Mark A. Corio wrote:
>
> The op-amp circuit that Octavio drew is one that got me in trouble once.

And if you look closely, you will see his is almost identical to the one I
 posted.

>  While the temp sensor is variable by temperature, the subtraction circuit is
> sensitive to power supply voltage.  If the supply varies, so does the
> temperature reading.

True. However, there are two things you can do to overcome that problem. First,
 you
can low pass filter the power supply voltage with say a 100 ohm resistor and a
 1uf
capacitor. Second, you can resistor divide this filtered voltage so that it may
 be sampled
by the same A/D converter that the temperature circuit is being sampled by. This
 way,
it is possible to compensate (software) for supply voltage fluctuations.


>                       Also, if the temperature sensor is not mounted
> remotely, care is required that all the resistors and other components don't
> introduce an error with temperature drift.

No doubt about it. 5% carbon comp resistors typically have temperature
 coefficients
exceeding 250 ppm / degrees Centigrade (ppm = parts per million, 1ppm = .0001%).
 So a
50 degree centigrade change can cause 50 * .025% = 1.25%! And, what's worst you
 don't
no apriori if it's + or - 1.25%. However, 100ppm/C 1% resistors are commonly
 available
(digikey has them) and if it is absolutely necessary, you can get 20ppm/C wire
 wound
resistors at about 2 USD a pop.


One more note on this thread. I recently received the latest Maxim selector
 guide. They have
a series of parts (MAX951 - MAX954) that include a rail-to-rail Op-amp, a 1.2V
 voltage
reference, and a comparator all in one 8pin package for 1.60 USD in qtys > 1k.
 Sounds like
an excellent part for this project!

Scott