# Heatsink design, factors to consider?

I'm trying to design an small but effective heatsink (DIY type), the design that I came up with so far is a 4x4x9cm aluminum square tube with a 12V fan under it blowing up:

poor designing skill...

The size of the heatsink is choosen based on the box space limit and I only have access to one side of the heatsink to install the regulators. I want to install 4 inefficient linear regulators on it, which in worst case will waste 20 watts of power in total.

• Does vertical heatsinks have any advantages over horizontal heatsinks?
• How can I calculate the cooling capability of this heatsink?
• That regulator which is going to dissipate more heat, should be at the bottom side of heatsink or top?
• Where should I install the temperature sensor to monitor the temperature and change the fan speed?
• Usually the transistors would be soldered on PCB and they would be 90 deg rotated. You do calculate with simple math, after you know the heatsink thermal resistance vs air flow, you have thermal resistance of the transistors, pads, and finally the ambient temperature, max silicon junction temperature, max power dissipation. Commented Jan 6, 2020 at 22:26
• Short fins are good for aluminum. However the back side will likely not even get warm, as all of the heat will be dissipated before it travels that far. For 20W, something a third that size (still using fan) would likely work just fine. (A 1990's CPU cooler is good for 20-40W.) Commented Jan 6, 2020 at 22:34
• @rdtsc I was actually thinking about having a flat surface instead of a tube but I wanted to have all the air directed at removing the heat. still I'm not sure about the shape of the heatsink but if as you say it can cool down the regulators even when the size reduced to 1/3, it seems that I'm in the right direction. thanks. Commented Jan 6, 2020 at 22:43
• Put one transistor on each side? Commented Jan 7, 2020 at 2:32

Does vertical heatsinks have any advantages over horizontal heatsinks?

For this small scale and low temperature delta, and with forced airflow, there will be no appreciable cooling advantage. Convection will be dwarfed by the airflow from even the weakest of fans.

How can I calculate the cooling capability of this heatsink?

"There's an app for that." Seriously, there are many good calculators online. Find one that you like. Many of them link to articles explaining how the calculations are done, in case you want to go down the rabbit hole. Here's a search link.

That regulator which is going to dissipate more heat, should be at the bottom side of heatsink or top?

I would place the hottest component near the middle where there's more aluminium to spread the heat in all directions. And perhaps place the less hot components some distance away from it. It will of course have to be a compromise.

Where should I install the temperature sensor to monitor the temperature and change the fan speed?

I would keep the sensor out of the airflow, but close to the hottest component, with good thermal contact to the heatsink.

Air velocity >> 1m/s from forced air fans can reduce Thermal resistance and reduce temp rise from 60'C to 10'C compared to stagnant air.** Your forced air vertical solution is adequate yet sub-optimal but may be slightly noisy.

Noise pollution is another factor with eddy current noise on the blades. This can be greatly attenuated with a laminar plenum between fan and heatsink then laminar plenum on exit for reduced loading .

The most effective heatsink design for removing 30W was a DELL tower with radial CPU heatsink and plenum hose to an inline fan to exhaust plenum removing all heat outside the box with no traditional eddy current blade noise buzzing sounds. It worked as a well tuned muffler and impedance matched (turbulent to laminar to turbulent fan to laminar exit. Optimizing the performance.

But I agree with @rdtsc . Any '90's CPU heatsink and fan will work, if it suits your application.

From my design experience ** on 1U (1.75"h) 19" racks with 180W PSU's , I used an intake air Mylar film spoiler for stirring up air to increase air velocity over open-frame PSU hotspot then inline dual 1.5" fans to push air out in a plenum without eddy current grid noise.

You have lots of room and even though your fin design is not optimal for max surface area/wind resistance, it should result in adequate case temp rise <20'C. It may be a bit noisy.