PIC16F1788 ADC reads full scale after pwr on

[Ted]

After power on, I have to do one dummy read of the ADC before I get, with the second read, a useful result. The first dummy read gives a fullscale result (65535, 2 bytes). This happens every time and is fully reproduceable. Has anyone ever observed this behaviour before and what could be the reason?
I use a PIC16F1788 and my setup and read commands are as follows:

[PCM programmer]

Post a full test program. ie., #fuses, #use delay(), #device statements,
main(), variable declarations, etc. But keep it short, and make it be
compilable with no errors. Post your compiler version.

Post a description of the circuit feeding the AN1 pin.

Post the Vdd voltage of the PIC. Post the voltage of the AN1 input signal.

[Ttelmah]

I'd wonder about one thing.

This chip is one that supports differential ADC operation. As such it has two parameters to the set_adc_channel function. The + and the - channel. You can either use two channels to get the differential value between them, or:

set_adc_channel(1, ADC_NEG_REFERENCE);

Which uses the channel in the 'normal' manner selecting the ADC negative reference for the second input.

[Ted]

first to answer the questions:
Vdd ist 5V
compiler is PCM 5.073
the AN1 pin has 2.2nF to GND and 5.6k to approx. 120mV (in idle mode)

I distilled a short test program from my project. It reads the two ADC's und diplays the results on an LCD. It works fine!! The first reading after pwr on is close to zero (2.2nF not yet charged), the second an further readings show the expected 120.

Then I got the idea to pull down the120 mV towards zero and below (negative). Surprisingly as you get below zero, the ADC results start from full scale (65535) downward again.

I did not expect this, and it probably explains my problem. I may have a couple Millivolts negative on the AN pin right after pwr on (due to voltage drop on a GND line) which gives me the full scale result from the first read_adc().

But there is still one thing confusing me: One would expect a delay (I tried up to 3 sec) before the first reading would cure the problem. It does not. Only a dummy read before the first useful read helps.

For the sake of completeness, here is my test program:

[Ttelmah]

OK. That makes sense. Unlike most of the PIC ADC's, this one supports differential operation. So potentially can give a -ve result, What is happening is with the -ve voltage, this is being seen, and takes a read or two to clear.
I'd suggest actually explicitly clearing it, by setting up the ADC to use the same pins for both inputs, then reading once, and then switching to normal use. So:

[Ted]

Yes I'll do. Thank you!!