57w Custom Multichip/COB Grow

Hello folks,


In this thread: Custom made 57w chinese multichip / COB with Epistar and Epiled


I have been growing a couple of years, mostly with LEDs except for the first couple grows. I want to thank the users of this forum, I have quietly gathered lots of useful information from here - there's so much info on here, one just has to look around not even a lot


After buying a chinese "Bysen Helios" grow light and closely inspecting the diodes I found what was sold as Cree in reality was epileds made to look like Crees. So rather than spending awful amounts of money on mediocre leds with noisy fans packed in gigantic cases I decided there has to be a better way.
This is what their "Cree" leds look like up close, clearly those are epileds made to look like Cree. Even so, I consider this a pretty good, if noisy, grow light - also had to deal with getting a replacement module, was no problem. A shame they lie, use noisy fans, use no thermal paste on their all-aluminium heatsinks and charge too much money, it could have been good.
Also, "If it's built to have parts replaced, it will need parts replaced"


So upon purchasing a cheap 3200k 30-chip multichip/COB from China with 50w driver and mounting it to a $7 intel CPU cooler I was pretty much sold, I was seeing better growth than 2x36w 6500k PL-L with way more reach. Now a few iterations of custom made multichips later the first grow is completed.


The single plant had 5 different lights throughout flower, most of the time 60w but 3 weeks of mid flower at 45w. It weighed approx an oz dry, not that impressive but lots of room for improvement. It was 70cm tall and the upper 2 ft were tight and ripened nicely! I consider 2 ft of penetration to be amazing for a 45w LED with no focusing lens or reflector. And that's at 450mA forward current and 26 w/sq ft


The most recent LED setup is 4 strings of:
6x 2700k, 2x 6500k, 3x 640nm

(earlier version pictured above had 3 strings)

2700k and 6500k are EPISTAR ES-CADBV45P
640nm are EPILEDS BH-R4242D-A1 bin 88

So I consider this spectrum to be a third way to get what EVO (Mr.X's grow) and Area 51 also try to - white/red combo.
EVO is/was 66% red, 33% 10000k
A51 is 66% "outdoor white", 33% red
Mine is 55% 2700k, 18% 6500k, 27% red

I simply try to get more of the red from warm whites, which also should give more IR. My understanding is chinese warm white phosphors have more red where US phosphors focus more on CRI. So far pretty happy, seeing about 1 week shorter flowering times compared to HPS. I haven't tried alot of spectrums, straight 2700k+640nm is a bit too stretchy for me (with the cheaper leds though), I try to aim for about 1 inch / 2½ cm between nodes - currently getting 1½-2cm so maybe too many 6500ks or use lower color temp.

At 450mA forward current that comes to around 57w. I like to do lots of experiments, one of the key things I have found with these multichips is that since the leds are positioned so close they act like one big led. Normally it is necessary to use focusing optics and run the forward current as high as possible to get proper penetration, but since multichips place the leds so close it is possible to lower the forward current - and thereby making each led more efficient.
So a 190 lm/w Cree is actually measured at 350mA / 1w, once the forward current is raised to 700-900 mA the efficiency drops but the penetration increases. I don't have exact figures, but I believe it would end up at around 120 lm/w (I know lumens is not the proper scale, the point is efficiency drops alot)


On the other hand, it's possible to use twice as many Epistars, get the same penetration or better and almost the same lighting efficiency at half the expense. That's the theory at least..

The cooler btw is a cheap 4-heatpipe chinese graphics card cooler (and is awesome for the low cost)

I am now growing the same strain again, Happy Amnesia, in an organic no-till setup, second run in this 10L airpot. Minimal input, no compost teas etc.
Soil mix is 4 part - organic potting soil (w/ 50% compost), ~10% coco coir, 4% wormcastings (purchased), 2% "Biobizz Pre-mix".

This one is a small former mother plant, recovered nicely over the last few weeks from abuse. Almost ready to flower. She's a little droopy, almost time for watering.



(White balance adjusted on camera, light is purple-ish with zero banding)

The tent is tiny, 40x40x160 cm - 1.7 sqft, with 57w LED comes to 33.5 w/sqft. I am considering another 57w to see what happens at 67 w/sqft.


My other setup is ~120w (2x57w) custom multichip + ~120w Bysen helios on ~90x60cm (3x2).. just growing some plants from seed to find females, boring, the idea is to replace that setup with ~500w multichip power in a 4x4 tent. In my opinion, the only way for current LEDs to beat a 1000w 400v Gavita on price, yield, running cost, lifetime and maintenance level is multichips. I believe this is possible today though will take more than 500w.


There's been alot going on, many things I forgot or haven't mentioned (such as single big LED point light source + lambertian radiation is win), apologies for the messy opening post

Any comments, questions, corrections, flames etc are most welcome, please

Very nice^^^......................you sure about the bysen panels having no crees??? love your thoughts about the easy "part replacements" lol so true

Thank you PSUAGRO and PetFlora

@PSUAGRO: I'm 100% positive. A normal chinese led even if high bin etc still looks exactly like the cheap ones

Notice the pattern on the LED die itself, this is the unique EPILEDS 42 mil pattern. 42 mil as in 2w diodes, max ~700mA. EPILEDS 45 mil is the other pattern in the photo of the Bysen.

Notice Cree's are just a little bubble on a green PCB and their pattern is very different:
View attachment 2865185

And this is how to build a fake Cree Theoretically, they may not be fake - the PCB could be authentic Cree.. so even if the "10% different" chinese knockoff law disagrees, to me this is still fake Can provide more pictures if needed



@PetFlora:
2700k+6500k = 4500k is true in some respects, however if you analyze the spectrums the result from 2700k+6500k will be more red, more IR and less green. Also potentially lower output from the thick 2700k phosphor. Aaand there's 3 2700k per 6500k, so it'd be more like 3500k anyway
EDIT: For me there's little to no cost difference between straight 2700k and white mix because the machine already coats each individual chip with phosphor rather than a big blob. If the chip didn't have red I'm sure it's cheaper to just mass-coat the whole area.

I like pointing them down. I prefer a single point source for maximum penetration, if I split the wattage in half or 1/4s to put them at the corners it would destroy the lights ability to reach to the bottom while at the same time giving too much light to the top buds. By using a single light source pointed straight down I make the inverse square law work for me rather than fight it..






Just getting used to this image uploading, pretty cool. Here's a bunch from the first grow:



And here's some from the current grow starting sept 27



Aaand here's a shot of the last grow in the 3x2, also Happy Amnesia

If you don't mind me asking, who makes your multichips and about what do they charge?

Small company called Yxyled, I'm not sure if they are made inhouse or they're having them made at macroled or something else. Either way, I'm a happy customer Paid ~$51 per chip and driver with shipping to EU, had to buy 5 though.. and pay vat and fee.

For reference, here is a mugshot of a recent 3W 660nm Epiled obtained from Satisled. The pattern is slightly different from the others.

Btw I wonder how efficient these are compared to Luxeon ES 660nm.

Btw I wonder how efficient these are compared to Luxeon ES 660nm.

Click to expand...

if i remember correctly, if you recalc their lumen Output for monochromatic into mW and compare it to oslon etc, those leds are the most efficient on the market today and 4 years ago (their data sheet is about 4 years old!lol), in other words BS

wimpy69 said:

For reference, here is a mugshot of a recent 3W 660nm Epiled obtained from Satisled. The pattern is slightly different from the others.

Btw I wonder how efficient these are compared to Luxeon ES 660nm.

Click to expand...

Strange, looks to me like you've got a EPILEDS BN-R4242E-A3 (I hope it's ok to link spec sheets), which would be 650nm at most. The only 660nm epileds I seem able to find is BH-C4242D-A1

I'd say, with no numbers for straight comparison, the epiled is about 1/3 to 1/2 the power of the luxeon, provided the epiled is top bin. It's also a 2w led so the forward current won't go as high as the luxeon.

Strange, looks to me like you've got a EPILEDS BN-R4242E-A3 (I hope it's ok to link spec sheets), which would be 650nm at most..

Click to expand...

Nice to see the exact same chip pattern in these specs, thanks for the link. I think (hope) that the document is outdated as it is over three years old. Either that or I was scammed... Satisled sells this LED as a 660-670 nm chip that can take 750mA. After a bit of testing they seem to tell the truth.

I'd say, with no numbers for straight comparison, the epiled is about 1/3 to 1/2 the power of the luxeon, provided the epiled is top bin. It's also a 2w led so the forward current won't go as high as the luxeon.

Click to expand...

I'm more interested in the efficiency as a ratio of what power goes in and what comes out. What power goes in is easy enough but what comes out as radiation is more difficult.

Interesting for sure! Looking forward to see what else you can come up with
I'm in the process of building 200W lamps using high quality top BIN Cree LEDs. I'm going to only use warm white, to test how it works.

wimpy69 said:

Nice to see the exact same chip pattern in these specs, thanks for the link. I think (hope) that the document is outdated as it is over three years old. Either that or I was scammed... Satisled sells this LED as a 660-670 nm chip that can take 750mA. After a bit of testing they seem to tell the truth.

Click to expand...

There may be a newer version, that was just the newest 42 mil I could find. It's just the spec sheet that's old - it's normal with some time between spec sheets and actually selling thousands of units (not three years though). Guess I'm saying it's not necessarily outdated If the newest epiled 660nm is three years old, they can't sell a newer chip

By the way, I've never seen a 45 mil (= 3w) 660nm from Epiled. You could always ask them (if they're using 45 mil I'd like to know the part number as I would love to replace my 42 mils with 45 mil. As far as I know no such LED exists from epileds or epistar for that matter)

wimpy69 said:

I'm more interested in the efficiency as a ratio of what power goes in and what comes out. What power goes in is easy enough but what comes out as radiation is more difficult.

Click to expand...

It's what I meant.. I don't think you should hope for more than a third of the luxeons light output per input watt. Maybe half (or more likely not). I seem to remember calculating my (similar age) epileds to around 3/4 of a Cree XP-G 630nm.

JMD said:

Interesting for sure! Looking forward to see what else you can come up with
I'm in the process of building 200W lamps using high quality top BIN Cree LEDs. I'm going to only use warm white, to test how it works.

Click to expand...

Thank you JMD. Sounds like an interesting project. I found chinese warm whites alone makes kind of stretchy plants, but didn't try flower. I hope you'll have better luck with your Crees

Lenny I measured the chip area and it's 1x1mm^2 which would make it 42 mil. Didn't know the rule about "42 mil = 2W, 45 mil = 3W". I have a 620nm red off the same size so this must then be 2W as well.

About the efficiency of the Epileds, you are saying that a luxeon is 2 or 3 times more efficient. Given an efficiency of about 48% for the luxeon, these Epileds would do 16 to 24%? That would be terrible.

Luxeons 48% is at 350 mA. It drops when run at say 700mA. But because the epiled is 42 mil, it drops off alot faster when run at higher current - check the graphs on the datasheet. If you compare them at 700 mA, none of them are anywhere near 48% but the luxeon should be like 3 times better if not more.

(Not giving hard fact numbers, just trying to explain the concept. I may be biased because I want to run many epileds at 350mA in a multichip to come closer to the efficiency of luxeon etc at 700 mA.)

Time for a little update

From 24th



From 27th (3 days later, day 2 since flip)
30 cm / 1 foot tall



Today - 5 days later, day 2 flower, day 7 since flip
37 cm - 1' 2"