Cool looking small copper heatsinks

[joecool85]

I found some Swiftech MC14 heat sinks that someone used on a small amplifier over at diystompboxes.com:



They are rated for 5watts, but I'd imagine that for our purposes they'd be good for ~10watts.  The guy at diystompboxes.com had two of them on his 20watt amp (one on a LM338 5amp voltage regulator) and one on the TDA7240A (20w audio chip).  I'm thinking it won't be enough heatsink for the TDA7240A if he drives it hard...but I could be wrong - I always over heatsink things.  I've never had a heatsink get hot enough I couldn't hold it in my hand indefinitely.

This was the circuit: http://musicpcb.com/pcbs/tiny-giant-amp/

[joecool85]

Since it is rated at 9 C/W I used this calculator: http://www.daycounter.com/Calculators/Heat-Sink-Temperature-Calculator.phtml

And found that for most chips (with a max temp of 150C on chip and 25C room temp) you would be good for just under 14 watts of dissipation.  When you factor in a warmer environment and the fact you don't want to run the chip at the bleeding edge of destruction all the time, 10 watts sounds good.  Apparently I have a knack for this since that was my original off the cuff guess :-P

Still some neat chips, and I would think it would look awesome using all 8 of them on a 1/8" copper plate for a LM3886 or something.

[DJPhil]

This didn't sound quite right to me, so I went and fiddled with the numbers. I apologize if you know all this already, I just had to give in to my inner science teacher.

With a typical TO220 you could use the listed value on the site of 3°C/W, but I guessed that the worst case for an chip like the TDA2040 would be different. They have to thermally bond two big transistor elements (the output push/pull pair) as well as other circuitry, so often they have a bit higher junction to case resistance.  I looked up the datasheet and they list a max of 4, though they messed up and gave the unit as volts. If you run the calculator with 4°C/W you'll get a case temp at 85°C, but the datasheet lists 16W for 85°C at the case, so they're assuming a shade better than worst case.

So far things are looking good on the calculator, 25°C ambient, 150°C max junction temp, R_{th jc} at 4° = 85°C junction at 15W output. By not adding anything else we'd be assuming a perfect, ideal case. To make it more realistic for a case with no heatsink we'd have to add a case to ambient resistance. It's not listed on the datasheet, but for something like this you can get a good idea from a voltage regulator like the LM317. It's datasheet lists junction to ambient resistance at 50°C/W for the TO220 package, and it's junction to case is the same at 4°C/W, so we can add 46°C/W to the calculator to give us a good idea what would happen if we used no heatsink. Yikes! By solving for wattage, we peak out the junction at 2.5W.

The next thing to consider would be the heatsink interface material. Delete the 46 and add, for simplicity's sake, a worst case of 0.3°C/W. Odds are you can do better, and a bad job would be worse, but it takes a lot of power to make the grease matter due to it's low thermal resistance.

Last we add the heatsink at 9°C/W. Finally! Now with a bit of tweaking you'll find that you're maxed out at about 9W. By upgrading the heatsink to 4°C/W you'll just barely max out at 15W. Heatsink packages for 3-4°C/W or so are larger beasts, like this series from Ohmite, which ranges up to about $2US per at Mouser. With the largest one there at 3°C/W, our calculations give us a junction temp of 134.5°C at 15W.

Ok, so this looks grim. There's good news though, the chip doesn't dissipate 15W at an output of 15W. Figure 9 on the datasheet shows the curve for dissipated power vs. output power at 4Ω, and it peaks at about 11W output. Keep in mind as well that this assumes a running voltage of 14.4V out of a max of 18V. The reason the curve looks like a hump is that class AB amplifiers work the hardest at half power, and the shallow downward slope at higher power levels might indicate the chip having difficulty dissipating heat (ideally it'd be symmetrical). Ok, so we only need to dissipate 11W, maybe a bit more, at least for a 4Ω load. Plug in the numbers with the massive heatsink and beep, bwoop, a 105°C junction temp. Much better, plenty of margin, longer lasting semiconductor, and we can even use a smaller heatsink if we like.

If you bump it all the way back to the 9°C/W of the original heatsink though it climbs back to 171.3°C. Solution? Run an 8Ω load, and you'll only have to dissipate half as much power! Checking the curve in figure 10 to be safe (again for 14.4V) we get a max dissipation of 5.5W. Plug in the original 9°C/W heatsink and you get a junction temperature of a bit over 98°C. Excellent!

There is a bit of wiggle with the numbers, as always. Bump up the T_amb to 65°C to simulate stage conditions (hot lights and a pile of hot gear) and things change a lot. Double the wattage and half the heatsink rating and see what happens. Poke around online for a 1°C/W heatsink (heavy metal).

I hope all that babble made some sense, I've had a lot of coffee tonight. I'm pretty sure I got the math right . . .
Hope that helps.  

Edit: I just realized I forgot to comment on the freaking heatsink itself after all that crap. How rude!
I couldn't locate it at diystompboxes' store, but I found it with google. The 9°C/W counts the 'TIM' (Thermal Interface Material), so our 0.3°C/W for paste was unnecessary. These are kind of neat really, and the built in sticky pad makes them easy to use. I'm not sure I'd trust the adhesive to much banging and jarring, but if you treat your gear like you would a computer it'd probably hold up a while. I'm not used to seeing forged copper pin style heatsinks, and I imagine they do a bit better in lightly convecting air (like near a video card fan) than most types would. I will file these away in the mental library, I'm sure I'll think of a use for them before too long. Good call.

[joecool85]

So basically I was right, it's good for about 10 watts of amplified power.  Although I do appreciate all the math, I'm a science geek too.   

[J M Fahey]

Congratulations, DJ Phil !!!
Good math and to the point  :tu: :tu:
I also find that long nights trying to understand something pays up.
Being able to concentrate without noise or interruptions for a few hours, digging deeper and deeper means a lot.
I'd only add:

the built in sticky pad makes them easy to use. I'm not sure I'd trust the adhesive

Me neither, specially because its thermal resistance is horrendous.
I see them as a gimmick intended to sell them to non-technical people.
Pull and drop the adhesive into the wastebasket and use real grease as a macho man.
I'd drill a small hole in the heatsink and mount the IC with a self tapping screw (no space for bolt and nut)
Thanks God IC case *is* ground so you need no mica.
Nice find !!!
I usually bolt everything that heats to the aluminum back panel (free heatsink) but sometimes in very compact designs I bolt a small CPU heatsink + cooler fan to the backpanel itself, to help it dissipate extra heat.
Be careful because some CPU heatsinks have two ridges along the border, which lets them sit flat over a square small CPU chip, but not against a flat aluminum sheet.