LM3886 Thermal Experiment (with data)

[tomchr]

This with color is interesting, you can paint it black if you want to lower down it's thermal resistance, but do not put excessive layer of color, just thin layer to become black colored (it is related to IR emission).

Black paint will add an insulating layer to the heat sink. If you want better IR emission, you need to have the heat sink anodized and dyed black.

~Tom

 

[tomchr]

Heat sink specified with for example 1°C/W or 1K/W have absolute same performance just our reference point are different.

Exactly.

~Tom

 

[benb]

I never knew there was such a 'controversy' over temperature notation (absolute vs. relative), but then the only temperatures I see are in weather reports and device datasheets. I would have suggested "delta Celsius degrees" which should be immediately understandable as well as distinguishable from an absolute temperature value, but no one asked me. I suppose you could likewise say a resistor has 28 "delta volts" across it.

...
"C degrees" although the Wiki Author claims as being non standard is a convenient way to write and speak temperature difference.

I wonder what the reference [28] has to say on it.

What's often as informative as the references is the talk page where these things are hashed out, often with even more references, and indeed there's this very discussion the talk page of that article:
Talk:Celsius - Wikipedia, the free encyclopedia

I often read through article talk pages (and even the edit history) and related articles ("transmission line" states you can ignore transmission line effects at cable lengths substantially below the shortest wavelength used, but "characteristic impedance" does not, as I mentioned in another thread). I also end up doing a lot more reading than posting on DIYaudio and other online resources.

 

[yu3ma]

@benb
I'm sorry for off-topic, but just to mention this very nice educational retro video from Tektronix about transmission lines, theory and some practical examples:
https://www.youtube.com/watch?v=I9m2w4DgeVk

 

[CristiGH]

well i dont know maths well but this worked over 20 minutes and didnt seem to get more hot ,at least to my measurments by hand , while chips back side was hotter then aluminium piece , aluminium part actualy got pretty warm,almost hot.

The epoxy used to protect the die is a better heat transfer material than air. Off-course given enough time the heat will get to the front side and be "sucked" out by the heatsink you connected. It might also reach an equilibrium depending on the thermal conductivity of the material, power dissipated by the die, material volume etc.
The heat will be drawn out of the front side of the package, even if less efficient, your method is better than no heatsink.
Thermal Resistance from die to back heat sink is waaaaay lower than die to front heatsink. But, the die to front heatsink Thermal Resistance is not that high as to not allow a front connected heatsink get hot. If the die is still within the maximum operating conditions then the process should work.
The kind of common sense that is telling you that it would help the chip to connect a front heatsink is biased by fact that you are judging from your point of sensing heat. If you keep your finger on the front of the chip and sense it is hot then yes, common sense would tell you that it would help. And as an absolute statement it helps. But in relation with the whole system, the kind of "help" in percent is so low that for practical purposes it can be safely ignored.
In an absolute sense, with the mention of perfect implementation, yes, it would help to have the chip surrounded in a perfectly connected good thermal conductivity material. For practical purposes it makes little sense if any.

There are many on this forum that have formal training. Some of them even excel at it. For them, precision of language and that of correct information transfer is of absolute importance. It comes with the territory. Yes, they understand what you are saying, but if you want to get all technical you should keep up, at least with the method of transmitting correct information.
Also a high degree of knowledge in a certain domain conveys authority in that area. If it is clear that you are lower in knowledge than the interlocutor you should at least respect/accept the other person's information. Especially when he/she is still relaying it to you! Lots of work went into that knowledge being correctly absorbed by such a person, and even if they at times come out as harsh/condescending it bears little importance in balance to what you get out of it. Get it?
Conclusion:
"Keep your head low and learn along the way until you can back up all of your claims with data".
That is the scientific way. And this is a scientific area. For philosophy (and also imprecision of language/information) you can hit the "audiophile snake-oil" forums.

 

[jneutron]

@benb
I'm sorry for off-topic, but just to mention this very nice educational retro video from Tektronix about transmission lines, theory and some practical examples:
https://www.youtube.com/watch?v=I9m2w4DgeVk

I loved the retro voice. I remember it from way back.

jn

 

[gazzagazza]

Don't accept what you can see is nonsense, even coming from a "manufacturer".

Read and sort what you see.

I have no trouble calculating thermal performance of heatsinks or the requirements for a given dissipation. I'm not sure the pedantry in your statements adds much. The industry widely uses °C/W as the unit of thermal resistance... being a practical person I just accept that and move on to make things work.

Attached example from Digikey heatsink listing.

 

[tomchr]

I have no trouble calculating thermal performance of heatsinks or the requirements for a given dissipation. I'm not sure the pedantry in your statements adds much. The industry widely uses °C/W as the unit of thermal resistance... being a practical person I just accept that and move on to make things work.

I think the "debate" of deg. C/W vs K/W has arisen because some people prefer to be very precise in their language (detail oriented, data driven) whereas others take a bit more "close enough, I understand what you mean" intuitive approach. It's just a difference in communication style, really...

Some vendors use deg. C/W, others K/W. I see them as functionally equivalent. It's just a matter of whether the freezing point of water or absolute zero is used as the system reference.

~Tom

 

[gazzagazza]

Tom I was just referring to Andrew's insistence on leaving out the degree sign on the degree C / watt usage.

 

[soongsc]

Yes. But it needs elaboration.

Copper has a better heat conductivity, and a specific heat capacity. As such, the velocity that the heat will take through the bulk material will be some value.

Aluminum has a lower heat conductivity, but it's heat capacity is quite a bit less than copper. So heat will travel faster through the aluminum.

If you compared two long rods, one of aluminum, one of copper...hold onto one end, and blast the other with a torch.... you will find that you have to let go of the aluminum first.

But in the end, the copper will transfer more heat..

The parametric being discussed is called "diffusivity".


Interesting, thank you for the information. The datasheet is very shy on actual information regarding the heat paths, so I'll have to do more research to figure out exactly what they are selling.



You'd have to measure the difference in heat flux through the top and bottom paths. That is very difficult to do accurately, as the instrumentation is generally beyond the DIY'er.


In the previous test technique (by yu3ma) of using ice water and partial immersion, I do not see how that can provide accurate results. The water is being heated by a constant power source, and does not care what the temperature of the chip is. Within the first 5 or 10 seconds of using that method, it will become impossible to discern the actual heat transfer coefficient.

jn

Thanks jneutron, you always have a straight and understandable explanation. Do you know there to look for more reference information about heat capacitance of materials?

 

[tomchr]

Thanks jneutron, you always have a straight and understandable explanation. Do you know there to look for more reference information about heat capacitance of materials?

Here's a bunch:
Heat capacity - Wikipedia, the free encyclopedia

~Tom

 

[soongsc]

Thanks, but it seems heat capacitance is larger for aluminium than copper? Now I am confused.

 

[tomchr]

You may be confusing heat capacity and thermal conductivity.

~Tom

 

[DPH]

Thanks, but it seems heat capacitance is larger for aluminium than copper? Now I am confused.

As far as we're concerned, volumetric heat capacity is the one to keep track of. Pound-for-pound, yes, aluminum has higher heat capacity. It's just MUCH MUCH less dense.

 

[soongsc]

Yes, at least according to the wiki page Aluminum 0.897 vs copper 0.385 for weight. As far as per volume goes, copper 3.45 vs Aluminium 2.422 Copper then does have a higher value.

So this does make some sense now. For the same external design, aluminium is a better heat sink. Thin copper might help spread the heat if a thin plating is used.

 

[DPH]

This is only heat capacity, remember. Thermal conductivity is another key component. A thick base does well here.

That said, you can buy an awful lot of aluminum for the cost of copper. That pragmatic effect swamps everything else.

 

[AndrewT]

........ For the same external design, aluminium is a better heat sink. ............

for the same external design, i.e. dimensions are all the same, then copper gives a better cooling effect, but it is much heavier and it is much more expensive.

If heating/cooling performance is critical, then copper is the preferred material. If you can manage with an aluminium heatsink then use that it saves weight and it saves money.

Look at some very high performance CPU coolers, they at very great cost have gone over to copper.
Look inside a power device. They use copper.

 

[soongsc]

For the same heat sink dimension, since copper has more heat capacitance, so it will take longer for it to reach a certain temperature. It will conduct heat away from the chip faster. Due to the larger heat capacitance, it will also take longer to cool down. The issue is comparing heat dissipation to air. How will different material effect this? What an interesting subject for integrated chassis design!

 

[AndrewT]

Due to the lower thermal resistance of the copper cf aluminium, the copper will be hotter farther away from the heat source. These hotter areas (the fins) will be capable of dissipating more power.

The copper is a better heatsink, than if made in aluminium.

 

[Mark Whitney]

Could this be correct?

Adding a block of copper to the front of an LM3886 could improve cooling because of the increased volume, surface area and thermal conductivity of the improved heatsink and less because of the extra heat transfer through the front of the device.