At first analysis to measure from -40 to +100 in 0.1C steps requires being able to resolve the input signal in up to:

(100 - (-40) + 1) * 10 = 1410 steps.

A 10-bit A/D converter with only 1024 steps will never get you to where you want to go. I suggest that as a minimum you will want to do one of the following:

a) Change your A/D converter to at least 12-bits so you get 4096 steps.

b) Change your expectations on resolution from 0.1 to 0.5C. This change lets you resolve your temperature range up into 141*2 = 282 steps.

It will be required that your A/D converter setup have more steps than the actual resolution range of your temperature scale for the following important reasons:

1) The thermistor will give nonlinear behavior and so in part of the reading range the amount of signal change per resolved step will be compressed over what it would be of the other extreme of the curve.

2) You will want to setup your circuit so that it is able to present signals to the A/D converter that are slightly wider than the end points of the readings you want to capture. This is necessary so that you can calibrate in the end points without finding one end or the other getting clipped.

3) Not every thermistor circuit that you build will offer the exact same range of signal level to the input of the A/D converter. So you need to allow for that characteristic via a calibration process.

You may want to consider an alternate means of measuring temperature instead of with a thermistor. A better choice would be to use a silicon diode or transistor's B-E junction. When biased at a constant current (1 mA suggested) the forward biased voltage drop across the diode junction is surprisingly linear over a wider range than you have for your goal. (You still have to calibrate your endpoints as each diode would gave a different response and the current source will vary some from circuit to circuit). If you use this diode junction idea you will need to use an opamp to provide some gain and offset to the signal. The junction also has a decrease of voltage with increasing temperature and so the opamp could also invert this behavior and permit a positive voltage change with increasing temperature at the A/D converter.

Another thing to consider for making temperature measurements would be to use a packaged I2C type device such as an LM75.

Michael Karas