I am currently design a high power system transistors in TO220 package.

I am planning to mount those transistors on a metal box that will act as a dissipator/heatsink for the transistors. The walls of the metal being thin, I can't just thread the holes, so I am planning on having a screw and a nut to mount it.

Most of the design I have seen online are having the head of the screw on the front of the transistor, and the nut at the back of it. (https://sound-au.com/heatsinks.htm#s11.1) However, I could not find any reason why it is recommended to place the nut at the back, and not in front. Does it really matter ?

  • 1
    \$\begingroup\$ I sure hope not. Because when I tap the hole in a heatsink I have no choice in the direction of the screw. \$\endgroup\$
    – DKNguyen
    Jul 20 at 22:26
  • \$\begingroup\$ Is there a typo in your question? Did you mean to say "The walls of the metal being thin, I can't just thread the holes"? If that is so then the answer is "no, it doesn't matter". My only concern would be that the thin metal wouldn.t dissipate the heat fast enough. As a general principal, all things being equal, the thicker the metal the better the heat dissipation. \$\endgroup\$ Jul 20 at 22:32
  • \$\begingroup\$ I fixed the typo. Thank you very much for the feedback, I'll check that the dissipation of the walls is good enough \$\endgroup\$
    – Johan B
    Jul 20 at 22:38
  • 1
    \$\begingroup\$ @JohanB I had a look at the article you quote and the possible problem of using the case / chassis as a heat sink is discussed in section 2 (sound-au.com/heatsinks.htm#s2) under "Flat Sheets" \$\endgroup\$ Jul 21 at 1:06

2 Answers 2


There should not be a problem. Having the screw head on top of the mounting tab is of course required if the package is going to be mounted on a heatsink with a tapped hole where the far side is between fins or blind. It also precludes any sort of clearance issues with a hex nut hitting the side of the package molding. A flat washer is recommended under the screw head and therefore should be present between the nut and mounting tab as well. The optional lock washer should be installed between the nut and flat washer. Additional information with recommended torque values can be found here (too much torque can be as bad as not enough): https://www.st.com/resource/en/technical_note/dm00238320-mounting-instructions-for-thd-through-hole-device-packages-stmicroelectronics.pdf

Keep in mind that the tabs are usually not electrically isolated.

Edit: I would also confirm that the walls of the enclosure are not warping and reducing the contact area.


Not electrically, no. I would insist on screw heads facing outward, so you don't have the jagged edges of the nut and spare thread sticking out.

In a... more professionally? made enclosure (or if you want to get the hardware to do it yourself!), the threads can also be press-fit inserts, or spot-welded studs, even in fairly thin material (0.8mm, maybe thinner?).

Obviously, mounting force isn't going to be great (the material tends to bow away from the component body, or vice versa -- like in Fig.1 https://www.nxp.com/files-static/rf_if/doc/app_note/AN1040.pdf but both sides may flex away!), and with thin material, the heat spreading power isn't very much. That is, you may end up with a hot spot on the enclosure, rather than dissipating heat evenly over its surface.

Offhand, I would guess, for 0.8mm aluminum, doing 10 W in a single device, mounted in this way, is about the limit I would consider.

Note that you can bolt on heatsinks to the outside -- which gives a stiff mounting surface again, so you don't have to worry about material flex (besides the device tab and such, of course). Grease all faces, of course (heatsink to enclosure, enclosure to device). But be careful that it's not awkward for the user (sharp corners, add perf/mesh covers?), nor blocked by them (e.g. a stereo system, with big heatsinks on its rear, tucked in the back of a cubbyhole, gets about nil circulation).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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