How to build your DIY LED array

Currently LEDs have reached the required perfomance to compete with HID lights, but the cost of comercial LED system does many people interested on trying them dont do it. This thread is intended to offer the knowledge required to build DIY LED grow lights.

Although the last target of this thread is to share precise ways of mounting and driving LEDs, some prior explanations about what are LEDs and how they work are required to understand what can be done, what not and the best way to do it.

LED stand for Light Emitting Diode. The light emitting part is clear, but the important part of it is LEDs are diodes. Diodes are electronic devices wich let the current flow in one direction and not in the reverse. This mean LEDs have electric polarity: they have 2 electrodes, one positive (cathode) and one negative (anode). It mean they must be carefully wired in the right way or they dont work or get destroyed (depending on current applied).

They are semiconductors. The difference between LEDs and other diodes is LEDs emits light when the current flows through it. How is the light they emits depend mostly of the composition of the semiconductor, as the atomic structure determines the wavelenght of light emited (its color). There are 3 main compositions of modern LEDs:

-AlGaAsP. The older one, is used mostly in deep red leds (>650nm) and IR (infrared) today. The name is based in the dopants added to the silicon sustrate. In this case, Aluminium (Al), Gallium (Ga) and arsenicum (As) Phosphate (P). Energy efficiency (power emited as light/consumed power) of this devices is very often lower than other newer technologies, and its just found in small LED's chip, running at 70mA as max, and mostly running at just 20mA.

-AlInGaP. Used in modern red, orange and yellow leds. It stand for Al, Indium (In) and Gallium Phosphate. High power LEDs of these colors are manufactured using this tech.

-InGaN. Stand for Indium and Gallium (Ga) Nitride (N). Its used on blue, green and UV chips, aswell as in white leds, wich uses a blue or sometimes a UV chip covered with phosphors. Most of the research is about this tech, wich goes improving very fast.

There are many others compositions used, but these are the most common and what have reached a good perfomance level.

One very relevant caracteristic of LEDs is, as any semiconductor, their perfomance is strongly affected by the temperature of the chip. Usually noted as Tj (temperature at the junction), it rules the degradation rate of the chip (light emission drop with age) and the efficiency, and affect to the wavelenght emited. Due this, keeping LEDs cool enough is a must in order to have a long lasting and efficient grow light.

The most know through hole LEDs, those with 2 long leads, have the problem that all the heat must be extracted from the chip through those leads, so is very difficult to cool them properly. Thats the reason most through hole leds works just at 20mA. Surface mount LEDs have all the bottom slug thermaly conductive and conected directly to the chip, and this allows to run them far harder than through hole ones. If you want a long lasting LED grow light, forget to build it with through hole LEDs. They are good to learn and are very easy to mount, but they arnt adecuate for lights running many hours a day (except when using some thousand of them on 1 cb ft).

The other main concept to be aware of when working with leds is they are current controled devices, and not voltage controled as many other electronic devices. The light a LED emits is directly controlled by the current flowing across it. When the LED is cold, this current is associated with a given Vf (forward voltage), but as the LED goes heating, Vf for that current drops aswell. If we keep using same Vf than when the LED was cold, then the current flowing across it is going to be way higher and this often lead to a friode So although voltage control is possible, its strongly adviced to drive LEDs by current control.

So the three things to remember about led when building an array are: they have polarity, they must be kept cool and they must be controled by current.

I decided to go posting the (Phillips) Lumileds K2 specsheet and explaining what means each relevant parameter included in it. This way, i wont forget to mention anything and people may see how are LED's specsheets and may learn to take the relevant info from it.

Ive choosed the K2 because it has a very complete specsheet, covering both AlInGaP and InGaN chips used on that platform (K2). Aditionally, although there is some better LEDs than this, the K2 is very versatil and easily avalaible to individuals (especially living in the US and Canada) for relatively cheap, wich makes it a good choice.

After the index, the specs starts with the luminous binning of the K2 family. First it explain what is the nomenclature used in their product:



I think it dont need further explanation, but ask for clarification if you still dont know how it works. Later comes each part named this way, so along the specs you could used to this nomenclature.

A bin is range of perfomance, wich inform more accurately about how a precise LEDs works beyond the generic perfomance of the model. There are bins of luminosity (light output), color and forward Voltage (Vf). The most used is the luminosity ones:



Each row is for a model (part number). They are grouped by colors (differents whites, red, blue, green, etc). Each part has an associated minimun lm emission (third column) and a typical one, wich is the medium of that bin range. The first part listed, LXK2-PW12-R00 (LXK2 is for all K2 models, PW12 for white led rated to work at 350mA and R00 the bin, being the R the bin from 39.8 to 51.7lm) has a typical lm output of 45lm, but a given led of this bin may emit anywhere into the range.

Notice that there is two parts, for two different currents. This is not common, but as i said, the K2 is very versatile and may be runned along a wide range of currents. The nominal rated current of this part is 350mA (defaulf for "1 watt" leds), so the binning is done based on this current. But Lumileds offers too the output for each bin at an higher current (700mA) just as informative note.

Notice how the header of the table starts saying its for Tj (chip led's junction temperature) of 25ºC. This is very important because almost always the normal operating Tj is going to be way higher so the true output at operating conditions is going to be lower.

There are many applied LED products that simply mutiplies this bin rated typical output by the leds used and gives you that figure as the output of the luminarie. This procedure is anything but realistic, as true output is going to be way lower than that figure. Be aware that calculating LEDs efficiencies just with the bin rating is very misleading and always overstate it.

Notice too how the bottom part, of Royal Blue, is given in mW. Its the true radiometric output, way more useful for us than the lm output. With this part, efficiencies of InGaN (white and blue mainly) leds of the family maybe calculated with decent accuracy.

The sheet continues with AlInGaP binning:


Notice how AlInGap leds (red, orange, yellow) are rated to work at lower current than most InGaN parts. This is pretty common. This two differents tech behave different with the increased current density at chip level. While AlInGaP leds increase light output near linear with increase current, up a point, InGaN leds goes losing efficiency as higher the current used (less aditional light output for each next increment in current). On the other hand, InGaN chips support very well high Tj, wich relatively low losses when heating, while AlInGap leds degrades strongly with the increased Tj. Thats why AlInGap leds are driven softer than InGaN.

It continues with an explanation about bins avalaibility:


The specs continue with the two esential parts of any LEd's specsheet: their optical and electric caracteristics:



The first optical caracteristic described is the wavelenght (or CCT for whites). It states the minimun and max of each color. These ranges are very wide and very often has some color bins covering each range. Later in the sheet there are some graphs of the typical SPD of each color (i wont copy them).

Next caracteristic is the the spectral half width. It says how wide/narrow is the emission of the led until half the light intensity emited, in nm (nanometers).

Next one is the typical temperature coefficient of dominant wavelenght. It inform of how the spectral emission is affected by heat: nm of increased dom wl for each degree (centigrade or Kelvin, its the same scale) of Tj increase. This is impotant because spectral emission changes when the chips heats. This spectral sift is more pronounced for AlIngaP than for InGaN leds, as you can notice. It would be better for us the info were given using peak wl and coefficient for peak instead of dominat wl. As said before, photometric info is less useful to us than radiomentric info.

The dominant wl describes how human sense a given light spectrum, similar to the CCT for whites (Kevin rating), while the peak wl is the physical nm where the led emits with more intensity.

This info is required to estimate the true emission at operating conditions, once Tj is calculated.

Next caracteristic included is the angle in wich the leds emits most of the light. Its given often, as in this case, for the angle in wich 90% of the light is emited (160º). The next caracteristic informs of the angle in wich is emited up to half intensity (than the max one in the axis). This definition of the "view angle" is the most used for other manufacturers, especially those having less equipment (this is a parameter way easier to measure), as most asians.

For growing, wide angles are useful when the leds are going to be placed very close of plants (max 2"), medium angles for slighty less closer and narrow angles when the leds are going to be used at some larger distance.



Electrical caracteristics are the forward Voltage (Vf) at the rated current. I only post the 350mA one, although its given for all the currents used u¡with the K2. The range is again wide, and often there is some bins wich cover those ranges.

Notice how AlInGaP leds run at lower Vf for same current. This is pretty typical too, and on this device the difference is smaller than on most ones: most InGaN runs about 3.3V at the rated current and AlInGaP about 2.5V.

Temperature coefficient of forward voltage informs of how much the Vf drops as the chip heats, for the same current. Often AlInGaP Vf drop with Tj increase is higher, but not in this case. Vf dropping with Tj increase have many practical effects on how to drive a LED array, wich is the thougher part when building a DIY LED array.

Thermal resistance is other of the esential parameters to look at. It determines how much is going to heat the chip according to the power burned. Almost always its higher for AlInGaP leds. The coefficient says how many ºC will raise Tj for each watt of power burned by the led.

When looking at thermal resistances, you must read carefully the path its refered, wich is given by the sufix after R(theta). In this case, J-C, junction-case (of the led). But on most specsheets, its given for J-B, junction -board (the soldering point of the led).

You must be aware that when given for J-B as usual, there is another thermal resistance to keep in mind, from board to ambient: the heatsink is usually hotter than the ambien air. Once the basic caracteristic of led are explained, calculating and reducing total thermal resistance (Junction to ambient) is going to be one of the main tasks when designing your DIY array, so one of the esential topics to be discussed in deep.

Through hole leds have usually thermal resistances over 300 K/W. Compare it with those 9 and 12 of the K2 and you will understand why those leds arnt adecuate for lighting. Newer leds still have lower thermal resistances. Last devices are about 3 K/W, as improving it is one of the main ways of improving LEDs duration and perfomance.

Before continuing, here is a direct link to download the K2 specsheet being analyzed in pdf format (im not posting all, just parts i think are relevant when doing a DIY LED array).

The other basic info offered in a specsheet is the maximun figures the LED is rated to work. They are often conservative, but for safe working of the array, they must be respeted:




Notice how the InGaN chip are allowed to work way harder, more than double DC current (except the SSC Ariche, all leds works on DC). The power drawn by a led is Vf*If, voltage x current (V*A=W). As InGaN leds works at higher voltages for same current, it results on InGaN drawing a lot more power from the same package: 3.85V*1.5A=5.775W is the max for the K2 InGaN while its 3.6*0.7A=2,52W for AlInGap ones.

ESD stand for Electric Static Discharge. These leds are well protected, up to 2000V of ESD. Not all led are well protected against it. InGan chip are especially sensible to ESD. Although not strickly required, wearing conductive shoes or a brazalet conected to ground is adviced when working with InGaN leds. Dont work with isolated boots on a carpet.

LED junction temp (Tj) given are the max allowed for the chips, but reaching those temps arnt adviced at all, as it results on a very reduced life (as reduced as less than 1000h).

Storage temperature alowed is very wide and it shoudnt be a concern, but excess humidity may damage LED when exposed to it before being soldered. Keep your leds on a sealed package until the moment of mounting them. InGan are especially sensible to humidity and when they were exposed to excess humidity or kept too much time off the sealed package, they should be baked on a owen to get rid of excess humidity:


Specs continue with the mechanical dimensions, reflow soldering profile, soldering layout reccomended and Spectral Power Distributions (SPD). I jump over it as most times we will mount them by glueing with thermal conductive adhesive, as the known Artic Silver.

Next are the graph showing how is the light output when Tj varies. This graph are esential, due they should guide us to choose the best current to be used with this led. Its different for each color:





There are two things very noticiable when you look at those graphs. The first one is the blue leds keeping its photometric output (lm) constant, and the rest losing lm fast, especially the AlInGaP ones, wich is the second thing to notice. Their graph uses a different scale, and lm output drop at same Tj is way higher than InGaN drop.

I posted the blue graph to show how tricky may be photometric measurements. A given lm output is given by the radiometric output (optical watts) multiplied by a coefficient derived from the spectral emission. Radiometrically, all colors drops in efficiency with the increased Tj. But remember how at the start i mentioned the temperature coefficient of wavelenght, wich measured how much the emission wl increases as Tj increases. So when a LED heats, if affect lm (photometric) output by two ways, as both radiometric output and coefficient varies.

It affect especially blue leds because at the wl this leds emits small changes in it result in a huge difference in lm (for same radiometric power). So as Tj raises, lm output is affected by two opposite ways: radiometric power decreases, but conversion of that power into lm increases. Finally, we didnt perceive the change, but be aware plants will do, as lm is nothing for them, and they uses the physical (radiometric) power avalaible.

Behavior of radiometric loss with the increased Tj of Blues and Royal blues is almost identical. Fortunatelly to us, the royal blue graph is given radiometrically, so we can distinguish between lm drop and radiometric loss.

But this effects works on the same trend for AlInGaP leds: when the chip heats, wl peak increases, and it result in lm drop (for same radiometric power), while radiometric power aswell drops. It results in the huge lm drop with increased Tj. Radiometric loss alone, wich is what we are interested when designing grow lights, is lower, although its still higher than of InGaN chips.

The practical consecuence of the radiometric loss with increased Tj is that we must try to keep Tj as low as possible in order to get the best efficiency. And this is the reason LED's efficiency cant be given generically, as it depends of the setup. Depending of mounting, cooling path and current used, efficiency of a led may vary strongly, easily emiting half photons per watt burned when the array has been bad designed.

The first purpose of this introduction is to show how important is to design properly a LED array if we want it works fine. The second, to teach how to design it properly and the practical ways for doing it.

Vf behavior vs If behavior.

When electricity flows across the LED, it do it with a differntial potential (voltage) and current. This two parameters of electricity goes by hand, as diodes let the current flow according to the voltage supplied.

So for fixed Tj, each current level correspond to a voltage level and viceversa. Vf-If curves describes this relationship:




Notice how as higher is either the Vf or the If, same increments on Vf correspond to higher increments on If. The opposite is valid too: same increment of If requires lower Vf increments as both goes increasing.

But this curves are built for a constant Tj of 25ºC. As in operating conditions Tj is way higher than that, it means in the practice each Vf associated to a given If is going to be lower. How much lower depend of the temperature coefficient of Vf (given at the start of the specs) and the increase on Tj.

This behavior is what advice to run LED by constant current drivers, instead of by regulating voltage (especially when its done with resistors). This way, the driver feed a constant current with enough volts avalaible, so each led sucks as many volts as it need to run that If. This avoid the concern of differential Tj along the array, wich results on differents voltage requeriments for same If between each led of the array. Aditionally, constant current drivers takes advantage of the reduced Vf to get an increased efficiency: as light output depends directly of the current (If) and not of the voltage, lower Vf for a given If means less power drawn for the same light output (remember, Power=Vf*If).

So now is time to analyze how the light output behalf versus If





First conclusion: increase of light output isnt linear with increase of If. Each new increment on If lead to a smaller increment of light output. It mean LEDs achieve their max efficiency al minimun currents, and goes lowering as we increases the operating current (If).

This effect is more pronounced for InGaN chips than for AlInGaP ones, wich behalf near linearly up a given current density. For the K2, its about 380mA (notice the slight change on curve's slope).

This effect causes the main dilema when designing a LED array. What to optimize, if cost (less leds) or efficiency (more light per watt burned). Both targets are always contradictory: if we choose to use less leds and run them harder, they will emit with less efficiency (so we need to install more total watts than using more leds but driven softer).

For example, with white leds. We have a target of 100uE as the light required in the space. If when we run the K2 at 1.5 A each led emits 4uE consuming 5W (0.8uE/W) and when we run it at 350mA it emits 1.68 uE burning 1.2watts (1.4 uE/w), then we will need to install 100uE/4 uE=25 K2 at 1.5A or 100uE/1.68=60 K2 at 0.35A. More than double leds installed for the second choice, wich mean more than double the cost, for more LED but too due more drivers are required.

But first choice (25 K2 at 1.5A) burn 125 watts to get same light output than the second choice (60 K2 at 0.35A) burning 72 watts.

But its possible to design the array at an intermediate point, for example 40 K2 running at 0.7A. The optimal choice must take into account the Tj achieved at each case, wich will determines the final efficiency and duration of the array.

I finish the introduction of this thread with some useful links for LEd arrays design:

Understanding power LED lifetime analysis and Thermal Design. From the Lumileds page.

From the Cree website
: LED luminary design guide
Thermal management

From Osram Semiconductor: Driving the Golden Dragon and Thermal Management

Happy construction,
knna

been wondering what site i saw led stuff on i remember seen good information by you on a few sites im gonna go read all your stuff on leds so i can get an idea of what im lookin at ,thanks

igrowwitmatches , welcome to the Garden

I started this thread looking for answer questions related to the technical side of DIY LED lamps. On my sig there is alink to a thread about lighting concepts alone.

Soon Ill continue this thread with a mounting tutorial for DIY LED lamps. Just I havent free time right now, but if you have any concrete question, do it, Ill answer it as fast as I can.


knna

ok now i need a link to where to buy the leds so i can build my own array

Sign me up! This is just the thread I've been looking for

Knna

I can't wait to get started with building a led array. The prices I've been seeing commercially are re-damn-diculous. Thank you for starting this thread. when your schedule opens up I'll be here waiting.

Thanks for taking this project on.
lookie_lou

Wow knna
I finally found you
I've read a lot about you and led lights
I'm a great believer in this new technology as I have 4 of my own leds and love them
AWESOME thread dude and I will come back to check it out
I'm a chick so will have lots of questions

dont know enough to have a question but thanks