Old CPU-cooler as an LM3886 heat-sink

[Thomeeque]

 Hello!

I'm building simple LM3886 power amplifier (50W/8Ω, I did already mentioned it here lightly), and I'm trying to use left-overs if possible (PT, PSU case..) to give them some use and to save some money. Now, my qustion is, can I use one of these old CPU-cooler heat-sinks for LM3886 cooling?

 

Left one is slightly smaller (height is around 30mm), but is significantly heavier thanks to copper core (height of the right one is 41mm).

My idea is to mount heat-sink on outter side of the box and LM3886 on inner side - case plate (0.6mm steel) will transfer the heat (?) or I can make opening there to connect heat-sink and chip directly, if necessary.

Does anyone of these look like it could do the job? I mean without original fan (there will be 80mm fan inside the box, but I don't want to put fan on outter heat-sink).. Or are these useless withou fan (too small, too thin spaces between ribs to ventilate the heat "naturally"..)? Btw. I would preffer to use left one (from the aesthetical point-of-view mainly, plus it could be maybe easier to mount and maybe I could manage some airflow between it's ribs from the inside fan..)

How much cooling will do the case itself?

Thanks!

T.

[R.G.]

I've pondered this myself. My conclusion is that yes, they'll be fine. I suggested using some of the older style pin-fin heatsinks in my article on Bulletproofing Thomas-Organ Vox amps.

If you're not using a fan, you'll simply have to arrange it so airflow can run through the fins in a vertical direction unimpeded. But even a tiny, slow bit of fan assist on the airflow will have huge benefits.

The sink on the right *might* be enough if you orient it so the fins run vertically. The LM3886 will turn itself off if it gets too hot (I know this for a fact!  ) so it's not dangerous to try this.

The fan on the left will almost certainly work, but if it were mine, I'd get a DC powered fan and run it on a lower DC voltage than it normally requires. This will run it slowly and quietly.

[BAARON]

Personally, I'd have the chip in direct contact with the heat sink.  If you put the chassis between them, the entire chassis becomes part of the heatsink.  And I agree with RG: putting a fan on that heatsink will make it much more effective.

[Mark Hammer]

....and if I have a water-cooled heatsink, can I overclock the 3886?

Actually, I'd been staring at all those leftover fan/heatsink assemblies I've accumulated from all the "Hey, you're into computers, right? Maybe you can use this." things that are currently filling up valuable storage space in my garage.  I was thinking they might be appropriate for just such a mission.  I guess the smaller 486 and P1 heatsinks would be decent for a TDA2030 type amp.

[Thomeeque]

I've pondered this myself. My conclusion is that yes, they'll be fine. I suggested using some of the older style pin-fin heatsinks in my article on Bulletproofing Thomas-Organ Vox amps.

If you're not using a fan, you'll simply have to arrange it so airflow can run through the fins in a vertical direction unimpeded. But even a tiny, slow bit of fan assist on the airflow will have huge benefits.

The sink on the right *might* be enough if you orient it so the fins run vertically. The LM3886 will turn itself off if it gets too hot (I know this for a fact!  ) so it's not dangerous to try this.

The fan on the left will almost certainly work, but if it were mine, I'd get a DC powered fan and run it on a lower DC voltage than it normally requires. This will run it slowly and quietly.

Thanks, R.G.!

I've got one pin-fin from the old mobo (chipset cooler), but it's too small I guess (50x50x36mm) and very light.

So primarily I'll try to find the way how to mount left one and manage some airflow for it - I'd like to avoid externally mounted fan mainly because I would have to cover it somehow, hopefully there will be way how to send some airflow from inner fan thru it.

Personally, I'd have the chip in direct contact with the heat sink.  If you put the chassis between them, the entire chassis becomes part of the heatsink.

But would it be bad thing, if the entire chassis becomes part of the heatsink? Why not to make the chassis part of it? Escpecially when it will be cooled by inner fan (so it should not become too hot to touch)..?

....and if I have a water-cooled heatsink, can I overclock the 3886?

Yep, you could play much faster thru it!

T.

[RDV]

The more heatsink material and density the better. I dare say it can't be 'too' big.

RDV

[R.G.]

But would it be bad thing, if the entire chassis becomes part of the heatsink? Why not to make the chassis part of it? Escpecially when it will be cooled by inner fan (so it should not become too hot to touch)..?

The real problem is that you can't have the entire chassis part of the heatsink. Or at least productively.

The issue is the relative size of thermal resistances.  When we play Electronicist, we get used to thinking of resistance as something that's almost infinitely controllable. We typically get into trouble where that concept fails, as in signals generated by currents flowing through "ground wires". With heat flow, the thermal resistance of a path is modeled as the temperature rise per heat flow, often in units of degrees C per watt. What's wrong is that it's almost impossible to make this resistance very small in any solid material. The thermal resistance of a chassis is such that only the metal nearby the device that's generating heat is effective in cooling the device in most cases.

Worse yet, any gap in the heat flow path is subject to having much of the gap filled with air; only the mountainous tops of the surface roughnesses actually touch each other in any contacting surfaces that are not lapped to almost optical smoothness. So the gap between two pieces of metal in contact will often have a higher thermal resistance than anything else in the thermal path.

For silicon, about the highest temperature you can let the junction get to is 200C. Higher than that and the junction starts degrading. There is some thermal resistance between the silicon semconductor junction and the package it's mounted in. This is often in the range of 0.5 to 5 degrees C per watt (C/W). For the very good MJ802 for instance, it is 0.875C/W. What that means is that if you had an absolutely perfect heat sink at 25C connected to the case of the device, it would reach the maximum junction temperature for that device of 200C at a power of P = (200-25)C/0.875C/W = 200W. You *can't* get an more power out of it - the chip fails inside the package. Other semiconductors may limit the junction to lower temps - 150C is common. And they mostly have higher thermal resistances inside.

In the real world, there are no perfect heat sinks, so the heat sink will have a thermal resistance from the mounting location to the air around it. And the mounting surface between case and sink will have a resistance. About the best you can do for the resistance between package and sink is something like 0.5C/W, by polishing smooth and filling with metal oxide loaded grease to exclude insulating air. So by the time your heat gets out of the chip and to the heat sink, it's already traversed 1.3C/W to maybe 5C/W or more. Only then can the heat sink begin transferring heat to air.

The metal has a thermal resistance, and that appears in series with the case and gap resistances, from the mounting area to the fins. A fair amount of math would show that you get rapidly diminishing returns for distances more than a few inches from the mounting surface. The trick in heat sink design is to do anything possible to get the fins started as close to the mounting surface as possible (that nice radial sink) or to put a lot of metal for conduction out to the fins (the thick bottom plate on the right sink). The worst thing is to just tie the package to a think metal plate - like mounting it on a chassis - or to put more gaps between the heat source and the fins, like mounting the chip inside the chassis and mounting the fins outside, with another gap.

I guess the simple explanation is don't use the chassis between the chip and the sink, it adds a gap.

[Thomeeque]

 Well, I'm knocked-out, I'm feeling almost guiltily to get such a huge portion of valuable information and education for free

Thank you very very much, R.G., again!!

T.

[head_spaz]

That radial heat sink will be fine if properly implemented. It's a very nice doohickey indeed!

I would strongly suggest mounting the LM3886 directly to the heatsink, and would NOT sandwich
the case inbetween. That will probably mean cutting a hole in your case.
Also, as R.G. stated, air flow is highly beneficial, if not absolutely required.
And if you can provide about a 1/4" air-gap between the heatsink and the case, you'll get much
better air circulation thru the heatsink.

Also, don't forget that the TAB on the LM3886 is connected to the voltage rail, so you'll need an
insulator of some sort. (or switch to using LM3886-TF)
You can insulate the device from the heatsink, but that will decrease thermal conduction a tad...
or you can mount it directly to the heatsink and insulate the heatsink from the chassis - but then
it would be foolish to mount it exterally where it becomes a shock hazard. (unless you plan to build
a cage around it.)

You might wanna visit DIYaudio.com
They are the premier chip-amp site, similar in popularity to stompboxes.com
They are very helpful, a creative bunch of guys including numerous electronic professionals that post
there on a regular basis. Guys like R.G.

On the downside, they recently "reorganized" the forums and put too many stickies up top for my taste,
which tends to get in the way. Plus with the downturn in economy... I've noticed that alot of the regulars
have left, assuming they've had to readjust their priorities. It would be to your advantage to look at some
of the older posts, maybe going back a year back or more - before some of the pioneers got bored and
moved on to other things.

[Thomeeque]

 Thanks a lot, head_spaz!!

I had made all the (hard) "metal-work" during this weekend (I did follow all hints I've got here including yours, so there's an openning in chassis and IC is connected directly to the heat-sink, there's a 1/4" air-gap etc.), I hope I'll get to wire it and test it soon (tomorrow evening possibly). I'll post some new pics and report results here than..

About TAB insulation - I'm aware of that. I had bought a kit for the complete LM3886 PA circuit, it contains some insulator already (you can actually see it on both pictures above - just locate pink color ) it should even do the "grease job" as well so I hope it won't make too much harm.

About DIYaudio.com - I had googled a lot before I have started, I could not miss it I believe it was there, where somebody created one huge cumulative post, containing a lot of previous posts (at least 50% by R.G. ) in very well arranged way - very informative (I just cannot find it now)!

Thanks, T.

[Paul Marossy]

So primarily I'll try to find the way how to mount left one and manage some airflow for it - I'd like to avoid externally mounted fan mainly because I would have to cover it somehow, hopefully there will be way how to send some airflow from inner fan thru it.

If you have to mount it externally, you could use a piece of metal screen or expanded metal as a covering. Or even a piece of sheet metal with a series of 1/4" holes drilled thru it. These would allow airflow and at the same time protect the fan from damage.

[scratch]

I'll have to keep an eye on this thread ... planning on building a stereo 20W amp (LM1875's) in a rack mount chassis along these lines. Planning on mounting the heat sinks (I also have some 'surplus' CPU heatsinks) inside the chassis, cut holes in the bottom and top plates and bolt in some screens for air flow ... i like the idea of running a fan at less than spec'd voltage for additional cooling.

[Paul Marossy]

i like the idea of running a fan at less than spec'd voltage for additional cooling.

Works well for cooling tube amps, too.

[Thomeeque]

 Hello again!

I've got better idea about placing heat-sink later so there was some more metal-work yet to do, but now it's finally finished and alive

Click the pic for hi-res. For more pictures (including "orignal" heat-sink placing) see full gallery. Now is heat-sink sunk into the box partly and most of other opennings are blinded by (transparent) plastic-sheets, so lot of fan-driven airflow is forced to go out thru it. So far I'm testing it at home only where I cannot drive it at full horse-power, so I cannot tell much about heat-sink efficiency yet, but it makes very good impression - I even believe that I'll be able to slow down fan little bit (now it gets full 12 volts and makes it little bit noisy). I'll report more once I'll know more. I may create artificial load, it could be handy even in the future - but now I'm in extra-lazy mood (I would use some cooling now ), we'll see.. Cheers, T.

[Ben N]

Nice, Tomas. (OT: What's the Vox?)

[Thomeeque]

Nice, Tomas.

Thanks, Ben!

(OT: What's the Vox?)

It's probably the cheapest Vox speaker-cabinet available (V212 BN; 2x 12" GSH12-30) - looks good & sounds OK, I use it at home

T.

[Paul Marossy]

Whoa, that's a cool way to recycle a computer power supply. 

[Thomeeque]

Whoa, that's a cool way to recycle a computer power supply. 

Thanks, Paul

T.

[Paul Marossy]

You're welcome. I personally love to recycle stuff that other people would just throw away, for a variety of reasons.

[dschwartz]

what are those big brown stuff ? is that the transformer? if so, it looks like it´s from the XIX century