I want to know if my thermal design is appropriate for dissipating 72W of heat while keeping the junction temperature below 150 °C (I am trying to cool an SFH154 that comes in a TO-3P package). My thermal circuit consists of a junction-to-case resistance of 0.61 °C/W, a silicon based thermal paste (grease) with a thermal impedance of 0.258 °C-in²/W (TO-3P package area is 15.5x20 mm²) so that gives us Rth= 0.537 °C/W) , and an old fan equipped CPU heatsink with no mounting holes and an unknown thermal resistance.

Text Text Text

My questions are:

  • Is drilling a hole in the middle of this heatsink (in order to properly mount the MOSFET) a sensible design choice?
  • Will this heat sink be suitable for this project?

Note: I've done some calculations and the maximum heatsink thermal resistance should not exceed 0.589 °C/W for a junction temperature of 150 °C. Also some data might not show up in the datasheet. if that's the case, just download the document.

  • 1
    \$\begingroup\$ If it were my experiment, I'd just clamp it; the more pressure, the better. But that datasheet is notably lacking in chart data - kind-of hard to gauge the device Safe Operating Area without it. \$\endgroup\$
    – rdtsc
    Commented May 3, 2022 at 12:02
  • \$\begingroup\$ @rdtsc You might have display issues, i'd suggest downloading the pdf since the datasheet does min fact have SOA graphs and all the necessary info for gauging if your design is safe. Also, how exactly can i clamp it? I cant see how i can do it. \$\endgroup\$
    – A.H.Z
    Commented May 3, 2022 at 12:09
  • \$\begingroup\$ @A.H.Z Did you find anything useful in any reviews of USB electronic loads? If the MOSFET is on a PCB, then you can use some nuts on those screws that are on the heatsink to attach the heatsink to the PCB with the MOSFET in between. It might be a good idea to use something like large diameter washers under the nuts to spread the load on the PCB. (I haven't actually tried it myself, otherwise I could confidently write that as an answer.) \$\endgroup\$ Commented May 9, 2022 at 17:20
  • \$\begingroup\$ @A.H.Z I found this review on YouTube which uses what appears to be the same heatsink as the one in this question: you can briefly see at 2:38 into the video that it looks like they have used the bracket from the back of a PC motherboard as the mount for the heatsink, which would be much better than a few washers. \$\endgroup\$ Commented May 9, 2022 at 17:37
  • \$\begingroup\$ @AndrewMorton I actually ended up using a metallic sheet with two holes for screws. I forcefully pushed those screws down the space between the laminations in the heatsink. As for the USB electronic loads I unfortunately did not find anything worth noting. I will check out the youtube video though. \$\endgroup\$
    – A.H.Z
    Commented May 12, 2022 at 8:08

2 Answers 2


Questions are:

Q1: Is drilling a hole in the middle of this heatsink (in order to properly mount the MOSFET) a sensible design choice?

Q2: Will this heat sink be suitable for this project?

Q1 & About drilling a hole in the heatsink
It is a feasible solution and I have done exactly that when I built an “Electronic Load” using a TO-220 transistor mounted on a CPU heatsink+fan.
However, there are some points of attention:

  • The TO-based transistor needs to be connected by wiring soldered on transistor’s terminals. Connectors or direct mounting on PCB are not advised IMO.
  • To maximize heat transfer, the hole for the bolt should not be positioned in the center, but with an offset, to leave the heat-generating part of the transistor in the center of the heat-sink plate. The exit path of transistor-wiring is influenced by the orientation/rotation.
  • Hole properly sized would accept an internal thread - then you don’t need a nut on the back (which would be practically infeasible).
  • After hole & thread are done, lightly countersink the hole - 0.5mm is enough.

So, yes for a DIY project and focused on the thermal dissipation it is not an heresy to do that. You can drill a hole, but not in the middle of the heatsink, studying beforehand the best terminal exit path. Try not scratching the already very-flat & polished surface.

Q1 & Preliminary TO-based transistor fixing to heatsink:
You can simply press/clamp/compress the TO-220 ou TO-3P or TO-247 against the heatsink.
Try using a metal (aluminum, steel) bar (retangular or L-shaped) and search your bolting box for self-threaded bolts (as those used for MDF/wood working) and (temporarily) tighten/clamp using two of those diagonal pass-through holes.
If testing seems good, then improve this press-mounting method to make it permanently adequate or go to the drill&thread&countersink method.

Q2 & Heatsink suitability
Pre-sizing the heatsink: Check to which CPU that Heatsink was designed for. Then search for its TPD (total power dissipation). For instance, an Intel Core i7-7700K, has TPD ~= 90W and the CPU continues to operate accordingly with the correct CPU-heatsink model - the temperature does not heat up so much (T>100°C) and maintains the CPU working right.

When you don’t have Specs: In this case, I strongly recommend you doing a simple test. Temporarily mount any TO-based BJT (as TIP122) in the heatsink (see above) - use thermal paste. Make the power transistor dissipate 10W (Vce = 10V x 1A, or equivalent), while running the cooling fan at known voltage/speed. Measure the temperature rise (before x after 10 min) in the heatsink surface close to the transistor.
Calculate the equivalent Thermal resistance:°C/W (Celsius degree/Watt).
With this method, you will definitively know
if this unknown Heatsink+fan will work for you.

Some issues:

dissipating 72W of heat while keeping the junction temperature below 150 °C

Issue #1 - Conductivity in Transistor: The semiconductor Max temperature is 150C but from your provided link, the SFH154 has a higher Junction-Sink thermal resistance of 0.85 °C/W (not just the 0.61C/W). For 72W, temperature increase will be +61C above ambient temperature, where I would anticipate a small working margin for the heatsink+paste.

enter image description here

Issue #2 - Paste Conductivity calculation: - It seems to be a conceptual misunderstanding in your text:

…thermal impedance of 0.258 °C-in²/W (TO-3P package area is 15.5x20 mm²) so that gives us Rth= 0.537 °C/W)…

The problem here is that you are not considering the thickness of the thermal paste layer and used equations incorrectly.
The thermal resistance of the paste is proportional to its thickness, where a common dimensional unit would be Watt / (meter x Kelvin).

From Wikipedia’s Thermal Paste, you reach this nice paper from Narumanchi et alli at NREL showing the following data for some heatsink paste/compounds, where I exemplified the comparison between experimental data at given test conditions (t = 150 μm) and the manufacturer’s nominal data:

enter image description here

Heat Flux Equation for Conduction:
Q = (kc / t).A.dT where:
Q = Heat Flux (W).
kc = Thermal Conductivity (W/m.K).
t = Thickness - paste (m).
A = Area under Heat Flux (m^2).
dT = Temperature differential (K).

Reordering for dT:
dT = (Q.t)/(kc.A)

Now, conservatively assuming the cheapest thermal paste (boxed in yellow), with a low-force pressing resulting in a (thick) layer with these following values:
Q = 70W.
t = 0.1mm = 1x10^-4 (m).
kc = 0.4 W/(m.K) (wrongly marked as “W/mK”).
A = 15mm x 20mm = 3x10^-4 (m^2).
dT = (70 x 1x10^-4) / (0.4 x 3x10^-4) = 58°C

Conclusions about thermal paste:

  • Poor paste: In yellow - summing 61°C (transistor) and 58°C (paste), at ambient = 25°C a this leaves just 6°C for the heatsink itself - for 70W, this gives a performance figure of 0.09 °C/W. Not feasible with poor paste.
  • Good paste: In blue - it is ten times better and has kc = 4 W/mK, dT = 6°C. Thermal margin is (150 - (61+6+25)) = 58°C. Then Heatsink = 0.83 °C/W. Feasible heatsink+fan +good_paste, but test if yours is good enough.

Some tips:

  • Unless an actual pre-test shows differently, do not use a cheap&poor thermal paste here (+70W). There is no need to be the fanciest paste, but for such power levels, avoid the cheapest paste with unknown thermal properties.
  • Use a thin layer: t ~= 0.05mm and “mind the gap”. Use just enough paste and most important: close/reduce the gap by tightening the bolt/cramping method accordingly. Avoid scratches, as even a small one can raise marks and increase the filling gap.
  • Heatsink in “Turbo mode” - Heatsink+fan dissipation capacity can be “overclocked” (as the CPU) if you apply a controlled overvoltage to the fan.
    Most fans use brushless motors operated at 12V Nominal and accept being 15V overdriven (max) without major problems - apart from increased acoustic noise level.
    I don’t recommend going higher than +25% (V<15V). Even better if this “turbo mode” is used only if Temperature becomes “Too hot”. So, if that 72W max is of intermittent/sporadic nature, this “turbo mode” may be enough even with less expensive or repurposed parts.

Closing notes:
Even for a DIY project, try to do some reduced-power tests - a “pre-test” with your heatsink+fan and a similar-sized transistor, but using a lower power setting - as 10W.
This will avoid bigger frustrations and problems when your project is completed.

  • \$\begingroup\$ First of all ,thank you for this very thorough answer. The CPU fan was designed to cool an E5200 Pentium Dual core that can produces 65W of heat flux. after testing the heatsink at three different power dissipations( 21, 11 and 6 watts) in an ambient temperature of 28C° (letting the heatsinks heat for about 4 minutes) with the fan operating normally (not overclocked) I got an average result of 0.291 C°/W for the heatsink. ( a thermo-couple with an accuracy of +/- 2.5% was placed inside the fins of the heatsink for the testing). \$\endgroup\$
    – A.H.Z
    Commented May 31, 2022 at 12:58
  • \$\begingroup\$ As for the mounting, I resigned myself to clamping the Mosfet using a drilled metallic sheet and a couple of threaded bolts going through its fins. If possible , I would like to know if there is a way to estimate gap thickness, since i do not have a vernier caliper at my disposal. Also, my HT-WT160 silicon thermal grease is advertised to have a thermal conductivity >0.6 W/m.k can I really trust this value? Note: This is just to confirm if im understanding correctly, dT is equal to the temperature between the case of the transistor and the heat sink right? \$\endgroup\$
    – A.H.Z
    Commented May 31, 2022 at 13:09
  • 1
    \$\begingroup\$ About Heatsinking capacity: considering your application allows a higher Temperature than a CPU, and HtSk nominal rating is for 65W, then it seems about right and your dissipation capacity seems to match. However, I would not be confident to use Temp.Sensor at fin side = it reads lower temperature due to intense gradient (conduction and convection). To play safer, I would repeat some tests (that 21W at least) with TC pressed close at the “corner” HtSk+TO220: use a foam rubber (or styrofoam) piece pressed: it accommodates and provides insulation. \$\endgroup\$
    – EJE
    Commented Jun 4, 2022 at 14:59
  • 1
    \$\begingroup\$ Your Paste thermal conductivity: the advertised value is lower(worse) than most brand advertised specs = so it seems right. To check that “at home” you would need to design a test apparatus using 1 Heat source (in Watts), 1 Measuring plate (with T sensor) and 1 Sink plate (with T sensor), where grease is applied between plates and pressed together. A pair of metal shims (as razor blades) assure controlled spacing. About “calculated” dT: I used dT to calculate the temperature difference for two cases of ThPaste under 70W, with A and t. So, it is dT between Transistor and HtSk surfaces. \$\endgroup\$
    – EJE
    Commented Jun 4, 2022 at 15:14
  • 1
    \$\begingroup\$ @A.H.Z see above replies and as an afterthought, your HtSk readings were probably masked by measuring errors caused by sensor pickup placement. So, HtSk may be significantly worse (higher) than 0.29C/W, but still might be Ok for Transistor at T.j < 150°C. The good thing is that Paste layer can be thinner than 0.1mm: better cases at 0.02mm, not only because pressure (Force/area) is high, but mostly because surfaces have good flatness match on smaller surface dimensions. \$\endgroup\$
    – EJE
    Commented Jun 4, 2022 at 15:27

First of all, there are heatsinks designed for TO-220. Second, your main concern is mechanical contact. A CPU heatsink generally has much better thermal resistance than a TO-220 heatsink due to the larger surface area but You will have to clamp it to the part somehow. Your other concern would be not to create any shorts and mechanical stress on your board. I would attach a thermocouple to the base of the heatsink, and see how high it goes, both with fan on and fan off. Drilling will not help you much, unless you can achieve tight tolerance and better contact with the part, that would require a CNC machine and filling the hole with thermal paste/glue.

  • \$\begingroup\$ Shorts and mechanical stresses aren't really an issue; I am only going to mount one MOSFET (as an active load) unto the heatsink. It has come to my attention that drilling a single hole to mount the transistor using a vias might not be a good option, in that case, is there any better way to clamp it? \$\endgroup\$
    – A.H.Z
    Commented May 3, 2022 at 15:21
  • \$\begingroup\$ @A.H.Z Could just try it and see, of course. Don't forget the thermal paste to fill air gaps. I'm not sure how you could get a better contact, other than perhaps using thermal glue, if such a thing exists, and clamping it tightly with a clamp while the glue sets. \$\endgroup\$ Commented May 3, 2022 at 16:19
  • \$\begingroup\$ @user253751 please excuse my english, but I don't really understand what you mean by a "clamp", because i cannot see how a clamp could be fixed on that heat sink, could you please explain what you're referring to? \$\endgroup\$
    – A.H.Z
    Commented May 3, 2022 at 17:22
  • 1
    \$\begingroup\$ @A.H.Z. something like this: aliexpress.com/item/33034978192.html that would hold the two parts together while the glue sets. It is a tool that is often used when gluing things together. It might not be necessary - you could also put something heavy on top. The purpose would be to squeeze the heatsink and the transistor together tightly, to minimize the amount of space filled by glue. \$\endgroup\$ Commented May 3, 2022 at 17:25
  • 1
    \$\begingroup\$ @A.H.Z again, I am not sure if thermal glue is a thing, or if it's better or worse than thermal paste. You are the one with the parts and you're the one doing the experiment \$\endgroup\$ Commented May 4, 2022 at 8:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.