Multiple Transistors on a Heatsink

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

gato

Well-known member
Joined
May 17, 2009
Messages
115
Location
Ecuador
Hey gdiyers,

I'm building a power amp and I'm not sure how to calculate the junction temperature of my transistors so as to determine how big a heatsink I need.

Right now I have 12" wide extruded aluminum heatsink from HeatsinkUSA. They claim that 3" will be roughly 0.85 c/w. Let's say its 6" long so Rha will be 0.85*.707 (inverse square law for length) = 0.6

I'm using 5 IRFP240 mosfets with a junction-to-case of 0.83 c/w, and 5 IRFP9240 mosfets with a junction-to-case of 0.83 c/w.

I have micas and thermal grease so maybe Rch is 0.5 c/w

The amp produces 350 watts.

Lets say it is just one transistor:

P*(Rjc + Rch + Rha ) + Ambient (20c)

35* (0.83 + 0.5 + 0.6) + 20 = 78.888 degrees

Ok that's fine.

But, if I need 10 transistors dissipating 350 watts vs 35,  do I just change P to 350?

Then the junction temperature rises to 608 degrees

Not good!

If the heatsink is shared then its ability to dissipate heat remains unchanged. Perhaps I divide Rjc by 10?

I still get 346 degrees which is still way too hot.

I must be misunderstanding something.

Thanks so much for any help!

Josh

 
The science for this is pretty straightforward but some variables are well variable or difficult to qauntify.

Indeed you need to understand the actual dissipation of the devices. For example the device dissipation varies hugely between Class A, Class AB, Class G/H, or Class D.

At 350W of amplifier power I am accustomed to typically seeing forced air (fan cooling) which introduces even more variables.

It would be instructive for you to look at a similar topology amplifier rated for a similar power point.  Not to copy their design but to get a sense of scale for how to move enough heat.

JR

PS: One of my patents at Peavey was for a (forced air) heat sink design, so there will be differences between otherwise similar designs.  (6,515,859 Roberts ,  et al. February 4, 2003)
 
> The amp produces 350 watts.

That is NOT its heat dissipation!

Class B rated 350W OUTput probably dissipates 100W-175W.

> 10 transistors dissipating 350 watts

Not 350W each. Not 350W total. 100W-180W total. 10W-18W per device.

> 3" will be roughly 0.85 c/w.

I don't know why you bring in inverse-square. That might be nearly true for ONE heat point, but I assume you will spread your ten transistors over the full 6-inch length. (Might be tight!)

100W-175W at 0.85 c/w is 85C to 150C rise. Case to Air. Not counting mica!

Conservative audio design likes 50C rise max. The parts can work all day at steady high temperature, true. But in speech/music the boom-ba-boom loud/soft/loud cycles accelerate thermal fatigue in the seals. This is old rule-of-thumb and can surely be exceeded, especially on short-warranty products, or in home use where the 350W amp may never touch full power. Your choice.

But my gut, based on decades of looking at both good and cheap audio power amps, says 12"x6" of fins is just-maybe-almost enough for a 350W amp, and I would like more.

I do see "350W!" plate-amps about this size. I see them blown-up a lot.

This assumes conventional linear amplification. Switch-mode Class D can be >90% efficient. Under 40 Watts total dissipation. I'd use heatsinks out of an old Sansui 40+40W stereo. Maybe less; these things have gotten very efficient. Because the reduction in sink-size is a major commercial advantage.
 
> Ecuador ....... heatsink from HeatsinkUSA.

Their stuff is expensive. Especially if you are in Ecuador.

Computer CPU heatsinks are usually very competitively priced and probably available almost anywhere. I did a lot of work with Intel Pentium Socket 478. The heatsinks are mostly simple (easy to drill). On a "smart" load (which can slow or shut-down) they would do 65 to 85 Watts without melting Silicon. I figure five of these (with fans) would be ample for a 350W amp dissipating under 200W in the transistors. New this would be $60-$100 US. If you find the right PC graveyard you might be able to get five very-similar heatsinks cheap or free.

https://www.amazon.com/85x70x50mm-Socket-Cooler-Heatsink-Connector/dp/B00006B8EX
 
Ok awesome.

Thanks everyone!

Huge oversight on my part using power produced for power dissipated!

There is a whole chapter in D. Self's Power Amps book on dissipation and the end has a walk through on calculating it for sizing heatsinks. General guidelines show that an amp like this running in Class AB will have 52% dissipated power.

My understanding on the inverse square law is that if you increase width you're increasing the number of fins so your reduction in c/w is 50%, whereas with length the number of fins is the same. Good point about spreading the transistors across the surface, though.

Thanks CJ for the calculator!

I want to use these power amps to run my monitors so I need as long a life as possible.

I'm thinking about putting these in my "machine room" (which is really just a closet), and putting fans across them. I can punch a hole in the wall and add ventilation.

If I were to do that I'd have to run the speaker cable about 8 meters. I was planning on having the amps on the floor right by the speakers. Probably worth the tradeoff?
 
gato said:
Ok awesome.

Thanks everyone!

Huge oversight on my part using power produced for power dissipated!

There is a whole chapter in D. Self's Power Amps book on dissipation and the end has a walk through on calculating it for sizing heatsinks. General guidelines show that an amp like this running in Class AB will have 52% dissipated power.
Class A/B is popular and commonly used for modest power points. Is that 350W per channel or 350W total from 2 channels? At 350W per channel it is borderline for using class G/H.

Note the more efficient amp classes not only reduce heat sink required, but place less demand on the power transformer, so save aluminum, copper, and iron. But the tradeoff is more circuit complexity.
My understanding on the inverse square law is that if you increase width you're increasing the number of fins so your reduction in c/w is 50%, whereas with length the number of fins is the same. Good point about spreading the transistors across the surface, though.
For passive cooling it is mostly about the surface area of the heat sink fins. Should be similar in both directions while perhaps somewhat lumpy when you gain or lose entire fins.

You may notice on better heatsink extrusions that they have tiny grooves in the surface of the fins. Perhaps not apparent but these small grooves significantly increase the surface area giving them more contact area with the ambient air.
Thanks CJ for the calculator!

I want to use these power amps to run my monitors so I need as long a life as possible.
Studio/control room monitors? Fan cooling may be a noise issue so locating in a tool room makes sense. You may also need to vent the hot air out of a small room. 
I'm thinking about putting these in my "machine room" (which is really just a closet), and putting fans across them. I can punch a hole in the wall and add ventilation.
Fans or forced air dramatically affect the heat transfer ( a little air movement does a lot of cooling). Some amps will have high and low speeds thermostatically controlled. I did one 24x7 always-on fixed install amp with 3 speeds (hi, lo, and off) for even better long term reliability.

Fan placement is not super critical, less critical than managing passive heat sink design, but you want all devices to receive similar air flow and cooling benefit to avoid hot devices (an amp is only as reliable as its hottest transistor). Amplifiers often include protection thermostats bolted to heatsinks to interrupt power at too-high temperatures.
If I were to do that I'd have to run the speaker cable about 8 meters. I was planning on having the amps on the floor right by the speakers. Probably worth the tradeoff?
This is pretty commonly done. Sometimes big console power supplies end up in the tool room too.  If your amplifier is located well away from the console you will want a differential or balanced audio input circuit on the amplifier.

JR

PS: I just sourced six channels of Hypex class D modules for my living room TV system, but haven't fired them up yet...Less power than you are talking about but no external heatsinks involved.
 
gato said:
Right now I have 12" wide extruded aluminum heatsink from HeatsinkUSA. They claim that 3" will be roughly 0.85 c/w.
So you can assume 1" of heatsink will have Rha=2.55°C/W

The square law does not apply here because the transistors are spead along the heatsink.

I have micas and thermal grease so maybe Rch is 0.5 c/w
Seems a lot to me.

The amp produces 350 watts.
But does it produce 350W of heat? A class AB amp used conservatively dissipates about half its rated power.

[/quote] Lets say it is just one transistor:

P*(Rjc + Rch + Rha ) + Ambient (20c)

35* (0.83 + 0.5 + 0.6) + 20 = 78.888 degrees [/quote] With the revised value for Rha I get
35* (0.83 + 0.5 + 2.55) + 20 = 135°C, which is very pessimistic.
 
abbey road d enfer said:
...
But does it produce 350W of heat? A class AB amp used conservatively dissipates about half its rated power.
...
Abbey, didn't you read all replies?

Gato, which monitors you are using?
 

Latest posts

Back
Top