The problem is the heatsink on the top covers any thermal image of the die. I need to be accurate to within 1deg. The heatsink on the top defeats this idea.
You can get that level of accuracy if you deposit a 100-ish ohm platinum resistor on the die, and then put the effort into calibrating your custom resistor
Alternatively, grind away the epoxy around the die and put the thermocouple right next to it on the side. Cut away the thermal film so the thermocouple bead does not touch the thermal film so it is not cooled by the heatsink.
You cannot put the thermocouple between the die and the heatsink because it will drastically change the thermal conduction, and you are still going to read the intermediate temp between the two. So if your chip is 60C the heatsink 55, the thermocouple will read 57.
A copper spacer with a thermocouple inside it would require two layers of your 0.2mm thermal film, which means you are reading the temp half way between the die and the heatsink.
I will have to confirm the details but a friend of mine had to send a dummy chip with a thermocouple epoxied in it, to an independent cal lab to prove that a 10 C temperature discrepancy (at -40C) was not imaginary.
Since i dont know the wattage i cant give you an order of magnitude estimate as to how difficult it becomes to get 1C actual accuracy.
You can do the math yourself, its actually really simple. But proving your 1.8mwk thermal film was actually 1.3... not so easy