I'm working on an application where temperature difference from initial/ambient temperature is measured. I'm using multiple TMP275 because among those in relatively robust cases, it's the most accurate as well as I2C with settable address. And extremely easy to use. Wonderful, really, but I'm having one odd problem:
In order to verify wiring, I2C addressing, and general understanding of the part software-wise before building the real physical prototype, I mounted 2 of the sample devices (TMP275AIDG4, in case that makes a difference) on opposite ends of an approx. 2"x3" SOIC breakout board. Connections between the two devices are made with thin wires and PCB bus traces; at one end of the PCB, pins plug into a breadboard, and I2C and power wires go another 4" from there to the microcontroller/display board.
I realize these sensors tend to measure PCB/trace temperature, but operating current should not be enough to heat 24AWG wire measurably like thism should it?
They both give reasonablem roo-temperature results, but the sensor on the far end of the wiring gives a result about 1-2 deg. F higher. (This is after plenty of time to soak with the room temperature.) Taking the circuit to a 10 deg. colder cold basement room has the sensors settling to similarly apart results. The offset is not necessarily constant or linear, but given enough time to soak, it tends to be between 1 and 3 degrees.
Potential differences:
-Airflow: given enough soak time, this should not be a problem. I've tried both one-shot measurements to reduce device heat, and continuous measurements to eliminate warm-up effects. Results are quite unaffected by that.
-Power supply: the fairly precise 3.3V supply from the uC board is fed to the sensors using wires with no other consumers attached. Initially, I had no local bypass cap; adding a smallish one on the breadboard (thus next to one sensor and 3" from the other...) had no effect. Could this be possibly be a problem? Does one sensor being half the distance from the uC board's caps affect the result? I have not yet tried putting smaller caps right at the sensors.
The datasheet speaks of absolute temp. error in a way that does not suggest an offset spread between parts. Is there such a thing?
Could any of the mechanical or electrical imperfections of my setup as described above have this effect, or are the other concerns that could affect the results? I do have additional parts which I will mount to compare more results.
