First look at Nichia 119D D340 sm505 R70

UPDATE 180626:
Beam shots added

- Clemence

Any news about the lux tests?

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 loss paying $80 for 119D custom clearance
I’m more interested in proving my theory that LED with 2 pads has higher performance potential than traditional (well, now it’s considered traditional) 3 pads with dedicated thermal bridge. And also to test if hot-rodders will benefit from pure Indium solder,

Found a broken aquarium cooler parts and decided to improve the test for faster readings and more consistent results. Total loss cooling was chosen rather than making a dedicated automatic Peltier cooling unit. Water temp is relative constant (within hours) and provide ample cooling with very few parts needed.

Measurement recorded manually anytime the voltage reading stabilized. With this massive cooling, I can cut down recording time to less than 2 minutes per current steps compared to normal air or passive cooling. My friend asked Nichia for 219D samples, he’s a street light manufacturer. Received 6pcs 219D sm505 D340 R70 samples from him. Here’s the specimens:

Much larger cooling area in 119D. Nichia is very thrifty when it comes to thermal pad unlike Cree which uses as large as possible thermal pad. I have a strong feeling that the larger pads in 119D affects current carrying and current heating characteristic. Smaller pad will be hotter for any amount current it passes through.

Before we begin please check the summary below.

And….crap….I forgot to use the same resolution as my earlier test!! I used the same 100mA resolution from 100mA - 1000mA, and but then 250mA from 1A - 6A. While previous test was done with 200mA resolution from 1A - 5A. So, I had to interpolate some numbers using 1A, 2A, 3A, 4A, and 5A as my anchor points. This is why later you’ll see dots rather than continuous plot in the chart result.
My first test result used derived data from Nichia data sheet and measurement lux. Now using Maukka’s calibrated lights I can measure the real output. The output numbers at 700mA barely meets the specified 340lm - 360lm but this is to be expected since Nichia use 25°C Tj and 10ms pulsed current which I can’t replicate.

And the result was really satisfying:
Inline with my previous test with 319A, aluminum VR16SP4 with normal 63/37 performs slightly lower than copper DTP above rated manufacturer maximum current. But the result with Indium surprised me, I didn’t expect such gain. I’ve been using Indium for all my nano-ceramic soldering just because it allows me to get cleaner and faster result while put less stress to the LEDs. This extra performance was unexpected.
I don’t think good DTP MCPCB would benefit much from pure indium other than reduced soldering temp required (160°C is all you need to reflow Indium). This because I suspect any good DTP MCPCB already near its peak performance limit. Pure indium will also benefit all non DTP MCPCB as long as the LED has relatively large pad(s).

LED remains:

CONCLUSION:
119D/219D
119D has higher potential output than 219D at extreme current
Both have very good output for its (die) size up to rated current (2A).
With their “ezy-dedome” feature is a good choice for small medium power thrower
They’re not good LED for extreme modifications beyond rated current

Extra result observed
With proper MCPCB 2 pads LED have higher potential performance than 3 pads LED because of larger cooling area and less current resistance. This simplifies MCPCB design. Currently, there are more high performance non DTP with very close performance to DTP previously impossible. Cost, while already vastly reduced compared to few years back still limits end product use of these MCPCBs. For example, to get twice the performance of VR16S1 there’s a steep 6x cost increase.
To get higher performance out of nano-ceramic MCPCB without spending too much in MCPCB technology, pure indium solder is a cheaper option affordable for hobbyist like us.

- Clemence

Wow Indium is amazing, but doubt manufacturers will be willing to spend extra for it in the very competitive flashlight world. Thanks for posting the very interesting and useful results.

I mean, if we could manage a bulk order of indium solder, maybe we could get the prices down.

And with LEDs like the XHP70.2/CFT-90 where power density is insane, or even the 1mm2/2mm2 White Flats, Indium could get a nice performance increase.

Nice result, thanks!
Though I wonder how much of it is caused by different thermal paste. I would expect neither solder nor paste to have such big effect…
Maybe it’s about solder microstructure? Voids?

Actually I think that indium could give higher benefit on DTP boards. And with LEDs that have small thermal pads.
That’s because there solder is a bigger part of total thermal resistance.

I would also expect it to be a bigger deal with LEDs that are more capable of being overdriven. There are a few LEDs that have > 8W/mm² of thermal pad.

Yes, the XH35 (HI) and the Osram White Flat LEDs would be a good for an additional test.

Solder paste vs bearing grease don’t make measurable different at all, minuscule if there’s any. I tested it before. Due to these factors:

- The LED cooling pads total surface area is very small compared to the MCPCB contact area.

- Both MCPCB and the copper water block lapped to almost mirror finish on a piece of thick glass using #320, #1000, and #2000 SiC powder. I could fine lap them with superfine diamond lapping paste but didn’t do it.

  • Only very little amount of thermal paste/grease applied, and most of them squeezed off by the overkill clamping force of two M10 bolts

You are probably right about DTP MCPCB benefits more from Indium than non DTP MCPCB. But factor these:

  • Most high performance DTPs has raised thermal pad to make very thin solder bond line. Sinkpad is not as good as Noctigon, Kerui, or perhaps L4P (haven’t tried L4P MCPCB) in that regards because thermal pad height is positioned lower than the anode/cathode pads (thicker solder joint). Very thin solder bond line creates small thermal resistance difference between solder materials (50 W/MK for Sn63Pb37 vs 80 W/MK for In100. But doubling the pad surface area will nett you much higher difference. With 15 microns solder bond line we only get 38% thermal resistance reduction with In100. With the same 15 microns bond line but doubling the area we get 69% thermal resistance reduction.

- Another bottleneck in MCPCB after the solder material itself is the surface finish plating. ENIG is very common these days. The heat from LED has to travel through a layer of nickel. Although very thin, Nickel has thermal conductivity <90 W/MK at LED operating temp.

Stay tune, more test to come!

- Clemence

An interesting read about indium solder in pdf format and how it can react with other metals.

https://www.google.com/url?sa=t&source=web&rct=j&url=https://www.cambridge.org/core/services/aop-cambridge-core/content/view/S1431927606066049&ved=2ahUKEwichK_L48PgAhW8VBUIHXXqBRU4ChAWMAJ6BAgIEAE&usg=AOvVaw2ZQq3-Jqz0qa2OkhYgTYjq

I didn’t know Sinpad had sunken thermal pads. That might have contributed slightly to poor thermal scaling of Luxeon Z ES in Koef3 test…

Thanks for the link, I read it somewhere else but basically the same study about Indium migration through gold. Fortunately all Virence boards except first few prototypes use OSP surface finish.

LEFT to RIGHT:

  • Unknown non DTP “Cree” aluminum MCPCB. Looks OK but the copper trace doesn’t spread good enough. Bubbled and delaminate at 240°C - 260°C.
  • Sinkpad. The first generation was made by punching small indentation from the back. Now the process is refined but looks like only better dies used. I prefer copper plating technique such as Kerui and Noctigon to make the DTP pad slightly raised.
  • Kerui. Although the thermal pad is not levels with the neighboring pads it’s only slightly less than Noctigon. The biggest problem comes from the too small anode/cathode pads.
  • Noctigon. So far this is the best in class, could have been better without ENIG finish both for the pads and the back of the MCPCB.


- Clemence

Thanks. :slight_smile:

IMO, any low melting point temp solder is not as good as it sounds, because diffusion of solder alloy atoms into soldering metal surface is greatly reduced compared to lead-free solder (diffusion increases exponentially with temperature). I've used some BiSn ~150C solder for prototyping before, and mechanically it sucks compared to PbSn, let alone SnAgCu. After soldering I can rip off components with relatively low force, basically every joint becomes like a cold joint, it's more like a weak glue than like a solder.

I suspect the low diffusion could be also bad for metal-to-metal "contact" thermal conductivity, indium has slightly higher thermal conductivity than Pb-free SnAg solder, but thermal conductivities are given for bare metals only.

I don't know what happens at boundary of solder alloy-base metal regarding the thermal conductivity, but my intuition tells me higher temp. solder would perform better even if it has slightly lower thermal conductivity because diffusion of atoms is better/deeper.

Sn52Ag48 is indeed sucks for LED soldering. I only use it for LED wires attachment to make soldering my boards (and those like yours too) easier. But preparing a low temp soldering blob for users is not as simple as it might look.
Sn52Bi48 is very brittle as itself and not a good choice unless mixed with a little bit Pb based solder. To get a proper mix without leaving unmelted 60/40 or 63/37 granules, you have to heat it to slightly above the melting temp of the highest melting point of the contents.
I’ll consider your explanation here for my next test. I can’t comment anything about it yet, have to go back to the classroom to study it. I’m still learning here. To be honest, I’m still not quite sure if it’s the Indium or something else that boost my last test result. I agree SAC305 is superior if we put thermal performance and tin whiskering aside. But for E21A, this is a nightmare. Almost all manufacturers prefer 63/37 than SAC305 for their E21A equipped products. My story with Indium started because of E21A. And never really cared for its superior thermal properties that much. I guess you also agree that thermal conductivity is not as important as thermal resistance. As long as we can make the solder joint very thin, the solder material used would not be too relevant.
Will get back to this after finishing my other tests.

- Clemence

SnBi solder is quite good if you have a lot of surface area, and you want ease of soldering, like with springs.

I’d never use it for LEDs and wires though. It sucks alone due to high brittleness.

You can use normal (low temp) solder for the plus and minus contacts of a normal triple-pad LED. Only the center contact needs the Indium.

Since lead solder is banned in EU and most other countries, it's strange that manufacturers are allowed to use it in their E21A products.

These kind of measurements definitely should not be done with LEDs as heat sources, thermal resistance, Vf, efficiency from LED to LED can vary +-5% for same bin, all this can make dissipated heat different enough to make such tests questionable. This problem could be eliminated only by testing larger number of LEDs, but even 3+ LEDs from same batch for each solder type measurement could give us a more valid clue.

Other solution would be to use small size power resistor in package similar to lfpak33 or something like that, in that case you exactly know how much power is dissipated (U*I), by measuring resistor top side temperature you can compare and see which solder or soldering technique performs better, and by how much. Also you eliminate light measurements which are one of the most inaccurate measurements in technology, temperature measurement is very simple and accurate (PT100 sensor).

Thanks for your input L4P

- Clemence

Currently, I don’t have all needed setup to make a valid experiments. And after some thoughts, it’s better if someone else do the test. People would think I’m biased to something that I sell. I’ll gladly provide some test specimens and materials to those who wants. Djozz, Koef3, Maukka perhaps?

In the mean time please hold all conclusions until a more valid test is done.

- Clemence

I know I am late to the party, This thread was just pointed out to me. I was not aware that a 219D was out.

How is the tint on it? Do we finally get rosy 219B tints with 219C output?

I have not been able to find a high CRI LED I like from any manufacture since the 219B 9080. They are all green!