Well, for limited temp ranges, you can fit a polynomial to it and get pretty
decent results -- I've done that for 0-50C.  The problem is that the curve
really is logarithmic, so no polynomial will fit very well, at least not
over any extended temp range.  Hint -- buried in Excel somewhere there is a
polynomial curve-fitting wizard, I just don't remember where.

> -----Original Message-----
> From: Eisermann, Phil [Ridg/CO] [mailto:peiserma@RIDGID.COM]
> Sent: Monday, May 22, 2000 12:47 PM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [PIC] Converting A/D into Temperature
>
>
> Damon wrote:
> > I've got an A/D circuit on my 16F874.
> > I have a couple of OP Amps hooked to some resistors a few Pots and a
> > thermistor so that I get a 5X gain on the voltage.
> (schematic available
> > if needed)
> > I have it giving me 2.5V at room temp. But for the LIFE of
> me I can't
> > figure out how to convert the 0-1023 value into a usable
> temperature.
> > I have an Excel spreadsheet that will tell me the readout
> if I type in
> > the voltage example: 0v=0, 5V=1023, 2.5V=512, 2.813782991V=576
> >
> > From my chart that last Voltage should correspond to a
> reading of 68F
> > (20C)
> > Maybe it's late and it's really simple but I can't figure
> it out.. any
> > help out there?
> >
> >
>         Well, using a thermistor, it's not exactly
> straightforward. The
> reason
> is the highly non-linear resistance to temperature
> characteristic of the
> thermistor. I have been working on something similar (in my copious
> spare time). This equation is from memory, i'll look it up
> tonight when
> I get home:
>
>         T(Kelvin) = 1/(A + B*ln(R) + C*(ln(R)^3))
>
> This equation is the Steinhart-Hart equation. R is the resistance
> of the thermistor. A, B, and C are material constants that
> you can figure
> out by measuring the resistance at three different temperatures. The
> constants are all less than 1 for my particular thermistor.
>
> Now you can see the difficulty with this equation in a PIC. While the
> fractional values can be taken care of by scaling everything,
> calculating the natural log function is a bit more difficult. On top
> of that, the temperature (in Kelvin) is actually the inverse.
>
> Of course, I admit I am not the world's greatest PIC
> programmer. Perhaps
> there is an easy way to realize the required mathematical operations
> on a PIC.
>
> In my circuit, I placed the thermistor in parallel with a
> fixed resistor
> in an attempt to linearize the resistance a little. Over a certain
> range of values, this works fairly well.
>
> A lookup table is one possible solution. You could put a lookup
> table for all possible values in external EEPROM. Or you could
> place a smaller lookup table in memory, and interpolate between
> points.
>
> If anyone has any other ideas, please jump in.
>
> Phil Eisermann
> H:(440) 284-3787 (mazer@ix.netcom.com)
> O:(440) 329-4680 (peiserma@ridgid.com)
>