Lpc2000

At 08:29 AM 9/29/04 +0000, you wrote:
>I'm having problems getting correct readings on the AtoD pins. I have
>3 inputs which I'm converting individually under software control. I
>drive one pin from a voltage follower with 2.26 volts. V3A is at 3.32
>volts. I expect an A/D reading of (2.26/3.32)*1023 = 696. The Atod
>varies considerably depending on whether I read the pin on it's own
>or cycle through the pins. Does anybody have schematic/details of the
>analog input pins on the LPC2292.
>I'd like not to use the voltage follower but am getting even less
>stable readings without it. My source impedance is 4k7. What is the
>input impedance of the pin?

That does sound like an impedance mismatch.  Assuming the 2292 uses the 
same A/D as the 2119, the data sheet appears to be missing several 
parameters, the impedance to the sampling capacitor, and the sampling 
time.  It seems especially odd to be missing the sampling time, that's 
pretty critical information, maybe it's hidden somewhere I've missed?

With the small sampling cap (1 pF) on the micro I would expect it to be 
able to handle fairly large source impedances.  I take it from your 
description that you have no capacitor on the A/D input pin, try placing 
something like a .01uF cap on the pin (to aGnd) that should provide a 
sufficient charge reservoir for the sampling cap to charge/discharge and 
not change the size of the signal significantly.  Besides you need a 
Nyquist filter anyway :)

There will of course also be gain and offset errors

I wouldn't be surprised by a +/- 10 count error before calibration.  If you 
get into that range I'd start looking at the detailed A/D error specs.

Now some elementary error analysis on your input

(2.26/3.32)*1023 = 696.

That suggests that the input could be from (2.25/3.33)*1023 = 692 to 
(2.27/3.31)*1023 = 702

Add in a 10 count error from the A/D and you get a possible range of 682 to 
712.

That's starting to get reasonably close to what you are seeing for the 
static case.  Have you 'scoped the reference and input signals to see how 
clean they actually are?  In particular check for short period spikes (in 
the nS range or less).

Robert

" 'Freedom' has no meaning of itself.  There are always restrictions,
be they legal, genetic, or physical.  If you don't believe me, try to
chew a radio signal. "

                         Kelvin Throop, III