How the heck can a TO-220 case MOSFET carry 80A?

Thread Starter

sagor

Joined Mar 10, 2019
903
Ok, time for me to post a question now....
I've been looking at some specs of N channel Mosfets like the RFP40N10 and the STP130N8F7. The RPF is rated at 40A continuous and the STP device is rated at 80A continuous at 25C. In my mind, the source leg of the TO-220 case would melt carrying at 80A, would it not?
So how the heck does one use a device like these that can carry huge currents? I can see the Drain, being the tab, handling high currents, but that current has to go out somewhere, out the Source leg. (or is it in the Source leg and out the Drain - no matter...)
Any of you experts out there ever pass something like 60A or 80A through a TO-220 Mosfet device? If so, why is it that the Source leg does not get red hot or even melt like a fuse? What am I missing here?
 
Last edited:

jpanhalt

Joined Jan 18, 2008
11,087

Thread Starter

sagor

Joined Mar 10, 2019
903
Yes, I've seen some datasheets state "Package Limited" by the current rating, but they do not state what the "limit" is in actual current.
Basically what is the current limit of the TO-220 leg itself, assuming short length and heavy PCB layer (or wire) soldered to the Source leg?
I'll read the articles, thanks.
 

Thread Starter

sagor

Joined Mar 10, 2019
903
Based on the article linked to by jpanhalt, the limiting factor seems to be the Source lead heat level and how well the heat is conducted away from that lead (via PCB copper or other means). The bond wires internally, based on that article, seem to be rated for the total current listed in the datasheets (in most cases), and the epoxy cases seem to provide some heatsinking of those wires. Manufacturers seem to be able to add more internal bond wires as required when the current rating goes higher.
Heat from the Source leg may eventually work its way internally, causing problems it seems.
The hard part is, to measure the Source leg temperatures as you pass more current through the device.

Well, that sort of answers the question, I think around that 40A range is ok provided you have heavy PCB traces or heavy wire soldered directly to the Source leg. I suppose this all assumes you have proper heatsink on the device as well.
 

Analog Ground

Joined Apr 24, 2019
460
From the TO-220 mechanical diagram on the RFP data sheet, the large part of the TO-220 lead is 1.14 x 0.48 x 3.5 mm. The AAC resistance calculator (https://www.allaboutcircuits.com/tools/trace-resistance-calculator/) gives this short piece of copper a resistance of 0.000124 ohms at 60C. At 80 amps this is 0.8 watts . With a reasonable heat path into whatever connects to the lead, I don't think there is a problem. This assumes the connection is to the fat part of the lead. It is probably a reasonable caution to keep the lead as short as possible.

Edit: The summary in the application note linked in post #2 basically says the same thing. It is all about the connection to the TO-220 lead and the thermal resistance.
 
Last edited:

MisterBill2

Joined Jan 23, 2018
18,167
I would not try a TO220 package at 40 amps for more than a very few milliseconds, no matter how good the heat sink was. Some data sheets are "rather optomistic" , or else the actual conditions are not typical. And that value may be with an infinite heat sink at zero degrees C. Certainly the circuit board would suffer if that current flowed for more than a very short time. So use 4 of the transistors with a 0.1 ohm resistor in the source lead of each one. AND don't believe everything that you read.
 

Analog Ground

Joined Apr 24, 2019
460
For the STP device, the Rdson is 0.0058 ohms max. At 80 amps, the power is 37 watts. The thermal resistance junction-to-case is 0.78 C/W. So, the temp rise is 27 C. Not too scary. As for the heat sink, again, you are only getting rid of 37 watts. Not scary. 80 amps on a PCB is doable but you probably need a specialized fabricator who can do thick copper like 6 to 10 oz. I did a system distributing 300 amps at 5 volts. You just have to engineer the heck out of it. BTW, it was a water cooled system which helped a lot.

EDit: That said, I would never operate a part at it's maximum. I agree with MisterBill2. Divide and conquer is the way to go.
 

jpanhalt

Joined Jan 18, 2008
11,087
@MisterBill2

If you check the IRF1010E datasheet ( a relatively old mosfet), you will see two measures of maximum current for the device:

1579985290168.png

1579984308101.png

And #2:

1579984241087.png

Several years ago, I used that chip x4 (parallel) in a controller for a 6-V Ford long-shaft starting motor (old Fords and tractors) operated at 12V from a Marine battery. Launch was at full power or as close to 100% duty cycle as my controllers could give. That was used in a winch to launch sailplanes of 5 pounds and more. We would actually stall the winch just prior to launch to build tension on the line to avoid a low-level stall, and although I never measured the current, I did know my winch was probably not legal for FAI (international) competition that limited current to 200 A or less.

The 4 mosfets were attached to a heat sink, connections were by heavy copper (4 oz PCB or 0.025 buss bar depending on the controller), soldered, and launch was for several seconds, not milliseconds. I never had a problem in several years.

I think your estimate of milliseconds for 40 A way underestimates what they can do.
 

BobTPH

Joined Jun 5, 2013
8,804
The thermal resistance junction-to-case is 0.78 C/W. So, the temp rise is 27 C
That just means the case will be hotter than the junction by 27C. Check the case to ambient to tell you how hot the case will get above ambient temperature without a heat sink.

According to wikipedia, that is typically 70C per watt. So, at 37W with no heatsink, the temperature of the case would be about 2970C higher than ambient and the junction would be 27 more than that, except that it melted long before it got that hot.

Bob
 

Analog Ground

Joined Apr 24, 2019
460
That just means the case will be hotter than the junction by 27C. Check the case to ambient to tell you how hot the case will get above ambient temperature without a heat sink.

According to wikipedia, that is typically 70C per watt. So, at 37W with no heatsink, the temperature of the case would be about 2970C higher than ambient and the junction would be 27 more than that, except that it melted long before it got that hot.

Bob
Whoa! You better provide a link. No way a properly greased junction to a heat sink is 70C per watt. As for the heat sink, there are some big ones and there is always water cooled, heat pipes, fans, etc. Link?

Edit: Using a heat sink is a given. Perhaps I should have made that clear. The 27C junction to case gives plenty of headroom for the additional rise of a properly designed heat sink.
 
Last edited:

BobTPH

Joined Jun 5, 2013
8,804
The junction-to-case thermal resistance of a TO-220 packaged device (which typically matters less than the case-to-ambient thermal resistance)
...
When a TO-220 package is used without a heatsink, the package acts as its own heatsink, and the heatsink-to-ambient thermal resistance in air for a TO-220 package is approximately 70 °C/W.
https://en.m.wikipedia.org/wiki/TO-220

Bob
 

BobTPH

Joined Jun 5, 2013
8,804
I read it. It contradicts nothing I said.

You said the 0.8 C/W for junction to case was not bad. My point was that it fairly unimportant. To get junction to ambient, you add 3 components, junction to case, case to heatsink and heatsink to ambient. It is the case to ambient or heatsink to ambient that dominates.
 

Analog Ground

Joined Apr 24, 2019
460
I read it. It contradicts nothing I said.

You said the 0.8 C/W for junction to case was not bad. My point was that it fairly unimportant. To get junction to ambient, you add 3 components, junction to case, case to heatsink and heatsink to ambient. It is the case to ambient or heatsink to ambient that dominates.
OK. I am going to lay it out in detail.

1. TO-220 to heat sink with thermal grease : 0.3 C/W. See: http://www.junradio.com/Discretes/TI/9701.pdf
2. Extruded heat sinks with forced air cooling easily attain <0.5 C/W. see lots of heat sinks here: http://www.wakefield-vette.com
3. 0.3 + 0.5 = 0.8 C/W. TO-220 case to ambient with the conditions in 1 and 2.
4. 0.8 C/W junction to case = 0.8 C/W case to ambient. The heat sink to ambient does not "dominate". It is easily about the same as the junction to case.
5. 30 C ambient + 27 C junction to case rise + 27 C case to ambient rise = 84 C junction temperature.
6. Max junction temperature is 175 C. If margin is needed, go to other heat sink arrangements.

Like I said, doable. Otherwise, why would anyone try to sell an 80A MOSFET in a TO-220 package? These people are not stupid.
 

BobTPH

Joined Jun 5, 2013
8,804
Whoa! You better provide a link. No way a properly greased junction to a heat sink is 70C per watt.
Here is the problem. I did not say anything like what you are objecting to. I said case to ambient was 70C/W.
You asked me to give a link, which I did. Do you know what “case to ambient” means, apparently you do not, since you go on about massive heat sinks and forced air cooling.

I did not say that it could mot work with a proper heatsink, which you also seem to think I said.

Bob
 

MisterBill2

Joined Jan 23, 2018
18,167
@MisterBill2

If you check the IRF1010E datasheet ( a relatively old mosfet), you will see two measures of maximum current for the device:

View attachment 197607

View attachment 197606

And #2:

View attachment 197605

Several years ago, I used that chip x4 (parallel) in a controller for a 6-V Ford long-shaft starting motor (old Fords and tractors) operated at 12V from a Marine battery. Launch was at full power or as close to 100% duty cycle as my controllers could give. That was used in a winch to launch sailplanes of 5 pounds and more. We would actually stall the winch just prior to launch to build tension on the line to avoid a low-level stall, and although I never measured the current, I did know my winch was probably not legal for FAI (international) competition that limited current to 200 A or less.

The 4 mosfets were attached to a heat sink, connections were by heavy copper (4 oz PCB or 0.025 buss bar depending on the controller), soldered, and launch was for several seconds, not milliseconds. I never had a problem in several years.

I think your estimate of milliseconds for 40 A way underestimates what they can do.
That value of 80 amps is a calculated value. That means that it will not happen in real life, only in a calculated situation. The term for that technique is "Specsmanship", which involves creating part specifications that only apply under conditions that can never be achieved in the real world.
 

MisterBill2

Joined Jan 23, 2018
18,167
Here is the problem. I did not say anything like what you are objecting to. I said case to ambient was 70C/W.
You asked me to give a link, which I did. Do you know what “case to ambient” means, apparently you do not, since you go on about massive heat sinks and forced air cooling.

I did not say that it could mot work with a proper heat sink, which you also seem to think I said.

Bob
"Case to ambient" is effectively case to the heat sink. It probably does not include the junction to case resistance.
 
Top