Getting heat out of a box

[markp]

Hi All,

I'm looking for a company that could help solve some thermal management
issues. Basically I've got a design that dissipates about 30W and need to
encase it in a sealed box (not hermetically, but to all intents and purposes
there can be no airflow through the box).

Current thoughts are an aluminium extruded box with large heatsinks on the
outside, and a method of thermally bonding or piping any higher powered
components directly to to the box inside. I also need an AC-DC supply which
I think needs to be a baseplate cooled type and bond that directly to the
side of the box that has the heatsink. I'm unsure of size of box needed or
even whether this is a viable solution.

I know it's a bit vague, but anyone know a company who could help
(preferably one with a vested interest in solving it, like an aluminium
extrusion box manufacturer)?

Thanks!
Mark.

 

[Joerg]

Hi All,

I'm looking for a company that could help solve some thermal management
issues. Basically I've got a design that dissipates about 30W and need to
encase it in a sealed box (not hermetically, but to all intents and purposes
there can be no airflow through the box).

Current thoughts are an aluminium extruded box with large heatsinks on the
outside, and a method of thermally bonding or piping any higher powered
components directly to to the box inside. I also need an AC-DC supply which
I think needs to be a baseplate cooled type and bond that directly to the
side of the box that has the heatsink. I'm unsure of size of box needed or
even whether this is a viable solution.

Sounds like the only viable approach, short of oil-filling the whole
thing. Try to do your best to reduce dissipation in the electronics. I
suggest to avoid electrolytic capacitors or similar components where the
MTBF significantly drops with rising temperature.

The AC-DC power supply presents a challenge here. Not because you can't
get enough efficiency but because it requires at least one electrolytic
capacitor. Any chances to at least send in DC, at whatever voltage?

I know it's a bit vague, but anyone know a company who could help
(preferably one with a vested interest in solving it, like an aluminium
extrusion box manufacturer)?

I know a company (client of mine) that can probably do the heat transfer
simulations but it's not going to be cheap. Would be strictly as an
engineering service, they don't make boxes. I don't think a box makers
would do that or even have the SW tools and engineering know-how to do that.

If interested send me an email. Clicking on the reply-to address works.

 

[markp]

Sounds like the only viable approach, short of oil-filling the whole
thing. Try to do your best to reduce dissipation in the electronics. I
suggest to avoid electrolytic capacitors or similar components where the
MTBF significantly drops with rising temperature.

The AC-DC power supply presents a challenge here. Not because you can't
get enough efficiency but because it requires at least one electrolytic
capacitor. Any chances to at least send in DC, at whatever voltage?



I know a company (client of mine) that can probably do the heat transfer
simulations but it's not going to be cheap. Would be strictly as an
engineering service, they don't make boxes. I don't think a box makers
would do that or even have the SW tools and engineering know-how to do
that.

If interested send me an email. Clicking on the reply-to address works.

Thanks Joerg. I'm going to do a bit of hunting first, if I can get into the
right ballpark with an aluminium extrusion manufacturer and get some sample
boxes I may even protoype something using dummy resistors and do some
temperature profiling myself. This might not catch hotspots in the real
thing of course, but it'll give me a good idea of how far off a solution it
might be overall. Thanks for the offer, I may take you up on it later

Mark.

 

[Joerg]

Thanks Joerg. I'm going to do a bit of hunting first, if I can get into the
right ballpark with an aluminium extrusion manufacturer and get some sample
boxes I may even protoype something using dummy resistors and do some
temperature profiling myself. This might not catch hotspots in the real
thing of course, but it'll give me a good idea of how far off a solution it
might be overall. Thanks for the offer, I may take you up on it later

I use the Dale RH-50 series for that a lot. You can even get resistors
in TO-220, should nicely mimic power transistors and large Schottkys:

www.irctt.com/file.aspx?product_id=259&file_type=datasheet

Another method that easily lets you "dial in" the dissipation without
needing a large variety of power resistors is to take a bunch of LM317
and then use them as constant current sources into ground. It takes only
one (smaller) resistor to set the desired current and thus dissipation.
The LM317 comes in TO-220 and all sorts of other packages, including
some that can sink heat into a chunk of copperclad so you can mimic a
circuit board that heats up.

 

[markp]

If the box has >30 square inches of external free-air surface, a
circulating
fan inside the box would do it. Or, does it have to be small?

Do you mean and area of the sidewalls of the box externally > 30 square
inches (that's equavalent to 5.48" x 5.48" for ease of visualisation), with
an internal fan, will get rid of 30W? What's the max temp rise above ambient
internally do you think, and would I need to put heatsink fins inside the
box as well as outside?

The box is going to need to be big for the components, probably 12" x 4" x
3" external dimensions or so. That's a good 120 square inches available if
you ignore the base and the two end plates. Those other free-air surfaces
can also have quite large surface area heatsinks on them too.

Having said that, I would really like it to be rugged and not have any
moving parts in it at all. What do you think the chances I could keep all
internal temps below, say, 20 degrees above ambient if I were to bolt the
high powered stuff directly to the box walls inside? The 30W is spread out a
bit component wise, but 20W of that comes from the PSU.

Mark.

 

[markp]

So if we understand you, you have a 10 watt load and a power supply that
is
loosing 20 watts in the conversion to DC?

That doesn't sound too "green" to me.

tm

No, this design is actually a custom power supply. The total heat
dissipation inside my box is 30W, there's an AC-DC converter in there which
is dissipating 20W of this, and is powering another circuit internally which
is dissipating the remaining 10W. This additional circuit provides an
output from the box and delivers power to another box. The overall
efficiency of my box is about 70% (series combination multiple of the
efficiencies of the AC-DC supply and my additional circuit), so basically it
is consuming 100W and delivering 70W to another box.

Mark.

 

[markp]

Hi All,

I'm looking for a company that could help solve some thermal management
issues. Basically I've got a design that dissipates about 30W and need to
encase it in a sealed box (not hermetically, but to all intents and
purposes there can be no airflow through the box).

Current thoughts are an aluminium extruded box with large heatsinks on the
outside, and a method of thermally bonding or piping any higher powered
components directly to to the box inside. I also need an AC-DC supply
which I think needs to be a baseplate cooled type and bond that directly
to the side of the box that has the heatsink. I'm unsure of size of box
needed or even whether this is a viable solution.

I know it's a bit vague, but anyone know a company who could help
(preferably one with a vested interest in solving it, like an aluminium
extrusion box manufacturer)?

Thanks!
Mark.

Well everyone, I've found this site which if I can learn to drive it might
be really useful. It's got some seriously interesting stuff in there,
including online calculators:
http://www.frigprim.com/

Mark.

 

[markp]

Something like this?

ftp://jjlarkin.lmi.net/L650_HS_proto.JPG

ftp://jjlarkin.lmi.net/L650_HS_pretty.JPG

John

Similar to that John, but with much larger heatsinks on both side of the box
along the full length of the box (unless I can get away with a large
heatsink on the top, but then there's going to be very little natural
convection so probably inefficient as it'll be mostly by radiation).

As a matter of interest, do you know how much power those hestsinks
dissipate, or roughly the temperature rise internally?

BTW the environment will just be still air externally, I can't rely on any
air movement other than natural convection.

Mark.

 

[Ian Iveson]

Configure a "endless" tunnel (with internal fins), fan
driven. Fins
on the outside. And pray a lot ;-)

Or thermally bond the culprit components to the top, attach
the potential victims to the bottom, stop air from
circulating inside, and blow on the outside if necessary.
That'd also relieve the inside from the heat generated by
the fan. Depends on how effectively the hot bits can be
heatsunk to the top extrusion, and how much the circuit is
open to layout changes, but generally it would be better to
isolate the heat generators from the bits that don't like
heat, no?

Just to be contrary.

Different strategies, horribly complicated situation for
analysis. Experienced thermal engineers are probably making
good money.

Ian

 

[Grant]

Something like this?

ftp://jjlarkin.lmi.net/L650_HS_proto.JPG

ftp://jjlarkin.lmi.net/L650_HS_pretty.JPG

John

On a smaller scale, this photo is of the inside of box bits from
a ~150W power converter.

The mains side power switch heat spreader to the left, the other
heat spreader is for the secondary TO247 dual schottky rectifier.

Insulation is that non-woven 0.4mm thick paper like stuff. The
aluminium plates are 2.5mm thick. One on left measures 86 x 56mm.

A screw each side of the power device applies pressure through an
elastic pad (blue, dunno what it's made of) and PCB to the TO220
or TO247 case onto the heat spreader.

http://grrr.id.au/image/heat-spreader-and-insulator.jpg

The box was lightly finned extruded channel with sheet metal ends.

http://grrr.id.au/image/heat-spreader-and-insulator-outside.jpg

No fan, sealed box but not airtight. Made in Taiwan.

Grant.

 

[legg]

Scary. A square inch of surface in still air can have theta over 100
K/W.

John

~160, actually.

RL

 

[legg]

Sounds scary. A square inch of surface in free air can have theta of
over 100 k/w. And that's just the outside of the box... ther would be
a similar theta on the inside.

Actually, in a situation where the major dissipators are more closely
coupled to the box wall, the box wall temperature can become a fairly
accurate indicator of internal air temperatures. Dissipation from
other isolated internal sources into this air has to be restricted
simply due to the increased internal anbient. The result is that the
sealed box wall becomes the dominant regulator, however backwards this
may sound.

RL

 

[markp]

Well, back to the question at hand, internally the OP should mount the
hot parts to the box walls, and externally he needs a heatsink that
gets him to ~1C/W.

Even better than getting the heat out of the box is... not putting so
much in in the first place.

I think that is what I'm intending to do. I'll mount the higher powered
stuff directly to the side walls, and probably use an aluminium backed PCB
thermally coupled to the side walls too so that lower powered devices have a
thermal path to the heatsinks.

The AC-DC converter will be a baseplate cooled type mounted probably
directly on one of the walls (or find some way of thermally connecting it to
both walls with a thick bracket maybe and mount it horizonaally).

Mark.

 

[Tim Williams]

Any others data points?

Hmm, I've got a large power supply that occupies a 5 x 10 x 13" aluminum
box (lots of free space). After a few hours, it gets maybe 20C above
ambient. That's 360 in^2 (not counting the bottom, which is against the
table). Measured power 100W, so the thermal conductivity is about 0.2
C/W, or 72 in^2*C/W. I may be grossly off with my power estimate, and the
top panel may not contribute much by convection, being in stall.

Eating mashed red potatoes. Not just redskin, these are red *all the way
through*. It looks like strawberry ice cream, and is exactly as
delicious, but savory instead of sweet.

Tim

 

[legg]

If the box has, say, 5 k/w to the world (30 sq inches, 150 K/w per)
and you dump 30 watts from inside, the wall temperature averages 150K
above ambient. That's bad enough, without the parts inside having
another bunch of theta between themselves and the hot wall. That's why
it's best to bolt hot stuff directly to the inside of the walls, and
not use convenction, even with a circulating fan, inside.

So I guess we agree.

Yes. 30 in^2 is a small package to dissipate 30W.

If this were a conventional isolated converter, you'd not find much
free air left inside this size of a package anyways.

RL