New VirEnce MCPCB for E17/E21/119/144/233U

339 posts / 0 new
Last post
clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Update171226: OP updated with VR16SP4 CCT mixing test results (scroll to the lowest pictures)

- Clemence

Jensen567
Offline
Last seen: 5 months 1 week ago
Joined: 04/10/2017 - 16:59
Posts: 491
Location: MD/DC Area USA

New test of E21A quadtrix. My second reflow attempt this time. Change to the test setup included an addition of a lux meter, though not a great one and not mounted well, but it did provide some insight into peak output.

Below is an image compiled by Clemence showing all of the IR images taken in the test, as well as some plots of the data taken. Emitter setup was 2S2P. Please note with the Lux data that I did not turn off the room lighting, so it started out at non-zero. The setup also means that the number could be changed quite a bit just by moving around the room, so they are very approximate.

On this test I went beyond the 150C phosphor temp based on lux meter readings. Peak output occurred at 5.5A!

Phosphor temps were imaged up to 5A, at 5.5A I unfortunately messed up and didn’t save an image, but phosphor temps were bouncing around the 200C mark. At 6A I did not take an image as I saw the temps at 210 and rising and the lux meter was starting to drop fast, so the test was immediately shut down.

This was the result of that 6A run, slightly browned edges right where the two hot spots are between the emitters. If i hadn’t shut it down quickly I predict it would have ended up like the emitters Clemence shows in the 1st post. This also proves the photon absorption, and I am fairly confident the Rev.2 boards with larger gaps between emitters will be capable of even higher output.

After this I tested a hypothesis from an observation of my last test. I re-ran the emitter at 4A for about 5 minutes, and then ran the full sweep from 0.5A to 4A again, with only about 30s per step, just to check Lux. Lux was basically unchanged, I could move my arm and change the number more. Visually the browning is now gone too. Not sure what causes this phenomena, but Clemence has seen the original images with their attached timestamp data and can verify.

It will be interesting to see if more people push these LEDs and also observe this strange healing effect of the phosphors. Will also be interesting to see how the new gapped MCPCB layout changes performance and the hot spots.

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Thanks Jensen, you’ve been extremely helpful in VirEnce board development and testing
As we discussed earlier last (my) night, there are many factors regarding this wolverine LED. It’s so interesting that the output didn’t change that much (you said the change was too small to be meaningful). I got the same finding but because I cleaned them for the second test, I assumed that was the IPA. I’ll ask Nichia today for surer answer or insights.
Another important thing is: E21A is VERY sensitive to solder quality. My results were on par with your first test (too much solder). I used leadfree solder with worse wetting hence the
Thicker solder joints. Like I reported two days ago, I got so much better result and way easier reflow using cheap 60/40 solder. Your 63/37 is the way to go. For extreme drive current, E21A must be reflowed with 63/37 or better. Lower reflow temp also keeps the solder mask white.

- Clemence

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Found this mother next to the board.




FYI: the boards have arrived. Will check it tomorrow in the custom office. Smile

- Clemence

mattlward
mattlward's picture
Offline
Last seen: 10 hours 57 min ago
Joined: 06/19/2015 - 09:20
Posts: 2684
Location: Illinois, USA

Testing with 2S2P so… 6 volts’ish in?

EDC rotation:
FW1A, LH351D 4000k (second favorite)
FW3A, LH351D 3500k
FW3A, SST20 FD2 4000k
FW3A, Nichia 4000k sw40 r9080 (favorite light!)
FW3A, Cree XP-L Hi 5A3
Emisar D4V2, SST20 4000k
S2+, XM-L2 T6 4C

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia
mattlward wrote:
Testing with 2S2P so… 6 volts’ish in?

Correct Matt
Please check the pictures in below link. I added some configurations diagram there
https://www.virence.com/blank-rwrlr/vr16sp4

- Clemence

mattlward
mattlward's picture
Offline
Last seen: 10 hours 57 min ago
Joined: 06/19/2015 - 09:20
Posts: 2684
Location: Illinois, USA

If I am looking at the pics correctly, for a 3V setup one needs 2+ and 2- wires?

EDC rotation:
FW1A, LH351D 4000k (second favorite)
FW3A, LH351D 3500k
FW3A, SST20 FD2 4000k
FW3A, Nichia 4000k sw40 r9080 (favorite light!)
FW3A, Cree XP-L Hi 5A3
Emisar D4V2, SST20 4000k
S2+, XM-L2 T6 4C

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia
mattlward wrote:
If I am looking at the pics correctly, for a 3V setup one needs 2+ and 2- wires?

Unfortunately yes. No space left on this tiny board to create jumper pads without compromising optic’s base or safe side margin.
I’ll make add on secondary jumper board for use on top of it in the near future.

- Clemence

djozz
djozz's picture
Offline
Last seen: 15 min 30 sec ago
Joined: 09/07/2012 - 17:04
Posts: 15952
Location: Amsterdam

Cool pics! Is that some species of paper wasp?

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia
djozz wrote:
Cool pics! Is that some species of paper wasp?

I think so Looks like its a digger wasp (Sceliphrinae). She didn’t care about anything but her egg safety. Filled two of the M3 screw threads on my testing block.

- Clemence.

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Specification:
Changes are highlighted in bold

VR16S1
- MCPCB thickness: 1,65mm ±0,1mm
- Dielectric thickness: 10µm ±3µm
- Dielectric thermal conductivity: 7,5W/MK
- Copper trace: 70µm
- Solder mask thickness: 15µm ±5µm
– Anode – cathode gap: adjustable from 0,5mm down to 0,2mm (was 0,5mm to 0,3mm) by scraping off the masking
– Brighter reflective white solder mask
– Tougher solder mask (untested): prevent yellowing in sustained leadfree 260C environment.
– Copper pads plating: OSP (was ENIG) to get every last bit of thermal performance
– Flat section for V-scoring 15mm maximum (was fully round): mass production consequence
- Routed MCPCB’s outline, not punched to get the flattest possible bottom.

VR16SP4
- MCPCB thickness: 1,65mm ±0,1mm
- Dielectric thickness: 10µm ±3µm
- Dielectric thermal conductivity: 7,5W/MK
– Copper trace: 35µm (was 70µm) for better reflow process
- Solder mask thickness: 15µm ±5µm
– Anode – cathode gap: 0,2mm (was 0,3mm)
– Final quadtrix footprint: slightly adjustable from 4,4 × 4,4mm to 4,6mm x 4,6mm (was 4,2 × 4,2mm to 4,4 × 4,4mm).
– Brighter reflective white solder mask
– Tougher solder mask (untested): prevent yellowing in sustained leadfree 260C environment.
– Copper pads plating: OSP (was ENIG) to get every last bit of thermal performance
– Flat section for V-scoring 15mm maximum (was fully round): mass production consequence
- Routed MCPCB’s outline, not punched to get the flattest possible bottom.

- Clemence

g_damian
g_damian's picture
Offline
Last seen: 2 hours 24 min ago
Joined: 08/20/2017 - 06:52
Posts: 461

Clemence, how hard would be to install vr16s1 with 144 in KF8 ? Seems like perfect host for that combination, except mcpcb size is 20 instead of 16…

Jensen567
Offline
Last seen: 5 months 1 week ago
Joined: 04/10/2017 - 16:59
Posts: 491
Location: MD/DC Area USA

You can always run a 16mm PCB in a 20mm spot with no problems. The light should have a centering ring for the reflector and emitter that will ensure the MCPCB is held in the correct spot.

However, jf there is no clamping pressure from the reflector to hold the MCPCB to the host, and instead the host relied on screws to hold the MCPCB down, it is likely you can just use washers on the existing screws to clamp the MCPCB.

steel_1024
steel_1024's picture
Offline
Last seen: 16 hours 30 min ago
Joined: 10/17/2016 - 14:06
Posts: 486
Location: Taiwan

Unable to checkout by PayPal?

Order No. 10131

Sorry for my poor english.

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

steel_1024 wrote:
Unable to checkout by PayPal?

Order No. 10131

Could be your ends. Looks like anyone could do it without problem. Try to clear your cookies and cache first.

- Clemence

steel_1024
steel_1024's picture
Offline
Last seen: 16 hours 30 min ago
Joined: 10/17/2016 - 14:06
Posts: 486
Location: Taiwan

again

EDIT
It may be a problem with my VISA card.

Sorry for my poor english.

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Great news!
The VR16SP4 alpha board test was a success! Although the test could not be completed due to inadequate power supply, the result was already very near it’s peak.

Reflow preparations:
Magnifier application and phone stand is invaluable when working with E21A

Dropping the solder paste. I’m done trying leadfree with E21A. Pb63Sn37 is the best.

Carefully even out the drops to match the pads and then placing the LED, try not to smear the applied paste too much. In this picture I dropped the LED too far away from the target.

Fine tip angled tweezer works best for me. Next to the ceramic tweezer is the press to open tweezer. You’ll need to decrease the tension if you want to use PTO tweezer, E21A’s sides are very soft/fragile.

Ready to go to my “deluxe” gas kitchen stove reflow oven… The final positions will be adjusted later, I tap each LED tried not to move them sideways to make sure the flux wet all pads evenly.

Done, cleaned thouroughly with IPA and toothbrush. Ready for the test

PSU, DMM, Lux meter, Digital thermometer + probe, Test block, big heat pipe fanned cooler, and notepad. Where’s Wally?

I found the heat sink in the last minute, it wasn’t screwed, just sat on top the block using my trusty Moly-graphite bearing grease. It was too late to drill and tap mounting threads, perhaps later.

THE RESULT

New record breaking!! 11,2A at whooping 35,6 watt for such tiny LEDs

The test ended prematurely because my PSU can only supply up to 48 watts. But the curve is already near it’s peak (I predict it should be around 12A). Single E21A can be driven up to 3,1A using my VR16S1 proto board. We assumed 4xE21A in 4P should be able to up to 12A, correct? But when they cramped together in gapless VR16SP4 proto board in earlier test, max current was only 4,8A. Stopped prematurely by the photon cross talks (burnt phosphor’s layer). Now without the photons heating the phosphor layers, the new board can unleash each E21A to their maximum.

Start test block temp was 27,3C, test ended with block temp at 35.3C. I held them at 11,2A for more than 5 minutes while texting and sending pictures to Jensen567. The lux, temperature, and voltage reading stayed the same! Here’s the after test picture, no visible damage, like new. The chipped phosphors caused by the ceramic tweezer when I positioned them in the oven.

Just an Idea to extract more lumens:
If we pot it with something like this 10mm spherical glass lens using optical silicone, more photons can be extracted while making the LEDs cooler.

Looks familiar?

Going for a deep sleep now.
- Clemence

MRsDNF
MRsDNF's picture
Offline
Last seen: 38 min 53 sec ago
Joined: 12/22/2011 - 21:18
Posts: 12983
Location: A light beam away from the missus in the land of Aus.

Amazing work Clemence. I dont understand it all but appreciate the effort put in. Beer

 

djozz quotes, "it came with chinese lettering that is chinese to me".

                      "My man mousehole needs one too"

old4570 said "I'm not an expert , so don't suffer from any such technical restrictions".

Old-Lumens. Highly admired and cherished member of Budget Light Forum. 11.5.2011 - 20.12.16. RIP.

 

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

MRsDNF wrote:
Amazing work Clemence. I dont understand it all but appreciate the effort put in. Beer
You’re welcome MR. Steve Grad
djozz
djozz's picture
Offline
Last seen: 15 min 30 sec ago
Joined: 09/07/2012 - 17:04
Posts: 15952
Location: Amsterdam

Nice test! So how does Cree manage to cram the dies so close together in the XHP35, XHP50.2 and XHP70.2?

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

Until someone test the BLF absolute max limit, I predict the limit for quadtrix E21A on VR16SP4 is as follow:
BLF absolute limits:
- 12A in 4p (3V)
- 6A in 2s2p (6V)
- 3A in 2s+2p/4s(9V/12V)

BLF limits (80%), for sustainable extended output:
- 9,6A in 4p (3V)
- 4,8A in 2s2p (6V)
- 2,4A in 2s+2p/4s(9V/12V)

Nichia absolute limits (to get the min 60.000 hours L70):
- 5,6A in 4p (3V)
- 2,8A in 2s2p (6V)
- 1,4A in 2s+2p/4s(9V/12V)

- Clemence

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia
djozz wrote:
Nice test! So how does Cree manage to cram the dies so close together in the XHP35, XHP50.2 and XHP70.2?

The same thing goes for closely packed 144A dies. With single uniform phosphors and silicone layer on top the dies, the photons cross talks is not there. Each time the photon crossing the silicone/air boundary some of it will be reflected back inside, adding more heat. In separated dies like the this quadtrix, adjacent exited photons get absorbed by the nearest walls, adding even more heat.
This is why Nichia uses some kind of reflective wall (soft white silicone) surrounding the E17A/E21A dies – to minimize photon cross talks (so they can cram more E17A/E21A in smaller LES). Without this barrier, we’ll need much larger space.
The barrier also what makes tint shift (from high angle emission, like those found in 144A and XPG3) is minimal with E series

- Clemence

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

With proper 0,2mm gap E21A centers easily on VR16S1.

ma tumba
Offline
Last seen: 1 week 1 day ago
Joined: 07/28/2015 - 19:09
Posts: 171
Location: Russia

They look great!

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

CLOSER LOOK OF THE BOARDS

Quick sanding using #1000,#1500, and #2000 sand paper revealed the narrower gap. But the copper trace’s edge is rounded, I have to dig deeper into the crevice to remove the solder masking which I didn’t do. From many previous tests we knew for this particular LED, the junction temperature is far lower than that in the phosphors layer. The top layer will be well done long before the die overheats. Djozz also did his test (E17A & E21A 5000K R8000) with only 1/3 of each pad sitting on the copper trace resulted in approximately same max output limit.

If I didn’t know about photon cross-talks, I would reflow them with no gap. These LEDs will replace those in my scooter (plus more powerful driver). They’re the same sm503 D220 R9080

Beauty!

Unlike 2 Oz copper trace in VR16S1, the thinner trace in VR16SP4 is more accurate with well defined sharper edges. This is why I decided to use only 1Oz. copper on it. I’m amazed at how accurate the pads aperture made for it’s tiny size (1,26mm x 0,86mm)

- Clemence

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia
ma tumba wrote:
They look great!

Those were yours Wink ! Don’t worry about them being not parallel to each other, I re-align the positions before shipping.

- Clemence

ma tumba
Offline
Last seen: 1 week 1 day ago
Joined: 07/28/2015 - 19:09
Posts: 171
Location: Russia

That is why I was so happy Smile

ma tumba
Offline
Last seen: 1 week 1 day ago
Joined: 07/28/2015 - 19:09
Posts: 171
Location: Russia

With these quad boards and each of the color temps of the E21 being equally useful in different circumstances I think it is time to revisit RGBW-type drivers. In terms of UI, the closest thing I’ve been able to find is this simple djozz’s hack. In case of quad boards each channel would go to individual E21. This is something I will certainly attempt to do when I have tested these boards with these LEDs.

What would be ideal to have though is a full featured programmable driver which would at least allow one to program current setting for each individual channel (the UI being the same, ie progressive switching to the next LED) but preferably the ability to also set specific switching for an arbitrary amount of modes. For example mode1 = LED1@10mamp, mode2=(LED1+LED2)@1amp.

So far I’ve been able to identify here only one RGBW driver more advanced that the djozz’s hack, which is this one, but the firmware went in a very different direction than what I am discussing here.

clemence
clemence's picture
Offline
Last seen: 2 hours 29 min ago
Joined: 07/12/2015 - 02:58
Posts: 2338
Location: Bali - Indonesia

UPDATE 180122:
- NVSWE21AT SM653-D240-R9080 shipped from Japan today

- Clemence

djozz
djozz's picture
Offline
Last seen: 15 min 30 sec ago
Joined: 09/07/2012 - 17:04
Posts: 15952
Location: Amsterdam

Interesting, what does 6500K do with almost 100 CRI?? Not directly my COT but I doubt that I can resist some Smile

I think that ma tumba has completed his quad with these!

Pages