Yes, case-to-heatsink thermal resistance is a bitch, with 4 transistors they are kind of in parallel, so you expect to see the same thermal difference in that juncture if using the same mounting technique. Anodized heatsinks from a respectable vendor might have enough anadization to be considered enough insulation, if you need them to be isolated.
Other improvement that could be made is to use a better thermal paste, any decent metal based PC grade is an improvement over regular silicon based ones.
As mentioned cooper plates might be helpful, as it's softer than aluminum might get better contact for the same pressure and then you have a greater area from the cooper to the aluminum and the cooper softness helping here again.
JS
Enviado desde mi LG-M250 mediante Tapatalk
As for CC Load #1 above, I have decided to mount the MOSFETs directly to the heat-sink, ie no mica washers. This means that the heat-sink will be 'live' but we are only dealing with low voltages <30v. I will be using thermal paste from Jaycar (Australia) intended for CPUs.
I have finished drilling and tapping the heat-sink and will post pix soon.
enut11
According to SOA on the datasheet IRFP4668 can handle:
500W @ 5V (not a realistic value, limited by Thermal Resistance)
30W @10V
10W @20V
So I don't think that is suitable for your application.
Yes, case-to-heatsink thermal resistance is a bitch, with 4 transistors they are kind of in parallel, so you expect to see the same thermal difference in that juncture if using the same mounting technique. Anodized heatsinks from a respectable vendor might have enough anadization to be considered enough insulation, if you need them to be isolated.
Other improvement that could be made is to use a better thermal paste, any decent metal based PC grade is an improvement over regular silicon based ones.
As mentioned cooper plates might be helpful, as it's softer than aluminum might get better contact for the same pressure and then you have a greater area from the cooper to the aluminum and the cooper softness helping here again.
JS
Enviado desde mi LG-M250 mediante Tapatalk
As for CC Load #1 above, I have decided to mount the MOSFETs directly to the heat-sink, ie no mica washers. This means that the heat-sink will be 'live' but we are only dealing with low voltages <30v. I will be using thermal paste from Jaycar (Australia) intended for CPUs.
I have finished drilling and tapping the heat-sink and will post pix soon.
enut11You can do that, even with riskier stuff as long that's hidden well within the case and marked.
If you intend to leave it outside, I wouldn't recommend it, as it might very well end shorting with simmering in your desk.
Regards!
Enviado desde mi LG-M250 mediante Tapatalk
According to SOA on the datasheet IRFP4668 can handle:
500W @ 5V (not a realistic value, limited by Thermal Resistance)
30W @10V
10W @20V
So I don't think that is suitable for your application.
Hmm, interesting data. I ran one again tonight at ~50 watts (4.4A at 12v) for about 10 min. This time I used in IR thermometer and measure 60C on top of the case. Junction temperature will be higher but no way to measure it. Heat sink was about 10C cooler. Ambient here is 18C.
Four MOSFETs pumping out 150W into the new heatsink should be interesting!
enut11
No room inside. The heat-sink and fans will be outside the box with a big warning sticker
@JS, to reduce thermal resistance I did not use any insulation between MOSFETs and copper or between copper and heat-sink. For the time being I can live with a live heatsink.
I will have to think about your thermal analysis. I take it that you are satisfied the 150W label is warranted?
I am now happy with the performance of this CC Load. When I purchased the original kit I thought that it was going to be an easy project! In the end I am glad it was a challenge. I got to learn about the behaviour of MOSFETs, a device designed for rapid switching, and how to force them to behave in a linear fashion. Also, the headline numbers on the spec sheet, Rds, Ids max and Vds max are relatively useless in linear mode. For DC operation I found you have to rely almost entirely on the SOA diagram and even downrate that! An interesting journey.
enut11