I'm working on a project that requires the use of a power transistor. The transistor is a BJT, and I'm running a heavy amount of current through it (about 20 amps). I bought a huge heat sink for the transistor, and have noticed that, when in operation, the transistor gets way hotter than the heat sink. If I place my finger on the transistor, it's too hot for me to keep my finger on it. At the same time, if I put my finger on the heat sink, right next to the transistor, the heat sink is cold.

My suspicion was that I wasn't using a good thermal paste. I had bought some thermal paste, from Best Buy, and it's advertised to have a thermal conductivity of 3.8 w/(m*k). I figured that, if I got something with a much higher thermal conductivity, I'd notice a big difference.

So, I looked up different thermal pads, on Digikey, and noticed that some of them can go as high as 1,800 w/(k*m). The one I saw, was made of graphite. According to what I've read about graphite, it tends to have a thermal conductivity that, often, is, at least, in the hundreds.

I have a thin sheet of graphene, and decided to use that. Because it's not a hard material, I figured it may squish fairly well -- like a thermal pad. To my disappointment, the temperature difference didn't feel any different than with the stuff that's 3.8 w/(m*k). On top of that, I've noticed that Dow 340; a thermal paste that seems to be popular for power amplifiers, only has a thermal conductivity of 0.67 w/(m*k).

Am I missing something? Can anyone explain whether or not the huge temperature difference, between transistor and heat sink, is normal? Also, shouldn't a material with a much higher thermal conductivity, work a lot better?

Thanks for your time and attention. Any insight, that anyone is willing to provide, will be greatly appreciated.

 

What's the power into the BJT? 20A @ 1V is 20W, @ 5v is 100W

 

Welcome to AAC.

Thermal paste—or pads—work because their thermal conductivity is better than air. They improve heat transfer to the extent they fill the gaps caused by machine marks and other imperfections with something more conductive than the air that would otherwise be in them.

Thermal paste is not an improvement on metal to metal or silicon contact and good performance relies on maintaining as much such contact as possible. This is why only a very small amount of thermal paste is effective, and why pads, while being much more convenient, do not perform as well as paste.

In the case of your graphene sheet, just how thin the clamping makes it is a critical factor. Its thermal coefficient only matters when it isn’t substituted for a better, native one. That is, where it replaces air it would be expected to improve heat transfer even if relatively thick. But, where it replaces surface to surface contact it would have to be very thin indeed.

I would suggest trying a reputable thermal paste applied very sparingly and uniformly to check the results.

Concerning your question, the rule of thumb that I know for a thermal interface temperature drop is 1-10°C.

 

Hello,

Have a look at section 7 of the following article on heatsinks:
https://sound-au.com/heatsinks.htm

Bertus

 

I suspect the thermal paste layer is too thick.
You just need a thin layer, with the transistor fastened tightly to the sink.

What is the transistor package?

 

I suspect the thermal paste layer is too thick.
You just need a thin layer, with the transistor fastened tightly to the sink.

What is the transistor package?

The transistor package is TO-247.

Based on what I've learned from Ya'akov's post, your suggestion seems to be spot on. Before reading his post, I didn't realize that the main goal of thermal paste is to be a better thermal conductor than air, in the places where there are gaps between transistor and heat sink. That knowledge changes my perspective, entirely.

 

What's the power into the BJT? 20A @ 1V is 20W, @ 5v is 100W

The transistor is absorbing about 91 watts of power. Also, as I've mentioned before, it's a BJT. For these reasons, it's critical that this part be heat-sinked as well as possible.

I learned, the hard way, about the issue of thermal runaway, with BJTs.

Posting my question, on this site, has shown to be an excellent idea. Ya'akov's post was a real eye-opener for me. Now, I realize that the goal is to use very little thermal paste; since metal to metal contact is best. Thermal pads are a bad idea for this scenario.

I greatly appreciate all of the insight that I've been receiving from all of you. This is extraordinarily helpful!

 

There is a remote possibility that You are dealing with a "Fake" Transistor, with really poor Quality-Control.
Just a thought.
.
.
.

 

Welcome to AAC.

Thermal paste—or pads—work because their thermal conductivity is better than air. They improve heat transfer to the extent they fill the gaps caused by machine marks and other imperfections with something more conductive than the air that would otherwise be in them.

Thermal paste is not an improvement on metal to metal or silicon contact and good performance relies on maintaining as much such contact as possible. This is why only a very small amount of thermal paste is effective, and why pads, while being much more convenient, do not perform as well as paste.

In the case of your graphene sheet, just how thin the clamping makes it is a critical factor. Its thermal coefficient only matters when it isn’t substituted for a better, native one. That is, where it replaces air it would be expected to improve heat transfer even if relatively thick. But, where it replaces surface to surface contact it would have to be very thin indeed.

I would suggest trying a reputable thermal paste applied very sparingly and uniformly to check the results.

Concerning your question, the rule of thumb that I know for a thermal interface temperature drop is 1-10°C.

Thanks so much! Your post has completely changed my perspective of the role of thermal paste. When you mentioned that the goal is for it to be a better thermal conductor than air, that changed everything. Prior to that, I never viewed thermal paste, or thermal pads, as any sort of liability to the design. As you've pointed out, too much paste can be a problem.

Thanks, again, for your extraordinarily informative response. It's a game-changer.

 

Yes, too much thermal paste can be a problem. It should not be a barrier between the transistor and heatsink. It should be there just to fill in the imperfections on the two mating surfaces. And presumably the transistor is tightly bolted to the heatsink. Also, use a quality brand such as Wakefield.

Does the transistor required electrical isolation from the heatsink?

 

There is a remote possibility that You are dealing with a "Fake" Transistor, with really poor Quality-Control.
Just a thought.
.
.
.

I seriously doubt that's the case. I've burned up two of these, throughout my process of trying to get this circuit to work correctly. The transistors, themselves, do an excellent job. They amplify the current to the necessary magnitude, and the resulting signal looks great, on an oscilloscope. Their performance has been consistent.

My biggest issue is with keeping them cool. I've only been able to run the circuit for limited periods of time, due to my fear of burning up the transistor. The heat sink is, certainly, helping. However, the temperature, still, slowly rises, over time.

This issue must be eliminated, since the circuit will, ultimately, have to run for a lot longer than a few minutes.

For anyone who's interested, here's the datasheet for my transistor:

MJW18020 power transistor

 

Yes, too much thermal paste can be a problem. It should not be a barrier between the transistor and heatsink. It should be there just to fill in the imperfections on the two mating surfaces. And presumably the transistor is tightly bolted to the heatsink. Also, use a quality brand such as Wakefield.

Does the transistor required electrical isolation from the heatsink?

Ideally, I'd prefer that all of my transistors be electrically isolated. However, in this case, it's not critical. I'm a lot more interested in keeping it cool.

Based on what I've experienced, they seem to be of good quality. They amplify my current, really well, and are very consistent with how they behave.

Here's the datasheet:

MJW18020 Datasheet

 

What is the transistor base current for the 20A output?

 

Hello,

Have a look at section 7 of the following article on heatsinks:
https://sound-au.com/heatsinks.htm

Bertus

Thanks for this link. It appears to be very informative.

 

What is the transistor base current for the 20A output?

The last measurement/calculation, that I was able to make, is about 750mA.

Ultimately, I'm using two BJTs. A smaller one drives the base current of this bigger one. After lots of experimentation, I've ditched the resistor that used to exist between the emitter of the smaller transistor, and the base of this bigger one. Because of that, it's been harder to measure base current. However, before ditching that resistor, I was measuring about 750mA, and that resistor value was quite low. So, I eliminated the resistor, and the circuit still seems to be working well.

 

Ultimately, I'm using two BJTs. A smaller one drives the base current of this bigger one.

So they are connected as a Darlington?

Can you then measure the collector current of the smaller transistor, which would be the base current of the larger?

 

Ideally, I'd prefer that all of my transistors be electrically isolated. However, in this case, it's not critical. I'm a lot more interested in keeping it cool.

Based on what I've experienced, they seem to be of good quality. They amplify my current, really well, and are very consistent with how they behave.

Here's the datasheet:

MJW18020 Datasheet

The datasheet does not state if the case is connected to the transistor collector. Check with an ohmmeter.

 

The datasheet says the current gain at 20A is minimum 5.5 typical 9 so
at 20A there should be about 2.2 amps of base current. For Ib of 750mA
takes a current gain of 26. Perhaps the transistor was redesigned
and it's better than the datasheet, however there is a maximum
current gain of 34 at 3A and 5V.

The current gain goes down as the transistor temperature goes up.

The Thermal Resistance, Junction−to−Case is stated as 0.5 C/W so at
100 Watts the transister junction is 50 C hotter than the transistor
case (assuming a "perfect" heat sink).

Assuming a heatsink in a 25 C environment, if it's also 0.5 C/W
then the junction is going to be 100 C hotter than 25 C so 125 C.
If the heatsink rises above 25 C it's easy to see the transistor
junction temperature exceeding 150 C.

We don't have a schematic nor know the source voltage. Since several
transistors died it's likely things are pushing the limits of the
transistor. Clearly there isn't much margin from the limits.

A good learning opportunity

 

Presumably you are using an aluminum heatsink? Copper has a much higher thermal conductivity than aluminum so using a slug of copper between the transistor and the aluminum heatsink will help. Once the area of the transistor is spread to an order of magnitude larger area of copper (not too thin) an insulating pad for electrical isolation between copper and aluminum could be acceptable