Keystone Mica TO3 insulators are all spec'd at 3 thou thickness:
https://www.mouser.com/datasheet/2/215/667-742167.pdf
https://www.mouser.com/datasheet/2/215/667-742167.pdf
I *think* it's linear: 3mil will have 3 x the thermal resistance of 1mil.
You could split the 3mil insulators to get a pair of thinner ones. It can be pretty fiddly, though.
You could split the 3mil insulators to get a pair of thinner ones. It can be pretty fiddly, though.
Thermal pads should have their thermal conductance in the specifications, usually in C/W
You can calculate conductances for any material from its thermal conductivity:
conductance = conductivity * area / thickness
( works like resistance and resistivity, basically )
With inflexible materials a thermal grease is needed to complete the "thermal circuit" - white zinc-oxide grease is suitable, or the grey aluminium grease.
Looking up a few materials:
mica 0.71 W/K/m
nylon 0.25 W/K/m
alumina 30 W/K/m
beryllia 330 W/K/m
zinc oxide 23 W/K/m
silver 420 W/K/m
You'll see that mica is only a little better than plastics - in fact you might as well use kapton tape as its available in thinner films than mica and withstands high temperatures very well.
The common silpads and similar are rubbery so they fill any air-gaps and conform to the device and heatsink well - air is not a good conductor of heat!
Despite containing metal particles (grey is the clue), the thermal conductivity is not brilliant, so use the thinnest types and make sure the heatsink and transistor have no rough spots.
The ceramic materials are brittle and beryllia dust is very toxic, so not well suited for thermal pads (and they have to be thick for strength defeating most/all of their benefit). Some RF power transistors use beryllia packages for good thermal performance and low RF losses.
The best approach I think is a very fine electrically insulating film on a graphite pad (not cheap). Or a diamond pad + diamond grease (very not cheap).
Spring-mounting is better than screw-mounting, BTW, as it maintains pressure over time and presses over the die, not to the side.
You can calculate conductances for any material from its thermal conductivity:
conductance = conductivity * area / thickness
( works like resistance and resistivity, basically )
With inflexible materials a thermal grease is needed to complete the "thermal circuit" - white zinc-oxide grease is suitable, or the grey aluminium grease.
Looking up a few materials:
mica 0.71 W/K/m
nylon 0.25 W/K/m
alumina 30 W/K/m
beryllia 330 W/K/m
zinc oxide 23 W/K/m
silver 420 W/K/m
You'll see that mica is only a little better than plastics - in fact you might as well use kapton tape as its available in thinner films than mica and withstands high temperatures very well.
The common silpads and similar are rubbery so they fill any air-gaps and conform to the device and heatsink well - air is not a good conductor of heat!
Despite containing metal particles (grey is the clue), the thermal conductivity is not brilliant, so use the thinnest types and make sure the heatsink and transistor have no rough spots.
The ceramic materials are brittle and beryllia dust is very toxic, so not well suited for thermal pads (and they have to be thick for strength defeating most/all of their benefit). Some RF power transistors use beryllia packages for good thermal performance and low RF losses.
The best approach I think is a very fine electrically insulating film on a graphite pad (not cheap). Or a diamond pad + diamond grease (very not cheap).
Spring-mounting is better than screw-mounting, BTW, as it maintains pressure over time and presses over the die, not to the side.
Thanks, very informative.
Going to pick up some 1 mil Kapton sheet and use that or nothing but
thermal paste and run the heatsinks hot.
Going to pick up some 1 mil Kapton sheet and use that or nothing but
thermal paste and run the heatsinks hot.
A brilliant idea for light duty applications but what sort of paper and "goop" are you using? Standard photocopier paper, as an example, is too thick for efficient heat transfer. If the paper has been laying around for a while, it will have swollen considerably and you may need to re-torque the clamp bolt(s) after a few thermal cycles to restore contact. That also makes it a candidate for spring clamping.
Perhaps Nomex paper also absorbs thermal grease? Its designed to last at high temperatures without degrading like ordinary paper does (through oxidation). Its used for insulation in electric motors and generators. Available down to 0.05mm