Battery voltage divider to get hundreds of µV levelse to measure ADC offset.

Hello.

I'm trying to determine the offset of the ADC in my arduino Due, 3,3V reference 12bit SAR ADC.
I can put on a small voltage without any transition from 0 reading but I want to know how little it is, 3300mV/4095=0,80586≈806µV LSB

My brain does not literally work always as it should and today I can't figure out the simplest of things, I've taken a 1,3V battery and solder 1k smd resistors in a long string one after the other. I've now got 13 1% tolerance resistors, am I correct in that when I measure the voltage over the bottom resistor I have a 1/13 full battery voltage with a 13% error margin?

If so that ain't nice, I'm looking for a way to scale down the voltage to a low enough level to enable it to be possible to adjust it with one of those small multi-turn potentiometers with a little screw on the top, mine are 100ohm, or 97,4 ohms. GND and Vbat to each outer leg of the pot and the viper to supply the voltage.

Also, when I have derived a voltage with a voltage divider. Does the voltage change in any degree when connected to the ADC input?

I don't care about the absolute accuracy very much right now but 13% is a little to much, in any case I'm going with the assumption that my DMM shows the voltage accurately. I have planes on a calibration board using a few very fancy voltage reference ICs but that's for later.

Why does the battery voltage slowly climb up when measured by my multimeter?

You are asking a lot of totally different questions.

1) The input impedance of your ADC has to be much greater than 13k (the impedance of your resistor string), or it WILL affect your reading. That's just a simple parallel resistance calculation. I would buffer it with an opamp.

2) Worst case error will be about 1.9%. Assume all your resistors are off by 1% in the same direction--> 990 ohms. Then the total error would be 990*13/1000*13-->1%. If by some fluke, all but one of your resistors were off by -1%, and the other was off by +1%, then you could have an error of

nominal ratio=1000/(13*1000)=.07692
worst case ratio=1000*1%/(1000*1%+12*1000*(-1%)
=1010/(1010+12*990)
=.07835

error=.07835/.07692
=1.019

3) Why does the battery voltage climb slowly? I have no idea. Charging a cap, maybe?

Thanks, its actually embarrassing that I figured it to 13%. Thanks for explaining.

I will have to check the MCU datasheet but I don't understand what they mean, it says that Zsource is a function of the ADC clock:
MHz: kOhms:
20.0000 10.0000
16.0000 14.0000
10.6700 22.0000
8.0000 31.0000
6.4000 40.0000
5.3300 48.0000
4.5700 57.0000
4.0000 66.0000
3.5600 74.0000
3.2000 83.0000
2.9100 92.0000
2.6700 100.0000
2.4600 109.0000
2.2900 118.0000
2.1300 126.0000
2.0000 135.0000
1.0000 274.0000

But what is Zsource... is it the impedance seen by a external circuit or is it the needed load on the input by the external circuit?

I have experienced that climbing with other measurements and I don't understand, like with this battery, there is nothing but the multimeter connected and I cannot understand.

- - - Updated - - -

the only clue I can find is "the allowed Zsource" so I guess that is the allowed load to put on the input...

But what is Zsource... is it the impedance seen by a external circuit or is it the needed load on the input by the external circuit?

Click to expand...

It's what the word says, the source impedance. More exactly the maximum source impedance to achieve 12 respectively 10 Bit accuracy. It doesn't hurt to drive the ADC by a lower source impedance, e.g. a "zero" ohm buffer, but increasing it above the limit affects accuracy.

The voltage divider you described should be fine.