I’ve been beating my head against this for a day or two, and I am stumped.

I am trying to measure car battery voltage.

I have a simple voltage divider with 47K and 20K resistors. I have a 1.5uF cap to ground as well to filter noise (I have a lot of noise at 1.5Khz and 3Khz). This is connected to an A/D pin.

With 12.7V input, I should be getting a reading of:

1024 * (12.7 / 5) * 20 / (47 + 20) = 776.

I’ve checked everything I can think of. I checked Vcc; it’s 5.03V.

No joy.

The voltage is steady (it’s connected to a 225Ah battery). Could the cap be pulling down the voltage? It doesn’t seem to make sense.

Could the cap be pulling down the voltage?

No.

I have a simple voltage divider with 47K and 20K resistors.

So what tolerance are these resistors? If each one was at the wrong end your divider could be off. I haven't worked it out because I don't know what tolerance you are using, but it feels like this is the problem.

Crap. Thanks for pointing out the obvious. I would have sworn I got 1% but I just checked the data sheet and they're 5%. Still, it seems to be a lot.

I also have a 5V zener to ground paralleling the cap. Could that make a difference? The leakage is on the order of 50nA, so it should be orders of magnitude less than the .2mA across the resistors.

I also have a 5V zener to ground paralleling the cap. Could that make a difference?

Yes it could. The knee on zenners is not so sharp so you could be getting a lot more current down it than you think. Leakage current is often for the reverse bias case, are you sure that it is so small for forward bias.

I can always build another board with no zeners. Or clip this one off and test. I'm going back and forth on the zeners.

I need something to protect against load dumps but zeners just seem to have a lot of issues.

Yes zenners are fine on a digital input but have issues with an analogue one. I would use clamping diodes to the rails.