Possible LED attachment improvements

clemence · 2018-12-17T16:41:47.000Z

Agro:

That’s a bit OT but I really like soldering leads with SbBi.

Only for leads, never use for this Sn42Bi58 for attaching LED. It has very bad thermal conductivity, approaching that of stainless steel. Somewhere around 16 W/MK. I usually mix some with Sn60/Pb40 to make melting point slightly higher. 138°C is too low

- Clemence

Agro · 2018-12-20T12:17:36.000Z

I stumbled upon another Al/Diamond composite maker:
http://www.mmmt.com/resources/thermadite.html

clemence · 2018-12-20T12:33:51.000Z

Agro:

I stumbled upon another Al/Diamond composite maker:
http://www.mmmt.com/resources/thermadite.html

540W/MK is outstanding! So does the price I guess. Its so hard that only EDM machinable which is very slow. I machine a 2mm, 45mm deep oil passage in a hardened race bike’s crankshaft few years ago. Precise control and diameter tolerance was an issue.

- Clemence

Agro · 2018-12-20T12:42:27.000Z

And another…:
http://www.rhp-technology.com

This one does only Cu/Diamond:
http://www.sumitomoelectricusa.com/products/heatsinks/copper-diamond

clemence:

Agro:

I stumbled upon another Al/Diamond composite maker:
http://www.mmmt.com/resources/thermadite.html

540W/MK is outstanding! So does the price I guess. Its so hard that only EDM machinable which is very slow. I machine a 2mm, 45mm deep oil passage in a hardened race bike’s crankshaft few years ago. Precise control and diameter tolerance was an issue.

- Clemence

The first one that I found would do 400 PCBs (that’s their MoQ) for mere $14K.
Maybe others are cheaper…anyway it shouldn’t hurt our record-chasers to ask for samples.

clemence · 2018-12-20T13:12:27.000Z

The Copper Diamond is the most interesting because it can be metalized while still electrically non conductive. With 650W/MK thermal conductivity the thermal bottleneck could the solder itself (assuming non gold plated contact used). In my opinion thermal resistance is more relevant since in the end this is what matters.
If an attachment system thermal resistance can be optimized to slightly below the thermal resistance in the LED (which we can do nothing about) then the rest of the system can use that number as the minimum resistance. In my opinion, as long as the cooling still creates negative delta temp copper or aluminum heatsink still doing OK. The higher the thermal conductivity of a material the less delta temp needed to flow the heat. We can make a superior cooling system using available budget friendly material widely available with good design and process. Manufacturers have to compromise the outcome to balance the design for a reasonable total cost.
For example:

- Use the best solder available (Indium is in the top of my list). And don’t forget to make the joint as thin as possible.

- Use the thinnest most rigid substrate possible. Metalized Aluminum Nitride can be made reliably down to 0,2mm without sacrificing rigidity. Although the thermal conductivity is only up to 180 W/MK in most cases, the thin substrate will make up the difference. This is demonstrated clearly in my MCPCB using “low” aluminum oxide dielectric with only 7,5W/MK thermal conductivity. Thermal resistance is what matters.

- Conventional wide and thick copper traces still works to further spread the heat over the entire board substrate as long as the radial thermal resistance still lower than through plane resistance.

- As flat as possible finish in the substrate and heatsink mating surface makes a big difference. Don’t compensate with thermal paste. Get the best possible mating surface first, use the least amount of thermal paste.

Bare metal with no paint or anodization at the heat sink joint.

- Clemence

Agro · 2018-12-20T13:37:27.000Z

The last manufacturer until I find some new search terms:
http://www.awindiamond.com/eng/04_application/Laser_HeatSpreaders_CDHS.html

Clemence, I don’t think that it makes sense for manufacturers to go beyond the standard SAC305. And definitely not now when even enthusiast haven’t demonstrated any improvements, only calculated some tiny ones. I’m not sure what budget materials do you mean but as far as I see budget light makers already do quite well when it comes to LED thermals. There may be tiny improvements here and there, but better materials don’t seem to be a budget choice.
Well, I imagine one possible improvement. That’s likely not a good idea anyway, otherwise saabluster would do it years ago. Eliminate MCPCB. Machine the shelf to leave a protruding thermal pad, add insulation on the rest of the shelf and make anode/cathode connections.
Or make the MCPCB several mm thick and press-fit into the head. Just like Nightwatch press-fits copper heat sinks.

clemence · 2018-12-20T13:59:20.000Z

Agro:

The last manufacturer until I find some new search terms:
AWIN DIAMOND TECHNOLOGY CORPORATION

Clemence, I don’t think that it makes sense for manufacturers to go beyond the standard SAC305. And definitely not now when even enthusiast haven’t demonstrated any improvements, only calculated some tiny ones. I’m not sure what budget materials do you mean but as far as I see budget light makers already do quite well when it comes to LED thermals. There may be tiny improvements here and there, but better materials don’t seem to be a budget choice.
Well, I imagine one possible improvement. That’s likely not a good idea anyway, otherwise saabluster would do it years ago. Eliminate MCPCB. Machine the shelf to leave a protruding thermal pad, add insulation on the rest of the shelf and make anode/cathode connections.
Or make the MCPCB several mm thick and press-fit into the head. Just like Nightwatch press-fits copper heat sinks.

Moving to Indium will only make manufacturing cost higher. Its inline with Pb Free regulation. Indium Corp. offers pre cut solder that can be used in normal fluxless nitrogen reflow chamber.
This approach still not feasible for real budget products but has its place in professional products. I know Saabluster idea and its already followed by Manker E01 but there’s still solder involved. I did an experiment with successful result with XPG back in 2012. I electroformed copper to cold weld the thermal pad directly to the copper heatsink. But the process was tedious and took too much time to make it feasible in any of my later built. It’s also permanent and can’t be undone without destroying the LED

- Clemence

Agro · 2018-12-20T14:05:58.000Z

I don’t know what idea you’re referring to…what interesting does Manker E01 do w.r.t. thermals?

I’ve seem solder preforms from indium, but didn’t think much about them. Yes, they surely seem doable for a serious factory.

I’d like to learn more about your electroforming experiment, is there a writeup somewhere?

clemence · 2018-12-20T14:31:41.000Z

Agro:

I don’t know what idea you’re referring to…what interesting does Manker E01 do w.r.t. thermals?

I’ve seem solder preforms from indium, but didn’t think much about them. Yes, they surely seem doable for a serious factory.

I’d like to learn more about your electroforming experiment, is there a writeup somewhere?

Typo, it’s Manker E02
Jerommel did the LED swap with his Manker E02 and the thermal pad was soldered directly to the copper slug. Manker uses hollowed PCB.
https://budgetlightforum.com/t/-/46012

Sorry, I didn’t document the electroforming modification. It was when I joined CPF. I didn’t bother posting since some of my previous posts were banned there. Now looks like my country’s IP address also banned by CPF. The result was somewhat messy because I grew copper starting at the LED’s thermal pad. It looks like a sci-fi stuff - an LED on a pink copper coral.
As long as you can get the right copper plating solution it’s a fun project. I plan to make copper electroformed flowers with E21A “welded” in the center. Should make a very nice ornament.

This for example was made using real flower:

https://www.instructables.com/id/Electroforming-an-Iris-Seed-Pod/

- Clemence

Agro · 2018-12-20T14:28:15.000Z

Thanks, that’s very very interesting.

kikkoman · 2018-12-20T15:41:33.000Z

clemence:

kikkoman:

A quick followup on the pure Indium:

I’ve really come to like the stuff, it makes reflows much less stressful especially with larger MCPCBs. You’re nowhere near the point where the emitters start to degrade and PCBs delaminate so you can take your time.
It also helps a lot with soldering wires where a large thermal mass sucks heat away quickly, which sometimes is unavoidable.

Yup, really like this stuff too, except for it’s steep price

- Clemence

Paid about 10$ for 20g and that’s going to last me forever. High quality solder e.g. from Kester is about half the price.

If used conservatively, the cost for one reflowed LED is less than a cent.

clemence · 2018-12-20T15:52:40.000Z

kikkoman:

clemence:

kikkoman:

A quick followup on the pure Indium:

I’ve really come to like the stuff, it makes reflows much less stressful especially with larger MCPCBs. You’re nowhere near the point where the emitters start to degrade and PCBs delaminate so you can take your time.
It also helps a lot with soldering wires where a large thermal mass sucks heat away quickly, which sometimes is unavoidable.

Yup, really like this stuff too, except for it’s steep price

- Clemence

Paid about 10$ for 20g and that’s going to last me forever. High quality solder e.g. from Kester is about half the price.

If used conservatively, the cost for one reflowed LED is less than a cent.

I bought 250gr 1mm 4N5 (99,995% pure) grade Indium for $180 (FeDex shipping) plus $62 import tax. That makes it $0,968/gr. Lower grade should be cheaper but I want to make a test using relatively pure Indium.

- Clemence

clemence · 2019-01-15T01:16:16.000Z

My Indium test report:

I always have problem soldering my MCPCB especially the VR21SP4. I designed this one with thinner and brighter solder mask to get better beam and better high current capability (thinner solder joint). The closer you cram E21As, the hotter the LED will be. The VR16SP4 has 0,02mm solder mask thickness, 0,4mm spacing, and less bright white color. VR21SP4 has 0,01mm solder mask thickness, 0,25mm spacing, and very bright reflective white color. But the thin reflective masking is very very fragile.

Soldering VR16SP4 and VR16S1 are easy as long as we use Sn63Pb37 solder paste for the LED pads. The flux residue easily cleaned by hot IPA and ultrasonic cleaner. The problem with VR21SP4 is its so fragile that even normal flux, ultrasonic cleaning, and hot IPA bath will destroy the masking. Reflective white masking ink contains special filler (higher content of Barium sulphate I guess) and less epoxy binder, reducing the toughness significantly.

One of the solution is to use low temp solder, compatible flux, and perhaps, no ultrasonic cleaning at all. Most low temp solders have worse thermal conductivity than Sn63Pb37. As bad as only 19W/MK for Sn42Bi58. And I can’t omit ultrasonic cleaning process, especially for high current applications. Without UC, it’s almost impossible to get the flux residue cleaned from underneath the LED. At high current the residue will thins and seep up to the LED surface and then gets burnt. Water based aluminum flux proved to be very good as long as soldering temperature does not exceeds 200°C. Aluminum fluxes are usually super aggressive and requires high temperature to work on aluminum. It easily destroy solder masks and eats the aluminum at normal soldering temperature (220°C - 260°C). But used at medium temperature (100°C - 150°C) its a very effective flux to clean any oxides on solder joints and copper traces. Being water based, warm water is all it takes to clean it. Dissolves readily even without brushing. Remember, always clean anything fluxed with aluminum flux!! At warm temperature it starts to be HIGHLY corrosive, I used Zinc chloride based aluminum flux.

So today I had a great result combining Indium and aluminum flux to solder VR21SP4. This is also the first time I can solder E21A with minimal manual LED positioning (with my sharpened bamboo stick). All those E21As align themselves just like “normal” LED. Solder dosing is now very easy to do without the need to use solder paste stencil. I just wet the pads with aluminum flux, heat it to 160°C. Then, put a tiny (really tiny) amount of Indium (the size of an ant head for the entire 8 LED pads). I moved the molten solder to distribute them evenly using flat shaped bamboo.
I let the whole MCPCB cool to room temperature, put a small blob of flux on the center. Then carefully placed the E21As at their approximate locations. Place the MCPCB back to the hot plate and slowly heat it to 160°C max. All the E21A, moved effortlessly to designated locations. If there’s slight misalignment, all I had to do was wiggle the LED from the side. After the MCPCB cooled down to about 40°C I soak them in 50°C ultrasonic bath for several seconds. I usually do the UC cleaning 3 times, replacing the water each time.

Only the LED pads soldered with Indium, the rest of the wire pads soldered with Sn42Bi58.

I haven’t test the performance with Indium solder yet, but it should be at least, slightly better than Sn63Pb37. As for now, Indium already saves me a lot of time in soldering department. At almost $250 (including $62 duty tax) per 250gr, this is the most expensive solder wire I have ever used. FeDex forced me to pay ridiculous duty tax because its classified as “noble material”, crazy!! I bought 5N Indium which supposed to be 99,999% pure. That’s why I couldn’t get any smaller than 250gr. The factory asked me for 1kg minimum order at first. IMHO, cheaper Indium with less purity should be OK for our uses. 3N purity and up are already enough and can be bought much cheaper via ebay or AE.
Note: you can see from the picture, I even salvaged the Indium balls - Cheapo!

- Clemence

Agro · 2019-02-01T08:31:18.000Z

Another option for thermal improvement: turn body into pcb. http://www.camnano.com/product-3D.html
Note that it would enable eliminating driver cavity, reflowing components on all internal surfaces, seems more elegant than Zebralight engines (which I have in high regard).
I wonder if MOQ and price would be feasible for flashlight manufacturers… and if yes - would it be only large makers of expensive lights (Olight f.e.) or smaller ones as well (Mateminco?).

clemence · 2019-02-03T18:49:39.000Z

Uploaded some videos about the easy way (with Indium) to solder E21A:

New VirEnce MCPCB for E17/E21/119/144/233U Flashlight Modding and DIY Parts

Easy Nichia E21A soldering on Virence VR16SP4 MCPCB . . . Part. 0 (forgot to make it, Kaybi made me shot this one) Pre-soldering the Indium to the MCPCB [Easy Nichia E21A soldering on Virence VR16SP4 MCPCB (Part. 0)] . . . Part. 1 This is a quick and dirty video I made with one of the order. Sorry for the quality, it was hard working with only one hand. The MCPCB was first pre-soldered with Indium at the LED pads. I simply add some flux (common water based Zinc Chloride) …

- Clemence

Barkuti · 2019-02-03T21:36:42.000Z

clemence:

… IMHO, cheaper Indium with less purity should be OK for our uses. 3N purity and up are already enough and can be bought much cheaper via ebay or AE.

Note: you can see from the picture, I even salvaged the Indium balls - Cheapo!

- Clemence

Thanks for your efforts in these regards.

Indium solder… just checked the Wiki article on solder, there are various alloys but, how many of them comercially available? And cheap, of course :-D cheap. What you bought is not what I'd call cheap, not to someone who pays €2,56 for a 100g gross weight reel of Sn63Pb37 shipping included with tracking. :-D

I love efficiency, even when spending money. ;-)

^:)

clemence · 2019-02-16T10:48:00.000Z

Performance test result:

First look at Nichia 119D D340 sm505 R70 Flashlight Modding and DIY Parts

UPDATE 190216: 119D vs 219D test result I finally received 219D samples with same bin to 119D samples received earlier. I supposed to do the test about 3 months ago but I lost my motivation due to disappointing 119D/219D performance. Yesterday I decided to take a one day break from making lights, out of boredom. I don’t care what the output these 119D/219D might be. And these LEDs really have no value to me since they’re limited to R70 and has no significant over current capability. That was a …

- Clemence

The_Driver1 · 2019-02-16T12:05:14.000Z

clemence:

Performance test result:
First look at Nichia 119D D340 sm505 R70 - #26 by clemence

- Clemence

Thanks! So Indium is indeed noticeably better, just like we thought. For extreme LEDs like the Osram Black Flat and White Flat, this is perfect.

clemence · 2019-02-25T22:28:36.000Z

@Agro, you’re right! When the bottleneck is not the solder, Indium doesn’t make much difference. So this should give boosted performance to those high power density three pads LED (DTP-able) more than two pads LED on normal dielectric material. With nano ceramic insulated board with very low thermal resistance the boost was significant. “normal” epoxy/alumina insulated non DTP board there was no gain at all.

I just finished testing two more insulated MCPCB types (four MCPCBs, and 12pcs E21A in total). The results supports your reasoning. Will put the numbers to the Excel first. BRB

- Clemence

clemence · 2019-02-26T20:45:36.000Z

The Indium gain is not as dramatic as in 119D test or perhaps other ceramic based LEDs. Because E17A/E21A has fragile phosphor layer and use polymer base. The phosphor layer burns way faster than ceramic base LEDs before the actual die cooked. I only did controlled test on 4pcs E21A to save time. The rest of the R70 tested at 2A starting point (60 seconds stops for current ramping) just to cross check/validate my finding. The controlled test done with 3 minutes for each steps. This what makes the burnt patterns looks different. Beyond peak current, E21A phosphor layer slowly degrade and burns at several hottest points.

- Clemence