On Wed, 6 Jul 2005 12:21:13 +0100, you wrote:

>As suggested, I've had a bit of a read up on the application notes for M=
AX197.  This is very purpose built IC and would be perfect (although it h=
as 8 different channels and I only need 1), but looking at the prices fro=
m RS, it's a bit impractical to use (=A320 for one IC!) shame.
>
>However, this set me wondering - IIRC (I am a newbie don't forget) there=
 are some PIC's with built-in ADC's could I not use one of these instead?=
  or am I missing something?  I think for 0.5C accuracy, I need a minimum=
 of 10-bit ADC.  Although I only want a temp range of 0C to 100C.  I coul=
d transform and scale the voltage to get better accuracy from 8-bit?
>
>I know I have to produce a reliable 1-2mA fixed current for the RTD's ex=
citation current.  Any suggestions on how I should do this would be welco=
me.
>
>I assume I can then map the voltage reading back to the resistance (with=
 some calibration factor applied?), and have the standard PT100 resistanc=
e table for lookup and interpolation.


To do PT100 on the PIC ADC, the best way is with an external dual opamp.=20

V+---Rlimit--A--PT100--B--Rref -----Ground

Rlimit sets the current - around 220R. RRef is a precision low-TC referen=
ce resistor - value depends
on temp range required, e.g. 1K

opamp configured as a differential amplifier with inputs A,B producing am=
plified voltage across the
PT100. Gain about 5, depending on temp range required.=20

opamp configurted as unity gain buffer from B to VRef of the ADC.=20
This provides a voltage proporitional to the exitation current, so your A=
DC reading is directly
proportional to resistance, with no need for a constant current source as=
 the effect of supply
voltage, and change in current with PT100 resistance are both cancelled o=
ut.

I have an excel spreadhsheet somewhere to help evaluate the effect of var=
ious component values
With a 10 bit ADC, my application did about 1 degree accuracy from -90 to=
 +180 - you could get
better if your range is lower.=20

As the PT100 curve is not far from linear, you can use a fairly small loo=
kup table with linear
interpolation. A table in 10 ohm steps gives a maximum error of less than=
 0.05 deg.C from -100 to
+350 deg.C 5 ohm steps gives better than 0.01 degree error. Unless you're=
 into serious ADCs these
errors are dwarfed by other factors.

>I assume there are some PICs that can perform floating-point arithmatic?=
  Or do you rely on the 'C' compiler to provide the functionality?  (neve=
r used the PIC 'C' compiler before, although I do know 'C').

You don't need floating point as the range is limited. Just scale it up &=
 use fixed-point.

=20

--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist