Introduction: High-Brightness LED Grow-Light

Note: This project page has morphed into a design discussion.  (Constructive suggestions and collaboration are still welcome.  See the comments section at the end.)
**There are no recommended construction-plans here.**
Instead, I recommend shopping around for a good deal on standard White LED Light-Bulbs.  Please try using them as grow-lights, and let us all know how it goes.  My calculations suggest they should should be economical, energy efficient, and great for plants.

After many hours of research I've concluded that it doesn't make sense to build DIY grow-lights, and I can't honestly recommend it to anyone.  This is partly because it is technically difficult (and making it easier makes it more expensive), but more-importantly because commercial products are much less expensive and higher quality in the end.

You're better off buying an off-the-shelf product.  And surprisingly, I wouldn't even recommend buying a red & blue "LED grow light", because the best option seems to be conventional white LED lights!  Sure, grow-lights are approximately 25% more energy-efficient for growing plants, however, they are priced 2 to 8 times more per unit of power.  LED grow lights cost $4/Watt-electric typically, or nearly $2/W-e if made in China.  White LED bulbs cost $1/W retail, and sometimes $0.50/W if there is a green-subsidy instant-rebate. 
White LEDs emit more green-light than typical "grow-lights", which unfortunately isn't used very efficiently by plants; however, paying for the electricity to emit that inefficient green-light actually costs less than buying a grow-light that doesn't emit green-light!  Even if hypothetically you have sky-high highest electricity rates of $0.50/kWh, and zero-percent financing, when I try to amortize the cost of grow-lights, I figure that the LEDs burn-out before accruing enough saved electricity to make up for their high up-front cost.

Step 1: Light-Spectrum Requirements

So what kind of light do plants need?  What is best?
Well, in terms of spectrum, the chlorophyll pigments are the main drivers of photosynthesis.  These are most active under blue light at ~430 nm, and red light at ~660 nm.  According to US patent #6921182, plant health is improved by the presence of light around 612 nm.  From the graph below, this corresponds to the peak for "Phycocyanin".

Some of the overall response curves I found, such as for "PAR" are graphed.  Also the spectral output of some common White LEDs are graphed.  It turns out that the peak sensitivity for chlorophyll from blue light (at ~430 nm) lines up almost exactly with the peak output from a white LED, so this is a fortunate coincidence.

Step 2: Light-Intensity Requirements

Suppose we have a plant with 1 square meter of leaf area.  The graph shows that when the light-intensity is about 200 "micro-moles" (umol) of absorbed photons (particles of light) per second, this causes about 8 umol/s of CO2 molecules to be converted into plant matter.  This ratio of 8/200 gives a particle conversion-rate of 4%.  With 6 times more intense light (1200 umol/s), the plant growth rate is increased by only about a factor of 3, since (24 umol/s of CO2) is "assimilated".

So we can see that plants grow more with more light, but not linearly.  There are diminishing returns.  This suggests that for a given amount of light energy: 1) a longer "solar day" (such as from artificial light) improves growth and efficiency, and 2) that it's better to have diffuse light that illuminates the foliage evenly, rather than unevenly, for example where there is greater light-intensity on the top leaves, and lower-intensity on lower-leaves (such as from shading of undergrowth).

Step 3: Obsolete Project Reminents That Was to Build-from-scratch

** What follows is an old discussion of a build-from-scratch concept, that I now consider obsolete.... left here as a reference only. **

Step 4: Circuit Schematic

For the power supply, use one with a somewhat higher voltage.  I chose 12v because it seems to be the most common "brick" voltage.
This allows you to put more LEDs in series.  You can always guarantee that all LEDs in a given "string" (or branch of the circuit) are carrying the same current, which simplifies matters.  Also fewer strings means fewer resistors. This reduces parts count, etc, and the whole system is likely to be more efficient, produce less heat, not overheat your plants, and keep your electric bill down.

Start the circuit design by arbitrarily adding LEDs to a string.  Add up the voltage of all the LEDs in a given string, and adjust the number of LEDs per string to get as close as possible to the power supply voltage without exceeding it. To do this it helps to mix and match LED colors in the string, since each color has a different voltage. 

The main concern is that you don't exceed the LED's current rating.
Ohm's law says that the resistor value "R" in units of "ohms" that you should use is given by this equation:
R = (Vs - V_LEDs) / I_LED

Where Vs is the power supply voltage,
V_LEDs is the sum of the LED voltages in the string, and
I_LED is the current in amps that the LEDs are rated for

If V_LED is just slightly less than Vs then you will only need a very low value resistor, like 1 ohm or less, assuming a 1 amp string. You shouldn't need to drop more than a volt or two across the resistor. If you're dropping over 2.2 volts, why not just add another LED instead?

Just calculate the proper resistor value for each string, and after you've built the circuit, you can measure the current through each string with an ammeter to make sure the current does not exceed the LED's spec, especially when it is operating at its highest temperature.  You can also calculate the current by measuring the voltage across the resistor and dividing by the resistance.  (again, ohm's law).

The figure below shows a simple schematic example.  12 volt power supplies are very common and you should be able to find an extra one lying around that you can re-purpose for this, or you can acquire one from your favorite surplus distributor, for example  In this example it should have a capacity of 2 Amps or more, or whatever is the sum of the current in each string. 

The resistor values may need to be adjusted to limit the current in each branch to 1 amp, particularly given the LED's negative temperature coefficient, which may be as high as -4 mV/deg.C (check the "datasheet" pdf for the device).  I'm also looking into using copper pcb-trace resistors (or "wirewound resistors"), mainly because copper has a positive temperature coefficient of resistance (about +0.4%/deg.C) which will help regulate current through the LEDs.  So far the approach looks promising.  Also, in theory this type of resistor is free (or under $1), simple (amenable to DIY), and high-power, which is ideal for our design goals.  In theory, if the LED and copper resistor are solidly connected to the same heatsink, and so are at essentially the same "case temperature" (although not necessarily the same "junction temperature", but I will gloss over that for now), then we can calculate what minimum value of resistance we need so that the overall temperature coefficient is above zero, so there is less risk of "thermal-runaway" that could damage or destroy one or more LEDs.  For example, for a 1 Amp LED, we would want at least 1 volt drop across the resistor, so that by Ohm's law, the +0.4%/deg.C resistor would yield +4 mV/deg.C, thus canceling the LED's -4 mV/deg.C.   Also by Ohm's law to get 1 volt drop at 1 Amp, requires a 1 ohm resistor.   This part can be bought for under a dollar, or can be made , for example with 2.5' of #36 wire, or 4' or #34 wire, etc.  (Table_of wire sizes)

Step 5: LED Selection, Bill of Materials

Use high-brightness LEDs that can handle at least 700mA to 1 Amp of current or more, because they put out a lot more light for the money.

The most common Red LED is made of GaAs and has a wavelength of 625 nm. This isn't bad, but it isn't ideal for plants. A "super-Red" or "deep-red" LED emitting a wavelength of 660 nm is better (FYI the LED material will have "Al' or "P" in addition to the "GaAs") . These are harder to find, and may cost a bit more, but in theory they are worth the trouble to obtain. Order from a site like or or even where they actually list the wavelength and other useful spec's.

For my LED selection, I used a Digikey (or Mouser) "parametric" search to narrow-down the list to the 2 or 3 ranges of wavelengths that are considered suitable for growing plants, and then used a spreadsheet to narrow this further to those having the greatest radiant output per dollar, although this is not necessary, especially as LED costs have dropped considerably.

One issue I found is that some LED devices will appear to be a better value, however if they are tiny, or "fine-pitch" devices (pins very close together), or non-standard component packages, then they will not be easy to heat-sink properly for maximum current without using a (aluminum) metal-core circuit board (MCPCB).  These can be tricky to source, and are too expensive for the hobbyiest's DIY one-offs.  It can also be rather difficult to properly solder the LED to such a circuit board, especially using only a soldering iron.  Preheating on a hot plate or small oven can help, but even still.  For this reason, it is probably easier to buy LEDs that are pre-assembled on a circuit board where possible. 

Step 6: Printed Circuit Board (PCB)

The image shows an example of the PCB layout for electrical connections and heat-sink requirements for an LED array.  Here the LED is a Cree XLamp component package.  There is an advantage to the configuration at left, where the heat sink can be a single continuous piece.  Considering for a moment the crazy possibility to "build-from-scratch", conceivably this single heat-sink could even be a piece of aluminum sheet metal, supposing there were a good way to mechanically assemble and secure the pieces.  I don't have a good way of doing this in my basement, so for now the industry-standard MCPCB is preferred, because it handles the wires and heat-transfer in one solid mechanical assembly.

If you buy LEDs that are pre-assembled on circuit boards, then no additional circuit board and surface-mount soldering are necessary.  This is much easier for DIY.


In an older design, the PCB layout was started using freeware from expressPCB
To keep the parts cost down, it can help to "panelize" the design.  In that case, 12 grow-lamps could be made from a single $51 order (plus shipping, and depanelization), thus holding the cost of the circuit board to around $5 per grow-lamp. 

Step 7: Heat-sink for LEDs

Here, high-temperature adhesive or heat-sink compound can be added to thermally-connect the MCPCB (or LED's heat-spreader) to a surplus CPU cooler, having a finned heat-sink.  Optionally, an attached fan can be used if you have a lot of high power LEDs and you're trying to use a small heat sink, or if you want to keep the LEDs and/or surrounding air as cool as possible.  Be sure to keep the LED operating temperature within spec, since their light output and the device life can be reduced significantly.  You may want to search the internet for information on sizing heat sinks for your application.

The following information I now consider obsolete for this project, since it is simpler and more reliable to buy an LED that is pre-mounted on a metal-core circuit board.

Now as a disclaimer, this concept is really "ghetto".  But if you're on a desert island, and don't have a MCPCB, you might consider this.
Else, eBay eschews this..

For the LED heat-sink, a short piece of #12 AWG solid copper wire, (as found in "Romex" cable of the type used for wiring an electrical outlets), could be used to make a solid copper through-hole via.  Cut a very short piece of wire and shape this until its length is the thickness of the circuit board, about 0.062". One of these pieces can be inserted into a large via in the PCB under each LED. This will conduct heat from the LED to the heatsink on the other side of the circuit board. This will keep the LEDs much cooler than a solder-filled via, and so will provide higher light output, higher efficacy, and longer lamp life.

The copper vias will probably need to be reflow-soldered at the same time as the LEDs, so that all voids are filled with solder.  (If this method of assembly is not possible for some reason then this thermal design probably won't work very well, and alternative designs will have to be considered; see below**).

The bottom-side (or "solder-side") of the PCB will have a copper pad around the copper via to act as a heat spreader.  As a simple rule-of-thumb, the heat-spreader will be efficient out to a length of about 100 times its thickness. plates to a thickness of 0.0017", so roughly speaking, the heat spreader will be effective out to a radius of about 0.17"

** If the above design doesn't work then one option for an alternative design would be to put the heat-spreader and heat-sink on the same side of the circuit board as the LED (the "component side"). One idea would be to combine the heat-sink with the reflector by making it out of aluminum.

Step 8: Luminare / Lamp Fixture / Lamp Optics

The low-cost LEDs that are commonly available tend to emit light over a very wide angle, for instance a 120 degree angle.  In order to use this light it will help to have a reflecting lamp fixture, or reflecting walls near the plant.  There are many possible options for this.  My favorite option for this is the aluminum reflector mentioned in the previous section. 

Alternatively, you may already have a reflecting lamp fixture you can reuse for this project.  You may have some shiny aluminum flashing, or aluminized-mylar, or a light-colored material of some sort with which to make a reflector.  Make sure it is non-flammable, and use plenty of electrical insulation and waterproofing where necessary to keep the electronics dry.