Watercooled Smart IC COB LED Build

ssj4jonathan said:

, to enlighten those still in the dark about linear driven COBs please watch these two videos:

Click to expand...

You've got a lot to learn. It is very rare for a linear driver to outperform a switcher. A current limiting resistor cannot beat a buck step down constant current driver.

There is a big loss in this chip for the AC/DC converter. Efficient AC/DC converters are expensive. Low current AC/DC converters are very difficult to design efficiently.

12V is an inefficient voltage compared to 48V. Look at any series of power supplies the 12V will always be less efficient than the 48V.
30 watt converters are less efficient than 600W. Look at the specs.

I only use voltages below 50V to circumvent electrical safety code.

The Chinese are not known for making quality LEDs. They are in serious financial trouble because no reputable light manufacturer will buy Chinese LEDs. This can be verified in the LED trade press.

The cost of electricity is more important than the cost of the chip. I guarantee you will be paying more in electricity than you need to if you had done the design correctly.

The amount of electricity you could save would more than pay for the more expensive LEDs.

Copper water blocks would have worked twice as good as aluminum. A CoB mounted to a thin copper plate which is mounted to a copper water pipe is a very inexpensive and effective heatsink.

Why is a vaiac so big? Because they can get very hot. This means they are inefficient. They are inefficient with small currents when they do not get hot too.

Your most efficient way to power a set of CoBs is to have one DC constant voltage power supply driving individual Buck step down LED drivers. The DC supply voltage should be set at just little bit over the CoB's forward voltage. The buck driver must be a switcher (NOT linear) the inductor should have minimum resistance, the switching FET should have very low "on state" forward voltage.

Lumens mean NOTHING. The only criteria for evaluation a grow light is how many photons are you getting to the plant's leaf per wall watt in. This does NOT mean a fixture with high wall watts is a good thing. Low wall watts with lots of irradiant photons (e.g. µmol/m²/s) is what you want. It is also what you do not currently have.

NoFucks2Give said:

You've got a lot to learn... Why is a vaiac so big? Because they can get very hot. This means they are inefficient. They are inefficient with small currents when they do not get hot too.

Click to expand...

This variac is big because it's rated at 20A. The smaller they are, the less copper they have and the less amps they can handle. Also, autotransformers are very efficient at regulating the voltage. They are more efficient than regular transformers, which every switching power supply uses.

Using a copper plate with copper pipe attached sounds like too much of a hassle to fabricate for a insignificant increase in heat transfer. And copper water blocks are very expensive.

Did you read any of the posts preceding yours?? What is going on in this forum? At least COBKITs does knows what he is talking about.

Anyways check out my video. 72 hours of smooth operation so far....

The one thing that concerns me about these COB's is that flicker they produced in that second video.

GrnMonStr said:

The one thing that concerns me about these COB's is that flicker they produced in that second video.

Click to expand...

It's nothing to be alarmed about. Fluorescent, and HPS/MH do the same thing. And by jove, HPS lights are top notch... lol

To the naked eye, and plants, the flicker is imperceptible. Because these lights are powered directly off the main power line they have a 60hz flicker. Cameras operate at 50-60 frames per second so they pick up the flicker and create a banding interference.

I was thinking after I posted that there had to be some explanation as to why it appeared as it did in the video I thought maybe he still had those two 1 Meg Ohm resistors in line. Do you have the link to the COBS you got?

GrnMonStr said:

Do you have the link to the COBS you got?

Click to expand...

Here's a link for the Full Spectrum IC COB

And the Cool White / Warm White COB

Full Spec run for $3, and the others run for $1.50. CHEAP! They're from the seller I mentioned in my first post. She's very professional and ships everything combine very fast. Other people have them too, so as long as they have a high rating you should be OK. Just make sure to choose 110V and 50W and the desired color spectrum.

Don't forget about the other MUST HAVE parts. A small 1 fan radiator will work for my setup. It's getting cold lately (mid 70's), and just running the pumps alone is sufficient to cool the rigs. I plan on creating another identical two rig setup down the road so I got the biggest of everything.

If you buy the bare minimum, you probably won't even spend more than $250 and you'll have enough left over stuff to build another for less than 140 bills.

One last note: DO NOT BUY anything from MALAYSIA. Packages take forever (1+ months) to get to the USA. Also take advantage of the BEST OFFER feature when you're buying bulk. I got the lenses down to $1.80 on the first offer.

Let me know if you have any more questions.

ssj4jonathan said:

They are more efficient than regular transformers, which every switching power supply uses.

Click to expand...

Switching regulators do not use transformers.

This is a schematic of my Buck step down CC switching regulator. No transformer.



This is the schematic of my 48V to 5V switching regulator, no transformer.




Both are about 97% efficient when simulated on Texas Instrument's Web Bench.

They are powered by four 100 watt solar panels with a max output of 72V in full summer sun.






.

ssj4jonathan said:

Don't forget about the other MUST HAVE parts. A small 1 fan radiator

Click to expand...

Nice little radiator at a nice price. Especially if the aluminum fins are anodized as they appear to be. Anodized aluminum is a better radiator than copper. Polished and non-anodized aluminum is not good for heatsinks/radiators.

ssj4jonathan said:

Using a copper plate with copper pipe attached sounds like too much of a hassle t

Click to expand...

The copper plate is not as difficult as you may think.

I would use something like #4 for a CoB. It is very effective and efficient and costs almost nothing.

The copper is 1.5" wide x 0.125 thick and costs less than $10 per foot at:
https://www.onlinemetals.com/merchant.cfm?pid=4265&step=4&showunits=inches&id=253&top_cat=87

I use a MAP gas torch to solder the plate to the pipe, very simple. Takes a few seconds.




#7 is a very hard stanless steel rod the same diamerter as a 1/2" water pipe. I used it with a hand sledge to bend the ends of copper bars #2, #5 and #6

______________________________________________________________________



#1 is what I used to use for strips of 3W LEDs.

Closeup of #1:





It is a 0.375 wide x 0.125 x 12" copper bar soldered lengthwise to the p[ipe. that works fine for small quantities. For higher production quantities not so great. So I came up with a much better way to clamp the pipe to the copper bar with improved thermal management capabilities.

I am waiting to get my new milling machine before I build a prototype.



______________________________________________________________________


#6 was an attempt to cost reduce #1. The results were okay but not good enough me.

.


______________________________________________________________________


This is how I would do a CoB:

In my thermal testing I found the heat would spread about 2" in each direction on a longer length of copper.

Attach the CoB to a 4" x 1.25" (or 1.5") and 0.125" thick copper bar and clamp it to the water pipe with two screws to a 0.375" wide x .0375" thick bracket. I may use the mill to cut a 0.3125 radius groove (same radius as water pipe) in the copper bar for better thermal conductivity and mechanical stability.


.

NoFucks2Give said:

Switching regulators do not use transformers...
The copper plate is not as difficult as you may think....

Click to expand...

Are you serious? DC regulators and SMPS are not the same thing. Step up/down converters operate off DC power; of course they will not need a transformer! Why is your device running off solar cells? because it needs a source of DC power. Try connecting your little toy to the mains and see what happens. WOW... I would love to know the crazy thoughts that run through your mind.

As previously mentioned, no one in their right mind will fabricate what you're doing. It's complicated, time consuming, very costly, and you need a milling machine and a torch welder. For the price of a milling machine, you're better off buying a Platinum LED grow light and save yourself the time.

The only fabrication needed for my set up is drilling some holes through a clipboard, and soldering leads to the COBs. Simple DIY kind of stuff; see the difference... It took me 4-5 hours to fully assemble everything and get it up and running.

Anyways, waiting for a power meter to arrive. Once I get that, I'll be able to report the amount of real power my rig is pulling at the wall. There's no way I'll be buying a PAR or LUX meter so I'll have to let someone else measure the light output of these COBs. Efficiency is a big deal and if each COB is burning up more than 8W to power 50W per COB LED, we have a problem. I'll keep you guys updated.

Even if you're only losing 8w to those little drivers, that's 42w consumed per chip. At 35% efficient (a generous guesstimate given the lack of info) that's only 14.5ish PAR w per chip. About 175 total for both your fixtures. That's roughly 30 more than my fixture, which was also built using simple diy style components, and consumes half the wattage of your fixture. Not to mention spectral qualities.
Not only that but I did not need any complicated water cooling system, which could fail at some point and fry your whole fixture.

I probably spent $150 more than you on a quality fixture, with top of the line components (and no clipboards),that will last me at minimum 10 years. Again, I respect your efforts, but those chips are just not worth it at this time. Maybe in a couple years there will be some breakthrough in the tech and I'll eat my words. Not at this point.

ssj4jonathan said:

Are you serious?

Click to expand...

You did not specify the type of switching power supply. Your exact words were " which every switching power supply uses."

ssj4jonathan said:

It's complicated, time consuming, very costly,

Click to expand...

It took seconds to solder and the cost is almost nothing as previously stated.

ssj4jonathan said:

a torch welder.

Click to expand...

No, a MAP torch is a $10 can of MAP gas with a torch head screwed on it. Looks exactly like a propane torch but with hotter burning gas.

ssj4jonathan said:

For the price of a milling machine, you're better off buying a Platinum LED

Click to expand...

I have a couple of those, and four Heliospectra RX30s, a couple of BLM Spydr, a KIND and on and on.

I am buying the milling machine to drill and tap holes. It's a glorified drill press. I wanted the table with a 23" x axis travel so I can move the table rather than the piece I am drilling. The milling machine cost much less than one Heliospectra RX30 which run about $3,000 each. I need the mill becasue I make a lot of LED fixtures.

You see I have been an accomplished electrical engineer for over 40 years. I am now a consultant to the University of Florida's Horticulture Department that does the LED grow research. I design LED fixtures for them. I was hired by Dr Thomas Colquhoun who is arguably the guy that knows more about LEDs and growing plants than anyone else on this planet.

Oh, if you want send me one. I have a radiospectrometer with dual NIST traceable calibration from 200nm-800nm with the NIST LED calibration protocols. I attached a sales PDF that describes the capabilities. I highlighted the calibration I bought.

This is my test setup for measuring lux, radiant watts, and µMoles. I also measure the wattage at the wall, the current flowing through the LED(s), the forward voltage, the temperature, and other stuff.




___________________________________________________________________________________________________




See the thing circled below. That is how you measure the LED current. It's a current shunt in series with the LED's power supply. It's a precision resistance with a very small value. You measure the voltage across the shunt to calculate the current. Then the forward voltage is measured between the shunt and the other power lead to the LED. an easy way to measure both current and forward voltage with out the hassle of changing the setup from measuring current to measuring voltage and back and fourth.

NoFucks2Give said:

Nice little radiator at a nice price. Especially if the aluminum fins are anodized as they appear to be. Anodized aluminum is a better radiator than copper. Polished and non-anodized aluminum is not good for heatsinks/radiators.




The copper plate is not as difficult as you may think.

I would use something like #4 for a CoB. It is very effective and efficient and costs almost nothing.

The copper is 1.5" wide x 0.125 thick and costs less than $10 per foot at:
https://www.onlinemetals.com/merchant.cfm?pid=4265&step=4&showunits=inches&id=253&top_cat=87

I use a MAP gas torch to solder the plate to the pipe, very simple. Takes a few seconds.

View attachment 3971420


#7 is a very hard stanless steel rod the same diamerter as a 1/2" water pipe. I used it with a hand sledge to bend the ends of copper bars #2, #5 and #6

______________________________________________________________________



#1 is what I used to use for strips of 3W LEDs.

Closeup of #1:

View attachment 3971422



It is a 0.375 wide x 0.125 x 12" copper bar soldered lengthwise to the p[ipe. that works fine for small quantities. For higher production quantities not so great. So I came up with a much better way to clamp the pipe to the copper bar with improved thermal management capabilities.

I am waiting to get my new milling machine before I build a prototype. View attachment 3971421



______________________________________________________________________


#6 was an attempt to cost reduce #1. The results were okay but not good enough me.

.View attachment 3971424


______________________________________________________________________


This is how I would do a CoB:

In my thermal testing I found the heat would spread about 2" in each direction on a longer length of copper.

Attach the CoB to a 4" x 1.25" (or 1.5") and 0.125" thick copper bar and clamp it to the water pipe with two screws to a 0.375" wide x .0375" thick bracket. I may use the mill to cut a 0.3125 radius groove (same radius as water pipe) in the copper bar for better thermal conductivity and mechanical stability.


.View attachment 3971428

Click to expand...

With a drill press and a milling machine, I would dispense with the pipe and flat stock entirely and drill multiple water passages in a 2x4x3/4 inch copper blocks, and mill end plates for it to connect the passages together..

A few months ago I hammered a 1/2" copper water pipe totally flat so I bet you could press that pipe in the area needed so it had a flatter profile for better heat dissipation along the chips but leave the ends round of course.

Also if I was growing say micro or small area say I only had a 4' tall box like 4'x4'x4' wouldn't the 30W COB be a better choice or say a few more 20W ones? I am thinking this would spread the intensity out more.

xX_BHMC_Xx said:

Even if you're only losing 8w to those little drivers, that's 42w consumed per chip. At 35% [in]efficiency...Not only that... complicated water cooling system, which could fail at some point and fry your whole fixture....those chips are just not worth it at this time. Maybe in a couple years there will be some breakthrough in the tech and I'll eat my words. Not at this point.

Click to expand...

The max power each chip will dissipate is 1W and the linear integrator chips have to reach somewhere around 130C? for them to begin thermoregulating the current. My guess is each chip at max power burns much less than one watt when properly cooled. The data sheet says: "The maximum power dissipation decrease if temperature rise, it is decided by TJMAX, θJA, and environment temperature (TA). The maximum power dissipation is the lower one between PDMAX= (TJMAX-TA)/ θJA and the number listed in the maximum table."

Let's say immediate temperature around chips is 38C (100F). Water cooling gets the chips down to say 68C? (warm water temp), (Note: BigClive's chips reach 92C with thermal paste-less heat sink) (68C-38C)/60C/W=.5W of heat dissipation per chip. Times that by 60. We have 30 watts of wasted power for 12 watercooled COBs. HPS lights use 30 watts alone just to power their ballasts and the bulb looses a big percentage of energy to heat. This is all guesswork here but these COBs just might be more efficient than a stand alone HPS system. SMART IC COBs FTW!!! And I'll shower with the heated water to add another 10% efficiency, like the coolmac... LOL

Joking aside, I really need some answers! It's driving me crazy.

NoFucks2Give said:

Oh, if you want send me one. I have a radiospectrometer with dual NIST traceable calibration from 200nm-800nm with the NIST LED calibration protocols. I attached a sales PDF that describes the capabilities. I highlighted the calibration I bought.

This is my test setup for measuring lux, radiant watts, and µMoles. I also measure the wattage at the wall, the current flowing through the LED(s), the forward voltage, the temperature, and other stuff.

Click to expand...

I was planning on sending some COBS to COBKITs, but it seems your testing equipment is top notch. Can you PM me your address, I'll send you some this month. Just post the results to this thread. I'm very limited in testing equipment. All I have is a shoddy volt meter, and soon will have a kill-o-watt style power meter. Do you have a variac? or something to that can modulate VAC? If so I'll get some sent to you ASAP.

GrnMonStr said:

Also if I was growing say micro or small area say I only had a 4' tall box like 4'x4'x4' wouldn't the 30W COB be a better choice or say a few more 20W ones? I am thinking this would spread the intensity out more.

Click to expand...

I'd still get the 50W versions and use a $20 1000W TRIAC dimmer to decrease the the power consumption if 600W of LED lights is too much. You're better off having more than enough, than not enough light. Watercooling these COBs makes all the difference, and you want to harness the most power onto these little waterblocks as you can. I'm running mine at around half power? (75VAC) and only need one fan to cool off the radiator during day time, and no cooling fans at night. I also have my setup next to a window so that helps a lot too.

Great thread. Everyone has valid points of view and there is heaps to learn.
I've been thinking about water-cooled cobs since before I built my DIY CXB rig, and it's great to see that someone has followed-through and published such a detailed report of the job.
Has anyone got any thoughts on the likely efficacy of an aluminum RHS as the basis of a water-cooled COB bar?
I considered 50mm x 20mm with 3mm wall thickness, but I didn't like my chances of accurately drilling and tapping all the holes, then attaching the cobs without leaks from the fixing holes.
Blind holes in 6mm ally was getting too hard in all kinds of ways. Maybe I should have invested in a milling machine (or just bought Bitcoin lol)

ssj4jonathan said:

Here's a link for the Full Spectrum IC COB

And the Cool White / Warm White COB

Full Spec run for $3, and the others run for $1.50. CHEAP! They're from the seller I mentioned in my first post. She's very professional and ships everything combine very fast. Other people have them too, so as long as they have a high rating you should be OK. Just make sure to choose 110V and 50W and the desired color spectrum.

Don't forget about the other MUST HAVE parts. A small 1 fan radiator will work for my setup. It's getting cold lately (mid 70's), and just running the pumps alone is sufficient to cool the rigs. I plan on creating another identical two rig setup down the road so I got the biggest of everything.

If you buy the bare minimum, you probably won't even spend more than $250 and you'll have enough left over stuff to build another for less than 140 bills.

One last note: DO NOT BUY anything from MALAYSIA. Packages take forever (1+ months) to get to the USA. Also take advantage of the BEST OFFER feature when you're buying bulk. I got the lenses down to $1.80 on the first offer.

Let me know if you have any more questions.

Click to expand...

Thanks for posting this info and all the relies! Question, how many COB's will that small 1 fan radiator cool? That is different than what you are using I think right?

nfhiggs said:

2x4x3/4 inch copper blocks

Click to expand...

I am thinking production quantities, not just making a single unit. I am trying to reduce labor and material costs.

If I bought a 4" x 2" x 12" bar of copper and cut it into 3/4" blocks, each block would be $16 per block of copper. The 12" bar costs $308. Aluminum would be $45.
Milling each block would take too much time. I'm concerned about the amount of time it takes to drill a hole and tap it. I may use a nut instead of tapping. I am going to try a tapping head. A tapping head basically has an automatic transmission in it. It detects the torque and keeps the tap from breaking and then when you back off it goes into reverse.


I will be using strips with multiple strings of individual LEDs. As of now my plan is to make some 16" x 0.35" strips some with 4 strings of 16 LEDs (white and blue) the others 3 strings of 21 LEDs (red). I want to match the voltages of the strings at about 45V, white and blue=2.8v and red = 2.1v

The strips will be screwed down to a thin bar of copper. The screw holes will have large copper pads surrounding them. The LEDs thermal pad will be soldered to that copper pad
Notice how the LED's thermal pad is soldered to the large copper pad. The unused is the foot print for a Cree XLamp LED (e.g. XP-3G) where its thermal pad is in the center between the anode and cathode pads. This strip was a prototype to test the thermal dynamics.

This copper bar is soldered to the side of the pipe. I want copper on copper without solder adding to the thermal resistance.
The pipe and bar are not straight enough



________________________________________________________________________________________________________


This is the thermal test setup in my garage. I use CoBs for thermal testing becasue they can concentrate heat in a small area. This is how I came up with the idea for a CoB cooler. I used C-Clamps to clamp the copper plate to the pipe. It worked VERY well considering how difficult it is to cool a 50 Watt CoB.

GrnMonStr said:

Thanks for posting this info and all the re[p]lies! Question, how many COB's will that small 1 fan radiator cool? That is different than what you are using I think right?

Click to expand...

I'd err on the safe side and meet in the middle. If you're after cost cutting, buy a 240mm radiator and use cheap fans. Wired up two of these fans to exhaust the tent, and they kick ass. I forgot to mention the fans do not come with mounting screws. So you'll need 4 - 10mm m3 screws to mount per fan. The 3 fan radiator, is a bit over kill for running the lights softly, but you can always add more rigs to the setup and wont have to worry about the radiator getting heat saturated.

Eliminating the big fans, big radiator and variac, makes for a savings over 100 dollars! This brings the costs down to around 25 cent per watt. Not bad.

I like the idea of squaring out the copper pipe and mounting the 3W LED directly on that. All you have to do is put 1x1 piece of wood in the pipe and hammer it flat. The sides and ends will remain roundish.

Mullumbimby said:

Has anyone got any thoughts on the likely efficacy of an aluminum RHS as the basis of a water-cooled COB bar?
I considered 50mm x 20mm with 3mm wall thickness, but I didn't like my chances of accurately drilling and tapping all the holes, then attaching the cobs without leaks from the fixing holes.
Blind holes in 6mm ally was getting too hard in all kinds of ways. Maybe I should have invested in a milling machine (or just bought Bitcoin lol)

Click to expand...

Aluminum and thermal paste works just fine for heat conduction. It's not as good as copper or gold, but it will do the job for a fraction less. How do you expect to attach tubing to the end of a square pipe? Also don't drill into anything that holds water! Use the lenses and bracket and just sandwich the COBs to some solid non conductive backing material. Also without fins in the waterblock, water won't soak up as much heat but it won't make a super big difference.

Yea... the bitcoin game is all over now. $2.6k for one coin is crazy. I bought a litecoin a month ago and almost doubled up already. Wish I'd spent my entire check on them. They're going to crash eventually but when that happens civilization won't far well either! Ripple coin has some good potential, I'm waiting for them to hit sub-20cents and will drop a bill on them.

NoFucks2Give said:

I am thinking production quantities, not just making a single unit. I am trying to reduce labor and material costs.

If I bought a 4" x 2" x 12" bar of copper and cut it into 3/4" blocks, each block would be $16 per block of copper. The 12" bar costs $308. Aluminum would be $45.
Milling each block would take too much time. I'm concerned about the amount of time it takes to drill a hole and tap it. I may use a nut instead of tapping. I am going to try a tapping head. A tapping head basically has an automatic transmission in it. It detects the torque and keeps the tap from breaking and then when you back off it goes into reverse.

Click to expand...

Yeah... Copper is pricy. That's why aluminum is generally preferred for sinks and water blocks, despite copper having superior thermal properties.

NoFucks2Give said:

I will be using strips with multiple strings of individual LEDs. As of now my plan is to make some 16" x 0.35" strips some with 4 strings of 16 LEDs (white and blue) the others 3 strings of 21 LEDs (red). I want to match the voltages of the strings at about 45V, white and blue=2.8v and red = 2.1v

The strips will be screwed down to a thin bar of copper. The screw holes will have large copper pads surrounding them. The LEDs thermal pad will be soldered to that copper pad
Notice how the LED's thermal pad is soldered to the large copper pad. The unused is the foot print for a Cree XLamp LED (e.g. XP-3G) where its thermal pad is in the center between the anode and cathode pads. This strip was a prototype to test the thermal dynamics.

This copper bar is soldered to the side of the pipe. I want copper on copper without solder adding to the thermal resistance.
The pipe and bar are not straight enough View attachment 3972184

Click to expand...

I would flatten the pipe (not completely of course) and bar together in a press to increase the contact area and make it more uniform, and put a high performance thermal interface material (3M makes some good industrial stuff) between them. Spot weld them every 6 inches or so while under pressure, instead of a continuous weld

How hot (in watts) do you run these diodes?

NoFucks2Give said:

This is the thermal test setup in my garage. I use CoBs for thermal testing becasue they can concentrate heat in a small area. This is how I came up with the idea for a CoB cooler. I used C-Clamps to clamp the copper plate to the pipe. It worked VERY well considering how difficult it is to cool a 50 Watt CoB.

Click to expand...

nfhiggs said:

How hot (in watts) do you run these diodes?

Click to expand...

The hottest would be the Deep Blue and White in strings of 16. The typical Vf of a 16 LED white string is about 45V with 1 amp current so 45 watts electrical. More than half of that is converted radiated light and less than half is heat.

When generating heat for experiments I use a Mean Well HDD-1500 (1.5 Amp) to drive the CoBs. Vf is 30-38V or between 45 and 60 Watts where more than half is heat.

The reason I use individual LEDs is for thermal management. When the heat is spread out it is easier to manage. I test with CoBs because they are very difficult to manage. When the heat is so concentrated there is too much thermal flux in a too small an area to significantly reduce the temperature. But CoBs are not all that sensitive to heat. You lose about 10% in a typical application where the Tj is 100° C. Whereas a red LED will lose 40% at that temperature. But then red LEDs Vf is less than white.

Results for no thermal management.

Results for water cooled red with minimal 40 gal/hr water flow and an air gap between pipe and copper bar.

The bar was soldered on the ends and in the middle. Did not use many screws. The solder added fair amount thermal resistance. there was not thermal flow through the air gap.