Thermally conductive adhesive

[John Devereux]

Hi,

I have a metal-bodied LED like this one:

<http://jp.hamamatsu.com/products/sensor-ssd/pd150/pd157/L9437/index_en.html>.

I would like to heatsink it via the PCB groundplane. Also it needs to
be fixed very solidly to minimise movement under vibration. I could
solder it down but would be worried about temperature.

I don't need to isolate the case, but I don't want to use that
electrically conductive silver paste either since will likely short
out the anode lead.

Any suggestions?

The "thermal" epoxy compounds I found seem to be designed with filler
particles that ensure insulation, but would limit heat transfer. I
guess normal epoxy would work, or maybe cyanoacrylate (low viscosity
so get thinner layer)?

 

[Spehro Pefhany]

Hi,

I have a metal-bodied LED like this one:

<http://jp.hamamatsu.com/products/sensor-ssd/pd150/pd157/L9437/index_en.html>.

I would like to heatsink it via the PCB groundplane. Also it needs to
be fixed very solidly to minimise movement under vibration. I could
solder it down but would be worried about temperature.

I don't need to isolate the case, but I don't want to use that
electrically conductive silver paste either since will likely short
out the anode lead.

Any suggestions?

The "thermal" epoxy compounds I found seem to be designed with filler
particles that ensure insulation, but would limit heat transfer. I
guess normal epoxy would work, or maybe cyanoacrylate (low viscosity
so get thinner layer)?

Suspiciously low storage temperature range-- only to 100°C.

I don't suppose you'd consider pushing it through a grounded ~4.7 mm
plated-through hole in the PCB and bending the leads 'round back?


Best regards,
Spehro Pefhany

 

[nospam]

I have a metal-bodied LED like this one:

I would like to heatsink it via the PCB groundplane. Also it needs to
be fixed very solidly to minimise movement under vibration.

Abs max power dissipation is 150mW so it is not like its going to cook
anyway.

Have you actually got one in your hands? The shoulder on the pin welded to
the case and/or epoxy/glass seal on the other pin tends to make such
devices not sit flat on a PCB anyway. A pad with a big enough hole to take
the pin all the way to the 'hilt' may end up shorting out on the case so
you may need a spacer. The data sheet says something about not having a
'step piece' but doesn't give enough information to know if there is or
isn't a problem.

If it does sit flat then you just have a tiny air gap from mismatch in
flatness of the case/PCB, and you don't have much power so I don't see you
need anything specially heat conductive. I would use a dab of ordinary
silicon rubber.
--

 

[John Devereux]

Solder conducts heat pretty good. Better than epoxy, and NO, you won't
be getting it so hot it reflows in use, so what is the worry? The only
worry with solder is mic-crystalline fracture over time, and solder
creep.

The worry is the maximum "storage" temperature of 100'C. What will be
the effect on the LED of raising its case to soldering temperature?

That would depend on how you place the epoxy, and what steps you take to
keep the lead isolated. Silver paste is a no go, but silver filled epoxy
IS the right animal.

The silver epoxy is best. Mainly because it doesn't flow into nooks and
crannies. It stays right where you put it, so a big bead outside your
metallic housing will not creep up under the shroud and invade conduction
paths if you are careful when you apply it.

Such applications are very labor intensive, so I hope you are not
talking about a production environment.

I should have mentioned this aspect. It *is* for a production
environment, quantities 1k or so initially.

EpoTek makes very good thermal bonding epoxies, and they are used by the
entire industry. Nearly all the heatsink capped chips in any of your
circuits use it. It is conductive enough to actually make traces out of
as well.

It requires an 80C 4 hour cure cycle though. I think it is the H20E
product.

Will have a look.

 

[John Devereux]

Abs max power dissipation is 150mW so it is not like its going to cook
anyway.

We need to run it close to the maximum temperature and current (at
least at the upper end of our products temperature range). Even a
small case temperature affects how we have to derate it. We have
confirmed with the supplier that the published derating applies to a
non-heatsinked LED mounted on its legs. So it should be possible to
run it at a higher current than would otherwise be possible, provided
we keep the case temperature below what it "would have been" at the
derated current, non-heatsinked.

Have you actually got one in your hands? The shoulder on the pin welded to
the case and/or epoxy/glass seal on the other pin tends to make such
devices not sit flat on a PCB anyway. A pad with a big enough hole to take
the pin all the way to the 'hilt' may end up shorting out on the case so
you may need a spacer. The data sheet says something about not having a
'step piece' but doesn't give enough information to know if there is or
isn't a problem.

No, it's not a problem. There is a step but it is on the cathode,
which we ground anyway.

If it does sit flat then you just have a tiny air gap from mismatch in
flatness of the case/PCB, and you don't have much power so I don't see you
need anything specially heat conductive. I would use a dab of ordinary
silicon rubber.

Remember we are also trying to make it immune to vibration.

Thanks,

 

[John Devereux]

Suspiciously low storage temperature range-- only to 100°C.

Yes, would have liked to solder it down.

I don't suppose you'd consider pushing it through a grounded ~4.7 mm
plated-through hole in the PCB and bending the leads 'round back?

Nice idea - I should have thought of that earlier. I bet I could
solder the base to a surrounding "pad" that way too (without frying
the led too badly). Or better some strategically placed bolt heads
could clamp solidly it in position. The case is the cathode and is
grounded anyway, so there would only be one lead to worry about.

For other reasons I think that is too big a change to do right now (it
changes the mechanical arrangement too much), but I will keep it in
mind for a future version.

Thanks,

 

[Archimedes' Lever]

Nice idea - I should have thought of that earlier. I bet I could
solder the base to a surrounding "pad" that way too (without frying
the led too badly).

Make a small daughterboard that has the hole. I am sure that the device
can handle one solder operation.

 

[John Devereux]

After seeing it, I question any need for sinking at all.

I explained it in another post. Basically we wish to take advantage of
the fact that the published derating curve is a "worst-case", based on
mounting the led up on its legs. So it assumes a bigger case
temperature rise at maximum current than would obtain with some
heatsinking. At maximum temperature it means we can put twice the
current through the LED, which directly affects the performance of our
product.

Nearly ALL electronic components are made to survive a single soldering
event. At the very least.

Yes, but I think it is designed to be soldered via its
leads. Soldering the can directly could melt the insides. There is
quite a thick baseplate on the device, so to solder it would raise the
whole thing to soldering temperature.

If the damned thing is that sensitive, I would consider finding a more
suitable device for the task.

Unfortunately it's by far the best one we found for our application
(it's not actually the one I linked to, but the case and temperature
specs are the same).

Thanks,

 

[Ecnerwal]

I explained it in another post. Basically we wish to take advantage of
the fact that the published derating curve is a "worst-case", based on
mounting the led up on its legs. So it assumes a bigger case
temperature rise at maximum current than would obtain with some
heatsinking. At maximum temperature it means we can put twice the
current through the LED, which directly affects the performance of our
product.

While you could certainly glue it down, and may want to for vibration, I
suspect you'd get better bang on the heat removal aspect by putting an
annular ring heat sink on it - and depending how you did that, you could
also get your solid, vibration-free mounting. Make use of that metal can.

ie, for glue down - glue it down, but add one of these (brim up):

http://www.aavidthermalloy.com/cgi-bin/stdisp_print.pl?Pnum=322400b00000g

For an alternate mount method, press fit, clamp, or glue the case into a
block of aluminum with fins milled into it, and screw (with loctite,
since you mention vibration) that down to your board - going nowhere.

 

[John Devereux]

While you could certainly glue it down, and may want to for vibration, I
suspect you'd get better bang on the heat removal aspect by putting an
annular ring heat sink on it - and depending how you did that, you could
also get your solid, vibration-free mounting. Make use of that metal can.

ie, for glue down - glue it down, but add one of these (brim up):

http://www.aavidthermalloy.com/cgi-bin/stdisp_print.pl?Pnum=322400b00000g

Thanks for the link, I had not seen those before. Another useful
option.

For an alternate mount method, press fit, clamp, or glue the case into a
block of aluminum with fins milled into it, and screw (with loctite,
since you mention vibration) that down to your board - going
nowhere.

We do actually have an aluminium block too, but it is mainly coupled
to the groundplane rather than the LED itself (because of
manufacturing tolerances). Fixed pitch font:


.. | LED
.. V
..
.. _
.. [XXXXXXX]| |[XXXXXXX] <--- 5mm Aluminium plate
..PCB =========================================
..
..

There are some vertical screws clamping the whole thing up to the
metal block of the product.

The critical thing is to cool the baseplate of the LED, so the thermal
interface to the groundplane is what I was concerned about.

I am liking Spheros idea of mounting it from behind the board more and
more:

.. [XXXXXXX] _ [XXXXXXX] <--- 5mm Aluminium plate
..PCB ===================| |=====================
.. `' `'

We could put some bolt heads either side of the led base to clamp
everything up really solidly without needing any adhesive. We would
need to carefully trim the anode lead and bend it around to a solder
pad. The cathode could just be be snipped off. A bit labour intensive,
but much better than gluing.

 

[Ecnerwal]

The critical thing is to cool the baseplate of the LED, so the thermal
interface to the groundplane is what I was concerned about.

I am liking Spheros idea of mounting it from behind the board more and
more:

. [XXXXXXX] _ [XXXXXXX] <--- 5mm Aluminium plate
.PCB ===================| |=====================
. `' `'

We could put some bolt heads either side of the led base to clamp
everything up really solidly without needing any adhesive. We would
need to carefully trim the anode lead and bend it around to a solder
pad. The cathode could just be be snipped off. A bit labour intensive,
but much better than gluing.

And if you have room, you could even put a flat heat sink (with grease,
rather than epoxy) on the backside as part of the "clamp" (with a couple
of holes for the anode connection). Some small heat-shrink on the anode
lead makes things simpler in terms of not shorting it. A short pin/post
soldered into the board might make for a less-labor intensive
(faster/easier for production) connection to the anode lead.

 

[MooseFET]

Hi,

I have a metal-bodied LED like this one:

<http://jp.hamamatsu.com/products/sensor-ssd/pd150/pd157/L9437/index_e...>.

I would like to heatsink it via the PCB groundplane. Also it needs to
be fixed very solidly to minimise movement under vibration. I could
solder it down but would be worried about temperature.

I don't need to isolate the case, but I don't want to use that
electrically conductive silver paste either since will likely short
out the anode lead.

Any suggestions?

The "thermal" epoxy compounds I found seem to be designed with filler
particles that ensure insulation, but would limit heat transfer. I
guess normal epoxy would work, or maybe cyanoacrylate (low viscosity
so get thinner layer)?

The "super glues" are all way to brittle.

Using a thin layer of epoxy to hold it down to a metal layer is about
the best you can do without adding an extra part.

Adding a via or two to take the heat all the way to the far side of
the PCB can help to get the heat away. Arranging things so that it
isn't very far through the PCB to get to the mounting hardware can
help to keep the temperature rise on the PCB low.

 

[John Devereux]

The critical thing is to cool the baseplate of the LED, so the thermal
interface to the groundplane is what I was concerned about.

I am liking Spheros idea of mounting it from behind the board more and
more:

. [XXXXXXX] _ [XXXXXXX] <--- 5mm Aluminium plate
.PCB ===================| |=====================
. `' `'

We could put some bolt heads either side of the led base to clamp
everything up really solidly without needing any adhesive. We would
need to carefully trim the anode lead and bend it around to a solder
pad. The cathode could just be be snipped off. A bit labour intensive,
but much better than gluing.

And if you have room, you could even put a flat heat sink (with grease,
rather than epoxy) on the backside as part of the "clamp" (with a couple
of holes for the anode connection).

No room unfortunately - just getting the lead bend in may be a
problem, will have to check exactly how much room there is.

Some small heat-shrink on the anode lead makes things simpler in
terms of not shorting it. A short pin/post soldered into the board
might make for a less-labor intensive (faster/easier for production)
connection to the anode lead.

Actually the anode lead seems to naturally bend around 180 degrees
with quite a nice smooth curve. I think it will only take a few
seconds in production.

Thanks,

 

[John Devereux]

The "super glues" are all way to brittle.

OK, thanks for that advice.

Using a thin layer of epoxy to hold it down to a metal layer is about
the best you can do without adding an extra part.

Adding a via or two to take the heat all the way to the far side of
the PCB can help to get the heat away. Arranging things so that it
isn't very far through the PCB to get to the mounting hardware can
help to keep the temperature rise on the PCB low.

Yes, we have done all that. Lots of thermal vias and one side of the
board is all groundplance.

 

[John Devereux]

Thoughts:

The Hamamatsu part looks expensive.

The datasheet directivity curves are beautifully centered on 0
degrees; they always are. Verify that. Small variation in chip
position or lens tilt can shoot that beam all over the place from part
to part.

Even if the parts themselves are perfectly aligned, your epoxy/solder
operation will need to be precise. That's a narrow beam.

It doesn't have to go *too* far; our testing so far indicates they are
OK for our purposes.

Epoxy+solder will be messy.

I am coming to that conclusion...

A T1-3/4 type package heatsinks pretty well through its leads. You may
not need heatsinking at all, at this sort of power level.

The last time I did something like this, the best solution was to use
a flat-face (no lens) T1-3/4 package. No alignment problems, nice
sharp point source, cheap, easy. It put more light in the sensor than
a typically-misaligned lensed part.

There are probably nice surface-mount parts that would work.

We need the intense spot with a fairly well collimated beam (anything
else would need an external lens). They are not too expensive for our
application, so we are happy with the part itself for now.

Thanks,

 

[MooseFET]

Whose rule is that? What army is going to enforce it?

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

Do I have to give the money back?

Where I work, we have these cute little spring loaded fixtures that
get put onto an assembly to hold the parts in place while the epoxy
hardens. Is this what you did?

 

[MooseFET]

OK, thanks for that advice.



Yes, we have done all that. Lots of thermal vias and one side of the
board is all groundplance.

I forgot, you can also solder down a bit of sheet copper on the top
side.

It is also worth, filling the holes. Nearly any material is a better
thermal conductor than air.

 

[David Lesher]

I bought some conductive epoxy from Tri_Con to embed a number of LM35
sensors in a system. While it was a better conductor than most epoxies;
it was later clear that that was better than air; it wasn't near what
metal-to-metal provided.

I seem to recall that the epoxy was about the same conductive
characteristics as heat sink paste.

 

[John Devereux]

It is an IR transmitter, yes? Would TWO LEDs side by side yield the
same energy and end up with the same transmission range?

Doesn't help in our application I'm afraid. It is the combination of
small spot and parallel light that we need. The target is only an inch
away.

Ever see those radial finned 'hats' for that package size from
thermalloy?

Not until "Ecnerwal" (Lawrence?") posted the same one

I hadn't seen it before. Yes, might have been a way to go.

This one is the simple variety:

http://www.aavidthermalloy.com/cgi-...2&SortBy=TR&NumPerPage=all&Device=7&x=34&y=13

They even have many with radial fins.

Thanks,

 

[John Devereux]

No. They are made to withstand a normal ramp up, and wave solder
session, generally. If yours is that sensitive, hand soldering is
required, and that by a VERY qualified person with some common sense.

If this device is really that sensitive to heat, I would question the
maker as to why. If hand soldering is required and that at the leads
only, you also then need to elevate it and put sinking clips on the leads
while soldering. This is because simply touching a lead with the high
temp solder iron feeds that heat right up the lead frame and into the
package, no ifs ands or buts.

I'll ask them for more information on this.

You could bend the leads surface mount style, and use the silver epoxy
for the electrical attachment as well.

Alas, without an active sink that actually has air moving over it, you
can temps are not going to be that much better no matter what you do
after the soaking takes place.

That's not a problem. The whole thing is in good thermal contact with
a metal block with fluid running through it!