I've had similar problems with the ADC on AVR AT90S8515,
Which is very similar to the 10bit ADC on PICs.
I then changed to the Microchip MCP3202 SPI ADC, which
Has got better resolution (12 bits) and it works a lot better.
Another thing to watch for in analog designs is the ground
planes and the decoupling capacitors on the power supply.
Microchip has got a very good appnote (AN688)in this regard.

Jeethu Rao


> -----Original Message-----
> From: pic microcontroller discussion list
[mailto:PICLIST@MITVMA.MIT.EDU]
> On Behalf Of Nick Veys
> Sent: Sunday, November 03, 2002 4:10 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: [PIC]: 16F870 ADC inaccuracy
>
> Had this problem a while ago and just didn't care because the
> application wasn't too important, but I'm coming up on a project in
> which temperature measurements will need to be within a few degrees...
>
> For this test, I am using an LM35 (Centigrade temp sensor, 10mV/C
> output), and a 16F870's 10bit ADC, and spitting the value out onto an
> LCD.
>
> I consistantly get around 10C (8-10 in the range I can test) less than
> the actual output (by probing).
>
> Say, for instance if I probe and get .334mV, that is 33.4C, but my
> output is around 22C.  That's just not right.
>
> I'm wondering if the built-in ADC's are just simply horrible?  Or am I
> doing something wrong?
>
> My scaling is as follows:
>
> 0-5V for the voltage range (Vdd as reference)
> 0-1024 for the input value range (10bit ADC)
>
> so it's simply (adcvalue * 5) / 1024 -> actual voltage, correct?
>
> The output does of course rise when I heat the sensor, and drop back
> down as it cools, it's definately getting *some* value from the ADC...
>
> I don't know of any offset register in the pic as there are only 4
> registers for the ADC in general, and none mention an offset i would
> need to set up.
>
> Now, I realize I could use a more accorate reference, a lower voltage
> for the reference to gain resolution, but right now I just want to get
> this working and move from there, and honestly, it should be much,
much
> more accurate than this.
>
> Ideas? Pitfalls?
>
> --
> Nick Veys | nick@veys.com | www.veys.com
>
> --
> http://www.piclist.com hint: To leave the PICList
> mailto:piclist-unsubscribe-request@mitvma.mit.edu

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics