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

Hi guys,

I'm going to make custom copper round CDTP MCPCB for a single Nichia 144A.

This new board will have several features:

  • Non insulated for superior heatsinking with direct soldering OR Insulated with options to use either thermal adhesives or pad
  • High reflectance white masking ink
  • High temperature masking ink (lead free/direct soldering friendly - less yellowing)
  • Hopefully will performs same or better than my previous 30mm x 30mm VirEnce board (will post the test result later)

The sneak peek of the design is in my avatar. The questions are:

  • Diameter (tell me your favourites, could be more than just one size)
  • Thickness
  • Optics (tell me your favourites, could be more than just one type)
  • Slots? (most used cable thicknesses)

Please post your suggestions. Thank you!

May the Photons be with us,

Clemence

-----------------------------------------

Update 170913:

I finally able to start the project. Thanks for all the support guys.

After many hours of sketching and discussions with some BLF brainiacs, I sent the final models up for prototyping today. The process alone will take 4 weeks, excluding shipping and document preparation. I'll report when I get the first 60 pcs of sample later.

Below are the two models undergoing processing:

16mm 5050 footprint quad [E17/E21 only] configurable for 3V, 6V, 9V, and 12V

16mm single [E17/E21/119/144/233]

--------------------------------------------------

Update 171016:

- 60 pcs prototypes shipped!

--------------------------------------------------

Update 171025: Finally the first prototypes received.

These are the most ultra EXPENSIVE MCPCB I've ever had. Here's the preview:

-------------------------------------------------

Update 171026:

Hooray! New record for 144AM.

The new TINY Aluminum NON DTP 1,5mm thick 16mm VR16S1, beats my older bulky 2mm thick 30mm x 30mm (CDTP copper).

This is a great achievement!

Peaked at 6,9 - 7,1A & 6,48V (shallow decline after 7A). Previous record was 6,4A & 6,2V (way hotter and drops rapidly after peaking). The quick test up to 8A didn't kill the LED yet. I didn't measured it since it's already dropping down, and was too excited that I accidentally moved the lux meter position. It must be brighter since the LED vF was lower.

I never thought the result would be even close to the early test. I couldn't measure the lumen output but the lux curve and voltage reading is a valid data to conclude this.

Early test by Djozz:

Almost 50watt of raw output. My estimation based on Djozz's test, this new board is capable to bring 144AM/AR E1300 R70 to output close to 3700 Djozz lumens and 3000 Djozz lumens for the R9050. The correct voltage shown by the DMM in the right.

T-connections to measure LED voltage

The poor guinea pig undergone many reflows and scrapings. The solder mask is indeed very thin and fragile. I asked for it to create the thinnest bump possible between pads. Even though the solder mask is fragile, the super thin 20nm (nano meter) dielectric withstands multiple 260C reflow cycles successfully with no delamination at all. Wire soldering was rather difficult due to excellent thermal dissipation. I abused my 30 - 200 watt solder gun extensively to create simple wire joint. It's like soldering the thermal pad of DTP boards.

The board mounted using only Moly-graphite automotive grease and two M3 x 0,5 screws.

That's all for today. Tomorrow I will test the quad E21A.

--------------------------------------------------------------------------

Update 171027: E21A single and quadtrix test results

Here's how to convert the VR16S1 to fit single E21A. The gap between pads was 0,5mm. By scraping the solder mask using sharp artists knife you can get 0,3mm gap. I asked for 0,2mm but the 2 oz. copper trace makes it difficult to maintain close tolerance. The test sampling from the manufacturer resulted in 2,9 - 0,31mm gap. Let's hope the next production stuff can make it closer to the target 0,2mm (+/- 0,05).

I used polyimide high temp tape (similar to Kapton tape) to keep the LED from twisting during reflow. You can see later in the picture that this tiny E21A is very prone to twisting. And the result was uneven heat spreading across the tiny thermal pads. I learnt that next time I must place the tape at the cathode side. E21A has different anode - cathode design. The cathode has larger surface area, hence stronger solder adhesion during reflow.

Various deaths of E21A:

top: twisted uneven thermal load between anode - cathode resulted in burnt phosphors started from the side. EDIT 171104: any slow death always started by side burns. The only path to cool the phosphor/silicone protective layer are: air and the die.

middle: good centered LED resulted in burnt mark started from the center (the hottest point) EDIT 171104: Quick current sweep to the known maximum resulted in center burnt mark. This happens because the phosphor and silicone protective layer still relatively cold at the sides.

bottom: combination of dirty (non visible at first) LED caused by burnt flux residue, recycled E21A's from previous reflow attempts, and of course tightly packed matrix. EDIT 171104: In cramped gapless config, the burns always start in spots farthest from the anode and cathode in the boundaries between LEDs. The crossing photons heats the neighboring LEDs.

I expected the problem when I check the board after reflow. I saw oily substance creeping out from the seams. Cleaned it with IPA but later during the test, the smoke was visible even at 1A. The trapped flux, liquify at high temperature and narrow gaps between LEDs acted as the capillary path.

The only cure is to open up the space farther. I plan to make at least 0,4mm gaps. It's needed to prevent trapped flux and photon crosstalks. The benefit of better beam shape isn't justify the reduced capability. Please keep in mind that VR16SP4 isn't designed for thrower. The beam pattern behaviour could be like those in XHP50

Here's why the phosphor doesn't have regular thermal path like "normal" LED. Usually the phosphor sits on top the AlN or SiC substrate separated only by thin reflective layer. The thicker silicone coating also aids in photon extraction better (less heat buildup)

You can see that even at very "low" 350mA current the phosphor already 30C hotter than the die. No wonder it got burnt way faster than the die. Above example are using generic performance MCPCB. The only thing we can do is to reduce die temperature as low as possible and prevent any unnecessary photon cross talks (by opening the gaps).

And here's the combined chart:

CONCLUSIONS:

  • This board is an excellent choice for many two pads Nichia LED such as 144, 119, 233 and could be used with many other brands too. The performance, in my opinion is very close and COULD even better than any DTP boards (2 pads LED has more exposed total area/footprint ration than 3 pads LED). I will try to test 119 and 219 LEDs using both VR16S1 and any decent copper DTP later.
  • E21A maximum drive current doesn't significantly improved by the low thermal resistance of these boards. Only 200mA higher than common copper MCPCB in VR16S1. The brightness below maximum current and lower junction temperature are the most interesting gain to be expected.
  • E21A maximum drive current is limited by the phosphor temperature. As can be seen from my earlier test (the Twig-light thread) with the burnt phosphor/encapsulant (the white reflective silicone blanket) removed, the bare blue light output increased steadily to ~ 6A region.
  • The quadtrix board VR16SP4 could be improved by opening the gap further to 0,2mm between LEDs to make reflowing easier. Now the total footprint is only 4,2mm with gapless version, still plenty of (0,8mm) margin to make it compatible with many 5050 optics. I will test it again using clean fresh LED later.
  • Watt per watt, the quadtrix E21A efficiency was higher than 144A even tested using "dirty" used E21As. I tested using the highest binned LED: D320 for E21A and E1200 for 144AM
  • in single setup E21A max drive current was 3A. I expect something like 8-9A total in a proper 4p (VR16SP4) config.
  • In 4p config in VR16SP4, the voltage for each E21A were very low. I think this is the combination of less resistance (Ohm's law) and hotter junction temp. The lower voltage recorded consistently across the current range (0,1 A - 3,3 A)
  • There's no point in using single E21A in most applications higher than 0,2A / LED (if luminous flux is what you're after). You can see from the chart above, there was HUGE efficiency and output gain by using 4p E21A compared to 1s E21A (~ same wattage).

I will make another test later and revise the above points if they're not relevant anymore. Meanwhile those are my latest findings.

-------------------------------------------------

Update 171107: some more tests and comparisons

For those curious about how the good old XPG performed on copper DTP....I couldn't find any XPG tested on copper DTP in BLF.

You'll also find how technology has gone very far back from 2006 to 2017.

First: the LED - much lower voltage yet almost twice as bright (more than twice if I were using newest Cree)

Second:the MCPCB - LED with thermal pad is not as critical as before for most normal applications

Those test performed rather unfair but should give you the idea. With dedicated "two pads" LED such as 119 or XBD series the result will be much better. FYI, VR16S1: 2 Oz. copper; Sinkpad: 1 Oz. copper.

The result with 319A D440f2 on DTP peaked at slightly lower current (5,75A) than what Texas_Ace did (6-6,25A). This, I believe due to the extra hot tropical temp (tested in 30-32C, 88% humidity).

Below are the LEDs after the test:

CCT mixing test results:

------------------------------------------------

Update: 171223 Some excellent tests from Godamn city performed by Rob-inn (Badman's equipments).

https://budgetlightforum.com/t/-/42898/127

https://budgetlightforum.com/t/-/42898/140

https://budgetlightforum.com/t/-/42898/145

and he said there are more to come....

------------------------------------------------

Update 171226: VR16SP4 CCT mixing test result

------------------------------------------------

Update 180119:

Latest test for the production VR16SP4:

https://budgetlightforum.com/t/-/42898/168

------------------------------------------------

UPDATE 180706:

New variants, VR21SP4 (quadtrix E21A) and VR21P4 (quad Optisolis/757). Same board material and technology as previous one.

https://budgetlightforum.com/t/-/42898/225

-------------------------------------------------

UPDATE 191104

New MCPCB

Detail: https://budgetlightforum.com/t/-/42898/320

- Clemence

That’s what I’ve been waiting for!

Diameter: 16 and 20 mm
Thickness: 1.6-1.8 mm
Slots: 2x3.0 mm

Wouldn’t 4 slots (2 a bit smaller for screws to hold it centered, and 2 for wires) be better?

Yes, definitely holes for screws in addition to the wire slots.

Are there any standard screw placements for each 16mm and 20mm?
The screw holes need insulated vias or additional insulated washer for these boards. For 16mm board, even 2,5mm screw would be too big. It has to pressed down either using the reflector, optics, or additional fastening method if thermal pad is to be used. Or we can use thermal adhesives.

I plan to use thermal pad with 12-16W/M.K underneath the base, far superior than the best Arctic Silver. There’s thermal adhesives with proven 30+ W/M.K conductance from Australian manufacturer. The bad news: it’s permanent and has only 3 months shelf life.
Checked the AIN but still too expensive for small batch quantity.

Thermal pad?

Any of you remember Aluminum Nitride?

#80 post in “Nichia 144A and 229A series: bigger dies, more output, 90CRI included,.......but no thermal slug :-(

Well, of course getting the MCPCB manufactured with a thin layer of that stuff somewhere would be handy.

Yet notwithstanding, it should be possible to buy some sheet/film. For example: http://www.goodfellow.com/E/Aluminium-Nitride'-Sheet.html

If you are willing to take a peek: aluminum+OR+aluminium+nitride+sheet+OR+film Google search

Cheers lasses/lads ^:)

I do remember it. AIN stands for Aluminum Nitride. I was referring to AIN coating to the base of MCPCB.
Nice find BTW Bark, will read them data sheets tonight.

Lurked into the price list and found the 0,64mm thick cost GBP 148 for a square piece of 25mm x 25mm!!! Usable for only a single 20mm board.
You can get lower price for 20 pcs but it still GBP 33,8/piece

Far more expensive than what I got. No good for BudgetLessFellow :weary:

Yes, for 20mm PCBs there are standards. I would use the same screw hole placements as Sinkpad and all the other companies do. Here at the bottom of the page are some pdf documents with measurements for 20mm Star PCBs from a German company.

The link is not working

Strange, it’s working now :stuck_out_tongue:

It works for me ;).
Generally this is a great idea! These will be much more practical compared to your first design. When can you have them ready?

Scheduled for January….
My first design wasn’t intended for flashlight, it’s for my personal specific use. This was a perfect size/shape for prototyping many general lightings with 144A/119 series.

I’m in the group buy for 2 of your “universal” boards and I’ll honor that commitment, but I’m definitely interested in some of these boards too. 16s and 20s seem like a great start.

Very clever design, Clemence! By the way, it’s really great to have you amongst us!

=)
Not possible without BLF & TLF supports

Let’s not hope much. CDTP design is highly surface area dependent, smaller surface area = higher Tj. That means bigger board will be better. But this new design utilizing maximum possible surface area possible for both anode/cathode. I also try to reduce current heating as much as possible. Using dielectric with as high as possible thermal conductivity I could get. I predict it would be a bit less cooler than the standard VirEnce S30 but in smaller 16mm package.

Fellows, you're venturing into new territory here with a cathode direct thermal path board.

The lack of electric isolation is going to be a real hurdle, I think. Something which could be fixed with a thin layer of some über-high conductivity material. If the cathode track pad is reasonably maximized in surface and you can get some thin layer of über-conductive electrical isolator below, the problem is solved, and the industry moves forward.

Pretty sure there's someone around there who can offer aluminium nitride/beryllium oxide sheet/foil at much less than those prices, clemence. Hell, you could get diamond foil for a lower quote, LoL!

Why don't you ask for some help from a specialized manufacturer? More or less recently some Shenzen Kerui Electronic Industry representative has been around here promoting their products. More info: https://www.linkedin.com/company/shezhen-kerui-electronic-industry-co-ltd-

Cheers ^:)

It’s indeed universal. I use some for my desk and workshop light with assymetric TIR optics. Plan to use it for my parking area lighting with Ledil Jenny CY optics.

Thanks Bark,
Even without super material CDTP can be designed to surpass neutral DTP approach as long as you have the required surface area. The idea is to lower the heat potential difference (does anyone know shorter terms for this?) thus less thermal resistance. This is pretty much why Nichia still manufacture 2 pads design.
In theory two pads design can achieve lower Tj compared to three pads design (ignoring the surface area required) since copper used in cathode/anode has much better thermal conductance than the dielectric material commonly used in LED. Combined together with very high conductance and thin solder joint such as Indium solder, the result should be even better.

Upstream just at the heat source, heat flux intensity is higher that downstream where the surface area is cooler and larger.

For example: If a 119 has to pass 10 watt of power through two tiny solder pads. The heat intensity is roughly 10 watt/11,9 mm^2 (0,84 W/mm^2). After heat flux passing the solder pads, the intensity lowers down to 10 watt/201 mm^2 (0,05 W/mm^2) using a circular 16mm MCPCB.

My simple logics:

- Upstream:
common Sn60Pb40 thermal conductance ~50 W/M.K
heat intensity ~ 0,84 W/mm^2

- Downstream
Thermal pad material with X thermal conductance
heat intensity ~0,05 W/mm^2
To get the same state, thus X= ~2,9 W/M.K

I still don’t have any valid data to stand up for this one yet. I rather make some and then test it based on that logic. The universal VirEnce board test result was higher than expected. And I think (emphasizing the: I think), solder choice is one of the most important thing in CDTP design, it’s where the bottleneck occurs (not downstream at the MCPCB’s thermal pad). As long as we use MCPCB with surface area reasonably LARGER than the footprint of the LED’s solder pads.

I’m using Sn96,5Ag3Cu0,5 for most of my newer project now. It’s harder on the boards (218C liquidus) than 60/40 and requires extra effort to get a good reflowed boards. The thermal conductance is ~58-60 W/M.K.
I have a spool of SnxxAgxx (lost the label) solder wire rated at 70W/M.K but requires 240C to a flowable liquidus. Much harder and stronger joint very useful for aluminum soldering.

The larger the MCPCB surface/mating area to the heatsink, the less sophisticated the thermal pad required.

That’s the ideas behind my idea….