Thermal Test Board Project Suggestions

[ransonjd]

I'm thinking of doing a project to test how various PCB factors affect surface mount component thermal resistance to ambient. I'm especially interested in how much nearby disconnected copper pours affect thermal performance.

My plan is to design an array of PCBs with different copper arrangements, but identical component arrangements. I'll use a surface mount diode with a thermal pad as the heat source. The diode will have kelvin connections to measure the voltage across it, and a current sensor inline with its power leads. On the PCB, I'll put down temperature sensors at various locations. I'll probably also glue a thermocouple to the top of the diode. The whole thing will be in a large box to prevent drafts, and I'll measure box's ambient temperature with another thermocouple.

My general goal is to keep the cost down without compromising the results.

I'm interested in suggestions in a few areas:

1. What't the best cheap way to monitor PCB temperature at various locations on the PCB? Diodes are cheap, but would require individual calibration. RTDs are expensive. Preferably I would just use a cheap, off-the-shelf temperature logger, or an IC that can be interfaced to a PC with minimal overhead.

2. What variables should I test? I'm considering copper area, connected vs disconnected copper, via size, via number and 1 vs 2 vs 4 plane layers.

3. Anything I should change in my experimental setup?

[ogden]

Consider NTC for temperature sensing. Any low cost temperature sensor will need calibration, so you shall plan two or three point calibration using external temp reference, for every PCB.

Thermal imager would solve temperature sensing.

[ransonjd]

  Right, NTCs, the obvious solution. There apparently was a hole in my head today.

Thermal imager would be nice. Dealing with emissivity errors might make it more work in the end.

[EmmanuelFaure]

1. What't the best cheap way to monitor PCB temperature at various locations on the PCB? Diodes are cheap, but would require individual calibration. RTDs are expensive. Preferably I would just use a cheap, off-the-shelf temperature logger, or an IC that can be interfaced to a PC with minimal overhead.

Definitely, some fine wire thermocouple(s). One of the wire of T types thermocouples is copper and is easily solderable with standard methods (Soldering iron + soldering wire).

As a guideline for your experiments : A standard 35µm copper layer is about 10 times more conductive than a standard 1.6mm FR4 epoxy layer.