Grow lights are growing (meh heh) in popularity, at least partially due to people growing pot legally for medicinal use. There are grow lights for sale online, some of them look pretty good. Some of them look pretty bad.

This grow light is similar to the Butterfly Sconce, but has many more LEDs, uses LEDs which are tuned for chlorophyll reception, and the dimmer circuit is replaced with a soft-start circuit so the lamp starts up slowly over a few seconds.  Just in case anyone reading this hates plants and wants to kill them, you can also load this up with normal white light LEDs to use while repairing chain saws, or maybe even ultra-violet LEDs for use in destroying our little friends by giving them sunburns.

LEDEngin makes their LZ1 package in many colors, including a 465nm blue and a 665nm ‘deep red’. The LEDs have different voltage drops, and balancing them will be a little tricky. These LEDs run as three times the current of the Cree XP-G’s, but they also put out a lot of light doing that.

NASA experiments and others posting about this subject on the innerwebs are using many more red than blue LEDs, sometimes as much as an 8/1 ratio.  This project will attempt to use 7 LedEngin LZ1 LEDs per string in four strings, each parallel driven by CAT4101 drivers.

The soft-start circuit will lessen the strain on the LEDs when first powered on.  Here is a demonstration of the soft-start circuit using Paul Falstad’s circuit simulator:

Simulation of the soft-start PWM circuit

$ 3 2.0E-6 382.76258214399064 63 5.0 50
R 72 176 72 128 0 0 40.0 5.0 0.0 0.0 0.5
r 72 176 72 240 0 10000.0
r 72 272 72 336 0 10000.0
g 72 336 72 368 0
r 216 240 280 240 0 10000.0
r 216 240 216 176 0 1500.0
R 416 136 416 104 0 0 40.0 5.0 0.0 0.0 0.5
c 368 272 336 272 0 4.7000000000000004E-8 0.6305849260682925
g 416 352 416 368 0
174 416 136 448 200 0 1000000.0 0.5 Start Speed
x 185 97 343 101 0 16 LED Lamp “Soft-Start”
r 216 176 280 176 0 1200.0
x 197 122 358 125 0 12
x 511 273 606 276 0 12 to CAT4101 PWM
x 153 347 334 350 0 10 Paul Falstad’s Circuit Simulator Rocks!
a 104 248 152 248 3 5.0 0.0 1000000.0
a 168 240 216 240 3 5.0 0.0 1000000.0
w 104 240 72 240 0
w 104 288 152 288 0
w 152 248 168 248 0
w 168 232 168 208 0
w 168 208 168 176 0
w 168 176 216 176 0
a 312 232 360 232 3 5.0 0.0 1000000.0
w 312 240 280 240 0
w 312 224 296 224 0
w 152 288 296 288 0
w 312 240 312 272 0
w 312 272 336 272 0
w 368 272 384 272 0
w 384 272 384 232 0
w 360 232 384 232 0
w 384 176 384 232 0
a 464 224 512 224 3 5.0 0.0 1000000.0
c 416 296 416 328 0 1.0E-5 4.0561105828521935
r 464 280 464 336 0 1.0E7
g 464 336 464 368 0
w 464 280 464 272 0
w 464 272 416 272 0
w 448 216 464 216 0
w 416 200 416 216 0
w 416 216 448 216 0
w 448 216 448 168 0
w 384 232 464 232 0
w 512 224 552 224 0
w 552 224 552 248 0
w 72 240 72 272 0
w 416 328 416 352 0
w 416 272 416 296 0
w 152 248 152 288 0
w 104 256 104 288 0
w 296 224 296 288 0
w 280 176 384 176 0
w 416 216 416 272 0
o 34 64 0 34 5.0 9.765625E-5 0 -1
o 52 64 0 34 5.0 0.0015625 0 -1
o 45 64 0 42 10.0 9.765625E-5 1 -1

Essentially the dimmer circuit I’ve been using has the ‘gas’ potentiometer replaced with a relatively fat capacitor charged through a resistor.  That will allow the lamp to start up over 1/4 to 14 seconds or so, adjustable with a trim pot. The fixed value hysteresis resistor of the pulse generator is undervalued to ensure that the lamp goes to full brightness when the capacitor is charged, and a high value resistor across the capacitor leaks the charge over a few minutes when the lamp is off. It’s a neat idea and might help with the life of the LEDs, but it will also look ‘intelligent’.

Here’s the circuit, it’s pretty similar to what I’ve been doing:

Here’s one solution for the PC board, this is going to have to use some serious thermal management, so I’m enquiring about having this fabricated on single layer metal core printed circuit boards.  Yes, note that there are very few no ground vias, but there are some SMD jumpers that are required.  They should be a lot cheaper than a second layer, especially if I use double or triple copper.

This board will necessarily be 80 mm x 100 mm, it has to be a little wider than the ‘butterfly sconce’ as it needs to fully support an LGA 775 CPU cooler mount.  Note the custom mounting holes to facilitate the 72mm (per side) mounting, it’s almost too big for the Light version of Eagle CAD.

The plan thus far is to use:

  • 16 LedEngin LZ1-00R205 Deep Red (2.8V @ 1A)
  • 8 LedEngin LZ1-00B205 Blue (3.6V @ 1A)
  • 4 LedEngin LZ1-00WW05 Warm White (3.6V @ 1A)

Each of the four strings will drop (4*2.8)+(3*3.6) = 22 volts, plus the drop on the power supply barrier diode (~0.4V).  I’ll also add in spaces for an array of additional diodes that can be optionally jumpered into series with the incoming supply to drop a few additional volts in case I want to use the same board with different LEDs in the future.  The whole circuit will use (22*0.95*4) = 83.6 watts of power for the LEDs, assuming I’m running the CAT4101 drivers at just under their rated maximum of 1A.

Holy knuckleballs!  This is some wattage, I’m going to actually have to calculate the thermal resistances.  Let’s assume I’m using a single layer of 2 ounce (double) copper on an aluminium (Righto!) MCPCB, in particular, Bergquist HT-04503.  I did not do well in calculus…


I’m sending the design to some proto shops that will make MCPCBs for quotes.  They’re coming back high, and since this project is pretty risky in the first place, the financing on this particular project may take a while.

A thorough review of LED grow lights on the web shows that there is starting to be some competition for this emerging market.  The LED panels are very expensive, or they’re pretty shoddy, and there’s a modicum of difference between the ends of that spectrum, so to speak.  Given the prices of LEDs at this time, I don’t really think that LED grow lights are really up to competing with existing technology, namely… low pressure sodium.  The bulbs cost $100, but last long enough, a whole lamp costs $250.  It would take a heck of a lot of LEDs to add up to one of these sodium lamps light output, so unless there is a marked change in LED prices, or in law (making some esoteric aspect of LPS lighting unpalatable), LED grow lights are a pipe dream.  It’s true that it would pay for itself in electricity savings over a very long term, but the market for indoor farmers who actually know how hard this is and plan on staying in the business long enough to realize the savings, that’s pretty small IMHO.  There are plenty of goods for resale from people who found out this isn’t for them.

This doesn’t necessarily mean I shouldn’t try this.  I’m not going to make a million dollars on it today, but that’s not why I started doing this.

The CAT4101 drivers won’t drive the LedEngin LZ1’s over an amp, and the LEDs can take 1.5 amps.  I’m investigating use of the STCS2 and STCS2A 2 amp drivers, they are virtually unadjustable after the current sense resistor is mounted, although I could blow it off with a hot air pencil afterwards.

Quote requested from Cambridge America LLC of Branford, CT.


The quote was too good to pass.  Boards are getting ordered, LEDs are ordered from Newark, other parts are ready to order, but can wait.

The sucker is going to overheat.  There’s absolutely no question about that.  But how fast?  Will the metal core PCB be enough to get the heat out until the fan can kick in?  It seems like it should.  I can start with a lowered current and move it up as the temperature becomes stable.  At a top end of 1 amp (okay, ~910mA), I still can’t top out the LEDs if for some insane reason this actually works, but then increasing the current won’t be the biggest of my problems.


The circuit boards came from Cambridge America last week, they were actually shipped from China, no surprise there.  Wo ai Zhong guo ren! The boards are perfect, perfect.  The aluminium isn’t too thin, the features are spot on, I really can’t find anything wrong with it.  I get the idea that the equipment could have handled MUCH more difficult jobs, my thinest trace/space was 16 mils.

I highly recommend Cambridge America for manufacture of aluminum metal core printed circuit boards.

Their customer service was also extremely pleasant, a lot of companies don’t care for catering to little pipsqueak prototypers like me because the setup takes the same amount of time as a big job.  It was expensive… as a hobbyist, this was pretty hard on the budget… but looking at the results, it was worth it.

The LEDs are here, the Christmas shopping season has slowed the delivery of the rest of the components, but they will be here next week.  I also have the stencil, a half kilogram of new Kester NXG1 no-lead solder paste, and a few more days off.



Stenciling is great.  You can swipe it a few times and ensure that you got it covered well, that saved me a lot of time over painting the parts on with a toothpick.  I dropped all the parts into place, I cooked it in the oven, and it melted just fine.  Even the big caps that never seem to melt right on FR4 worked out.

The big power diodes don’t fit, and I checked them twice to ensure that they would.  Oh well, if I lean them on a side I can make them fit my pads, I got some really big diodes, but they’ll handle the load.

The metal core PCB is amazing, I didn’t have any trouble with it in the oven, if nothing else I think it worked better.  However, it’s extremely efficient at absorbing heat, and so using a soldering iron is really difficult.  I had trouble soldering the power wires onto the supply pads.

The fan power supply works, no trouble there.  There was a lot of solder bridging on the quad op-amp for the slow start circuit and I’m suspicious that I fried it while trying to fix it.  I also have one bad blue LED, it’s acting as if it’s shorted, so I should be able to remelt it… I hope!   Getting thermal with this board is an invitation to futility, I should be able to hot air blow it off, but it’s NOT going to be fun.

It’s bright.  Very bright.  I wish I had a light meter.

It’s taking 90 watts at the plug.  That’s a little hot, but still right around where I thought it would be.  I’m using the 24V 100W power supply, and it’s too high voltage.  I need to install two more power diodes, they each drop about a volt and I’m running over at the CAT LED pin.  The excess voltage causes heat at the CAT drivers, but they’re mounted on the board very well and they’re not staying hot.

The most interesting thing to me at this point is that the heat pipes aren’t even getting hot.  I was very suspicious that a fan driven heat sink would be many, many times more efficient than just the convection cooling that I’ve been using, and all the reading I’ve been doing says this too, but what I’m seeing is pretty amazing.  Using my Fluke 62 Mini infrared thermometer, when the lamp is on and has been running for a few minutes, the center of the board is up around 70C, getting right up there near the red line.  The CAT drivers are around 48 – 52 C.  The power diode is 68C.  None of these is surprising, and the circuit is not optimally adjusted yet.  However, move half an inch from any of these components and it’s barely warm, or ambient.  The heat pipes in the back are cool to the touch, there’s not even a hint of heat anywhere behind the printed circuit board at all.   I suspect that as the thermal compound (I used Artic Silver 5) becomes cured, it will work even better.  I am definitely going to have to reconsider use of fans.  It makes very little noise, it’s crazy efficient on the cooling.

I have to fix the bad blue one, and oscilloscope the bad slow start circuit.  Pictures soon.


Whew!  That bad blue one was SCARY to fix, but between the electric skillet and my hot air pencil, I mashed a little tin ball out from underneath it, and it worked.  The ‘slow start’ circuit does NOT work.  It kinda worked, but only took the dwell up to around 50%, and any adjustment to try and fix that didn’t work.  The circuit has been disabled, and I won’t bother to try it again unless I breadboard it or find a definitive reason it’s goofed.

Here is a short movie of me banging it around while trying to figure out how to hang it:

I found that a 7 inch to 6 inch HVAC galvanized sheet metal duct ‘reducer’ made a pretty good fit for the CoolerMaster GeminII S heat sink.  Here are all the parts, the power supply, a tube of Arctic Silver 5, and a US quarter dollar coin (Puerto Rico).

Here’s the lamp sitting on it’s output.

And here it is on with the overhead light out:

Here is a shot of the LEDs, you can see the 4 strings of red/blue/red/warm white/red/blue/red.

The lamp is currently in testing, and preliminary results are “Wow, you gotta come see this”.

Creative Commons License


Copyright (c) 2010 MDVE.NET. Some rights reserved.
This design is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License .

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

Copyright © 2010, Multi-Dimensional Visual Echo. All rights reserved.