ADC biasing question

Using Arduino Sensors
#1

I have a voltage read for battery but I would like to increase resolution by removing the bottom half of the battery voltage. Basically change the ground point. Is this possible?

Battery is 2 cell lipo so approx 6 - 8.6v

Is it possible to read only 5 - 8.6v and ignore the bottom? The voltage divider changes the reading to 5v max.

This way instead of using only 1/3 of the ADC resolution I should get closer to 2/3 the range to empty.

Can I tell analog read to say read from 2.5 to 5v and translate that into 0-1023?

#2

Depends on the rest of your circuit.

If the Arduino is powered independently from that battery you could connect the Arduino GND to your battery's mid point, making that your GND reference.

The biggest problem may be that this requires a voltage divider, which by nature leaks power. This may be an issue for battery powered projects, if it's a small battery that has to last long.

Anyway, you merely double your resolution. No great gain. If that is make or break get a higher resolution ADC such as the ADS1115.

#3

That’s what I thought. Can’t be done without isolating completely. Yeah you’re right doesn’t help the accuracy issue. Trying different things...

#4

Some AVR processors support differential reads.

That still leaves you with the problem of bringing the values down to less than Vref (Vcc unless you changed that, to e.g. the internal reference).

#5

I was hoping I could live with the error with higher resolution.

#6

You need a voltage divider to bring down the battery voltage to something you can read (best is to take it down to 1.1V and then use the internal reference for stability).

Then you're reading your 7V battery at 7 mV steps using the 10-bit ADC. Is that really not good enough? In other words: does it really matter if you see your battery voltage at 6.445 V or 6.452 V, when the actual voltage is 6.449 V? For most practical uses it's good enough to display this as 6.45 V.

#7

I just need to operate an auto shutoff at 6v. Even with the errors I think I can get close enough. 2.8 - 3.2 per cell

It’s a matter of accuracy not so much the resolution. I’ll play around in excel but I don’t think it matters.

#8