Without ultrasonic cleaner, the only way to properly clean the residue is to heat the module to 80C (on hot plate), let the gunk thins and flows - soak in hot IPA bath, brush it - reheat to 80C - resoak - rebrush - etc….until you don’t see any flux residue coming out when you warm the module.
Unless you want to run it hard, simple brushing in hot IPA is enough. I did super clean cleaning for VR21SP4 because it has narrower gaps at only 0,25mm. In full power hot run, without proper cleaning the flux will “climbs” to the phosphor layer. With VR16SP4, the gap is 0,4mm and won’t cause too much problem. With crowded E21A arrays, anything closer to 2A/LED requires meticulous soldering and post cleaning.
This is my first attempt with gapless design. Check the bottom picture
Yes, if you want to show the colors differences better, the camera needs to be calibrated on a same (true) white point for each different light CCT before shooting the picture, or you can put a true white chart in the corner of the picture and correct the white point afterward.
The 219B SW45K has a bit more chroma and hue shifts than the E21A which has better colors fidelity.
I guess that the 219B is saturating the wood colors to make it looks a bit warmer than in reality. I better trust the E21A despite its cooler aspects, the wood probably has those palish/grey-ish colors, being discolored by sun’s UV.
Jeez, I never thought that soldering E21A will be this complicated. I guess I have to buy ultrasonic cleaner then. In the description under your video(“Easy Nichia E21A soldering on Virence VR16SP4 MCPCB (Part. 1)”) you mention that you heat MCPCB until E21A gets stuck to the pads then cool it and after that you add more flux(on the E21As?) and reheat it to finalize the soldering. Is this two step process really needed? Can I just wet the pads with the soldering alloy add some flux place E21 and heat it till it is soldered? Or do I need to add more flux after that and reheat it? Won’t zinc chloride damage the led surface?
P.S. Thank you very much.
I would gladly buy a presoldered one from Clemence or anybody else if I could. But the MCPCB that I need - VR16SP4 is out of stock and probably will be out of stock for a long time.
I’ve been soldering with Sn63Pb37 paste for probably 10 years, but I let Clemence solder mine since he must be skilled at it. He didn’t disappoint, they look great. I figured it would have taken me at least 2 quads to perfect my technique, and 2 is all I needed. Money well spent.
Edit: We posted at the same time, haha. I see. I understand.
If soldering is easy then you should have seen many E17A or E21A in the market already. From what I’ve heard from Nichia, they don’t want to suggest this LED for mass production unless the factory has very good reflow soldering system (actually, it’s not the reflowing, explain later). IMO, for small production scale, manual soldering E17A/E21A yield the highest solder joint quality with far less defect.
The easiest method to solder multiple E17A/E21A in tight array such as quadtrix, hexatrix, etc… is to use Non Solder Mask Defined (NSMD) MCPCB. In this case, it’s the copper trace that define the LED position rather than the masking. It’s an open trace design without any solder mask near the LED/LED array. But performance would be lower since there is no over sized copper trace to spread the heat.
I did the soldering in two steps because I used Indium and aggressive water based flux. Since Indium only available in solid wires or ingots I have to first solder some blobs to the MCPCB. Later flux dropped to the solder points and LED carefully placed on top. This flux reacts violently with copper at slightly more than 100C and because it’s water based, it boils at around 100°C. The reaction produce lots of tiny bubbles and easily flings LEDs off the MCPCB, in most cases they’re just flipped upside down. To overcome this, slow heating is the answer. But slow heating creates more thick flux residue before the solder welded completely to LED pads. The first step used to temporarily fix the LED to the MCPCB to prevent them jumping under fast heating. The second step with more flux added, faster heating won’t cause the LEDs to jump off the MCPCB because they were fixed before. This second heating meant to perfect the positioning and finalize solder joint.
You can do the same method with any solder/flux, but cleaning the alcohol based flux residue is harder. I chose water based flux because it can be cleaned so much easier. There’s a trade off for water based flux: it’s so aggressive it will brown (not just yellow) the solder mask above 200°C. That’s why I used Indium with very low (157°C) melting temp (and added performance benefit). Don’t worry about cleaning the flux from underneath E21A, it’s easy as long as you have access to low power ultrasonic cleaner. My proven and cleanest way to clean E17A/E21A:
With water based flux:
- Use only <100 watt 40kHz ultrasonic cleaner, below that frequency it’s too harsh and would damage the LED instantly.
- Submerged in minimum 50°C water + 1% dish washing liquid mix. Never use use laundry detergent or anything contains chlorine or sodium bicarbonate. Run the UC while shaking the module in LESS than 30 seconds. Blow dry the LED with low pressure compressed air, I use 15 psi max.
Repeat UC cleaning in distilled water TWICE. Any leftover zinc chloride flux will corrode your module
With alcohol based “no clean” flux
- Soak the soldered LED in 50°C - 80°C IPA (any 70% - 95% works equally as good) for at least 10 minutes
- Use only <100 watt 40kHz ultrasonic cleaner, below that frequency it’s too harsh and would damage the LED instantly.
- Submerged in hot IPA as above, run the UC while shaking the module in LESS than 30 seconds. The solder mask in Virence MCPCB (especially in VR21SP4) is so thin it will be damaged in prolonged exposure to UC agitation.
Test: heat the module to 80°C and inspect if there’s any flux residue melt off from beneath the LED under a magnifying glass.
There will be VR16SP4m (4x E17A), redesigned VR16SP4, and (probably) VR35SP16. All drawings already sent to the factory last week. Should be faster than 2 months. I dropped the idea of 20mm triple E21A because it’s less universal and omit the color mixing feature. Small triple is also limited to Carclo which has very limited beam options.
I’m wondering if there’s any advantage of using 4x E17A instead of 4xE21A in the 16mm board? I would think 4xE17A would be ideal for a 13mm board 3V board where the host is too small to fit 16mm.
E17A quadtrix is a better MCPCB for those small lights. The main reason is optic compatibility. Quadtrix E21A (0,2mm gap) can’t be used with standard 5050 TIR lens, the LES is 6,22mm. E17A quadtrix (0,1mm gap) LES is 4,95mm.
Single E21A is indeed the best from beam control perspective but, it can’t output more than 900lm at R70 (700lm at R9080) temporarily. And 2,5A is the max reliable current.
4x E17A array can take at least 7A (and triple the output of a single E21A). And lots of CCT/color mixing option
Optic compatibility is a good thing, but how are E17As color-wise and efficiency-wise compared to E21As? Personally, these are the characteristics I find the most important.
E17A is the smaller clone of E21A. For white color, it has the same options in all CCT and CRI. Being smaller also means slightly less efficient at the same current rating and obviously lower max power.
Is optic compatibility worth the efficiency losses? I am asking because I intend to use quadrix E21A with 5050 TIR optics and afaik It should fit just fine, with the exception that not all emitter surface is properly covered with lens.
Did you know that as the LES gets bigger you need to use larger optic to get the same Out-The-Front efficiency? And it’s not all about fitting the LED under the optic, it’s about beam control. Smaller LES means you more freedom to shape the beam with less light loss. As the light source moving farther from the LES center, more of the light “leaked” from the TIR optic sides rather than projected to intended direction.
Even with modification by trimming the optic base so more LES covered we still lost control (can’t shape the channeled beam).
Yes, I know that some lumens are lost in E21A/TiR setups due to incompatible optics, but I don’t know how many. My biggest concern about moving to E17A is what system(E21a+TIR or E17A+TIR) will have better OTFE as a whole. Do you have any data regarding luminous flux for quadrix E21A and quadrix E17A at the same current without optics?
It is also important to mention, that we are speaking here only about TIR optics, typical XM-Lx or XHP50 or other 5050 insulation gasket + OP reflector systems will be compatible with E21A and there shouldn’t be any light loss.
I have the same feeling about 4xE17A vs 4xE21A. Despite the potential optic loss of the 4xE21A, it will still end up much brighter than the 4xE17A. Also on all of my modded 4xE21A lights, I need to use DC-Fix to smooth out the beam anyways and I believe the same is required for 4xE17A so I don’t think beam tuning is a big concern in most cases. Therefore, I think for any host that fits a 16mm mcpcb, the 4xE21A will always be brighter and more efficient than the 4xE17A version.
I think where 4xE17A would shine is if it is used in a smaller mcpcb such as 13 or 14mm 3V only boards. There are a huge number of 16340, AA/14500, and some 18350/18650 lights where 16mm mcpcb is too large to fit and that is where the 4xE17A would be best put to use.
Well, there’s 4xE21A VR16SP4 for you guys don’t like E17A. You won’t believe me until it’s tested, totally understandable.
Just think of it as XHP35 vs XHP50 and you would perhaps, get the idea.
Then there will be no more 16mm quad MCPCBs for E21A? :cry:
