LM3886 Heatsink

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

I will soon have a stereo amp running with one LM3886 on each channel. They are powering one 4ohm speaker each and being ran from a +-28v supply.

The chips will likely be bolted to a thin sheet of aluminium (less than 1mm think) which will be bolted to the base of this heatsink:

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=180384944264&ssPageName=ADME:X:RTQ:GB:1123

My questions: is this heatsink going to be suitable, if not where is goot to look within the UK and how much with thermal transfer compound actually help?

Thanks :D
 
You have about 40 W power dissipation per chip. Thermal resistance for 25 C is 1.9 C/W. I wonder the CPU heatsink is sufficient. You can visit http://www.wakefield.com/ to get an idea for some CPU heatsink performance.

How many CPU heatsink do you use? How far does the heatsink from LM3886? Can you put 3886 on the heatsink surface? The aluminium sheet is too thin to be a good heat conductor.

Always use thermal junction compounds (thermal grease) even you have a huge heatsink.
 
The above heat-sink was made to cool this wattage of CPU thermal dissipation, but there is a catch. It was made to do so with forced cooling, provided by the fan attached to it. Most CPU coolers are pretty damn good, but they are rated to do their job with a fan attached.

Do not try to run 2 chips attached to one of those heat-sinks. It will not work. You can probably get away with it, using one per chip, without forced cooling. If there is reasonable airflow in the case (or outside of it, depends on how you're mounting them) it should be working reasonably fine.

As for attaching them to the heat-sink, with an intermediate step, it's probably worse, no matter what you do.

If CPU heat-sinks are getting your attention look for the following or similar:

- "Silent" heatsinks - Zalman, thermaltake, Akasa and the likes.
- Intel Xeon copper heatsinks
- Heatsinks where the fins have good gaps between them and can be mounted with the fins vertically ( similar to the second one you posted)
 
ormo said:
Would something like this be more suitable? It's 80mm x 80mm, not sure of the thermal resistance
If you are working with the T package, it will be too small. Consider Atilla's advice to add a fan.
Even the optimistic Overture Design Guide shows a possible design problem for the TF package due to its limited dissipation capability, and that is for a single channel.

What ambient temperature do you expect around the heatsink? Will the heatsink be mounted inside the case or with the fins sticking out?

ormo said:
I'm not allowed to edit posts
You can edit posts during a limited amount of time after posting.
 
I am currently using an old slot-A heat sink to cool two lm3886's with absolutely no problems. I do not run hot, and I don't press the power limits of the chips, but I have never had heat issues, even with a chassis that is not the best in terms of air flow.

I definitely second the motion that you should avoid tightly spaced heat sink fins for CPU coolers unless you are going with forced air cooling. Just glancing at the heat sinks you mentioned in your first post, for the cost, I would go with one sink per chip and you should be fine if you follow the mounting instructions everyone keeps repeating.

Good luck!
 
You can also mount the amp chips to an aluminum plate and then attach four or more CPU heatsinks to the plate.

The thermal conductivity losses with the added "spreader plate" should be more than offset by considerably more fin area.

Same construction rules apply: Thermal paste, clean mounting surfaces, use the correct screw torque when mounting the chips, and mount the CPU heatsink fins vertically for improved convection cooling.

Forced air (fans) will improve cooling efficiency, but can be noisy and require an additional +V supply.
 
dfdye said:
I am currently using an old slot-A heat sink to cool two lm3886's with absolutely no problems.
Do you use the T or the TF package?
Do you use ±28 V rails in combination with 4 Ohm speakers?
What is the ambient temperature around the heatsinks?

DCPreamp said:
The thermal conductivity losses with the added "spreader plate" should be more than offset by considerably more fin area.
It does not work that way. If the spreader plate cannot transport the heat away from the IC fast enough, the additional fins make no difference.

With the heatsink right on top of the IC on the spreader plate, there is little loss. In that case the spreader plate will rather improve the thermal resistance due to its additional dissipation. If the heatsinks are in different locations, the spreader plate must be thick enough to carry the heat away from the IC.
 
Ormo,
why are you looking at fan cooled CPU sinks?

Have you looked at the 3886 datasheet where it tells you the size of sink needed to run the chip at it's maximum temperature?
Go to page14.
Find 56Vdc (=+-28Vdc) and go up to the intercept with the 4ohm plot.
Go across to the Ta=25degC. It shows that one 3886 needs 1.9C/W to run at maximum temperature.
To run a pair of 3886 you need 0.95C/W.
To run a pair at a sensible operating temperature you need~0.5C/W

Now go and search for 0.5C/W passive sinks and they are big.
If it runs inside the case you may need to use Ta=40degC or even higher. This makes your sink requirement even bigger.

Download that datasheet and start reading.
 
Originally posted by pacificblue Do you use the T or the TF package?
I used the "normal" chips with metal tabs (I think the T)--not the insulated ones (I think those are the TF, but I could have this backwards.) I used the typical setup with a mica insulator and some thermal paste since I didn't have silicon insulators that were large enough for these chips.
Do you use ±28 V rails in combination with 4 Ohm speakers?
My speakers that I typically run with this amp are ~6 ohm, but I have tested the amp with 4 ohm speakers with no problems. My rail voltages are actually closer to 26 V.
What is the ambient temperature around the heatsinks?
Do you mean what is the temperature of the heat sinks when I am running? If so then it is warm, but not terribly hot. If you are really talking about the ambient temperature of the air around the heat sink, then it is slightly above room temperature, probably around 80 degrees. I never saw a need to measure temperatures any more than that.

The great thing about the 3886's is their internal thermal shutdown if something does go wrong, like overheating. I typically don't run my speakers that hard (this chip isn't really designed to be a paint peeler), so if you are really planning on pushing a set of 4 ohm speakers HARD you will probably need a different amp design, not a different heat sink.

Here is my take: If you can't cool a dual mono LM3886 amp with a slot A heat sink, you are trying to get too much power out of it and should switch to a different amp design that has more headroom. If you to then a) it will sound better since you won't be pushing the limits of the chip and b) you will have to cool it differently ANYWAY since it should be a higher power amp.

Just my $0.02
 
AndrewT said:
Now go and search for 0.5C/W passive sinks and they are big.
If it runs inside the case you may need to use Ta=40degC or even higher. This makes your sink requirement even bigger.

Download that datasheet and start reading.
Andrew, you are killing me! I am about 99% sure you have built amps with these chips before, but you are recommending massive heat sinks when they really aren't warranted. You and I both know that reasonable operating temperatures can be achieved with smallish heat sinks! There are a ton of examples of heat sinks being the aluminum chassis in which amps were built!!

Math is great for working in theory world, but practically, National just covered their tails when providing these guidelines.

Again, if you are actually trying to run these chips at their max output, you have the wrong amp design for your system.

PS Here is a bad picture of the heat sink in the amp that I am talking about. It has been running for a few years now with no issues. It may not be the pretties thing on the block (it was actually my first homemade amp and was built from spare parts I had lying around), but it has had no issues and sounds really nice. http://www.diyaudio.com/forums/attachment.php?s=&postid=919642&stamp=1147980386
 
a) I have just finished 2 stereo amps, each using 2x 3886s for a pair of 4 ohm speakers with +/- 25V rails. They have been running for a month for continuous periods of up to 6 hours.

b) I have a room of 20 m2 and I ran them at normal music listening levels.

c) For this first month, I used lab PSUs just to be safe, with voltage and current indicators and variable current limits.

d) As measured by the lab PSU, I never need to supply more than 0.2A continuous for 2x LM3886s. The transients are supply by local decoupling caps.

e) In this one month, I have never tripped the lab PSU current limit set at 0.6A.

f) I use ONE heatsink for both LM3886s, measuring 150x100x25mm. It is not even warm, at thirty something degrees Celcius.

This is my experience. Fresh out of the box.

You may also wish to have a look at Peter Daniel's Patek amp, and decide for yourself whether his housing design would dissipate 10Ws without getting warm. A friend has bought a Patek and uses it as kitchen radio all day. It is not even warm to touch. Hardly over thirty degrees.


Cheers,
Patrick
 
dfdye said:
Math is great .....................National just covered their tails when providing these guidelines. .................... it has had no issues and sounds really nice. http://www.diyaudio.com/forums/attachment.php?s=&postid=919642&stamp=1147980386
Dfdye,
National did not cover their tails in the manner you suggest.
They show the chipamps with the data to make their product look as good as they can. They specify Ta & Tc =25degC. They specify Tj = 150degC. They specify that Rth c-s =0.2C/W.
We can never and should never assume these extreme values for our practical amplifiers.
The maths allows us to model sensible operating conditions.
It is not just theory.

Now let's take an 8r0 example.
Look at the graph of dissipation vs output power of the 3886 chipamp on page14.
Note, that at maximum output power (85W into 8r0) the Dissipation (33W) is less than maximum (47W).
Note, also that at quarter of maximum output power the chipamp is dissipating ~ the same dissipation as at maximum power. i.e. @ -6dBWmax the dissipation >= dissipation at 0dBWmax.
Now look at -12dbWmax, 5.3W output. The dissipation has fallen to ~30W.
Even @ -20dBWmax (0.85W into 8r0) there is still significant chip dissipation. I would guesstimate that National are showing between 5W and 15W of chipamp dissipation.

Now if you are playing music at -20dBWmax average level with regular peaks rising to -10dBWmax then the chip is operating at dissipations of ~5W to 33W.
This heat must be got rid of.

National show us how to calculate the minimum size of heatsink to get rid of the maximum dissipation using all the most advantageous figures to make their amps look economic.
They also go on to show how to adjust the calculations for more reasonable operating conditions.
We can see from the last paragraph that even when average levels are -20dbWmax that high power must still be dissipated. On reactive loads, that are all speakers without exception, the heat dissipation is even higher.
Now to be practical we should use sensible figures for our actual installation.
Ta<=30degC in temperate climate with an external heatsink,
Rth c-s ~=0.6C/W,
Tj<=100degC,
all seem a reasonable set of operating conditions.
I advise that rather than calculating for these practical and reasonable operating conditions that you adopt a simplification of using double the minimum size that National show in page14.

If you have not made your self aware of what is required and cannot read what National tell us then you are hardly qualified to specify real heatsink requirements.
Yes, do tell us what you have found works for you, but please do specify the whole set of operating conditions so that any enquirer can weigh up your case with his/hers and then go on to make that informed decision.

Finally,
you show a two channel chipamp heatsink.
I can confirm that for my operating conditions and finding the National recommendation and then doubling it, I end up with a sink that looks very similar to what you show.
 
AndrewT said:
If you have not made your self aware of what is required and cannot read what National tell us then you are hardly qualified to specify real heatsink requirements.
Yes, do tell us what you have found works for you, but please do specify the whole set of operating conditions so that any enquirer can weigh up your case with his/hers and then go on to make that informed decision.
This will be my last comment since we seem to be pretty much on the same page here and are just arguing approaches.

Andrew, I am perfectly capable of doing the math. My "real" job involves building chemical instrumentation. My contribution here was an attempt to relate what I have had success with to THIS particular question--do computer heat sinks work for an LM3886 amp? I have stated over and over again that if you can't effectively sink your amp with one standard cpu sink per chip, then you probably need a different amp design rather than a bigger heat sink.

I am NOT offering exact specifications for sinking chips operating at maximum operating currents, and I am NOT implying that one cannot effectively calculate that! I AM saying that there have been numerous examples in which much smaller than recommended heat sinks have been used for LM3886's quite effectively. I am basing recommendations off of empirically gathered data, and you are basing your recommendations off of math (albeit with some quite reasonable, common sense conditions) from the data sheet.

So, that being said, I'll ask you, Andrew, a direct question that actually relates to the original topic: Will one of the heat sinks described in the original post effectively cool one chip of an amp? Isn't that the question we are trying to answer? My answer, as I have said, is to put one sink per chip. I'll leave the rest of the discussion to you and others.

David
 
but you are recommending massive heat sinks when they really aren't warranted. You and I both know that reasonable operating temperatures can be achieved with smallish heat sinks!
seems very different from
we seem to be pretty much on the same page here

I do not think that fanned CPU cooler will be adequate in the no fan condition for the normal use that we put our chipamps amps to, i.e. powered to be able to produce ~60W into our chosen speaker load with sensitivity<=89dB/W/m listened to at 2.5m distance and at full summer temperatures.
The national Page 14 table specifies 1.9C/W and I say double that to ~1C/W. That sink will not achieve 1C/W without a fan.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.