Hi Jim Promised to post on weld conditions. Wish hadn't - this is about being driven mad by not being able to check conditions which worked on a particular day, given covid19 pandemic and laid-off from work. ......
======================================================
I've done a lot of temperature, heat flow and contact resistance measurement to confirm engineers' designs, and largely found that the most significant source of variation is the exact conditions of metal to metal contact. You can predict the heat flow but you may have to make measurements and redo the connections repeatedly to achieve it. Blacksmithing gives a good visual example, a small crack in the glowing iron bar reveals itself by a large step difference in color. Presumably you are long past seeing that in your welds.
For serious temperature measurement Type K thermocouples are a nice balance of reasonable cost, decent accuracy, wide temperature range and ease of use. They should be welded with acetylene rather than just twisted together at the working end. KX extension wire is a reputedly cheaper version that only works near room temperature.
IR meters may be handy but my experience with them shows they can be thrown off by emissivity variations, so I use them to find hot spots but check them against thermocouples I've calibrated in slush and boiling water.
The TM902C is the cheapest thermocouple meter I've found.
formatting link
I buy my Type K wire and connectors from Omega for their quality and accuracy, which you may not need.
formatting link
24 gauge tolerates brief yellow heat exposure and isn't as annoyingly springy as the larger sizes. 30 gauge is rather too delicate outside the lab, but better to slip under a refrigerator's door gasket. The braided fiberglass and cheaper ceramic fiber insulation can be upgraded to tolerate 1000C by rubbing wood stove gasket cement (sodium silicate) into it.
Thermocouples need to be in a small drilled hole to be accurate. Tightly clamping them to the surface under a wad of insulation may come fairly close, perhaps 10C low. Ideally the metal surface being measured shouldn't short the wires away from the welded junction but the error isn't great if it does, coming from heat lost down the wires. The extra metal to metal voltages theoretically cancel out, which we had to prove to ourselves from first principles in a homework problem.
I use these to log temperature readings to a PC when I have more to do than watch and write down the display.
formatting link
The datalogging program compensates for thermocouple nonlinearity that the meter display ignores. They are optically isolated and can float at power line or welding lead voltage, so you could use them with a shunt or current probe to record your actual welding voltage and current. The separate datalog files can be combined and aligned by timestamp in a spreadsheet.
With constant heat input, temperature will rise exponentially to a final limiting value that it may take half an hour to approach if the heatsink is large. You can raise the heatsink temperature close to the rectifier's limit with a blowtorch and then see if it decreases under normal load to quickly determine if your heatsinking is adequate.
formatting link
In the lab we measured the actual junction temperatures of diodes and transistors by their forward voltage drops at low current. That's the junction temperature you see on datasheets. In normal operation it could easily be 10-20C higher than the heatsink. Exceeding the temperature rating decreases the life of the part. I've heated a power transistor above 175C where it failed in minutes. The surprising test was dripping liquid nitrogen on an unpackaged integrated circuit, under a microscope. The outer silicon nitride passivation layer shattered like a windshield but the part kept working. Those maximum ratings come from techs like me who let the electron smoke escape under controlled conditions.
formatting link
Good luck