Cooling LEDs?....Bigger volume of metal, or bigger surface of metal? 3

[grigson]

Hello,

We have four narrow rectangular LED PCB's, each one comprising three 1W LEDs. These are double sided PCB's with thermal vias to the bottom layer.

Each PCB is mounted on one of the four faces of a vertical, solid aluminium bar...so we have light shining North, East, South & West from the aluminium bar, so to speak.

Now, unfortunatley, the LED PCB's totally cover the vertical faces of the aluminium bar, so unfortunately the only bare metal surface of aluminium that can contribute to heat transfer by convection is the top 1 inch square part of this vertical aluminium bar.

As you can see, this big aluminium bar is mostly a waste of time, since most of its vertical sides are covered by the LED PCB's themselves....which impedes convection from the metal sides of the aluminium bar.

Anyway, do you believe that a better heatsinking solution would be to simply remove the solid , vertical aluminium bar, and have the four LED PCB's standing vertical, with their copper clad bottom layers providing the convection surface?

(...i.e more metal surface area but far less metal volume)

(The bottom copper is obviously covered in solder-mask, as it must be to prevent oxidation of the copper......do you think that the solder mask layer will ruin the effective convection of heat from the bottom of these PCB's.?)

Is there any other material (other than solder mask) that we can cover the bottom copper layer with to aid convection more?....what about covering it with solder?,

-eg as in some EMS's will cover bare copper pads with solder (as in HASL = Hot Air Solder Layering).




Anyway, to summarize, for cooling LEDs, what do you choose.......A Big volume of metal, or much, much smaller volume of metal but more metal surface area?

 

[VEBill]

There are plenty of inexpensive 5 watt LED flashlights and 12-14 watt LED household light bulbs kicking around. It might be time for some 'gnireenigne'.

 

[mcgyvr]

Square tubing is NOT more expensive than solid bar stock.. In fact it should be roughly 1/2 the price..
Example..(Mcmaster Carr pricing)
1"x1" 6061 solid bar is like $11/ft
1"x1" (.125 wall) 6061 rect tube is $6.5/ft..

A "smart" design might actually allow the inside of the tube to be exposed/open to outside air on each end...

Are you using any thermal grease/pads on the mating surface of the bar/PCB? (That can help significantly too)

Thermal clad PCB's should be looked into also. (like the star mounts OperaHouse was talking about.

With proper engineering this should not be a tough challenge at all..
However its seems like virtual NO "engineering" has gone into this anywhere in the process. Do you even have a thermocouple?

 

[cranky108]

So what is the difference between LED cooling and that of power transistors?
My stero power amplifier had a solution that works, so if you lookup what they do for power transistors.

There should be solutions in the world for power semiconductors from 0.1 watt to several hundred watts. The industry already has solutions and I am willing to bet they have are economical for the given conditions.

 

[grigson]

-thanks

Sorry Operahouse, this product was designed by another engineer in the company who is now working on a different project.
-I was simply asked to change a transistor to a cheaper variety on the schematic, -but looked at the product, and noted the lack of "bare" , uncovered aluminium surface for heat convection....this worried me and hence this post.

The vertical , solid aluminium bar described above actually sits on top of a circular , horizontal, aluminium base. (Diameter about 4 inches and thickness about 1/8th inch.
-The vertical aluminium bar is screwed into this base. -There is no heat paste at the interface of the vertical Alu bar and the circular Alu bar.

Unfortunately, all of the top surface of the circular aluminium base is covered by a 1.6mm thick double sided PCB which houses the smps led driver circuit. (apart from obviously that bit that is underneath the vertical aluminium bar)

Anyway, as i said, the circular alu has a pcb sitting on top of it, so its metal surface has its convection path to air (internal air) impeded.

The bottom of the alu circular base is bare exposed metal, but it faces downwards, into the base of the product.


mcgyvr i find your post amazing where you say that hollow alu extrusion is cheaper than solid bar....but i'll check that out...as it is only cost that prevents us from suing hollow alu bar.

Operahouse i appreciate what you are saying about foil on a pcb not being sufficient, but i am asking.....is it better than being connected to a solid aluminium bar which is virtually completely covered by the LED PCB's?

I also appreciate suggestions of external heatsink, but again , that is more expensive.

Again, thermal paste/glue between LED PCB's and the alu bar woul dbe nice, but thermal grease products are hideously expensive.

VE1BLL...... "GNIREENIGNE" ... i must have had a long day, because it took me about 35 seconds to figure that out....good idea too.


* * * *
Is any reader going to be brave enough to say which is the best of a bad bunch.?....i.e...which of the following EVILS is best?.....

1. Keep the solid alu bar, with LED PCBs stuck to it and obscuring it.
2..Throw away the solid alu bar and let the copper clad backs of the PCBs do the heat convection

 

[grigson]

sorry forgot, by the way, there is NO thermal test data on this product. It has past EMC testing, but no thermal testing has EVER been done.

 

[mcgyvr]

Grigson,
The company you work for is doomed...get out QUICK
Now I remember you are the one that works at the place with the "engineers" (or luckyneers) who don't know why LED's in parallel is a bad move.

How can a company actually produce/design LED lighting fixtures and perform NO thermal testing.. What a joke...
Oh yeah that's right.. you just turn them on for 20 minutes and if they don't catch on fire out the door to a customer they go.

Please..oh please post your companies name so we all know to STAY CLEAR of that garbage..
And please tell me this is the same product with the parallel led strings..

I guess these fixtures don't get UL listed/recognized either..
Doomed...

 

[IRstuff]

1. Keep the solid alu bar, with LED PCBs stuck to it and obscuring it.
2..Throw away the solid alu bar and let the copper clad backs of the PCBs do the heat convection

Come on, no one besides you keeps harping on this. This "convection" you keep referring to is a myth, since you've already stated that the top is closed off.

Do the math on the aluminum. 0.1" thick ~3in width, 5in tall --> ~1.2°C/W So, even if you only heat sink the base, the aluminum core can conduct a sizeable amount of power away from the board. So, given that calculation, that explains why the original designers thought they could get away with, since the solid core would have even better performance.

A small SilPad is only about $0.36 in quantity, but you can certain machine the aluminum to make better contact with the thermal pads on the PCB.


TTFN
faq731-376

 

[IRstuff]

oops, swizzled a couple of numbers. It's more like 3.3°C/W

TTFN
faq731-376

 

[grigson]

IRStuff:

......there's still going to be convection into internal ambient....which transfers heat through the plastic & perspex enclosure to external ambient....i admit its not a great heat transfer path, but its all we have.

IRStuff

....When you say that, do you mean that we should make the base of aluminium so that we have aluminium with surfaces internal and external to the enclosure?


OK, so when i finally convince people here to do some thermal testing on the produict in its current state (we need to buy an environmental chamber first, and it'll likely be second hand, or more likely one likr you bake cakes in)....what are you predicting that the junction temperature will be if we set the oven to max ambient (40degC) and have it on max power (8W) with the enclosure completely fitted (except a tiny hole to bring out the thermocouples)

...so thats a question for us now, everyone's slated the design as it is.......so what then do you say the junction temperature will be?

 

[OperaHouse]

"we need to buy an environmental chamber first"

Kids these days. The real world isn't pretty or controllable. Go to the back room and get a cardboard bos. Cut a couple inch hole in it and slide a book over it. It should heat up and you will get some data. Maybe something will fail. Maybe the sky won't fall. Don't go to managment till you actually know something. We have all had our fun here. One thing I have learned is not to believe anything that is told to me. For all I know this could only consume 2W.

 

[IRstuff]

You continue to ask for answers to question for which you've failed to give parameters or drawings.

Perspex's thermal conductivity is about 1000 times worse than that of aluminum. Air's thermal conductivity is over 8000 times worse than aluminum's. You do the math.

TTFN
faq731-376

 

[grigson]

Thanks,

My math is in pounds and pence, and we have done it the cheapest way.

Our product range is for architectural use and often only used for one TV set, or for one poccasion/dinner/party etc.

..but still we'd like to have longevity in the product as well as their prettiness.

The power consumption is definetely 8W.

With the above set-up, it is going to be extremely difficult to find the junction temperature mathematically...

We'll have to just measure it.......anyway, all have be-rated the thermal design and rightly so..........but it would be nice to hear some junction temperature estimations for operation in bright sunlight and 35degC external ambient.

Are we talking:

80 degC ?
100 deg C ?
150 degC ?
200 degC ?

 

[mcgyvr]

1.21 gigawatts

 

[DRWeig]

In your last post, you mentioned "in bright sunlight" for the first time in this thread.

You've made a greenhouse, you know? Ambient inside your unventilated plastic box could reach 70°C.

Good on ya,

Goober Dave

Haven't see the forum policies? Do so now: Forum Policies

 

[Skogsgurra]

grigson, grigson...

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[LionelHutz]

And here's another little detail....

How to answer a question when the details are given in bits and pieces?

The foil board vs Al block question was answered on 6 Jun 12 4:53 by OperaHouse. If the foil won't work as a heatsink, then the block is the better solution between the 2.

As for the rest of the product. Just go do some testing already.

 

[OperaHouse]

While it seems unconsionable, you can run things really hot and still survive. Chinease designs have all taught us just how far you can get to the edge. With a big bar you can get the junction temp pretty close to the bar temp. The circuit board may prevent a lot of front radiation if it totally covers the bar. I imaging it does totally cover the bar because they probably edge solder the boards to each other to save money. I think some have overthought this. I kinda doubt automotive tail lights have a lot of heat sinking. I do wonder where the 40C environment spec came from.

 

[IRstuff]

40C is only 104F; That's not even that hot, in Vegas...

TTFN
faq731-376

 

[mcgyvr]

LED lifespan/lumen maintenance is directly related to junction temperature.. To achieve anywhere near 50,000 hours you need to keep LED junction temps below 80 deg C..