Hello all,
I'm designing a very simple resistor-ba...
# analog-designj
Jorge Marin
09/14/2022, 2:26 AMHello all,
I'm designing a very simple resistor-based temperature sensor, using P-well and poly resistors, taking advantage of the simulated temperature coefficient difference, please see the image attached.
How reliable would this be? i guess process variation can generate big differences, but I'm considering some calibration mechanisms. Can I trust somehow these simulations? do you see any other potential issues in doing this?
👍 1
l
Luis Henrique Rodovalho
09/14/2022, 7:57 AMIs it linear enough for your application? How will you calibrate it? How does it behave considering process variation?
h
Harald Pretl
09/14/2022, 10:23 AMTemperature sensors based on temperature coefficient of resistors are well established and pretty accurate (see attached paper). Since you plan on using two different types of resistors you might want to build a Wheatstone bridge (which is also suggested in the attached paper snippet). You need to plan for at least a single-point trim. The nonlinearity of the reading needs to be corrected in a post-processing step, ideally digitally.
Harald Pretl
09/14/2022, 10:25 AMIf you want to go deep, here you go: https://link.springer.com/chapter/10.1007/978-3-030-91741-8_12
j
Jorge Marin
09/14/2022, 8:48 PM@Luis Henrique Rodovalho the resistors are inside a time-based interface loop, so the nonlinearity is attenuated
Process is hard to say, since there's no chance of simulating process in resistors for SKY130...
Jorge Marin
09/14/2022, 9:34 PM@Harald Pretl thanks a lot for the references
I will take a look at the wheatstone bridge idea
The idea is to have a system which linearizes the resistor characteristic up to a point in which two calibration points are enough to calibrate to a linear function with a decent nonlinearity error
when you say single-point calibration, you mean that there's no gain error and therefore you just need to calibrate the offset error, right? or I'm understanding wrongly?
h
Harald Pretl
09/15/2022, 5:30 AMCorrect. Just a trim for offset, but no gain error. Of course, two-point trim will be better. In commercial ICca one-point trim is a huge advantage in mass production, so people go for that if possible.
Regarding spread: Assume +/-10% for poly res, +/-30% for well res.
j
Jorge Marin
09/17/2022, 2:14 AMthanks once more, @Harald Pretl
3 Views