Stable ADC reading

Assen · Joined: 18 Oct 2006 Posts: 8

I'm using PIC18F8722 and PIC18F6722 for my last projects and I had really hard time measuring the analog inputs in 10 bit adc mode. Even using Vref+/- inputs couldn't help a lot. I have tried many accumulation methods but they gave me not so stable result for measuring of slowly changed values for temperature or battery voltage readings for example.
One method that I wrote and I found as stable enough for these micros is this one:

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

1. Make a very simple A/D test program. Get rid of all that array stuff.
See the short program in my post near the end of this thread:
http://www.ccsinfo.com/forum/viewtopic.php?t=28887

2. Increase the delay after you change the channel to 20 us.

3. What is the output impedance of the device or circuit that's driving
the A/D pin ? The 18F8722 data sheet says it must be 2.5K max.

4. Add a .01 uF capacitor between the A/D pin on the PIC and ground.
Put the cap very close to the PIC pin.

Assen · Joined: 18 Oct 2006 Posts: 8

It's not a problem of aditional capacitors, low resistors or time to change the channel. This is an internal noise problem of the chip. I bought the original demo board from Microchip and it is the same - when you put the potentiometer to any position and leave it like that for minutes you can get floating result with +/- 2-3 digits, never less than +/-2 even with 10nF+100nF+1uF close to the pin. I changed the software with no change of the adc channel - the same result. Errata for these micros is clear enough - internal noise, +/-1.5 digits is the error when using Vref pins and up to 3 digits for the other pins.

Ttelmah · Guest

This is not what the erratum says.
The fault in the errata, is that the offset error is out of spec, _not_ a noise problem. Your system is showing a noise problem, and it sounds as if there may also be a long term drift in the reference.
Most of the PICs only give their noise floor ADC bhaviour, if sleep mode is selected for the ADC reading. However your drift, sounds as if your Vref, may not be adequately stable, or smooth.

Best Wishes

w2drz · Posted: Sat Dec 30, 2006 10:13 pm

Hi,

Try 2 things in your code.

1. replace the value (5) you have and replace it with a (25) us in the:
set_adc_channel(ach);
delay_us(5);
This settling time can be reduced in -5- us steps at a time to suit your results.
In the read loop add a: delay_us(6); delay before each read time.
this can be adjusted up or down, I find the us is fine.

I find this helps or cures problem in some chips I have used.

tom