Here's one for the analog gurus :)

I have a couple of GE (as in, General Electric / Thermometrics)
ZTP-135SR temperature sensors. This is a non-contact infra-red
temperature sensor, with two outputs:

  * Resistive output: a thermistor which measures the case temperature
of the sensor. 947.9k at -20C, 324.1k at 0C, 100k at 25C. Beta(25C/50C)
is typically 3960 K.

  * Thermopile sensor. Typical outputs at 25C: -1.25mV (object at 0C),
0V at 25C, 2mV at ~47.5C, rising to 7.5mV when pointed at a 100C object.

Output resistance is about 60k-ohms typical but could be as high as 78k
or as low as 42k.


Here's the documentation I have:

  GE product page and brochure (the Cliff Notes):
    http://www.ge-mcs.com/en/temperature/infrared-ir-sensors/ztp-135sr.html
    http://www.ge-mcs.com/download/temperature/920-159B-LR.pdf

  Technical datasheet (the Meaty Chunks):
    http://www.realchance.com.tw/pdf/CdS_Led/ZTP-135SR(218001).pdf

  Appnote: "Thermometrics thermopile sensors"
    http://www.thermometrics.com/assets/images/thermopile.pdf


There are four output pins:
  * Thermistor and case ground
  * Thermopile ground (isolated from the case)
  * Thermopile output
  * Thermistor output

Essentially, you measure the voltage output from the thermopile and the
resistance of the thermistor. Once you have those, you convert to two
temperature values and add the case (thermistor) temperature to the IR
temperature.

Now here's the thing.. I'd like to connect it up to a PIC. Assuming I
use parts I have in my junk box (and the PIC's internal A/D), Vref will
be ~4.096V.

As I see it, I'm going to have to do two things here:

  * Amplify by about 400 times to get a reading in the range of the
PIC's ADCON. This would mean an input of 10mV would get me an output of
4.0V.
  * Add an offset to allow measurement of temperatures lower than ambient.

This seems fairly easy:
  1) Take Vref and divide it by two using a resistive divider. Buffer
with half a dual opamp chip. Call this Vref/2(buf).
  2) Use a non-inverting amplifier circuit to amplify the thermopile
voltage. Connect the 'ground' side of the voltage divider to the
thermopile ground, then connect that junction to Vref/2(buf).

For the thermistor, I'd be tempted to use a resistive divider and a
unity-gain buffer amp, then linearise in software (Inverse
Steinhart-Hart or Beta-parameter). I need one measurement per second at
most (after averaging and so on).

Now on to the actual questions...

1) For this application, what type of op-amp should I be looking for?
Just a general-purpose FET-input amplifier, or something a little more
esoteric?
I have these in stock but am open to other suggestions:
  LM358 (who doesn't have a drawer full of these?)
  LM7171B1N (appears to be a high-speed voltage-feedback amp)
  OP37GPZ (low noise, precision, high speed -- 17V/us, 63MHz gain BW)
  LMV321M5 (general purpose, low voltage, rail to rail)
  LMV791MK (low noise, CMOS input, 1.8V, PowerWise)
  LT1207CS (dual 60MHz 250mA current-feedback, 900V/us)
  NE5532 (dual low noise)

2) How can I keep the noise level in the sensor and amplifier to a
minimum? Obviously I can average the A/D values in software, but I'd
prefer to stop the noise being generated in the first place!

Would it be worth adding a lowpass filter to the amplifier, or am I
better off doing that in software (averaging)?

3) Is there anything else I should be aware of when designing something
like this?

Pardon the "fifty questions" treatment, but these are expensive sensors,
and I'd like to get this right!

Thanks,
--=20
Phil.
piclist@philpem.me.uk
http://www.philpem.me.uk/
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.