Building LED lights from facts, no theories

[knna]

Oldmac, ive been searching too for the Reflex tubes in the US and it seems Phillips dont sell them there. But there are other brands which does them, as Hagen, the same of the tubes you are using. They sell an actinic tube for marine acuariums with a 160º window. Ive seen it on acuaria webs for around 20$.

Unfortunatelly it seems that Reflex tubes are hard to find in the US.

Im sad to know what happened you and your greenhouse. I gave up growing outdoors after being robbed two years straight. All the worst karma for those sons of a bitch

Be aware, its very probably they will try to do it again next year.

Coming back to LEDs, i think is still too soon to replace large HPS bulbs. probably on about 2-3 years its going to be doable, but currently the extra cost is difficult to recover by electric savings. Spectral tunning, micro grows and suplementary lighting for lower areas seems the areas where currently LEDs may be halpul and cost effective.

Veggi, take care. All the best.

I dont see any advantage on using full spectrum bulbs. MJ likes a lot of red light, at least 60% of total. Full spectrum bulbs emits too much green, IMHO.

Redline, i was trying to calculate the thermal resistance of your heatsinks. 15w on each seems correct, although maybe some less ( about 10) would work better, specially if you are going to run your LEDs at 700mA. Are they about 10*10cm and 2 cm thick (4*4", 1")?

I try too to avoid as much as possible the use of fans. Given the large surface area we have avalaible on a grow room, i think using passive cooling is easier and more efficient, as far as heat dissipating surface is well calculated.

I have a spectrografic chart of a red Cree XL lamp running at 700mA. Peak wavelenght is at 642-643nm. Half intensity at 630 and 651nm. Not ideal, but good enough. Emiting spectrum of red leds shifts noticiably with the increasing temperature. Datasheets spectrums must be taken as orientative. Spectrum emited on operating conditions is always longer.

Although not designed for hand soldering, its possible to do it with Cree Xlamps. Ive done it several times without problems. Some leds have been soldered and desoldered some times and are still working fine. The problem is you need to do it on the top sides, where there is very little space avalaible. I solder the electric contacts directly on the side of the led, where there is no problem if there is too much tin (no shorts with the metalic ring of the center).

I use to put a piece of Kapton tape below the electric contacts, and solder the side of the led to the wire over the tape. Kapton tape do well up to 300ºC. Ive tried to put the solder tip on the tape directly for more than 10 seconds without damaging it

[attachment=o209743]

[redline]

KNNA,
I am right at the 9 square inches/watt for heat sinks that Luxeon recommends for passive cooling. In practice, I have found that the figure is cutting it a little too close. I am working on ducting a concentrated flow of air across the tops of the heat sinks. Hope that does the trick.

Any reason to go with the Crees over the Luxeon IIIs or K2s?

[knna]

Wit 9 sq inches per watt, you get a thermal resistance about 20 K/W. Plus the thermal resistance of the LED package itself and any aditional dielectric layer (a thin film of thermal adhesive is 0-9-1.3 K/W), it put the thermal resistance of the system near 30K/W. Although its enough to keep decent efficiencies, it may result on unaceptable LEDs duration when running LEDs hard:

At 700mA, a blue leds draws about 3.5V or more. Still at 3.5V, its 2.45W. At 30 K/W, it result on a increase on junction temps of 73.5ºC. So with ambient temps around 25ºC, the LED chip is going to work near 100ºC. Its good for many applications, but on ours, with many hours on a day, that result on shortened life. Not dramatically shortened as if Tj were 150ºC, but shortened.

Most manufacturers only "guarantee" (based on extrapolating short term measurements by stadistical analisis) a lm mantenience of 70% at 50Kh when Tj is kept at or below 80ºC. For the typical depreciation curves, its about 35Kh for 80% of initial output, which mostly is the criteria for horticultural lamps replacement (while human barely notice a drop below 30% on light intensity, for plants it means at least 30% less production, that is usually inaceptable).

For Tj over 80ºC, lm degradation is exponential with the increased temperature. So Tj about 100ºC are aceptable for many applications, where 25Kh until reaching an emission 30% lower of initial is aceptable. But for ours, if you want to recover initial higher investment by electrical bills savings, we want a lm manteinance of 80% at 30Kh minimun, and for that is a must to keep Tj closer as possible to 80ºC. It requires a thermal resistance from chip's junction to ambient temp of 22 K/W.

So when using 9 sq inches per watt instaled of heatsink dissipating surface, if you run your LEDs at 700mA you get a somewhat reduced expected useful life of the array.

Probably the use of airflow on the heatsink will help reducing further those 20 K/W of the same heatsink on open air. So probably with it, you are going to be able to expect 30Kh of useful life. But i would choose to run LED slighty below 700mA to guarantee it (about 600mA). Aditionally, it enhances LED's efficiency.

When running the LEDs at 350mA, that heatsink is more than enough, without the need of forced airflow.

Any reason to go with the Crees over the Luxeon IIIs or K2s?

About blues and whites, way better energy efficiency. Currently Cree is doing much better blue leds than Luxeon (as better as double light output per watt).

For red leds, benefit isnt so clear. I prefer the Crees over the Luxeons because the reduced Vf. But differnce on efficiency, although still better for the Crees, is small, and the easy avalaibility of Luxeons and price (when buying small amounts and not full reels) often does the Luxeons a good alternative.

One think you must take into account when using Luxeons and controling voltage, and not current, is their tendency to reduce Vf for same current after some hundred (or less) hours of operation. It mean that you design the system initially to run, for example, at 600mA, but after 200h, its running at 750mA.

Curently, im going to try the Seoul SemiConductor P4 red leds. If specs are correct, they are way more efficients than both Cree and Luxeons (always talking about top bins avalaible: M for K2 and S for Lux and SSC).

[redline]

Appreciate the advice. I am an electronics hobbyist and don't work in the field, so I can always use some good help.
I bought a bunch of P4s, blue and cool white really cheap, so I am stuck with them. I was planning on using them with red K2s, but will probably need to buy some more reds. I might try the Crees and see if I can get a comparison.

Regarding Cree Vs Luxeon: Isn't a higher VF better? Since you will be running a higher wattage at same current.

I try to be pretty conservative on the current. Don't like to run at more then 60% of max.

I do a couple weeks of break-in during array building process.

The array I am building will have several panel current meters, so I can monitor overall current plus each string.

What are your thoughts on where to operate on current curve? What do you think is the sweet spot where you are trading off efficiency for better penetration? I am setting my rig up so I can play around different current settings and see what produces best yield/watt.

It wouldn't surprise me if we have to run these guys as close to max as possible and use aggressive cooling.

I think with LEDs we are going to have to start tailoring the grow to fit the light. We need to start thinking a little more 3D about productive growth area then we do with HPS. Maximizing available light in the Z axis is the real limiting issue with LEDs. You probably only have a couple of inches of decent penentration with LEDS. If you can somehow increase that to 4 or 5 inches you should be able to double yield in the same square footage.

I am currently playing with growing horizontal /flat plants, trying to keep the canopy depth within 6 inches.

[redline]

My crude method for heat management:
I look for hotest spot on array with IR thermometer and make sure it is not more then 10 degree F above ambient temp after it has been running for a few hours. Max ambient temp is 80 degrees.

I know there are some pretty good ways of estimating junc temp with basic equipment. I just haven't got around to doing it.

I would like to set my test array up so I can continously monitor temp of each array. Any ideas?

[knna]

Regarding Cree Vs Luxeon: Isn't a higher VF better? Since you will be running a higher wattage at same current.

With LED, light output is determined by current, not wattage. Any curve of light output is given vs the current level, independent of the Vf associated to that current.

So for a givent current, a given light output. As higher the Vf, higher the consumption, for the same light output: lower efficiency (more watts burned for same light output). This analysis is valid when comparing different Vf bins of same model.

When comparing different LEDs, obviously they havent the same light output at a given current level, so its necessary to compare energy efficiencies of both. But in general, for leds having similar light output at a given current, the lower the Vf, the more efficient. As manufacturers always report light output at a given current (350mA for 1W leds and 700 mA for 3W), when you compare them, the lower the Vf, the better. As higher Vf increases the wattage burned but not the light emission.

We need to forget using wattage to estimate light output of LEDs, because the wide differences between brands and models, and still between different bins of the same LED (both of flux output and Vf).

That way has worked pretty well with HID lights, because efficiencies of them are very, very similar. Saying "i have 50w of HPS per sq ft" gives a very good indication of the light used, because HPSs have near constant energy efficiencies (400-1000W).

But it dont happen with LEDs. 10w of a given LED may gives 1.5 PAR watts while others (or same with way superior bins) may give perfectly 3 PAR watts (double light for same watts burned). So watts burned dont say nothing about light emited when working with LEDs. At least, currently.

I try to be pretty conservative on the current. Don't like to run at more then 60% of max.

I think its the right way: easier to cool. more efficients, more light points. But its more expensive (more drivers or components and heatsinks, more leds for same light).

The array I am building will have several panel current meters, so I can monitor overall current plus each string.

I agree with that way. Its the only way to take conclusions of how the array works better.

I think current adjustable systems are almost a must. Way more versatility, specially if each color has its own control. Once you do a constant current driver, doing it supporting current limit adjusting is as simple as adding a variable resistor to the current sense circuit.

What are your thoughts on where to operate on current curve? What do you think is the sweet spot where you are trading off efficiency for better penetration? I am setting my rig up so I can play around different current settings and see what produces best yield/watt.

For the increased efficiency, the lower the better.

From cost/performance perspective, it depends of each LED model. On most cases, below 600mA. I think 400-500mA is a good compromise on most cases. Some top LEDs support higher currents keeping decent efficiencies, but currently they are little. I believe in just 1-2 years LEDs released are going to support 700mA very well. But not yet.

It wouldn't surprise me if we have to run these guys as close to max as possible and use aggressive cooling.

If you find that you need more light, you can only do that or add more LEDs. Anyway, id dont think we are going to use our experimental arrays for many years, so running them hard is doable.

I think with LEDs we are going to have to start tailoring the grow to fit the light. We need to start thinking a little more 3D about productive growth area then we do with HPS. Maximizing available light in the Z axis is the real limiting issue with LEDs. You probably only have a couple of inches of decent penentration with LEDS. If you can somehow increase that to 4 or 5 inches you should be able to double yield in the same square footage.

I am currently playing with growing horizontal /flat plants, trying to keep the canopy depth within 6 inches.

Fully agree. I noticed this many time ago. Thats why im insisting about the need of use side and/or intracanopy lighting. That of only grow SCROG. As i dont like SCROG at all (just personal preference), i tend to seek for ways of allowing the grow of tall plants with LEDs. And BTW, increase the light's productivity.

My crude method for heat management:
I look for hotest spot on array with IR thermometer and make sure it is not more then 10 degree F above ambient temp after it has been running for a few hours. Max ambient temp is 80 degrees.

I know there are some pretty good ways of estimating junc temp with basic equipment. I just haven't got around to doing it.

Measure temp at solder point or just next to the led base when heatsink temp is stabilized.

You must know the thermal resistance of the LED package you are using. Its a figure on K/W (or ºC/W, its the same) that says how many ºC junction temp raises for each watt burned by the led. Multiply it (add 1 before if there is a thermal adhesive layer) by watts drawn by the LED (If*Vf, current multiplied by voltage (A*V=W)).

Add the result to the base temp and you get a close approximation of the junction temp.

More accurate measurements implies using very short light pulses (microseconds) and compare Vf of them to Vf on continous operating condition. But this requires to know accurately how much drops Vf for each ºC more, and manufacturer's data often is not precise enough.

I would like to set my test array up so I can continously monitor temp of each array. Any ideas?

Using a thermocouple attached to each heatsink, always on the same position (preferably, the center). But if all arrays uses same LEDs, same power and same heatsink, differences shouldnt be large enough to justify measuring each one, IMHO. But doing it together with current control is going to give very valuable info. :thumbsup:

[physicsnole]

Knna, to the best of your knowledge what would be the wavelengths that would most efficiently and productively enhance marijuana growth? I was hoping you had tossed a few out.

[physicsnole]

Also Knna, here is a program that simulates LED junction temperatures, max temps, whatever. Very useful. Future Lighting Solutions | Making LED Lighting Solutions Simple

[gioiapura]

hey knna first of all thank you to share your experiences with us...i really need some facts to start in the right way my first DIY Led array.

Like you, english isn't my first language so please don't look at the form but the meaning!
I'm working with commercial 90w led arrays(R/B 8:1) like UFOs and i've already made some grow tests mixing Leds with HPS and CLFs with some good results.
I'm a "perfect led light" thread follower since physicsnole(thanks god) put the first post so after a while i decided to do my perfect led array but i already know how difficult is to make a efficient led light...and finally my techincal skills are really poor in assembly an array by myself...lukily a friend works in this field and can do it for me.
From what i've seen regular commercial arrays like procyon and ufos are good for veggi but lack in potency and spectrum in the flowering stage.
My ideal light is a "bloomer"... something to add to the 2 90w red/blue led grow lights in my room when i turn the light 12/12!

Yesterday...after a long waiting i received the Ledengin Leds and 4 35v 1050 mA drivers:

LLZ1-10R105 LEDENGIN 5W RED ON MCPCB PZ 2
LLZ1-10UA05 LEDENGIN UV ON 5W MCPCB PZ 1
LLZ1-10WW05 WARM WHITE 5W MCPCB PZ 8
LLZ1-10B205 BLUE ON 5W MCPCB PZ 3
USLEDLPC351050ALIMENTATORE LPC-35-1050 PZ 4
LLZ1-10R205 DEEP RED 5W MCPCB PZ 8

I still waiting for the 2 5w 720nm Leds that should be here in a week but i need a good projet! The glutec's antibrick idea is nice IMO but hard to move.

I've got a mover and i'd like to move this array between the 2 90w Grow Lights but now i'm thinking about at the heatsink.
I've got 20 cpu coolers with fans and i'd like to keep colors divided to play with spectrum at different blooming stages!

Any help is gold for me at this moment!

[knna]

Knna, to the best of your knowledge what would be the wavelengths that would most efficiently and productively enhance marijuana growth? I was hoping you had tossed a few out.

Hello, physicsnole, welcome to this thread (and welcome back to the board).

I preferred to not go that way in this thread because this issue has been well addressed on your perfect led grow light thread. I agree with conclusions there.

If any, i would like to point out two things:

-Red part of the spectrum, from 625 to 685nm must be at least 50% of all PAR watts used. And preferably close to 70%. Using 640 or 660nm leds makes little differences. Ideally, both would be used, but only one works fine. Although all the red of 660nm obliges to use higher amounts of blue and far red.

All the rest of the spectrum may share the remaining 25-50% of the PAR watts, with special weight of blue, preferably about 440-450nm, which need to be at least 10% of the PAR watts and preferably 15-20%. More than 30% of blue often harm more than help.

Green and yellow need to be at least 10% and preferably 15% (but how much of this range is optimal is what im working now, and share between green and yellow).

Out the PAR range, some far red, preferably peaked about 720-725nm, is required. Along most of the time, on small amounts. On flowering induction, it need to be higher, in order to shorten it. Ideally, far red must be independent adjustable in order to control plants morphology and phenotype expression.

Im not sold on the benefits of UVB (and im against UVA), but im thinking on carry some experiment with it right now. Obviously, i dont have any idea if it worth or at what amounts.

But we still need to learn a lot of how enhance MJ growth by far red manipulation, and the best way of doing it.

-Selection of wavelengths used are more dependent of the efficiency of the leds at each wl than to the peak wavelength:

For example, 660nm peaked leds produce about 5-10% more than 640nm peaked leds when using same amount of photons. But 640nm leds currently emits at least 50% (up to more twice) more photons per watt burned the 660nm. So currently 660nm are clearly not profitable.

This situation may change on the future, as in fact max radiometric efficiency of AlInGaP leds happen on those peaked about 654nm (almost none Al used on the chip composition). But manufacturers choose to do 633nm red leds, that still with lower radiometric efficiency, gives more lm per watt burned.

But its probable than on a near future industry need longer wl red leds for color compensation on white leds so they will be available for us. So i experiment too with 660nm leds, but only for in case they are profitable on the future. Currently, they are not, with efficiencies below 20% (21% the ledEngin ones), while there are 635nm leds with efficiencies well over 30% (50% more energy emitted as light per watt burned).

So we must differentiate between the perfect led grow light and the perfect led grow light available (or possible to build) at a given moment.

Doing this distinction was what move me to open this thread.

gioiapura, welcome to the thread, too.

I think you are right on track. I just can to cheer up on your project and ask you to share your progress.

I had to order 400Kg (about 890lb) of heatsinks of anodized aluminium. It exceed my needs, so probably ill sell it at ebay, together with the excess of current adjustable LED drivers (100-270VAC input, 50VAC output, 100-900mA current range) i had to order. Ill had them on about two months. But if i sell them ill link them here.

Im thinking on ordering a reel of the new Edixeon 740nm leds too, and for sure ill have excess of them if i finally do it.