1

I am in the process of designing a data acquisition device that will be used outdoors, with temperature swings of around 30 degrees.

For this i need to scale the input to a voltage suitable to the adc and reference i am using, the initial offset of the circuit is a non issue as it will be calibrated out, however what concerns me is the temperature drift.

One way to be sure the drift won't be a problem would be to use resistors with a very low Tc, these usually are pretty expensive if i want a drift of less than 0.02% over the temperature range, so i was wondering if it was reasonable to assume that by using resistors of the same series and keeping them at the same temperature one could assume the drift will be reasonably equal in both hence canceling out or if that assumption is completely erroneous.

Thanks

diegogmx
  • 785
  • 3
  • 14
  • 1
    Same series isn't as good a bet as a combination of the same **value** resistors off the same reel, but they're both bets. – Spehro Pefhany Aug 13 '20 at 21:51
  • the only problem of that is that in that way you can only do 1/n dividers where n is an integer bigger than 2 – diegogmx Aug 13 '20 at 21:54
  • 1
    What ratio do you actually need? Series/parallel combinations for each resistor can do just about any ratio. And once you get down to 99% of the way there, you can use some other kind of resistor without much fear. – Spehro Pefhany Aug 13 '20 at 21:55
  • was thinking in 1.32 – diegogmx Aug 13 '20 at 21:56
  • yet i could probably change the reference so that 1/2 can do – diegogmx Aug 13 '20 at 21:58
  • 1.32 = Vin/Vout? – Spehro Pefhany Aug 13 '20 at 21:59
  • exactly, would need 7 resistors for that if im not wrong – diegogmx Aug 13 '20 at 22:04
  • If you want resistors with close-as-possible temp co's and temperature tracking, use a multi-resistor array such as: https://cdn.sparkfun.com/assets/9/f/e/3/6/515ddfc0ce395f9a58000000.png This is probably about as good as you can do w/o buying the crazy expensive stuff you mentioned. You'd want to be further careful to do things like make sure they all had similar copper traces coming off the PCB (because they'll sink heat), put it in an insulated box so differing air currents can't cool one side more than the other, etc... – Kyle B Aug 14 '20 at 00:51

1 Answers1

0

You could do it like this (resistor values can be scaled to whatever, obviously, just R and 8*R required):

schematic

simulate this circuit – Schematic created using CircuitLab

Since R5 is 24x the value of R1||R2||R3, its tempco will have very little effect on the tempco of the divider.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842