LED grow light

[tooki]

For years, I've had problems with my houseplants (mostly chiles, but also some citrus) perishing over winter indoors. I had chalked it up to just having a black thumb, but my roommate thought it was insufficient light. So we started just turning on a desk lamp at night to shine on the plants. Lo and behold, that seems to be it! The plants are now beginning to grow new baby leaves to replace all the ones they shed. So I began thinking about building some kind of grow light, probably using LED strips as the light source.

Anyway, I stopped at a local electronics shop to pick up a broken LCD monitor I had intended to salvage optical films from for a different project. The 32 inch TV that was all they had in their junk bin ended up not having the optical films I needed. But its old-school CCFL backlight box is the perfect size and shape to make a light panel to illuminate the plants, and has a lovely thick, robust diffuser panel. Unfortunately, the power supply board in this TV appears to be shot (I suspect that's why it was thrown out to begin with), so I can't use its CCFL lighting to hold me over until the LED strips arrive.

So my plan at this point is to rip out the CCFL tubes and glue a bunch of LED strips inside. The Tim the Toolman Taylor in me of course wants to line the thing completely with LEDs, but a back of the envelope calculation said that'd be about 240 W, which would probably set the plants on fire. A more realistic plan is to line it with two different types of LED strips: white LEDs for when humans are inhabiting the room, and red and blue ones as more energy-efficient grow light for when it's just the plants. (I actually found some neat LED strips on eBay that contain dual-color warm white/cool white LEDs so you can tweak the color temperature.) I was thinking that somewhere in the ballpark of 30 W each for white and red/blue would be more than enough.

I know that in theory I could use RGB LEDs to cover both use cases, but I really hate the sickly white of RGB, which is why I want to have the option of true white LEDs.

Other than making sure I have an adequately specced power supply, is there anything I'm missing? I haven't decided whether to use off-the-shelf remote control modules, a bank of manual power switches, or even an Arduino to control the brightness. I'm open for suggestions if anyone has experiences to share.

On the subject of power supplies: obviously you can get cheap Chinese made power supplies, but that's probably a bad idea. The local electronics distributor sells both open-frame and laptop-style power supplies from both TDK-Lambda and Mean-Well. The cheapest models don't have great efficiency, and they don't specify there no-load power draw. Would I be correct to assume that it's higher than the <0.5 W no-load power draw specified in the better models? Are there any advantages or disadvantages of one brand over the other, or of open-frame versus laptop-style?

Edit: Added bolding to show people that I am very aware that white LEDs are inefficient for photosynthesis, since people keep reminding me of this despite me explaining repeatedly that a) I know this, and b) the reason.

[nowlan]

Easiest would be to get some T8 led tubes.

I wouldnt worry about red/blue leds. White lets are blue with phosphor. There is a lot of information about corals/weed using led lights. I would just look at someones build. Get some crees and a meanwell power supply. Supplement the day hours.

[rs20]

Plants reflect green (source: I just looked outside), so red and blue is ideal if all you care about are the plants:



White LEDs are questionable because they're typically blue LEDs with yellow phosphors. Whether the yellow light is within the "red" range that plants accept is not something I know. It's certainly not reasonable to just assume that'll work well though.

The T8 led tubes will have this same potential problem. Red/blue LEDs, or real fluorescent lights (like, nice $8 high-CRI ones, not the $2 crap at the supermarket), or indeed red/blue fluorescents are preferable. 600mm and 1200mm both available.

Plug: fluorescents don't flicker either if you use a proper electronic ballast. Did I mention that I think fluoros are underrated? 

[Siwastaja]

White LEDs are questionable because they're typically blue LEDs with yellow phosphors. Whether the yellow light is within the "red" range that plants accept is not something I know.

You are on the right track; the yellow phosphor isn't inside the desired range. Quite the opposite, the "red" that the plants want is on the longer side, which is why "deep red" 660 nm LEDs are used. White LEDs optimize by not providing these long wavelengths which give too little visible lumens to the eye per watt.

White leds make absolutely no sense. If something is going to be DIYed, then it's just easiest to pick those deep blue and deep red (i.e., 445/450 nm and 660 nm) LEDs. There are specific "grow light" LEDs which are like white leds, but with red phosphor, but I can't see the convenience in using just one type of LED.

[tooki]

Easiest would be to get some T8 led tubes.

Why would LED "tubes" be preferable to LED strips that are so easy to work with?

Cree LEDs (I assume you mean the high-wattage COB types) would certainly work, but they'd be a lot more work to assemble and ensure cooling for, and would ultimately create less even light than the strips. The white LED strips I'm looking at are about 10W/meter, with a 5m strip costing a whopping $7 or something, and are already evenly spaced on adhesive and don't require additional cooling. If I get the special grow light strips (660nm deep red + 445nm blue), they're around $35 for 5m.

Plug: fluorescents don't flicker either if you use a proper electronic ballast. Did I mention that I think fluoros are underrated? 

Fluorescent can be good indeed, but given that I want to be able to dim them down so they're not blinding me when I'm in the room, LED is probably much, much easier to deal with. (Dimmable fluorescent ballasts cost a fortune anywhere, and here in Switzerland, 10 fortunes.)

Besides, the lightbox is only about 2cm thick, so T8 tubes wouldn't fit inside anyway.

But speaking of flicker: If I did decide to use PWM dimming (either via an off the shelf controller, or a DIY Arduino solution), and I found flicker to be objectionable, is there any reason I couldn't use some electrolytic caps to smooth it out? I've done that on a small scale (LED backlight dimming on an LCD module) but don't know how well that would scale up to a multi-watt scenario. If I don't do PWM, I'd simply have switches to switch x amount of the strips on or off. (Assuming I installed the full 5m, it's enough for 7 strips across the panel.)

White leds make absolutely no sense. If something is going to be DIYed, then it's just easiest to pick those deep blue and deep red (i.e., 445/450 nm and 660 nm) LEDs. There are specific "grow light" LEDs which are like white leds, but with red phosphor, but I can't see the convenience in using just one type of LED.

As I said, the reason for having switchable white and red/blue LEDs is that since the plants indeed cannot use green light for photosynthesis, it means that theoretically 1/3 of white light is simply wasted. (And that would be assuming white LEDs had a fairly even spectral energy distribution, which they don't necessarily.*) So the idea is that when nobody is in the room, the light panel can be set to photosynthesis-optimized red/blue, and when humans are in the room and don't want to be inside a giant magenta chamber, switch it to white, which works as both room lighting and inefficient (but better than nothing) plant lighting.

*There are three types of white LEDs:
- RGB (i.e. three chips, seldom seen today except as the multi-leaded type intended for multi-color use. There were for a while RGB LEDs that had just two leads, to be used only as white.
- blue LED with yellow phosphor (apparently the most common)
- UV LED with white phosphor (and of course, the phosphor here could have any of a number of compositions, just as in the white phosphors of fluorescent tubes)

[Zero999]

But speaking of flicker: If I did decide to use PWM dimming (either via an off the shelf controller, or a DIY Arduino solution), and I found flicker to be objectionable, is there any reason I couldn't use some electrolytic caps to smooth it out?

No, it will increase the power consumption and could even damage the transistors.

Also note that PWM is not a very efficient way to control the brightness. LEDs are more efficient at lower currents so it's more efficient to just reduce the forward current. PWM is often used because it's more convenient and the spectrum of some phosphor coated LEDs can change with the forward current so sometimes it's better to keep it constant and reduce the duty cycle.

[Fungus]

Easiest would be to get some T8 led tubes.

I wouldnt worry about red/blue leds. White lets are blue with phosphor. There is a lot of information about corals/weed using led lights. I would just look at someones build. Get some crees and a meanwell power supply. Supplement the day hours.

White is the totally wrong color for photosynthesis (it will work but you're throwing energy away...)



The lamps they sell on eBay seem to have 2 orange LEDs for every blue LED 

eg. http://www.ebay.com/itm/131585886673


Edited to insert the proper quote...

[tooki]

White is the wrong color for photosynthesis (t will work but you're throwing energy away...)

I guess you didn't read either my original post nor the reply... 

[Fungus]

White is the wrong color for photosynthesis (t will work but you're throwing energy away...)

I guess you didn't read either my original post nor the reply... 

You're assuming that was what I was replying to.

(I messed up the quote... now fixed)

Do you need a different color for when humans are in the room? Won't it be enough to just switch on the LEDs when the humans are away? That would simplify things a lot (you even could have a light sensor and turn off the LEDs when there's enough normal light in the room)

[tooki]

But speaking of flicker: If I did decide to use PWM dimming (either via an off the shelf controller, or a DIY Arduino solution), and I found flicker to be objectionable, is there any reason I couldn't use some electrolytic caps to smooth it out?

No, it will increase the power consumption and could even damage the transistors.

Also note that PWM is not a very efficient way to control the brightness. LEDs are more efficient at lower currents so it's more efficient to just reduce the forward current. PWM is often used because it's more convenient and the spectrum of some phosphor coated LEDs can change with the forward current so sometimes it's better to keep it constant and reduce the duty cycle.

Thanks! I've seen reference to using caps (well, a simple RC network) to convert the PWM to analog, and then using that to drive a FET to control the load. Would that work here, or would I just be burning off the excess voltage as heat? (Cuz indeed, my intent was to make this thing brighter than needed in everyday use, such that it would normally be driven well below its maximum.) Or is there a better approach I'm not thinking of (dimmable CC source??)?

[tooki]

White is the wrong color for photosynthesis (t will work but you're throwing energy away...)

I guess you didn't read either my original post nor the reply... 

You're assuming that was what I was replying to.

(I messed up the quote... now fixed)

Ah, okay! 

Do you need a different color for when humans are in the room? Won't it be enough to just switch on the LEDs when the humans are away? That would simplify things a lot (you even could have a light sensor and turn off the LEDs when there's enough normal light in the room)

It's a sort of nice-to-have, easy-to-do thing. I'm often awake late at night, so there's significant overlap between the times the room is occupied and when the plants need extra light. (Here in Switzerland, at winter solstice there are less than 8.5 hours of daylight, so even if I weren't a night owl, adding artificial pink light to the end of the day would still be an issue.)

One could, of course, have a motion sensor to detect human presence and automatically switch from the pink grow light to white when people are around. :p

In any case, my questions aren't really dependent on the color of LEDs in question. We can think of this as two entirely distinct LED panels that will happen to share a single housing for convenience reasons. Besides, this being an engineering forum, I think we can all appreciate "Because I can!" as a reason to do something unnecessarily complicated.

[Fungus]

One could, of course, have a motion sensor to detect human presence and automatically switch from the pink grow light to white when people are around. :p

You mean like one of those new offices where you have to stand up and wave your arms every ten minutes to keep the lights on and not freeze to death?

this being an engineering forum, I think we can all appreciate "Because I can!" as a reason to do something unnecessarily complicated.

Sure.

The motion-sensor part in particular could get easily become...  "interesting".

[tooki]

One could, of course, have a motion sensor to detect human presence and automatically switch from the pink grow light to white when people are around. :p

You mean like one of those new offices where you have to stand up and wave your arms every ten minutes to keep the lights on and not freeze to death?

Oh dear god, perish the thought!! (I chill easily when seated doing office work.) On the other hand, forcing me to stand up and wave every 10 minutes would be really, really good for my repetitive strain injury (my docs have ordered me to take very frequent stretching breaks from desk work)!

Of course, here in Switzerland, people are obsessed with airing out rooms ("It's stuffy!"), so in the dead of winter when it's -5C out, they'll fling open the ultra-efficient triple-glazed, rubber-gasketed windows and air out the office for 20 minutes, so then I freeze anyway. (Not hyperbole. I've experienced this countless times.) Of course, when it's summer and warm and finally a cool breeze decides to gently blow, they complain that "it's drafty!" and shut the windows... [/rant]

this being an engineering forum, I think we can all appreciate "Because I can!" as a reason to do something unnecessarily complicated.

The motion-sensor part in particular could get easily become...  "interesting".

Aye. But it would first require settling on a dimming method.

[Fungus]

Aye. But it would first require settling on a dimming method.
[/quote]

If they're cheapo 12V LED strips with built-in resistors then use an Arduino and a logic-level MOSFET.

If they're high-power LEDs then use an Arduino and an LED driver which has a PWM input.

The Arduino gives you all sorts of possibilities for sensors, remote controls, etc.

[mazurov]

White LED are less efficient for grow but they are so much cheaper, especially when you want to build something bright, like this -> https://www.circuitsathome.com/diy-2/high-power-led-grow-light-a-build-log . This light is now ~3 years in operation; combined with hydroponics it gives outstanding results. Peppers are very happy.

Another thing I'm experimenting with are lightweight fanless fixtures made with small 100 mA emitters. I hang some of the lights vertically around the plant to provide intra-canopy illumination.

[tooki]

White LED are less efficient for grow but they are so much cheaper, especially when you want to build something bright, like this -> https://www.circuitsathome.com/diy-2/high-power-led-grow-light-a-build-log . This light is now ~3 years in operation; combined with hydroponics it gives outstanding results. Peppers are very happy.

Another thing I'm experimenting with are lightweight fanless fixtures made with small 100 mA emitters. I hang some of the lights vertically around the plant to provide intra-canopy illumination.

$20-30 for a 5m strip of 660/445nm grow light LEDs (totaling 72W) doesn't seem that expensive to me. Yes, it's a lot more than white LEDs, but given how much more efficient it must be, probably makes sense in the end.

[mazurov]

$20-30 for a 5m strip of 660/445nm grow light LEDs (totaling 72W) doesn't seem that expensive to me.

You also need 620nm. Then you need to look at specific emitters - I have some Chinese 10W "620" which are in fact white with red filter in front of the phosphor. Real ones are quite a bit more expensive. For example, quarter watt LXM3-PD01 costs ~$1.5/pc today when you buy 1000pcs; for 72W fixture you'll need 24 of those which gives you $36 just for 660 band.

Yes, it's a lot more than white LEDs, but given how much more efficient it must be, probably makes sense in the end.

This depends of where you place "the end".  When I was doing cost analysis the break even point appeared to be around 70 years from now, at current rate of change of electricity prices where I live.

Prices in US dollars.

[tooki]

If they're cheapo 12V LED strips with built-in resistors then use an Arduino and a logic-level MOSFET.

Hence my still unanswered question: Could one use an Arduino to create PWM, smooth that to analog, and then use that to drive a MOSFET or whatever to allow flicker-free control of the LEDs? Or were you suggesting PWM all the way to the LEDs?

If they're high-power LEDs then use an Arduino and an LED driver which has a PWM input.

I've stated consistently throughout this thread that my intent is to use 12V strips. That's because individual high-power LEDs a) are mechanically/thermally more complicated, and b) would not be as diffuse a light.

[dfmischler]

I know you know all this  .  It's for everybody else reading this.  Plant Productivity in Response to LED Lighting

[rs20]

But speaking of flicker: If I did decide to use PWM dimming (either via an off the shelf controller, or a DIY Arduino solution), and I found flicker to be objectionable, is there any reason I couldn't use some electrolytic caps to smooth it out?

No, it will increase the power consumption and could even damage the transistors.

Also note that PWM is not a very efficient way to control the brightness. LEDs are more efficient at lower currents so it's more efficient to just reduce the forward current. PWM is often used because it's more convenient and the spectrum of some phosphor coated LEDs can change with the forward current so sometimes it's better to keep it constant and reduce the duty cycle.

So one solution is to use a smoothing capacitor in combination with an inductor, right? That way you're basically making an open-loop buck converter; the transistors are OK because the inductor prevents current spikes, you get the efficiency/current boosting effect of a buck converter, and the LEDs are running on a nice constant current.

Incidentally, there are specially designed LED buck converter chips which do all this for you, although you can just PWM it from a MCU too.

[rickey1990]

Well a project of mine last year, also been experimenting with cheap super caps and a hand crank for instant bright light. Using a modified camera flash circuit out of a disposable camera. Four led bulbs being driven from a 1.5v battery. Obviously the battery doesn't last to long but running off a solar charger bumped down to 1.5v's there not too bad.  60LEDS on each bulb.

[Fungus]

If they're cheapo 12V LED strips with built-in resistors then use an Arduino and a logic-level MOSFET.

Hence my still unanswered question: Could one use an Arduino to create PWM, smooth that to analog, and then use that to drive a MOSFET or whatever to allow flicker-free control of the LEDs?

You could do that but it's tricky and there's no need.

Or were you suggesting PWM all the way to the LEDs?

PWM all the way to the LEDs.

The standard Arduino "analogWrite()" function outputs a 1kHz PWM signal so you're not going to see any flicker.

[Siwastaja]

Just PWM the LEDs. Filtration doesn't make any sense.

LED efficiency is practically constant unless you are near or exceeding recommended maximum current rating. Maximize efficiency, lifetime and ease thermal design by running the LEDs at slightly below recommended maximum current. PWM at high enough frequency so that you don't visually see the flicker. Anything above, say, 200-300 Hz, works. I guess the Arduino library uses something around 1 kHz which is optimal.

[tooki]

If they're cheapo 12V LED strips with built-in resistors then use an Arduino and a logic-level MOSFET.

Hence my still unanswered question: Could one use an Arduino to create PWM, smooth that to analog, and then use that to drive a MOSFET or whatever to allow flicker-free control of the LEDs?

You could do that but it's tricky and there's no need.

Or were you suggesting PWM all the way to the LEDs?

PWM all the way to the LEDs.

The standard Arduino "analogWrite()" function outputs a 1kHz PWM signal so you're not going to see any flicker.

Unfortunately that's incorrect. It's around 490Hz, which at low duty cycles causes very obvious flicker. If one is willing to sacrifice accuracy of internal timers, it's possible to raise it, though.

Can you summarize the trickiness that would be involved in doing the smoothing -> MOSFET route?

Just PWM the LEDs. Filtration doesn't make any sense.

LED efficiency is practically constant unless you are near or exceeding recommended maximum current rating. Maximize efficiency, lifetime and ease thermal design by running the LEDs at slightly below recommended maximum current. PWM at high enough frequency so that you don't visually see the flicker. Anything above, say, 200-300 Hz, works. I guess the Arduino library uses something around 1 kHz which is optimal.

Because LEDs react instantly to being turned on or off, the flicker thresholds that worked for CRTs and other things do not apply to LEDs. At low duty cycles, PWMd LEDs flicker perceptibly at frequencies far, far higher than people think. That's why LED driver chips use PWM frequencies from the tens of kilohertz all the way beyond a megahertz. LED-backlit LCDs also use PWM at over 200KHz (or analog dimming).

I know from personal experience that I readily see annoying flicker with Arduino's 490Hz PWM at duty cycles below about 30%. I also routinely notice (and am annoyed by) flicker in dimmed LED lighting at bars and restaurants here, which I assume are using COTS LED dimmers. My concern about flicker, and willingness to go to significant lengths to avoid it, are because I know I am sensitive to it and know that it annoys the living shit out of me.

[Fungus]

The standard Arduino "analogWrite()" function outputs a 1kHz PWM signal so you're not going to see any flicker.

Unfortunately that's incorrect. It's around 490Hz, which at low duty cycles causes very obvious flicker.

Unfortunately that's incorrect. It's around 490Hz on 8-bit timers but 980Hz on 16-bit timers.

Ref: https://www.arduino.cc/en/Reference/AnalogWrite

And ... you can easily have frequencies much higher than that. It's one line of code: Just subtract a value or two from the clock divider on the corresponding timer (Warning: timer0 generates the millis() interrupt so millis() will run faster if you do this on timer0).

I know from personal experience that I readily see annoying flicker with Arduino's 490Hz PWM at duty cycles below about 30%. I also routinely notice (and am annoyed by) flicker in dimmed LED lighting at bars and restaurants here, which I assume are using COTS LED dimmers. My concern about flicker, and willingness to go to significant lengths to avoid it, are because I know I am sensitive to it and know that it annoys the living shit out of me.

If you can see flicker the Arduino hardware isn't the problem, the problem is with the code. The Mega328 chip is capable of much more than Arduino's "easy-for-total-newbies-to-use" libraries expose.

(nb. make sure you can drive your MOSFET gate hard enough at the chosen frequency - that's the real limiting factor to the highest frequency you can drive the LEDs at).