages will be the most critical.

I assume the hardest connections to keep thermally equalized would be the 
terminals connecting your reference/ADC to an external device. If your 
voltmeter is limited to low voltages, optimizing this suggests the smallest 
and most closely-spaced connections possible, embedded in an insulating but 
thermally conductive matrix (ceramic?). Standard banana jacks with 3/4 inch 
spacing and surrounded by plastic, seem far from "small" or "closely 
spaced" or "well thermally coupled".

* The "20-bit DAC" app note mentions this technique:

John Beale

More information about the volt-nuts mailing list