How to reflow solder an LED emitter on a PCB or MCPCB.

That’s great. Never done this before, but it looks easier than with a skillet.
Any advise about the temp setting? Cotton?

I think you all are talking about two different type “irons”. :wink:

Clothes vs. Solder

I think you are right. I was wondering 60w was one of those small traveling irons.
Now I think about, LEDs get very hot. Hotter than a clothes iron I guess.

I have a Weller solder station.
I’ll scrape the corn bread out of my skillet and use that :slight_smile:

Tnx teach :wink:

We were indeed talking about different irons, but, coincidentally, I have also used an ironing iron before! That also worked, it was a different project though. My solder paste has a melting point around 175° (measured with IR temp).

you can use skillets, mug warmers, hot plates, anything you have that gets hot enough.

:+1: … I love it when a plan comes together…. :wink:

its easy to overheat the LED qwith little thermal mass and high power heat source
slow and steady heat up and cool down is important especially for flip chip LEDs

and same as for food if its too hot the LED gets crispy, may have a bad taste when it is dead or quite working after very short of time

With the standard reflow procedure used here (melt, let auto-center, tap the dome, cool), what is the resulting solder bond thickness?

Hi Argo,

I don’t have a measurement I can give you, but I will say it’s quite thin. In most cases the LED will appear to sit perfectly flat with the mcpcb. If you can see a gap, the emitter is probly sitting to high.

Thanks. It tells my I botched my first reflow which I did yesterday. :wink:
But the purpose of my question was to get better understanding of thermals of solder joint.
With XHP35 running at full power we’re talking about temperature delta exceeding 20 °C per 0.1 mm of joint thickness. And that’s assuming no voids, which is probably unrealistic.
I see that Osram / Cree recommend reflowing with stencils resulting in 70-100µm joint. But I think that standard reflow doesn’t involve applying pressure to the LED, does it? So maybe what we do here is better? Or maybe no, I’ve seen some post here on BLF telling that manual reflow can’t result in a joint as thin as that with a stencil. I also found a joining process from Indium Corp which leads to 40 µm joints (though ones of poor thermal conductivity, overall not better than good paste).

OK, I spent far too much time seeking about how to best do a reflow.
I’d like to dump here some info that I found. I did not organize the search well enough, so I won’t quote all the sources, sorry.

  1. The thermal shock causes damage to the LED. It is normal. The only hard data that I found is here:
    https://www.nichia.co.jp/specification/products/led/ApplicationNote_SE-AP00032-E.pdf
    It shows that 260°C reflow for 10 seconds reduces output of Nichia NS3W183 by 10%. I’ve seen somewhere that flip-chip LEDs are less susceptible to being damaged like that. Also, normally you heat the LED twice. To reflow it and to solder the leads.
    Conclusion: low-temperature solder should improve output (use it for all LED pads and leads). A conductive glue should be even better. There’s also a soldering process from Indium Corp. called NanoBond that reduces thermal stress too.
    But I’ve seen no LED manufacturer suggesting really low-temp solders. Why? I don’t know. Maybe normally the effect is less pronounced than with NS3W183?
  2. Some LEDs don’t have high power density in relation to the thermal pad size. Some do. The worst affected is XHP35 HI. Osram Oslon Black is quite bad too. Using a highly-conductive solder may bring some real improvements for them. So can the use of liquid metal.

Overal glueing anode and cathode + pasting thermal pad with Conductonaut seems like the best attachment method to me.
But:

  • pads suck solder, but not glue. You need to cover them carefully. And then place the LED precisely because it won’t self-center. And then tap it without moving it. Overall much harder to do. I’m not about to try any time soon.
  • I’m somewhat afraid of thermal paste pump-out. Laptop forums recommend using CLU / Galinstan instead of Conductonaut due to it being thicker. This would provide somewhat different tradeoff, but probably not full safety. Maybe some adding non-conductive glue around the thermal pad would be a good idea?

The second best looks to be using a low temperature high-conductivity solder. Not too low temperature to avoid the LED unsoldering itself during work though. There’s one option that really stands out. Indium

However:

  • it is very expensive
  • it is not available as a paste

You can buy indium scraps for a moderate price . Is it really pure though?

Thoughts?

On new thought already.
Regardless what you do with the LED, it is easy to avoid heating the LED while soldering leads with the use of conductive glue. All the difficulties mentioned with glueing LED don’t apply here. It may even be fairly inexpensive.

Excellent video vestureofblood! I learned a lot. If I ever need to reflow an LED, I’ll remember slow heat up (and not too hot), and slow cool down. Makes sense to not damage the LED.

Conductive glue? That is C.R.A.P

I have it home and it is ridiculous. I have no absolutely single use for that and it is fragile like glass when it hardness.

Why would you want to avoid heat? Isn’t solder paste good enough for you? It has very low melting point.

Don’t complicate things. Do Vestureofblood method or Old Lumens soldering Iron method and call job done.
I use soldering Iron method since it is virtually impossible to reflow certain emitters with VOB method.
Sometimes you need to manually place and tap emitter because or it is to small or it does not getting sucked in pads.

I find 12% output loss due to heating the LED twice to be extreme and would like to avoid it, as simple as that.
And that’s with not-so-bad temperature profile that they use, not semi-random as we do.

In fact even if the output loss on chips other than the Nichia is far lower than that I would like to avoid it as I simply like to optimize stuff. I’m not fully aware of the tradeoffs involved (like glue selection which as you indicate may not be trivial), but half bin upgrade is worth a lot for me.

You will not upgrade anything. Gluing the led is ridiculous. Glue maybe is electrical conductive and it could be used for some tiny wires repair or something like that but what about thermal properties of glue? How it will transfer heat from led to mcpcb?
It will crack like glass once when you turn on the light(I give it 30 seconds - 1 minute before cracking).

Your idea would be possible if glue would have thermal and mechanical properties like solder but that kind of glue sounds like sci-fi to me. Yes I would also like if possible. “Cold re flow” sounds great but with today conductive glue impossible :slight_smile:

Yes, at this moment I find the all-glue way challenging, there’s quite some R&D involved in making it right. And actually in validating whether it’s worthwhile too, the singular data point that we have is very promising, but my guts tell me that if it was so important, manufacturers would try harder to remove thermal stress themselves. I don’t find it impossible, actually I’m certain it’s possible.

There are many conductive glues. Some are based on polyurethane, some on acrylic. Some are filled with silver, some with graphite, some with diamond. They have various additives. There are likely other bases and fillers too. Are all of they brittle? I doubt it.
saabluster used conductive glue with XR-E. But he did so for structural reasons because he thinned down the ceramic substrate in order to shorten the thermal path.

How would I glue the the thermal pad? I wouldn’t, I would use liquid metal thermal paste instead.

And in the meantime I found a potential problem with indium as well. Will the bond be strong enough at near-melting temperatures?

So Agro you tested this loss when you soldered the driver leads to the contacts on a DTP mcpcb, that you heated the LED enough to cause damage to the LED?

” Also, normally you heat the LED twice. To reflow it and to solder the leads.”

I didn’t. Nichia tested their LED, heating it twice and measuring output before, after the first heating and after the second heating.
I posted the link a couple of posts ago.

In the part that you quote simply I notice the fact that a similar double-heating occurs in what we do.
It’s not exactly the same because you heat up pads several mm from the LED. Heat from it goes either through board traces or through a layer of insulation and then the board core (which has minimal thermal resistance).

In the meantime I found glue that works up to 190°C:

A followup…Previously I didn’t realize pure indium was so weak mechanically. Over 26 times weaker than SnPb. Is it bad enough to cause problems?
http://alasir.com/reference/solder_alloys/

20 kgf/cm² would be 1.4 kgf for XP footpring LED. Assuming a perfect reflow with solder covering entire pads (unrealistic). Sounds good enough, but some drop tests would be interesting….

Give up… Nothing will beat solder. :smiling_imp: