I have a 3 volt direct from battery power system using Vcc
as Aref to perform ratiometric measurements from 2 adc channels.
A third channel is used to read the internal 1.1 volt bandgap
(mux select "e") to determine battery voltage. Each of the 3
inputs is read sequentially, their readings placed in an array.
The two "external" channel readings are rock steady, but the
bandgap channel reading bounces around sometimes more than
+-5%. I improved the situation by taking two bandgap readings
in a row and using only the second one, thinking if the bandgap
was a "weak" source, giving the sample cap more time to stabilize
might help. This helped a lot, but does not completely eliminate
the problem. As a test, reading *only* the bandgap (not
doing any mux changing) results in a solid reading. The data
sheet mentions nothing special about selecting and reading
the bandgap. Anyone else notice this problem?
FWIW, I'm using a 2 mhz cpu clock, and a 125 khz adc clock.