Monday, June 14, 2010

Infared -vs- Thermal Junction Temperature Sensing Part 1.

Recently I came upon a situation where I needed to take very detailed temperature measurements across a large number of locations to validate and environmental control system. Alas this was a subject I know very little about. In my scenario I was attempting to validate the spread across numerous points using a series of thermocouples to evaluate 2 optical temperature sensors in an environmental control system. So I had to do some research.

In my previous experience there had been 2 scenarios in which you measured temperature. For embedded devices a digital temperature sensor that came calibrated (like a DS1820 from Maxim/Dallas Semi) was usually easy to use, and could be easily bit banged with any old micro controller, meaning code the bus in C as opposed to needing hardware support like I2C/TWI etc. In IC design situations Bandgap Voltage/Voltage Proportional To Absolute Temperature (VPAT). In my experience they were usually used to give "perfect" reference currents for delicate analog circuits like the Optical chips Luxtera Makes. Of course that's until process varaiation, but that's outside the scope here.

So I will start with the fun stuff. Infrared Optical Temperature sensing is basically a practical application of the Black Body Radiation theory that was a first big step from classical physics to the quantum era. Nice! In a Blackbody, all light is absorbed, and re-emitted as a thermal spectrum. To me this is basically turning order into entropy. We know that the spectrum of a flame changes depending on its temperature. In fact the spectrum, and spectral peak of emitted light correlate well with the temperature of an object.

You can use the Stephen Boltzman Equation, add in an emissivity factor, and get pretty good measurements.
Emissivity is basically a measure of how ideal a material is for this type of measurement.

Typical Emissivity Values

Aluminum Paint
Asbestos: Fabric
Plastic: acrylic, clear
Plastic: Black

As you can see there is quite a range. Nickel is terrible. Graphite is almost ideal. In our case clear and black plastic also happen to be pretty good. There are other factors (good application note) like sample to spot ratio ("dot" from beam small compared to sample size), and getting the beam close to the sample to reduce stray radiation. With a good material the results are very accurate assuming theres no dirty air, smoke, stray light, and some other pretty obvious things. The part we're using claims 0.5C accuracy which is identical to the DS1820, and the results are instant (or basically the speed of light and a look up table), while the 1820 has thermal mass which must equalize with it's surroundings.

In part 2 I will examine Thermocouple based temperature sensing in comparison.

No comments: