i have an extruded aluminium case that i wish to mount a pair of the insulated lm3886 chips to. im concerned about thermal conductivity and overall heatsinking performance.. i have no data as to the capacity of the case to shed heat..
its this case:
Little Power Amplifier Aluminium Chassis 03 [chassis03] - $75.00 : DAC,Audio Amplifier,Hifi Amplifier,Tube Amplifier,Hlly Audio, Hlly Electronics
i already have it. (it doesnt have the vents on the top as shown there)
my issue (apart from overall heat shedding ability) is the base of the case is only 2mm thick, not the 3 indicated,and im worried its not enough metal under the chip to transfer the heat to the rest of the chassis.
question is, would it be better to bolt directly to this, with a good heat transfer compound, or to first mount the chips to say a 1cm thick aluminium plate, and then mount that to the case floor?
the plate could be large, say 10cmx10cm, so it would have a large thermal contact area with the case, but it would of course mean 2 thermal junctions instead of 1 between the case and the chip.
also is there much benefit to adding an aluminium plate across the tops of the chips too, with a small "s" profile so it steps down to bolt to the case too? then id be transferring heat away from both sides of the chip..
i know this should all be calculated, but without measurements for the case it seems hopeless..
any tips/ experiences most appreciated.
its this case:
Little Power Amplifier Aluminium Chassis 03 [chassis03] - $75.00 : DAC,Audio Amplifier,Hifi Amplifier,Tube Amplifier,Hlly Audio, Hlly Electronics
i already have it. (it doesnt have the vents on the top as shown there)
my issue (apart from overall heat shedding ability) is the base of the case is only 2mm thick, not the 3 indicated,and im worried its not enough metal under the chip to transfer the heat to the rest of the chassis.
question is, would it be better to bolt directly to this, with a good heat transfer compound, or to first mount the chips to say a 1cm thick aluminium plate, and then mount that to the case floor?
the plate could be large, say 10cmx10cm, so it would have a large thermal contact area with the case, but it would of course mean 2 thermal junctions instead of 1 between the case and the chip.
also is there much benefit to adding an aluminium plate across the tops of the chips too, with a small "s" profile so it steps down to bolt to the case too? then id be transferring heat away from both sides of the chip..
i know this should all be calculated, but without measurements for the case it seems hopeless..
any tips/ experiences most appreciated.
I have a commercial amp using similar chips. It has a 800W rating. The chassie is 19", 3 units high, and the size, vents and fans really helps to keep it cool. The box you have is best suited for preamps, dacs, or t-amps. Can you even fit a heatsink in there? That small air vent will not be enough.
Last edited:
Although your plate might be 10 x 10, that is nothing compared to the surface area of even a modest heatsink.
I read Katie's 10x10 comment first and assumed 10" x 10" and agreeing that it is not large in terms of heat dissipation capability.
But then read post1 to discover 10cm x10cm. That is tiny !!!!!
But then read post1 to discover 10cm x10cm. That is tiny !!!!!
I would suggest having a read of Rod Elliot's pages to understand how a heatsink works.
1) You've got to get the heat into the atmosphere (physical area).
2) You've got to get the heat away from the chip (physical design).
If you have a look at a lot of heatsink profiles you will notice that they are thicker in the middle than they are at their extremes, this is to help the heat flow into the radiating fins.
YES, you could get away with a thick plate if it were mounted with additional fins but you've got to make sure that the heat is being effectively transferred from the source to the atmosphere. Mounting a heatsink inside a sealed box is next to having no heatsink at all. You can use smaller heatsinks and smaller boxes if you use a fan to blow the air across the fins. BUT, ultimately the air must be exchanged with cooler air.
1) You've got to get the heat into the atmosphere (physical area).
2) You've got to get the heat away from the chip (physical design).
If you have a look at a lot of heatsink profiles you will notice that they are thicker in the middle than they are at their extremes, this is to help the heat flow into the radiating fins.
YES, you could get away with a thick plate if it were mounted with additional fins but you've got to make sure that the heat is being effectively transferred from the source to the atmosphere. Mounting a heatsink inside a sealed box is next to having no heatsink at all. You can use smaller heatsinks and smaller boxes if you use a fan to blow the air across the fins. BUT, ultimately the air must be exchanged with cooler air.
hi.. i had assumed that the 18x30x6 cm aluminium case would provide a reasonable amount of heatsinking capability.. ive even seen pictures of a 3886 amp in one of these cases with the chips bolted to the floor of the case.
http://bluegourd.img.jugem.jp/20110329_2027219.jpg
i am concerned about dissipation, and from your replies it sounds like im right to be.. but since i already have the case id like to try.. they will be running an 8 ohm load and i dont really crank the volume very much.
i was basically wondering if adding a thicker plate between the chip and the 2mm thick case would aid in dispersion of the heat, since at 2mm thick id likely end up with a hotspot under the chip. however with a plate between chip and case id have 2 layers of thermal compound to get through instead of one (although the junction between the plate and the case would be large so have lower thermal resistance)
if this is a no-go i have 6 lovely big copper heatsinks from rackmount servers that i could fashion into a custom case with some sheet aluminium..
- on a related note how to clean up copper heatsinks so they are shiny?
http://bluegourd.img.jugem.jp/20110329_2027219.jpg
i am concerned about dissipation, and from your replies it sounds like im right to be.. but since i already have the case id like to try.. they will be running an 8 ohm load and i dont really crank the volume very much.
i was basically wondering if adding a thicker plate between the chip and the 2mm thick case would aid in dispersion of the heat, since at 2mm thick id likely end up with a hotspot under the chip. however with a plate between chip and case id have 2 layers of thermal compound to get through instead of one (although the junction between the plate and the case would be large so have lower thermal resistance)
if this is a no-go i have 6 lovely big copper heatsinks from rackmount servers that i could fashion into a custom case with some sheet aluminium..
- on a related note how to clean up copper heatsinks so they are shiny?
You shouldn't have layers of thermal compound. It should be almost transparently thin, it's only there to iron out any air pockets between the two metal mounting surfaces. In theory, if the two surfaces were mirror perfect you wouldn't need goo at all.
The LM3886 is a Class A/B device so you can try it at low volume as long as you have sufficient dissipation to get rid of the residual dissipation.
ok now im quite confused.. from the posts here it seems a fairly substantial heatsink is required.. however the documentation from the chipamp.com kit i am awaiting states that the kit does not require large heatsinks and a 3" x3" x 1/2" aluminimum plate in free air should be sufficient.
surely a 30cm x 16cm x 6cm aluminium extrusion (the case in question) is superior to this spec? i appreciate only the external surface is in free air but its still several times more surface area than both sides of a 3" x 3" plate.
i guess ill just have to find out.. better than spending loads of time and brainpower considering alternatives if i dont really need to.
surely a 30cm x 16cm x 6cm aluminium extrusion (the case in question) is superior to this spec? i appreciate only the external surface is in free air but its still several times more surface area than both sides of a 3" x 3" plate.
i guess ill just have to find out.. better than spending loads of time and brainpower considering alternatives if i dont really need to.
download the data sheet for the LM3886 and have a read, it gives a fair bit if information on the dissipation you can expect as a function of load impedance and supply voltages.
A 2mm thick base is capable of dissipating a modest amount of power. I am not initially inclined to feel that adding a 1cm thick spreader will do much, as the issue you face is the surface area dissipating power into the air. Given the primary surface you are putting power into is the base, there is not actually going to be a lot of airflow across that! So it will be pretty inefficient.
Without a sausage of analysis, so beware, I would say that a case of that size can probably dissipate something in the region of 20 watts or so before things start getting crazy hot. Dont ask me to justify this, it is gut feel from experience - and I have surprised myself with my stupidity before.
What am I saying then?
If it were me trying to shoehorn a pair of LM3886's into such a modestly sized case, and to use the case as the heatsink I would:
- Set my expectations very carefully
- Design the amplifier such that it will not get into a situation where it gets too hot. Do this by using a low supply voltage
- I would use probably +/- 20 volts
- Put a thermal cutout on the base (though these chips do have protection).
- Think long and hard about using a different case!
A 2mm thick base is capable of dissipating a modest amount of power. I am not initially inclined to feel that adding a 1cm thick spreader will do much, as the issue you face is the surface area dissipating power into the air. Given the primary surface you are putting power into is the base, there is not actually going to be a lot of airflow across that! So it will be pretty inefficient.
Without a sausage of analysis, so beware, I would say that a case of that size can probably dissipate something in the region of 20 watts or so before things start getting crazy hot. Dont ask me to justify this, it is gut feel from experience - and I have surprised myself with my stupidity before.
What am I saying then?
If it were me trying to shoehorn a pair of LM3886's into such a modestly sized case, and to use the case as the heatsink I would:
- Set my expectations very carefully
- Design the amplifier such that it will not get into a situation where it gets too hot. Do this by using a low supply voltage
- I would use probably +/- 20 volts
- Put a thermal cutout on the base (though these chips do have protection).
- Think long and hard about using a different case!
Low Power supply rail voltage gives less heat in the chip.
High load impedance gives less heat in the chip.
Low SPL replay levels gives less heat in the chip.
You are in control of all these variables.
Look up the 3886 datasheet. It details heatsink design.
BUT !!!
I recommend you double the heatsink requirement that National guide you to.
i.e. if you put in your data and National specify 8C/W for one channel, then use 2C/W for two channels. A 2mm thick steel plate sitting face down on an insulating surface is not going to get anywhere near 2C/W
Standing the plate upright so that air can flow up the surface will perform much better.
High load impedance gives less heat in the chip.
Low SPL replay levels gives less heat in the chip.
You are in control of all these variables.
Look up the 3886 datasheet. It details heatsink design.
BUT !!!
I recommend you double the heatsink requirement that National guide you to.
i.e. if you put in your data and National specify 8C/W for one channel, then use 2C/W for two channels. A 2mm thick steel plate sitting face down on an insulating surface is not going to get anywhere near 2C/W
Standing the plate upright so that air can flow up the surface will perform much better.
ok i have one final idea that doesnt involve me ditching the case..
if i were to bolt the chips to the underside of the top of the case, and securely bolt one of these on top of the case:
Fischer Elektronik SK-44-200-SA Extruded Heat Sink from Conrad.com
it would almost entirely cover the top of the box and look quite neat.
that looks highly oversized (0.9 deg/watt) but i dont know how much performance would be lost having the 2mm aluminium plate of the case between the chips and this bad-boy.
i promise ill soon give up and get a new case.. its just it has all connectors, toroid, power sw, volume pot etc already done.. its from a dead amp which im supposed to be "repairing" getting a new case will basically mean its a new amp and not a repair job.. a much harder sell to her next door.
if i were to bolt the chips to the underside of the top of the case, and securely bolt one of these on top of the case:
Fischer Elektronik SK-44-200-SA Extruded Heat Sink from Conrad.com
it would almost entirely cover the top of the box and look quite neat.
that looks highly oversized (0.9 deg/watt) but i dont know how much performance would be lost having the 2mm aluminium plate of the case between the chips and this bad-boy.
i promise ill soon give up and get a new case.. its just it has all connectors, toroid, power sw, volume pot etc already done.. its from a dead amp which im supposed to be "repairing" getting a new case will basically mean its a new amp and not a repair job.. a much harder sell to her next door.
From a practical point of view, your idea will work if :
1) you stand the case on tall rubber legs, 15/20mm will be fine, so air can really move away.
2) you build a reasonable amp, such as, say, 40+40W/ch (+/-28/30VDC rails, 8 ohms speakers) and listen to recorded music at home (not DJ , MI or PA duty).
In fact, you will listen to loud music (within home parameters) and you won´t feel it more than warm to the touch.
Anyway, add a 75ºC cutoff bolted to the heatsink, which opens power line (110/220V) and turns everything off.
Don´t trust just the internal chip thermal cutoff.
I´ve built and sold thousands of Musical Instruments amplifiers with similar cooling system (transistors bolted to aluminum chassis, with an extra heat spreader aluminum sheet in between).
An MI duty is infinitely more grueling than home use.
1) you stand the case on tall rubber legs, 15/20mm will be fine, so air can really move away.
2) you build a reasonable amp, such as, say, 40+40W/ch (+/-28/30VDC rails, 8 ohms speakers) and listen to recorded music at home (not DJ , MI or PA duty).
In fact, you will listen to loud music (within home parameters) and you won´t feel it more than warm to the touch.
Anyway, add a 75ºC cutoff bolted to the heatsink, which opens power line (110/220V) and turns everything off.
Don´t trust just the internal chip thermal cutoff.
I´ve built and sold thousands of Musical Instruments amplifiers with similar cooling system (transistors bolted to aluminum chassis, with an extra heat spreader aluminum sheet in between).
An MI duty is infinitely more grueling than home use.
If you mount the LM3886's to the inside of the lid of the case and bolt those heatsinks to the outside, using some heatsink grease of course, then you will have heaps of cooling. having the lid of the case in between the LM3886 and the heatsink will not be a problem at all provided everything is nice and flat.
3" x 3" x 0.5" is a reckless recommendation. They obviously didn't do the math on that one...
I've been writing a page on the LM3886 for a while now. I still need to fill out some of the sections, but the one on thermal management is done. You can have a look here: Neurochrome.com : : Audio : Taming the LM3886 Chip Amplifier
~Tom
...and not painted or coated in any way. If the chassis is painted or powder coated, you're better off cutting holes in the chassis and either use a heat spreader or poking the LM3886es through the holes directly.
~Tom
Have you tried looking at the datasheet to find out how much heat you need to get rid of or are you just guessing ?
Ive seen the datasheet and now the very informative neurochrome site (thanks tom) and both seem to suggest extremely large heatsinks are necessary... much larger in fact than i see on most of the 3886 amps ive looked at.
F.e.
http://www.audiosector.com/chassis_patek_amp.shtml
Im sure the maths is spot on, but i wonder how a real world music load compares to a sine wave constant signal.. id imagine average load would be much lower.
In any case if i get a taller heatsink of the same size, 40mm instead of 20, finer fin pitch..i can get close to the suggested .04 deg/watt mentioned on the neurochrome page.
I also wonder if it would be cheaper/as effective to buy some uninsulated 3886's than a larger heatsink.. ill actually have to do some maths i guess 😉
Appreciate all the guidance here..
F.e.
http://www.audiosector.com/chassis_patek_amp.shtml
Im sure the maths is spot on, but i wonder how a real world music load compares to a sine wave constant signal.. id imagine average load would be much lower.
In any case if i get a taller heatsink of the same size, 40mm instead of 20, finer fin pitch..i can get close to the suggested .04 deg/watt mentioned on the neurochrome page.
I also wonder if it would be cheaper/as effective to buy some uninsulated 3886's than a larger heatsink.. ill actually have to do some maths i guess 😉
Appreciate all the guidance here..
I posted a report recently detailing the approximate temperatures of a pair of 3886 chipamps on a 1C/W sink while delivering ~2W and ~1W to resistor test loads.
They got too hot to touch for longer than a couple of seconds.
I now have an IR temp gun and could repeat the test.
Now go back to the datasheet and look at the power dissipation when the chip puts out 10W. Can you estimate the dissipation when the output is 5W and 2.5W?
Those datasheet figures are for resistor loading.
ALL ClassAB amplifiers run hotter when driving a reactive load (your speaker).
That's partly why I recommend doubling Nat's heatsink advice
there is obviously something (s) fundamental wrt. heatsink specifications that i dont understand.. probably a newbie thing but surely a 1C/watt heatsink should only rise in temperature 1 degree with a 1 watt load? obviously im missing something big if a heatsink that size was too hot to touch when the 3886 was barely ticking over.
back to school for me.
back to school for me.
- Status
- Not open for further replies.