1
\$\begingroup\$

I have some high-power (~5W) SMD LEDs that I will be running for extended periods of time. I would like to use a PCB mill to make breakout boards for these chips for easy prototyping. Ideally I would like to reuse these boards in the final circuit since the chips are expensive.

The chips have legs which I can solder by hand, but the thermal pad on the bottom will be a bit trickier since we don't have access to a good oven or similar equipment. Can I use thermal paste instead of solder, relying on the soldered legs to hold the chip down?

\$\endgroup\$
2

4 Answers 4

2
\$\begingroup\$

The board, pad, paste, insulation, second layer of paste and component in this case must be as close together as possible, really tight together, but you should be able to use it fine provided the pad is large enough to dissipate an adequate amount of heat.


Typically, if the pad size exceeds the component size, you can use a high temperature iron (400C+) and a generous amount of solder to 'tin' the pad until there is a thin layer of solder on it, then you place the legs of the chip down firmly and reheat the solder until it takes onto the underside of the chip.

You will know that it's done when a fillet is formed in the small gap that is left behind when you place the component on the solder layer, seen closely from the side under light.

Either way, just make sure you have very good contact. Solder makes the 'best' contact, paste is kind of terrible in comparison, but still very good compared to nothing at all.

In any event, make sure the heatsink and its pad on your board are isolated from everything else, unless it happens to be the case that the heatsink on your component isn't connected to anything.

\$\endgroup\$
6
\$\begingroup\$

Adding some actual number work to your question:

If the Pad is hard to reach, I am inferring you mean it's a slug-pad under the LED, hidden entirely in the plastic casing, such as on the Avago AJSD 1W/3W types.

These slugs are very small, in the case of the mentioned type they are about 2mm in diameter from memory, which is Pi mm2 in area (A = pi*r2, with r = d/2 = 1).

Assumption:

For the next bit I am going to assume it's a 3W LED, because it is unfair to use the slug size of a 3W LED and then sneakily input 5W into the rest of the equations. I'm also going to assume that 1W of electrical energy is converted into light, is it? Depends on the quality of your LED, but... eh.... That leaves 2W of heat to be transferred.

Numbers game:

If your solder paste has 2W/mK, this means that a ten micron thick layer of paste will get you a thermal increase of 3.1K across the barrier: That is probably fine in most occasions. But then you have to guarantee 10micron layer thickness.

So with the same thermal paste, if you use 50um thick, you get 16K of thermal increase between the slug and the PCB. I expect with a little pressure you can guarantee 50micron layers of paste when you are careful soldering the pins, but I'm still not 100% sure without knowing the quality of your PCB and type of LEDs. But don't push too hard, because then the LED will "relax" after mounting and suck little peaks and valleys into the paste, causing reduced contact area.

But consider if you could live with 16 degrees C of temperature increase. I don't think I'd like the challenge of keeping the PCB cold enough to guarantee a junction of below 120degrees C. Not to mention that you should prefer keeping the junction at or below 80 degrees Celcius for a decent lifetime of the LED.

Assuming the Junction to slug is about 5K/W (a number I'm just plucking from the air as an estimate, refer to your datasheet), that's another 10K lost with the 2W in heat generated. We're already at 26 degrees heat up, without considering the transfer properties of your PCB. Will your immediate environment be at or below 25 degrees, always? Then your task is till quite easy. Can it be 50 degrees sometimes? Well, you're sure to not achieve 80degrees Celcius for your junction in that case.

Compare the thermal conductivity of reasonable quality thermal paste with solder:

  • Tin/Lead - about 50W/mK
  • Tin/Silver - about 78W/mK
  • Other Lead Free types - between 55 and 70W/mK generally

That means that Tin/Lead solder can conduct 25 times as much heat away at the same temperature difference as the previously mentioned Thermal Paste. Which means that the slug to PCB in the case of the 50micron distance would only give a temperature rise of 0.6K. That pretty much means that this specific example LED would fare reasonably well with even 50 degrees Celcius environment, where the thermal paste one might not always.

Will it work for testing? Most likely yes, it will transfer quite a lot of the heat away and you'll probably notice it getting a bit hotter, causing the colour of the light to change slightly, it may change a bit in brightness too, but they won't instantly die on you.

Is it a good idea for production? Probably not. You should always account for worst case when building a device for production and your worst case isn't a nice smooth full-contact thin layer of paste under a LED, as I assumed to save myself some annoying maths.

Put simply, your worst case with solder is more than an order of magnitude less bad than the one with thermal paste. Not to mention, if you get the soldering right, with the right temperatures and flux levels, the chances of bad thermal contact are significantly lower as well. (It expels air-bubbles and moisture when it's hot and it likes to flow out across all the metal surfaces)

Quick Maths section:

The formula for thermal conductivity is explained here (why type it again if I don't have to): http://www.engineeringtoolbox.com/conductive-heat-transfer-d_428.html

Which you can rewrite as dT = (qs / kA) - units same as on the linked page.

The assumption of 2W/mK for thermal paste comes from 0.5W/mK for non-brand leaky types to 3W/mK for normal electronics high-end or early graphite types. There are pastes with beyond 10W/mK, but they are messy and sometimes have other corrosive properties that you have to be very careful of. My most expensive, safe in all uses paste is a graphite-silicone composite at 4W/mK. It could be that the Gellid Ultra's are safe enough though, I believe they go at about 7 or 8W/mK? But still, much less than solder.

Data about thermal conductivity of solders came from: http://www.electronics-cooling.com/2006/08/thermal-conductivity-of-solders/

That should allow you to do your own feasibility study with the sizes and parameters of your own LED.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ +1 good answer. Datapoint only: Many Cree LEDs are now specified at die temperatures of 105C with lifetimes of maybe 50,000 hour at these temperatures (surprisingly0. \$\endgroup\$
    – Russell McMahon
    May 27, 2015 at 2:22
  • \$\begingroup\$ @RussellMcMahon Fair enough, but what about their colour continuity and brightness? In the basics if you heat the substrate its characteristics change, if you can control the temperature well, you control your light better. But less risk is always better. \$\endgroup\$
    – Asmyldof
    May 27, 2015 at 9:34
  • \$\begingroup\$ Asmyldof - Surprisingly (again) this is the rated running temperature for these Cree parts and is where the main specifications are given. Some specs which I am interested in at lower temperatures are not given. Fortunately (for me) the Royal blue version (which is used as a light engine to drive external phosphors or to do eg dental curing) seems to use the same die and has lower temperature specs available. (Not the ideal way to have to design things). Cree will give detailed data if asked but do not publish it otherwise. \$\endgroup\$
    – Russell McMahon
    May 27, 2015 at 16:48
  • \$\begingroup\$ @RussellMcMahon Bizarre! Good to know though. \$\endgroup\$
    – Asmyldof
    May 27, 2015 at 19:05
3
\$\begingroup\$

That depends on the specific chip. Some of them may have thermal pads labelled "N.C.". These do not (and should not) be connected to any part of the circuit and so thermal paste may be acceptable for use here, but make sure you do the thermal calculations first of course.

Other chips may require their pad to be connected to ground or some other part of the circuit, and it would be inappropriate to use any paste on them since even conductive paste can dry out and therefore the connection may be broken.

\$\endgroup\$
2
  • \$\begingroup\$ Thank you, that's exactly what I wanted to know. I'll check the datasheet for more info. Since the final circuit needs to operate over the course of several years with minimal maintenance, should I avoid this strategy altogether? I don't like the idea of the paste drying out after a year and having my LEDs burn out. \$\endgroup\$ May 26, 2015 at 22:06
  • 4
    \$\begingroup\$ I would strongly recommend using solder in the final product even if it takes a little more work. \$\endgroup\$ May 26, 2015 at 22:11
1
\$\begingroup\$

Can I use thermal paste instead of solder, relying on the soldered legs to hold the chip down?

No. You cannot use thermal paste.

You must have a thermal pad on the PCB and use solder paste. There is no other way to mount the Cree part.

The thermal pad on the bottom of the LED package (labeled "heatsink") must be soldered to the thermal pad on the PCB.
enter image description here

The

enter image description here



You should use solder paste and a squeegee with a solder paste stencil to apply the solder paste. An alternative is solder paste sold in a syringe.

enter image description here

Put a small bead (the thickness as it comes out of the syringe) across the PCB thermal pad. Put a small dab on each pin pad.

enter image description here



Heat an oven up to 500°F.
Put PCB in oven.
Close oven door
Turn off oven.
Remove PCB from oven within 2 minutes.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.