TO-3 vs TO-220 power dissipation

[Tioleco]

Which one has more power dissipation capability?Some voltage regulators have a high price for the TO-3 versions.

[ivan747]

The TO-3 is a little higher without heatsink, but I prefer to stay away from them because you have to desolder them to get the heatsink off. Also, They lack of any pins connected to the case itself, requiring a heatsink for every one of the devices or having them mounted in a way they don't conduct to the case but can be bolted into the PCB.

Here are the power dissipation ratings for some surface mount packages for comparison. The TO-3 dissipation ratings are hard to find, at least for me. Also, I feel lazy right now. Go figure. 
http://www.irf.com/technical-info/appnotes/an-994.pdf

Ivan

[gregariz]

Here's a table of thermal resistances
http://cds.linear.com/docs/Packaging/Linear_Technology_Thermal_Resistance_Table.pdf

[Bloch]

Which one has more power dissipation capability?Some voltage regulators have a high price for the TO-3 versions.

That easy :-)
TO-3 have the best power dissipation / transfer

TO-220 JC °C/W = 3
TO-3     JC °C/W = 11
This was a wrong a simple copy paste error. Thank you tecman

TO-3     JC °C/W = 3
TO-220 JC °C/W = 11

11-3=8° "wasted"

One more example is a  3055 npn transistor 115 / 90 watt = 28 % more

But as you and Ivan they have mentioned is the cons  Cost and not so easy demounting.

So is it now up to you if you need that extra power.

Did you see David video ? http://www.eevblog.com/2010/08/15/eevblog-105-electronics-thermal-heatsink-design-tutorial/

[Harvs]

A big advantage of TO-3 is they're a metal package.  The 150C temp limit on TO220 packages is because of the package itself (being plastic) not the die.  So lots of high power TO-3 packages are rated to 200C.  So your not only increasing your thermal transfer, but also the temp you can potentially run your heatsink at (if your physical design allows that.)  That's why they are still often used in large high power linear amplifiers where typically the heatsinks are far more expensive than the components.

[tecman]

As stated (but Bloch got his numbers reversed), the big difference is the junction-to-case thermal resistance.  3 deg/watt for TO-3 and 11 deg/watt for TO220.  So the thermal flow in the TO-3 almost 4 times better in a TO-3.  This corresponds to a higher power capability.

paul

[Harvs]

I don't know where you got 3 degrees C per W for a TO-3 from.  Every TO-3 I've ever used has been somewhere between 0.6 to 0.7 degrees per watt.  3 Degrees per watt is terrible junction to case! ON-semi 3055 is 0.657 degree/W for JC.

Anyway, I can't say I'd use a TO-3 purely for JC thermal resistance.  There's almost always a circuit to parallel devices, where you can spread the load across more than one package very cheaply (and in doing so even out the heatsink heat transfer).  And for that reason, in our present age, I don't believe JC is the big difference/reason for use.

On the other hand, as I was saying about the package temp rating.  If your in an enviroment that has a very high ambient temp (e.g. industrial appllications, automotive, military), it doesn't matter that you've got a low JC resistance, if your heatsink can't effectively dissipate to free air because the free air temp is very high.  This is where the TO-3 packages win hands down.

But as a side note, back the original question.  If you need to disipate more than what a TO-220 package can handle, you're probably better off looking at one of the many easy circuits to couple a linear regulator with regular cheap NPN transistors to get your desired rating, or even consider a switchmode supply

[tecman]

I don't know where you got 3 degrees C per W for a TO-3 from.  Every TO-3 I've ever used has been somewhere between 0.6 to 0.7 degrees per watt.  3 Degrees per watt is terrible junction to case! ON-semi 3055 is 0.657 degree/W for JC

Pulled that off the LT sheet referenced in a prior post.  I agree that it is off.  I just did a quick look and posted the numbers.  Looking in a Motorola data book, the TO-3 packages vary from about 0.8 to 1.5 deg/watt.  So there is atill a 2 to 3 improvement over a to220, which averages 3.25 deg/watt

As for using smaller devices in parallel, nice if you can do it.  Paralleling requires more heat sink/board area, usually additional components, such as current balancing emitter resistors, additional base resistors, etc.  You also increase drive capacitance.  So it depends on the application for more than thermal reasons.  With linear regs (original question), paralleling them is usually not a very easy task if you want to maintain performance.

paul

[Zero999]

A big advantage of TO-3 is they're a metal package.  The 150C temp limit on TO220 packages is because of the package itself (being plastic) not the die.  So lots of high power TO-3 packages are rated to 200C.  So your not only increasing your thermal transfer, but also the temp you can potentially run your heatsink at (if your physical design allows that.)  That's why they are still often used in large high power linear amplifiers where typically the heatsinks are far more expensive than the components.

I don't think that's true. Some TO-220 packages are ceramic and are rated to higher temperatures than 150^o, for example the IRL540 (TO-220 package) can run up to 175^oC and the 2N6896 (TO-3 package) is only rated to 150^oC.

http://www.vishay.com/docs/91300/91300.pdf
http://www.datasheetarchive.com/dl/Datasheets-312/178273.pdf

[alm]

But as a side note, back the original question.  If you need to disipate more than what a TO-220 package can handle, you're probably better off looking at one of the many easy circuits to couple a linear regulator with regular cheap NPN transistors to get your desired rating, or even consider a switchmode supply

Note that you give up the over current and over temperature protection with an external pass transistor unless you build a somewhat more elaborate circuit. If cooling a TO-220 is an issue, switch mode sounds like the right solution unless you require low noise or extreme simplicity.

[Harvs]

@Alm - Good pickup on the over current, didn't think of that one. 

@Hero999 - I could be wrong, I got this origonly from one of Douglas Self's power amp texts, and it has generally proven to be the case when selecting power transistors for linear audio amplifier applications.  Granted they generally never use TO220's though, usually something bigger like the TO265 which seems to have the same JC as the TO-3, though usually the 150C limit.

The problem with generalizations is you can always find whole familys of devices that for some reason or another that is often unclear seem to go completely against the trend.

@tecman - your right on a lot of those points, however I often find mounting 4x TO-220 devices a lot easier than dealing with a signal TO-3 package (straight drill line to a heatsink vs the odd footprint).

[Bloch]

And a easy way to "cheat" the TO-220 limits

Eks 5Volt 1,5 A from 25 volt

Put 2 LM78 in serie

First  L7815 ((25-15)*1,5 = 15Watt)
Then L7805 ((15-  5)*1,5 = 15Watt)

[Conrad Hoffman]

I've never been too sure on this one because you still have to do a good thermal connection to the heat sink. Which is better and with which package- Silpads, Super Silpads, mica and grease, anodized aluminum, Kapton, alumina, beryllium oxide, or how about adding a copper spreader in between. I swear the published numbers are are set up as to be non-comparable unless you're a physicist and mathematician. IMO, the TO-3 has a lot of advantages, especially if you use a good socket so you can easily pull the device. Still, I use TO-220 for almost everything for cost and size reasons.

[Zero999]

And a easy way to "cheat" the TO-220 limits

Eks 5Volt 1,5 A from 25 volt

Put 2 LM78 in serie

First  L7815 ((25-15)*1,5 = 15Watt)
Then L7805 ((15-  5)*1,5 = 15Watt)

Another option is to simply put a power resistor in series but you should check there's still a high enough voltage at the input under full load conditions when the input voltage is at the minimum.