Quick guestimate:
Thermal difference between heatsink and chip;
Assuming 1W/mºK
Assuming 7x7mm surface
Assuming 2.5W/chip
Assuming 0.2mm epoxy compound between chip and heatsink
Delta T= (2.5/(0.7*0.7) * 0.02 / 0.01 = ~10 degrees Kelvin or Celcius
Roughly losing 10 degrees from chip to heatsink (in PCB here) is not nice... (plz check my calculations, it's late and I'm sleepy
)
Same formula for solder with 30 Watt/mºK would result in;
Delta T= (2.5/(0.7*0.7) * 0.02 / 0.30 = ~0.34 degrees Kelvin or Celcius
Isn't the m in W/mK in meters? 7mm would be 0.007m but also the thermal pad on the ASIC is only 5.4mm (and for closer accuracy the power dissipated by the ASIC should be 1.86W only). And a thickness of 0.2mm would be 0.0002m. I guess they cancel out anyway.Thermal difference between heatsink and chip;
Assuming 1W/mºK
Assuming 7x7mm surface
Assuming 2.5W/chip
Assuming 0.2mm epoxy compound between chip and heatsink
Delta T= (2.5/(0.7*0.7) * 0.02 / 0.01 = ~10 degrees Kelvin or Celcius
Roughly losing 10 degrees from chip to heatsink (in PCB here) is not nice... (plz check my calculations, it's late and I'm sleepy
)Same formula for solder with 30 Watt/mºK would result in;
Delta T= (2.5/(0.7*0.7) * 0.02 / 0.30 = ~0.34 degrees Kelvin or Celcius
Checking with Fouriers Law,
dT = q*s / (K*A) where q is Watts, s is thickness, K constant, A area.
Using 1.75 for 4 hour silver epoxy,
1.86 * 0.0002 / ( 1.75 * 0.0054*0.0054) = 7.3 C
vs. solder,
1.86 * 0.0002 / (50 * 0.0054*0.0054) = 0.25 C
So pretty close to what you said. To be viable you would have to be able to use a much thinner layer of epoxy compared to solder. If a 3mil plastic stencil is used that would be 0.07mm thick, which gives a temp difference of 2.5 C.
Using a taller heat sink or more air flow may be able to offset that. It's an interesting idea for those who don't want to get involved in reflow soldering ASICs in multiple passes but would require some testing to see how well it works. I don't plan to do that any time soon.