A voltage is developed along the temperature gradient in a wire. That could be the entire length, but not necessarily.
Say you have 1m of wire. The hot end is at temperature T. The other end at room temperature. Suppose that at 10cm from the hot end, the temperature is also at room temperature such that from 10 cm to 100 cm of the wire is at room temperature. The voltage will appear only from the hot end down the wire for 10cm. Assume that there is no voltage drop for the remaining 90cm, the the voltage from the hot point to the 10cm point will be the same as from the hot point to any point along that 90cm.
A common installation in temperature controllers is to place a 1MΩ resistor from one thermocouple wire to a positive voltage rail. That way if the circuit breaks, the indication goes to a high temperature, shutting down any heaters. That prevents a runaway condition.
tomii, presumable, if you wre to build this circuit, you've calibrate it using some other instrument that is NIST traceable. At the very least, you'd compare the measurement circuit against something you trust, NIST traceable or not.
Calibration is all based on trusting the device you're using as a standard. I've been to stadarnds labs where they people compare the latest temperature-measuring instruments, with 4-wire RTDs and 20+ bits of resolution against glass thermometers. Sounds crazy, right? But, the glass thermometers was the trusted instrument. The high-end digital device was an unknown.
Hi Caleb. I don't have an engineering degree either - so I am glad I am in such good company!! but I digress.... Temperature measurement..... I have been doing some articles for Max recently on the PICAXE MCUs. The first one is posted, the second will be up soon but in the third, which I am working on at present, I am dealing with the Dallas-Maxim DS18B20 digital temperature sensors. Accuracy of 0.5 deg C, from -55 to +125, and the interface to the PICAXE, and reading them, is REALLY easy, Not in the same league as the thermocouple system in the EDN article, but for a quick easy temp measurement I don't think you could get easier. I'll try and ping you again when the relevant article is up.
I'll say that the posted article is quite a bit shorter than the original .. Part of that was handled by the link to the older thermocouple article.
I think every one of us has gotten it "wrong" before (if not still). Extra sources of error also come from the calibration curve of that thermistor (the PT1000) and the junction of diddimilar metals at contact points (e.g. thermocouple lead to copper PCB trace). Dissimilar metals in contact with each other present a "galvanic" potential (this is how a battery works, right?), where other impurities/humidity act as the electrolyte. (thus the source of "galvanic corrosion")
Also, the variance in "on channel resistance" over temperature of the analog mux may well be a significant percentage of error in this circuit (from the part's datasheets)