How to build your DIY LED array

I have actually done a spreadsheet and XM-L have the "lowest lifetime cost per lumen" at 1 and 1.5A, IIRC. YMMV because the lifetime cost per lumen depends on 3 things: cost per LED, cost of electricity, and time cost of money (aka internal rate of return, in engineering economicspeak). Those 3 variables are different for everybody, but in general my spreadsheet says if electricity is cheap - < $.11/kwh, then run them full out (3A). $.11 or more then run more LEDs at lower current. That means a higher startup cost, but over the life - and I'm assuming 50K hours even though they are rated to 100K hrs - it will be MUCH cheaper.

There's also the certainty that LED's will get cheaper and more efficient, but we don't know the rates at which they will do so. I didn't account for that in my model -- it's too complicated!

They drop 3.3V @ 3A - so 10W total! Your heat is a fraction of that, and I think we're using roughly 70% now. Meaning only 30% of the energy becomes light(!), and the rest is waste heat to be dealt with.

As for driving them, lower current will give you max efficiency and lifetime, but you will also spend more on LEDs for an equivalent amount of lumens. Also, I've had trouble finding pre-made drivers that can push 3A, though 700mA and 1A drivers are readily available. The best binned XM-L's have over 140lm/w at 700mA, which I think is super cool.

I just want to thank everyone in this thread for an awesome information regarding leds. This is the most well written thread about led out here for sure!

As for the current to run XM-L, they dont show a noticeable efficiency drop (called current droop) along the full range up to 3A, just a continuous but slight drooping. This makes XM-L suitable to run up to 3A if efficiency is not the main concern.

However I generally advice against running LEDs at their max specs, due the efficiency loss (not big in this case) and thermal management. When running LEDs at their max specs, you need the thermal pad is very good if you want the best reliability of the system and lifetime, and usually it is something difficult to achieve on DIY setups. So I wouldnt run XM-L much over 2A in DIY setups. 2.5A as max, preferably close to 2A, say at 7W power (~2.2A).

When you analyze performance/cost, you need to do it first with the emitter alone. As imnotawino posted, then running them at max spec is optimal. But including full cost of the system gives a more complete picture. As most red LEDs available today (anycase, the ones which best performance/cost) run up to 1A, using XM-L requires to use a separate channel and driver for them. It is more difficult today to find high constant current drivers, and a little more expensive.

From the optical viewpoint, getting a good spectrum mixing when using such high power whites is more difficult too.

For those reasons, I prefer XP-G, with same current range and power than the rest of LEDs used in a grow light, simplifying electrical, thermal and optical setup.

On the other hand, and specially with a self made driver, using XM-L may achieve a nice initial cost reduction.

In order to have a reference of the current level, think on XM-L as having the chip area of 4 "3W". So 1.4A gets the current density of a XP-G at 350mA, and 2.8A equal to standard 700mA, give or take. Actually is a little less, but this way you can have a reference.

People runs the XML at minimum1.4A, but they are most used at 1.75-2A. And as said before, it can be used well past 2A. How much past 2A is more dependent of the thermal path. There is very little gain of running them below 1.7A. In order to reduce initial cost, running them at about 2A seems a good compromise with small trade off in terms of efficiency, with the possibility of running it harder and reduce cost still further.

When do we get the info and PICs for a DIY driver? Pretty much a given factor that no one makes a driver to suit our purposes.

I avoided this question because the best solution is usually a switched driver working directly from the mains, and it involves many small components so the only practical way of building it is on a PCB. And the cost of ordering an small number of PCB plus the components usually sum more than getting an off the self driver. As time passes, there are new drivers which match better with our application.

But there is a solution for DIYers simple to build and robust: its a classic scheme, consisting of working from a high power Constant Voltage PS and add a linear current controller to each string. In order to get a good efficiency, we need to use a CV source as high as possible. 48V is the standard higher CV supply, although there are 54 and 60V out there. Due safety standards, there is no typical CV supplies over 60V, because this is the called very low voltage range which dont need special or double electrical isolation.

A 48VDC supply allows to run a string of power a little over 30W (700mA) and losses about 1W, so about 96% efficient (but the CV PS has its own losses).

Implementation is very easy, you just need to use a constant voltage IC working in CC mode (a resistor fixing the current). I often suggest to use dedicated ICs, I like the Infineon BCRs, as they are cheap, have a low voltage drop and with an external transistor can feed at any target current, still very high current devices (although losses becomes higher as higher the circuit current, and past 1A makes sense to use a switched driver), and have built in protection (open circuit, overvoltage and inherent overtemp, because BCRs tend to lower output current as temp increases). Schematics are available on the linked page, very simple and straightforward, 3 components (BCR, current sense resistor and transistor, about 1$ for each string).

Ideally, a DIYer driver should include the conversion from the mains. This task is way more problematic for a DIYer as noted before.

However Ive been thinking on a simple way to achieve this working from rectified mains voltage. With a bridge rectifier we can get directly an high voltage source and then adjust current linearly.The problems here are two, the sinusoid DC obtained (which can by filtered with a cap) and mains voltage fluctuations. The later is a huge problem, as it may driver the rectified voltage under the string requirements or worst yet, over the IC allowed range. We could use a linear CC system by using directly a pair of transistor, resistor and diodes, and it should work, but likely losses could become large if the mains voltage is not very stable.

If we find a simple way of limiting the rectified voltage, this scheme could work pretty well, as linear losses when working with 165V strings (120VAC mains) or less yet, 300V stings (for 220VAC) are low in percentage.

Any ideas, TBM?

Ive found on same page the TDA 4682/3 nice preconverter, not too complex to build, and cheap components (8 resistors, 4 caps, one inductor, one MOSFET, for CV output up to 420V). My concern about working with high voltages is people unaware of the dangers dont use it with the required precaution.

The MeanWell constant current series is a good off-the-shelf solution.

I don't agree with your statement that noone makes a driver for our needs. The MeanWells are adjustable and hit 90% efficiency. Plenty good for me.

Or use some voltage regulators wired for constant current. If you know nothing about electronics or Ohm's Law, I would stick with the MeanWells.

How has your light been working?

Are chinese blue leds ok?

Edit: There is also similar red chinese available on Dealextreme

Joined forum for this one thread

Hello all,

First off Knna such an amazing thread.

I have some questions about LEDS and was hoping someone could help me. It has taken me 2 days to read the entirety of this thread. Most of it has sailed over my head. Im a stoner mushroom grower I have no clue as to electronics. I read in the beginning of the thread that HPS lighting was missing the blue spectrum. Wouldnt it help to buy one of those cheap dealextreme 10w blue lights to supplement a small HPS light? A small driver for it and some heatsinking and your light is maybe $20 from dealextreme in the right supplemental spectrum.

Now I know these cheap lights are no good for building an array but what about these 1W lights at 1.60 a piece? eLED.com, your online source of LEDs. Products include: LED Lamp, LED Display. SMD LED, SMT LED, Surface Mount, Ultra Bright LED, White & Blue LED, Circuit Board Indicators, Dot Matrix Display, Light Pipe, IR & Phototransist www.dotcom-monitor.com

I wanted to use the drivers from dealextreme on page 42 I think they were listed. For the heatsink Ill just source some aluminum from the scrap yard or use old computer heatsink's

I'm going to take the easy route - hire an electrical engineering student to figure it out for me

Seriously, I was checking out the new LED lights at my local grow shop and they are cool but mad money - $600 for something approximating a 500w HID. Wowza!

I know of a wholesale supplier that provides high density LED strips - kind of like the candy on a paper tape they used to sell back in the 1970's. The LED strips are super cheap - like a few bucks per linear foot. I guess the money is in the circuitry to drive it (and the know-how to make it work).

I'm going to look into this. Thanks for the hard work on this thread. Unfortunately it's over my head... I'm a nuts-and-bolts kind of engineer

not to derail thread, but the easiest way would be to pick up some cheap CFLs from the dollar store. buy different brands if the k isn't on the package, you'll find some are 'cool'