Sound like potential for a DIY lamp?

So I have been looking at stuff on fleabay and have seen http://www.ebay.co.uk/itm/390304162652?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649 and other similar goodies ranging from 10W @400lm to 100W @10000lm with some nice white ones in the middle.

I propose this:
4* 10w @ 460nm = 1200lm
4* 10w @ 630nm = 1200lm
This left holes though so:
5* 10w @ Cool white (6000k) = 4500lm
5* 10w @ Warm white (3000k) = 4000lm
and then just because I can:
2* 20w @ Warm white (3000k - 3500k) = 3200lm

Totaling ~220 Watts of power drain (Much less actual eg: Red Led is 8v - 1amp so more like 8 Watts but we'll assume substantial loss for the supply)

Uhh yeah and a filthy great bit of 6mm Aluminium for a back, maybe some chunko resistors that kind of thing.

From my experience in building led panels. A thin sheet of an aluminum is not an effect heat sink. You should try buy a regular heatsink (min 30cm * 15cm) and slap a couple of 120mm fans.

Really? Aww hell, I was hoping that I would get away with running it passively I was rather hoping for better performance than 220W of light AND heat.

The plan (probably a bit hopeful tho) is to mount the LEDs and the resistors, a total of 4 large 100-150W aluminium ones (for max heat dissipation etc) onto an aluminium sheet 320x300x6mm using thermal epoxy.
In an ideal world the surface area would be enough to compensate a little for the lack of ridges because tbf' a heat sink or set of heat sinks with that kind of face would cost £££ it is a hefty old bit of metal however. Plan B is to mount one or two (depending on need) 120mm fans at the back blowing down onto the plate and resistors, I have a vague fantasy about mounting one of these http://uk.shopping.com/NorthQ-NorthQ-Silent-Ice-UFO-80mm-CPU-Cooler-Alu-copper-Socket-478-775-754-939-940/info because it would look cool and I have one laying around, it wouldn't actually "cool" but it would definitely look the part hehe. As they say - an object must be either useful or beautiful, the best ones are both.

Unless you're massively under-driving them, running the LEDs directly with resistors has large potential for thermal runaway, followed by burned-out LEDs.
As the LED gets hot, the Vf goes down, and the current goes up. As the current goes up, it pumps out more heat, the Vf drops further, and the current keeps on going up.
You can get away with it using low-powered LEDs (0.5W diodes running at half their rated current are okay on resistors) but I wouldn't run a 10W chip on resistors at any current for which you would want a 10W-capable chip (or worse, array).
Spectrum should be awesome.

Thanks for the response! Heat is my main concern (actually my flammable possessions are) so I actually plan to run everything at it's lowest rated voltage already. I deal with computers for a living so Thermal runaway has a meaning to me, you are coming from a different angle but the principle is the same - active cooling ftw then.
I didn't quite get "I wouldn't run a 10W chip on resistors at any current for which you want a 10W capable chip" though, do you mean that the heat output from scratching all that current would just be ridiculous? I am suddenly concerned as I was rather hoping that the use of resistors would actually lower the overall temps. I should point out that I have no physical experience of this bar creating a multi fan rheostat some years ago with resistors and a switch block.

From what you are telling me I need to be able to dump 220W of heat off that board? That may be tough but PCs can dump 180W of heat into the size of a postage stamp - if I can facilitate that then I can probably cope with it spread so wide.

Here's a couple of pics of what has been delivered thus far, hopefully it will give you an idea of the scale of my back plate against the individual LEDs - you can also more or less see my layout, I'm waiting for 1 more blue and the ten whites (3000k and 6000k) Here you can see (left to right) 10W Red 630nm 10W Blue 465nm and one of the two 20W 3000k ones.

This is the ali plate with some of the LEDs on there for scale, you can see that I have marked out where the others will go - essentially 2 strips down each side.


I am still sorely tempted to source a bunch of these GPU coolers http://www.ebay.co.uk/itm/80mm-PC-Computer-Graphics-Card-Heat-Sinks-Cooler-Cooling-Fan-2-Pin-/180937872832?pt=UK_Computing_Other_Computing_Networking&hash=item2a20bcd5c0 instead of the plate, each of them should cope with 10W fairly easily but that's like $50 over my budget. When the driver arrives I will do some extensive testing.

Active cooling then.

So I've been thinking today about heat sinks and how to get them cheaper than cheap. I have a suspicion that the real answer is thrift stores. Now a thrift store is gonna have a few items that contain epic heatsinks, the two I've thought of so far are:
Mini fridges, the 12v ones for chilling beer


Old school amplifiers (Preferably not working)


I'm sure there are loads. As I no longer live in the part of town crowded with such things I shall have to go for a forage this weekend - can anyone think of other "junk" items that may have useful heatsinks in them? I have an old Lian Li aluminium computer case that I could cut up but that's just gonna give me more sheets.

Ok so an exciting day today, I got my resistors through! I can now start really mucking about with some LED action and as a result I have some pictures for you!

This is thermal and concept test number one, included are:
50W resistor x 2 @ 2.2 and 3.9R they have aluminum housings in a fetching Green color

LED specs:


The White LED ran at a flat 12V

It looks like I'm going for a Darwin award but I have it all under control I promise. I also had my faithful and trusty assistant (well out of the way before I flicked the switch)
View attachment 2339840

For tools I didn't have anything other than a temperature gauge with inbuilt alarm
View attachment 2339842

So I flicked the switch
View attachment 2339843 ------ View attachment 2339851

No fizz no pop no fire - sweet this is going well - so I wired the third LED
View attachment 2339844

And had to go and find some shades because that shit is blinding
View attachment 2339848View attachment 2339849

I left this running for 20 Minutes, there was an obvious increase in light as the resistors got hotter along the way and applying the temperature gauge clearly showed that the resistors had become far warmer than estimated BUT they had not been thermal taped down. The LED temps didn't rise anything like so quickly and in fact remained well within tolerance of the gauge.

Temps at end of test following a 20 Minute burn:
LED - warm but certainly well within tolerance.
Aluminum back plate - did a grand job of leaching the heat off the LED and was warm in the area but I am unconvinced of it's potential as this test used only 18% of the total wattage for a 20 minute burn.
Resistors - Way out of tolerance, temperature was approximately "Ow fuck that's hot" and the inbuilt alarm on my gauge went off.

All told a good start, I am however going to have to seriously think about how I intend to achieve this - on the plus side I now understand the statement

• I wouldn't run a 10W chip on resistors at any current for which you would want a 10W-capable chip​

More research required but one thing is for certain, ACTIVE COOLING specifically for those resistors but also for everything else.

As always I am semi begging for advice from people that actually know this stuff, especially if it saves me a Darwin award in the process!

A resistor is not a current limiter. A resistor is used to drop voltage or for logic isolation.
So your working current is 220w. You know what that means, you'll need a resistor that is rated at > 220w. At least double.
The really easy way to do this is just go buy a (a few) constant current drivers. Mean Well is a top brand.
You'll probably have to spend 60 bucks or so.
Go to www.rapidled.com and sniff around.

tenthirty said:

A resistor is not a current limiter. A resistor is used to drop voltage or for logic isolation.

Click to expand...

years of my life i was wrong!
thank you so much

So your working current is 220w.

Click to expand...

current is W(h)att

and a good old rule for wiring

Red is Blue and plus is minus

tenthirty said:

A resistor is not a current limiter. A resistor is used to drop voltage or for logic isolation.
So your working current is 220w. You know what that means, you'll need a resistor that is rated at > 220w. At least double.
The really easy way to do this is just go buy a (a few) constant current drivers. Mean Well is a top brand.
You'll probably have to spend 60 bucks or so.
Go to www.rapidled.com and sniff around.

Click to expand...

Thank you so much - I have been scratching my head over this one and here you have brought me the answer + rep I say! Payday tomorrow so I'm going to screw the $100 budget and make myself a monster light I know that if I spend another $100 then I may as well have bought a pre-made but would it be this much fun? I think not.

Do you think something like this http://users.telenet.be/davshomepage/current-source.htm would work with an adjustable voltage regulator would be the trick? Yes I spent all evening trying to understand

guod said:

years of my life i was wrong!
thank you so much



current is W(h)att

and a good old rule for wiring

Red is Blue and plus is minus

Click to expand...

Thanks for chipping in, it was a lot your thread that made me start this project Guod! Thanks for the rules of thimb - If I remember correctly when defining + and - they got them the wrong way round right? I love the idea that all of our technology is based on a mistake lol.

Jahulath said:

I didn't quite get "I wouldn't run a 10W chip on resistors at any current for which you want a 10W capable chip" though, do you mean that the heat output from scratching all that current would just be ridiculous? I am suddenly concerned as I was rather hoping that the use of resistors would actually lower the overall temps. I should point out that I have no physical experience of this bar creating a multi fan rheostat some years ago with resistors and a switch block.

Click to expand...

Yes, I wasn't very clear.

What I mean is that I would be quite happy running a chip capable of 10W on resistors, only if it were at a tiny fraction of its rated current. If it's rated for 3 Amps (like the Cree XM-L) I would see no problem setting the resistors for a current of 60 to 100 mA and it would probably run forever. If they get overdriven by four times the expected current, big deal - as long as they have adequate heatsinks they'll be fine. If I were doing that, though, I might as well use chips rated for 700mA or less, which are much cheaper.

As for resistors reducing total heat: if they're providing a significant slice of the total resistance, they're converting a lot of power to heat. If they're not providing much of the resistance, the current will fluctuate significantly with small temperature changes.

Using a mosfet-based constant current driver (as few as four components), I'm happy to put 19.3V of LEDs on a 20V power supply, with 0.6V used by the driver, and an extra 0.1V left as a buffer. Using resistors in series with LEDs (even at low current), I would want the resistors to eat at least 10% of the voltage, meaning only 18V of the 20 can go into producing light. Building the drivers is a bit of work, though, and they're only cheap if you mean to build 50-100 of them (and then they're not cheap in labor).

Also, consider how a small change in Vf can create a problem:

Three white diodes in series, 700mA at 3.2V each gives us 9.6V. The resistor is going to take up 2.4V, an excessive 20% of a 12V supply.

R=E/I = 2.4 / 0.7 which is about 3.43 ohms. Just say we make up exactly that much resistance.

Now it's a stinking hot day and the LEDs get a bit hotter than ever before, and their Vf for a current of 700mA goes down to 3.1, so if they were still at 700mA they would only eat 9.3V, leaving 2.7

I=E/R
I = 2.7 / 3.43 = 1.11 Amps

To keep the current at 700mA, we would want

R=E/I = 2.7 / 0.7 = 3.86 ohms

1.1 Amps isn't actually the current it will draw (we need to test to find out the real number), but it will be somewhere significantly upwards of 700mA. This just shows that a small change in Vf can make our resistor calculation badly wrong, and even using a massive 20% of the power for the resistor, the LEDs can be overdriven.

Jahulath said:

Do you think something like this http://users.telenet.be/davshomepage/current-source.htm would work with an adjustable voltage regulator would be the trick? Yes I spent all evening trying to understand

Click to expand...

Yes, BUT if you're building your own linear regulator CC drivers from components, then this is what you want instead:

http://www.instructables.com/id/Circuits-for-using-High-Power-LED-s/step6/The-new-stuff-Constant-Current-Source-1/

Those LM317s are neat, and I'd be using them myself if their dropout wasn't so huge, 2.4V I think. Even on a 20V supply, that's 12% of your power converted to heat before it even hits the LEDs. With the MOSFET driver, the dropout is set by the Vbe of whatever transistor you get - usually 0.6 or 0.7V.

EDIT: Although there are "LDO" equivalents of the LM317, their dropout is still up around 1.8V and they're more expensive - a little more expensive than the mosfet plus other components.

You could do it that way as an experiment. It's not very efficient and will get hot.

For the lazy.
http://reefledlights.com/how-to-diy-led/driver-faqs/

These things can get really fancy if you are willing to put in the time.

Hi all, thanks for the replies! I may actually be getting a half an understanding! I've looked through and costed up some components to make this:
http://www.instructables.com/id/Circuits-for-using-High-Power-LED-s/step6/The-new-stuff-Constant-Current-Source-1/ (Many many thanks Viewer 1020!!! and thank you for taking the time to elaborate, it's very helpful)

It came to too many $$ so I started comparing to pre-made devices like this:
http://www.ebay.com/itm/DC-StepDown-Converter-Constant-Current-LED-Driver-Buck-Voltage-Regulator-Circuit-/261096541994?pt=UK_BOI_Electrical_Test_Measurement_Equipment_ET&hash=item3cca910f2a

It looks like it will do the business, Input is good, efficiency is ~90%, they are adjustable (one size fits all and I can dim!!!) it costs $30 less overall than buying all the components! Any thoughts before I order?

Thanks in advance

Yuppers, now the question is where are you going to get 3A of 30V DC to plug into the wall?
3A x 30v=90w

Hehe that ones easy, I bought a 250W (12v @ ~21A) transformer for like 40 bucks (or rather ordered, not arrived yet dammit) back when I had convinced myself that it would work with resistors only (It won't) That should do a sweet job of powering all these AND I get the extra protection on my crappy cheap Darwin award contender

Yes over confidence is an issue for me lmao.

Thanks for the yay

That'll work for 4 of those 10w leds at about 2v each.
"transformer"??? or wall wart??
If it's a transformer, you'll need a full wave rectifier and some caps.

Sorry I need to get more pro with my words, a 250w, DC, 12v & 21a LED Water resistant switching power supply box - and I'll run 2 LED per PCB (with heatsinks) so the load will be 8-12v and 2a per unit, all powered off the switching power supply (clearly NOT a transformer then). The supply is designed to run outside LED arrays...

I'm hoping to leave some overhead on each pcb (5-12w unused out of the max 25w) and some of the power supply unused (~30w) so nothing is too stretched and so is more efficient than it could be with a bunch of in line resistors only...

Am I way off track? It was all seeming pretty peachy there...