15

I have a doubt about heat-sinking that I could not solve by surfing the Web. My doubt arised when mounting a TO220 package on its heat-sink using these dirty cheap pads, but actually it has quite a general scope.

There are a lot of articles out there about comparisons between thermally conductive pads vs. thermal grease (and most say that grease is better with regard to thermal conductivity), but I found almost nothing about whether a thermal interface pad is needed at all when you are not concerned about electrically insulating the tab from the heat-sink.

Wikipedia says:

Thermal pads and thermal compound are used to fill air gaps caused by imperfectly flat or smooth surfaces which should be in thermal contact; they would not be needed between perfectly flat and smooth surfaces. Thermal pads are relatively firm at room temperature, but become soft and well able to fill gaps at higher temperatures.

So it seems to imply that a thermal pad is always a good thing to place between the TO220 tab and the heat-sink in order to improve thermal coupling. But is it really so? References are a bit scarce and tend to focus on CPU/GPU cooling setups.

Moreover I remember having seen some equipment where TO220s were attached to their heat-sinks without either thermal grease nor thermal pads. I can well understand why one would avoid thermal grease (more complicated and costly building procedure), but thermal pads are dirty cheap and don't add much effort when you already have to screw/bolt the metal pad to the heat sink.

Bottom line: if I don't care about electrical insulation between TO220 and heat-sink and I don't want to use thermal grease, is putting a thermal pad between the two always useful, from a thermal coupling point of view?

  • 1
    The thermal pad in your situation will probably make things worse not better.You could think about a copper washer to take up the surface roughness .Copper will have more thermal conductivity than any thermal pad . – Autistic Feb 04 '17 at 10:38
  • @Autistic thanks for the tip, but I'm not interested in substituting the thermal pad with something better, but to know if it is worth putting it there in the first place. I.e. will I be better off *without* the pad? – LorenzoDonati4Ukraine-OnStrike Feb 04 '17 at 10:41
  • 1
    It probably depends on how much power you are dissipating with the component to whether you need to consider these things. – crowie Feb 04 '17 at 11:46

2 Answers2

21

The thermal conductivity of a dry vs. thermal grease is shown here for TO-264 package (from ON Semi application note AN-4166)

![enter image description here

Looks like roughly 0.7 to 1.1 °C/W depending on the torque for dry mount.

Your particular thermal pads have no specifications for performance, so you really can't be sure what they are going to do. Maybe they are counterfeit made of some inappropriate material. The thickness isn't even specified! Buying such a product is a total crap-shoot, and not recommended. You can buy thin electrically conductive thermal pads with real datasheets for $18/1000 through distribution- they have excellent performance.

But anyway, if they were as good as the silicone pads used in the same application note, the performance would be as follows:

enter image description here

As you can see, the performance with the pads will be as bad as several times worse than dry mount, even with high quality silicone pads. If they are cheap cr*p and leave out the boron nitride (BN) filler from the silicone rubber to save a few pennies (likely, IMHO) they could be much, much worse.

Note: I deliberately did not use the data directly from the manufacturers of thermal pads for two reasons- the link is to unknown pads of unknown characteristics, and secondly the manufacturers have reason to specify their products in such a way as to make them look better than they may be in practice. The graphs are from actual measurements with semiconductors mounted to heat sinks, not bulk material measurements, and use the same methodology for each mounting type so A-B comparisons are more likely to be valid.

TL;DR: Don't use the pads, or get different ones that are thermally conductive and from a quality source.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • 1
    +1; Thanks! I'm going to grok that nice application note immediately! – LorenzoDonati4Ukraine-OnStrike Feb 04 '17 at 11:04
  • A doubt about those graphs: could it be that they made a mistake when writing the Y-axis label? It is labeled as Rth(jc), i.e. "junction to case". Shouldn't it be Rth(ch), i.e. "case to heat sink"? – LorenzoDonati4Ukraine-OnStrike Feb 04 '17 at 11:25
  • @LorenzoDonati I think it is junction to heat sink. That's what they can measure. – Spehro Pefhany Feb 04 '17 at 13:13
  • Thanks again, also for that AN. I know I was probably buying crap, but for hobbyist work or "throwaway" school projects that's good enough. Sadly, here in Italy reputable distributors tend to charge quite more for their goods than in the USA, and there is also a VAT surcharge of 22% and a minimal order of ~100EUR (unless you want to pay expensive mailing fees). Therefore I buy from them only the devices I absolutely need to be reliable in the long run. – LorenzoDonati4Ukraine-OnStrike Feb 05 '17 at 10:27
  • BTW, I din't know that thing about the BN filler for thermal pads. Is there an easy test that can reveal if it's there? If I cut one of those pads could I see something that reveals if that filling is there? – LorenzoDonati4Ukraine-OnStrike Feb 05 '17 at 10:35
12

The pads you link to are insulator pads, to be used when the transistor must be electrically isolated from the heatsink.

If you need isolation, you can't do without one (or something similar). And additionally you'll need something to make sure the screw/bolt doesn't make the electrical contact.

If you don't need isolation, you are better of without, but use (a little) thermal grease to make sure the transistor and the heatsink make good thermal contact.

Wouter van Ooijen
  • 48,407
  • 1
  • 63
  • 136