Project aborted. LEDs are still too expensive at this point.

Many people have speculated, few people have tried, less people have created plogs, and almost none have succeeded in creating an LED based projector. I plan to change all 4!

First, let me talk about a few of the advantages LEDs offer over MH bulbs...

Much longer life (MH might be up to 25,000 hours, LEDs are suppose to do 50,000 - 100,000 hours)
As close to instant on as anyone could care about.
No waiting between "restrikes"
Not as much of a detriment to turn on and off (it decreases the lifespan of a MH bulb to turn it on and only let it run for a few minutes)
LEDs typically only emit light in one direction: unlike MH none of the light is emitted in the exact wrong direction
They're a little more efficient (more lumens per watt)
Low heat: no need for costly heat resistant optics, LCDs don't need as much cooling, etc.
The heat they do radiate is easier to manage (heatsinks)
Virtually no UV and no IR - I got a sun burn from my MH bulb
You can dim LEDs
Low voltage - no 20,000 volt strikes here!

work on your projector with the LEDs turned down low
save electricity if/when you want

And some of the advantages of the small form factor

Scalable - Small LCDs to large LCDs, it's all the same to LEDs - Want to make a 30" 1600p projector? Fine, just find a large enough fresnel and triplet with a wide enough field of view
Reduce vignetting (dark corners) by using an array of LEDs
Small bulbs = smaller enclosures are possible

Of course, LEDs have their down sides too, here's a few...

Lots of soldering (not everyone's forte)
If you're using more than one LED you need to do everything multiple times (like aligning the bulbs and optics)
High upfront cost
They're kinda delicate
Like MH they don't necessarily have very good color rendition
Also like MH they require special circuitry (ballasts, drivers)
I'm sure there's more, but I want to get on with this post

There's tons of choices when it comes to LEDs...
What LEDs to use?

White
Red, Green, Blue

High power
Regular (5 mm, 10 mm)
Super high power (100 watt, 30 mm x 30 mm, giants)

How to get the LED light to the LCD screen / projection screen

Just use bare emitters
Reflectors / Custom LED optics
Lenses (traditional or fresnel)
Diffuser film
Some combination of the above

I'd like to touch on my opinion of each of the "choices" I've listed...
White: I'm leaning toward using white LEDs. There's so much competition from various manufacturers that prices are low, quality is high, supply is readily available, and new products are coming out all the time.
RGB: This makes a lot of sense, but there is added complication from using 3 different colors and by using 3 times as many LEDs. I might change my mind and use these at the last minute

High power: To me, this is the best way to go at the moment. Mostly because of the shortcomings of the other choices...
Regular: I haven't tried these to be honest. But they seem like the "old way" to manufacture LEDs. There's very little consideration in their design for heat dissipation, they're not that bright, and the manufacturers use lies/deception/fuzzy numbers to artificially inflate their specs. No thanks, I'll pass. Other's on the other hand have used these and are satisfied.
Super High Power: I haven't tried these either. Right now they aren't as efficient, cheap, or as available as the High Power alternative.

Bare Emitters: The main problem with bare emitters is that the primary optics on the LEDs aren't optimum for use with an LCD. Because of this, some other light spreading method is virtually required.
Reflectors / Custom LED optics: IMO these are good for flashlights, bad for projectors. Some people have used these and are satisfied.
Lenses: I've tried a few lenses, they've worked for the most part, but some major flaw has forced me to reject every one I've tried.
Diffuser Film: Using this is my current plan. It's certainly not the most efficient light dispersion method, but it's also probably the easiest and will produce the least vignetting.



I haven't finalized my design yet, but here's where I am now...
I'm probably going to use white high power LEDs, almost certainly "Cree" brand and hopefully their "R2" "WH" bins.
I'm shooting to use about 60. The R2s can do 250 lumens each (minimum I think) so 60 of them will be about 15,000 lumens. I hope to be left with about 100 to 200 lumens on screen.
I've done a standard horizontal build, this time I'm going to go much more advanced. My current design is a double-folded, pro, vertical, WUXGA, with lens shifting. I hope I can strip the antiglare correctly this time (I pretty much ruined my last one). This new enclosure should be about half the size of my last one.
Once again, my primary goal is a "high quality" projection, this time I'm going to try harder to make my box look good.

Well, I may have jumped the gun a little with my decision to use the diffuser sheets. I guess the search for suitable lenses continues.

I have to say, I think the most practical way to do an LED projector would be to modify an old slide projector with a small LCD screen and one of these.

Im interested! I kinda gave up on leds for awhile. I decided I want a working projector before I fool around with very expensive led array. but i will definantly be watching. I thought your fresnel experiment turned out ok, but I dont think i would want to watch tv on a projector that had 8 dfiferent shades of light throughout the screen. thats why I think diffusers (even though they loose a lot of light) are the only way to go right now. did you ever look into how much the 100watt edison leds will cost?

I think you're talking about this...

I'm right there with you on the different shades! Just to be clear, these are from 2 different brands and 2 different tints of LED. Left = Seoul P4 "SV0" tint, Right = Cree R2 "WH" tint. I'd say within the same brand/tint the color is consistent.
Oh, and the bottom right one is different because I have one of those LED optic/reflector things on it, just for testing.
I'm going to contact Fresnel Tech tomorrow to see if they have a lens array where the center rings aren't so thick/widely spaced. If they don't, I have tried a little with the fresnels and a very transparent diffuser, and the results are promising.

As for wanting to get something that works, instead of fussing with LEDs and stuff I can't say I blame ya!

P.S. Oh yeah... no I haven't looked into those 100 W Edisons any more. I think they're kind of a step in the wrong direction (they're technically an LED, but I don't see much of an advantage over MH).

I applaud your efforts, and hope to see you succeed.

I've also been considering it, and unfortunately I just don't have the TIME to build on my theories, but I'd like to share some thoughts on RGB vs white.

With white LEDs, you're stuck with the colour temp that they come with. Youc an make SOME adjustments with the LCD, but these are really a stop-gap.

With RGB, you can literally tune your output whites to whatever you want, so long as the LCD will let it pass. There are resources to EXACTLY colour match your projector to give truely real colours, and you can use these resources.

The PCB is a bit more complicated, you need to have 3 distinct circuits, and you need to ensure that your RGB clusters are correctly placed in order to work correctly. I designed one that can be infinitely repeated, I may print it.

Power would then be delivered through a set of emitter follower circuits allowing a near infinite variability for each of the red, green and blue outputs. 3 small trimpots could then tune your colour spectrum. This is even better than what the commercial projectors currently offer, and thus becomes somewhere that we can excel over them. The emitter/follower pair is basically the same basic idea as an amplifier, allowing a 1/4W trimpot to control a circuit that might be handling hundreds of watts.

Anyway, that's my hypothesis. I think that it's worth some consideration.

The counterpoint to that is that white LEDs are reasonably cheap, and the circuit is much easier to build and wire.

Sorry y'all if this post is rambling...

I'm a total newbie when it comes to circuits, which is a big reason I'm not using RGB. I mean, I know that negative goes to positive, and can tell the differences between series and parallel wiring, but that's about it. Right now I'm using these drivers to control the power delivered to my LEDs, but at $45 each I'm trying to limit how many I have to use (right now it's looking like I'll use 6).

From what I can gather; high power LEDs are best wired in series. This is because each individual unit is likely to have a different voltage drop depending on many variables (temperature, manufacturing differences, silicon quality(?), age of the LED, etc.). I'm not certain, but I gather that combining 2 LEDs with different Vf on the same leg of a parallel circuit is not a good idea for some reason.

It's also important to regulate the current, I'd like to run my LEDs at as many mA as possible without fear of overdriving them, even for a couple seconds. So what I'm left with is using power regulators like the CCHIPO (Constant Current High Power) (linked above) or something from luxdrive or something. I chose the CCHIPO because it's got the highest power output that I could find (up to 39 volts and up to 45 watts) and I've also done business with the TaskLED guy before (with a project I did before I even found Lumen Lab) and am very pleased with his products and service.

Anyway, the major limitation as far as I can see is the 39 V output limit. This means I can pretty much only connect 10 to 15 LEDs to a single driver (regardless if they're white, red, green, or blue). So, by using 60 White LEDs I can limit myself to 6 drivers, but with RGB I'd pretty much be stuck with using... like... 13 or something (AKA $585 worth!). That's bad enough, but that would mean I'd have to also get larger (or more) AC/DC adapters, which start getting expensive pretty quick. Lastly, all these additional components mean my enclosure will have to be bigger. Oh, they'll make more heat too, so more fans will be needed... it's a vicious cycle.

But, like I said, I'm a super noob* at all this, maybe there's a better way? If I can take AC and convert it to current regulated DC that would be great! SupraGuy if you know about such things could you give me a hand with this?

* OK, I've read a lot of theory and stuff and can repeat it, but I still consider myself a newb because I haven't actually produced much of anything tangible...



But, your mention that not even the commercial projectors let you control the red, green, and blue separately gets my brain churning (maybe that's my ego?). I did intend to do RGB tests, but I ended up breaking most or all of my blue rebels, and now I cannot find any more...

I can help you with that, I study Electronics and I'm also very buzy with this project

Btw what do you mean with Diffuser... You give up on the fresnells?

Using a fresnell for a led array is impossible isn't it... that's why you used a fresnell array...

Grtz Keires

Hmm based on your assumption that you will drive your LEDs at max while watching video, then the only option would be for you to dim the RGB LEDs individually to adjust the color ... which seems like it would have the same dimming effect as adjusting your LCD colors? So is there really any reason not just use the white LEDs?

But why would we use RGB leds, the light output of these leds are not that high as white leds... (I'm sorry but I'm not RGB up to date)

Well, I would build my own drivers, because that's the kind of thing that I would do. Spend $600 on commercial units? Bah!

Partly your information is correct, however, there is a trick to it...

LEDs will always have slightly different Vf, so you tend to put them in series strings. The trick is this: You can put more than one series string in parallel, if you use an equalising resistor. If you can have 45W of power, you can have a bit more than 1A at 39V, or, you can have a bit under 9A at 5V. This is more useful a circuit. You can then put the LEDs in strings of 3 with a small resistor (Offhand, I'd say about 100 Ohm, but I'd need to do the math for sure.) The resistor acts as the current limiter for each string of 3 series diodes, and ensures that they don't fry. This also means that when (not if, but when) an element fails, only 3 lamps go out. With your massively series strings, any LED that fails will take out 1/3 of your light engine. Plus that's going to be an absolute NIGHTMARE to troubleshoot. With my circuit, each string is only 3 LEDs, so at worst you replace all 3 lamps and get on with your life.

Now since a regulated 5V with large current is a relatively easy power supply (Heck a PC power supply can literally supply HUNDREDS of watts at 5V) this makes the whole thing cheaper yet. There is the additional expense of a bunch of resistors, but since this is likely to be less than $10, I think it's affordable. The LEDs themselves are the expensive element.

I don't know what you know about transistors, but basically a transistor is a miniature amplifier. Apply a small current to the base to get a large current across the collector/emitter. This means that you can use a 1/4W pot to control a large wattage current with a lot of precision.

Sure, you are "turning down" some of the LEDs to get the right colour, but you leave at least one of them at maximum, so what you get is the brightest possible light at the desired white balance.

If you like, I can spec out a power supply for you that you can use for either white or coloured LEDs.

I'll definitely take all the help I can get on this subject! Once I start delving into exactly what LEDs I'm going to use I'd love some help with choosing the best power supplies and stuff. Hopefully it wont be too long now.

Sorry for being confusing about the diffusers. I tried one with my current setup (ruined a perfectly good 19" LCD to get the diffuser BTW) and thought it was giving decent results. Then I did an "A : B" comparison between the diffuser and my fresnel array. The fresnel array blew the diffuser out of the water entirely! I think it's possible a very transparent diffuser may be helpful with smoothing out the rings and masking the seams between the fresnel array elements. I'll try to get something put together to test this...
So, the quick answer is: a diffuser alone probably isn't going to be enough. Some kind of lens is going to be required.

It's going to be impossible to use a single fresnel with an LED array (because any LEDs outside of the focal point of the fresnel will be wasted), but matched sets of LEDs and fresnels in an array will work.


You're right for sure, but...
As I just recently learned as well, the balance between red, green, and blue lumens isn't even in white light (or at least in humans' perception of "white" light). The ratio is something like 2:6:3 (don't quote me on this). So the red and blue are already turned down a fair amount. For instance with the Rebel LEDs you'll want to buy the brightest greens available (or use 2 of them) then match the red and blue by limiting their current. I figured it out once, and I have forgotten the exact numbers, but the power ratio would be something like this: 400 mA:1000 mA:680 mA (R:G:B) using LEDs that put out 40:80:23.5 lumens each when driven at 350 mA.
Quick answer again: we already have to turn the red and blue down quite a bit to get white, so turning them down a little more/less shouldn't be a big deal.

You can change the colors with the LCD, but that is really just artificially darkening or lightening (are those words?) that color. The adjustments reduce the range of different shades of that color the LCD can put out.
I'm not saying I can tell the difference between adjusting the lights or adjusting the LCD, I'm just saying the difference exists.

They're not as high, but they're close. The theory is something like this...
The white LEDs are putting out a bunch of different colors (yellow, purple, cyan etc.), those colors are blocked by the LCD (or they should be). So, if you start with 100 lumens of all the different colors it stands to reason that by cutting out the colors the LCD cannot use you're left with less lumens; so it's possible it would even out.

The other advantage is that by the LCD doesn't do a perfect job of filtering out the other colors, those other colors will muck up your picture quality. By eliminating those other colors by using RGB LEDs we should see an improvement of picture quality. (BTW, that is why that other thread is called "Rgb Leds As A Light Source? A Spectrum Advantage?!?"

A. That's usually my case.

B. One could use the 12 V rails of a computer PSU also right?

C. I don't know anything.
To dumb it down even farther: So I could use transistors to fine tune my LEDs, sounds good to me! Are there down sides? (Maybe they produce a lot of heat, they're expensive, inefficient, or they fail easily?)

D. I'll take you up on that, later though. I'll eventually get around to wiring a couple of my remaining green and red LEDs together to see if I can get them both collimated by the same element of my fresnel array, if I can't I may scrap the whole RGB thing.

Thnx Hirundin, I understand completely now why RGB leds are an advantage and why we have a low lumen output when we use white lights! So the big reason why they haven't tried the RGB leds yet, is because it's not easy to drive them?

Btw, I have searched on the web, but I can't find the difference between a fresnell and a diffuser...(sorry I'm still a noob in optics)...

I can give you an answer on your electronic questions, but maybe Supraguy will love to do that

But if you can't wait, here are some answers...

B. You could use a power supply of a PC, but the supply has to provide enough current...9A (Supraguy has calculated it, 5V at 9A). Ofcourse you will need a circuit to down size the 12V to 5V, but that's no problem.

C. Don't worry about the costs of a transistor... it isn't high at all... it works perfect, normally it never fails... (except if you blow it ) and the big advantage is that you don't lose power... with a regulable resistor you will lose power... (I'm sorry if it's not clear enough, but my knowledge of english isn't that perfect so I missing important words to explain it.)

If the current through the transistor is high, you will need to cool it... normally with a heatsink...but you already using that, so that's np, right? Or you can use more transistors, so the dissipation (P= U.I) will be divided over the tranisistors...

It also has other advantages... if you want to know more about it, np just ask maybe it's easier for you to add me on msn, if not np


I hope I have helped you, greetings Keires

Transistors are awesome little devices. They're very reliable, as long as you stay within their operational parameters for power and temperature.

Yes, they'll need heat sinks for larger power applications, but this really isn't a problem, since those can be placed literally anywhere, they don't need to be (Nor really should they be) close tot he LEDs.

I suggested 5V since this is actually the largest current source in a PC power supply. You're not limited to 9A, you can usually get more than 60A at 5V from a typical 450W PC power supply. That's more than enough to power your whole LED array, hopefully. If not, well, it's not all that expensive to do more. Additional advantage is that the 12V lines are left available, so you can use that to power fans, and possibly even your LCD.

So what you need is a circuit to take the 5V supply available at the power supply, regulate it into 3 channels, one for each of the red, blue and green LED sets of circuits. The other reason for the 5V source is that now you are dealing with a circuit of 3 maybe 4 LEDs per string. As I mentioned this makes troubleshooting much easier when you have a lamp burn out. They may have a MTBF of 10,000 hours plus, but I've seen LEDs fail with only a couple dozen hours on 'em. It happens, and when you're dealing with larger numbers, it becomes more likely.

Hmmm... Now I'm giving this some consideration for myself...

Hirundin? Any new results?

I'm still kicking... nothing new to report.

Not much anyway. I've been in contact with the fresnel company a little more. I even visited a local optics shop that had been doing business with Fresnel Tech for years.

I'm also trying to get the LEDs situated... There is that group buy going on, but the brightest LED that I could get through that is a Q5. I asked Cutter if they have an ETA for more R2s and they said "late February or March". Too many choices... I'd like to help with the "group" part of the group buy, but it seems that not many people are interested. Of the people who are interested, I'm the only one who wants Q5s (so far the other 2 participants want P4s) so it's possible that my contribution wont matter... Then there's the tint issue. The Q5s are only available in "WG" tint, and the R2s are only available in "WH" (or that'll probably be the case). I'd like to try all the tints for myself, but the 2 weeks it takes to ship across the world is excruciating! Why can't 7-11 sell state of the art LEDs?

Also, new LCD controllers are emerging. I'd like to consider the new one from johnzo1995 but first I'd need to know things like "how much" and "when"...

Very interesting, the R2S, what kind of LED is that? Is it brighter than the Q5s?

The new controllers, are that the controllers of Chen?

If I can help, let me know. I have posted something in the CREE groupbuy topic aswell, check it out

Greetings, Keires

See my new post XGA LED projector maybe you will have some answers about LED projectors and help me with some suggestion's since this LEDs that I used are not enough.

Welp, I put in an order for 4 different tints of Q5 Crees.
Click to view attachment

This'll give me a little more practice with mounting and soldering ye olde Crees as well as test how the different tints look. I hope they come within 2 weeks!


I don't know anything about those 5 mm or 10 mm LEDs... or commercial projectors. Sorry.

Welp, I'm about to give up on this frigging Fresnel Tech company! I'll call them again today, if they can give me an immediate answer that would be great. If not I'll probably tell them to just forget it...

Oh yeah, I also got word that my crees are on the way.

I may put this project on hold for a while...

I just put in an order for the RoGR bearings and plan to buy the motors as soon as they are released. I'd like to focus on getting this thing built before proceeding with the projector. IF I can actually get it built it'll help immensely with building a quality LED projector...

Absolutely it will.

Another use for a CNC router is doing some low-depth milling which can be used on copper clad board for making printed circuit boards. Once you can do that, then you can take a drawing of a circuit board, and make it a reality. This will make building even fairly complex circuit designs much easier, and allow you to do things that you might not otherwise.

I will miss watching your progress in this thread for a while though.

Yeah! Making PCBs would be awesome! I've looked into the chemical method a couple times, but I guess the whole process is so foreign to me I'm scared to actually attempt it.

Welp, my tint-test LEDs came in, um... on Monday, but I haven't hooked them up yet. I'm not exactly sure what I'm waiting for... Maybe I'll get em onto a heatsink tonight...

Threads going off topic? Here? Does that ever happen?

LOL. I actually deleted the whole post, since I see no need to have a post that just says "*deleted*"

Anyway, I should look at those LEDs myself, I was thinking that it could be a good thing to know something about them, and I'd love to help you design a driver board for them, so that you can drive the LEDs from a relatively low voltage power supply.

I'll have to follow your RoGR progress, too.

This RoGR thing is getting complicated (read: expensive)! Apparently someone told the world that all CAM software should be $1000 or more... or free. There's not a lot of middle ground. I'll prolly start a GRog when I actually get some of the parts in... I'm very conflicted as to whether I should actually attempt it or not...

A quick update regarding the projector...
I got a whole buncha lenses from crazychen and have been practicing gluing them to pieces of acrylic ("gluing" isn't actually technically correct, the stuff I'm using actually "welds" acrylic). I'm thinking I'd like to have a single piece of acrylic with all the lenses glued to it, this way I'll probably be able to seal off the light chamber and have only a single, flat surface to keep clean. I don't think the welding process I'm using will effect the efficiency of the lenses, but I haven't really tested it yet.

I'm having trouble deciding between using 228 (WOW!) P4 Crees (~80 lumens per watt) or 135 Q5 Crees (107 lumens per watt)...

Ignore the difference in the size of those circles, the smaller circles are just the lenses without their holders.

Pros for 228 P4s

Lighting will be more even, I should be able to place the LEDs closer to the LCD without hotspotting
Will be brighter at 350 mA (18,376.8 lumens)
A little cheaper, at least for the LEDs alone (but only $20 less)
Possibly more tint choices from crazychen's group buy

Cons

More soldering
More gluing and other stuff
More power required (263.34 watts, or so)

Pros for 135 Q5s <- I'm leaning toward using these (mostly for the power (and therefor heat) reduction)

More efficient (107+ lumens per watt)
Less power required (155.925 W)
Less soldering 'n' junk

Cons

$20 more for the actual LEDs (but maybe I can save on the power supply)
Not as bright @350 mA (14,445 lumens)
LEDs will probably need to be farther from the LCD (which means bigger enclosure)

SupraGuy, I think I'll take you up on the offer to help me with driving these LEDs starting... NOW!
It looks like for the same lumen output the Q5s will need 220 watts (compared to 260 for the P4s)
@350 mA the total of all the Vfs for all the Q5 LEDs will total 445.5 (compared to 752.4 for the P4s)
At full power the Q5s will require 499.5 watts (compared to a whopping 843.6 for the P4s)
--Are any of these factors going to be impossible/hard to pull off? For instance, finding a 500 watt computer PSU is pretty easy (in fact I have one), getting up to 800 watts is a little harder (but then I could just buy 2 right?).
Will any of these factors increase my cost/hassle significantly in some other way? Like, if each transistor can only do 50 watts or something the cost difference between 10 for the Q5s and 17 for the P4s would add up quickly it would seem.

2 PC psus would work just fine. LEDs typically work just fine so long as their forward voltage is met, and their max current is not exceeded.

You will have to check the documentation on the PC PSUs, though. Most of these have separate limits for the 12V and 5V paths. In most cases, it is the 5V path that has the higher wattage limit, but that's not a problem, so long as the LED circuits is designed for the proper voltage.

For the LEDs, you basically ensure that you have enough voltage to activate the LED, then you tune the circuit to have the appropriate resistance to limit current to an amount that won't blow the LED. You set up the voltage by putting LEDs in series until your power supply is just a bit over the forward voltage. Computer PSUs are cheap enough, so we'd design using hte regulated 12V output to limit the number of parallel connections.

Transistors can be had that will handle very large amounts of power, after all, that's what they use in the computer PSU. I only have a few high power devices around, but they're capable of a couple hundred watts each. The trick is that they need a heat sink. I'm thinking that this may not be needed, however, if you aren't going to want/need to fine-tune the amount of light that you're putting out, however. If you just want to have a "high/low" setting, then it's probably easier just to provide an alternate power path for the different current requirements.

Circuit resistance, on the other hand, could be a bit more of an issue. The resistors to limit current need to be able to handle the amount of power that they will dissipate, or else they fry. Fortunately, we're only dealing with miliamps, here, so can use nice 1% tolerance resistors, though we might have to put them in larger parallel configurations.

I've got this pretty bad ass Corsair PSU that'll put out 41 amps on the 12 V rail (28 A on the 5 V**). Lets see... 41 * 12 = 492 watts... about perfect for the Q5s. While messing with projector designs on the computer I've had trouble fitting even a single PSU in without increasing the enclosure size significantly.

If I'm using the 12 V output, and have ~3.3 to ~3.5 Vf on the LEDs it sounds like I'll be limited to 3 LEDs in series at a time (I don't want to hook up any LEDs in parallel, unless each parallel line has it's own power regulation). 135 / 3 = 45 <- does this mean I'll need 45 transistors?

Another option is there are ~36 V power supplies on eBay (like the one crazychen is using (I think there might be some ~48 V ones too). With those I could probably hook up 9 per series line, 15 lines for 135 LEDs... Another advantage of these is that they're smaller in size. I do worry about the quality though.

**I've read a few power supply reviews and I think the typical power output of PSUs these days is weighted more toward the 12 V side than the 5 V. It sounds like newer processors need more 12 V (hence the 12 V plugs on motherboards) where as older ones needed 5 V.

i'd use Q5 because of the power requirement.

for my part less power requirement than mh is the big advantage in led technologie and with 260W there isn't much advantage

For kicks I made a lil' Cree LED 3D model. The measurements are very accurate, but I fudged a couple of the dimensions (mostly the LED die).


Here it is... (it's a Rhino 3DM file)

Zero transistors., or, if you want to have some fine control over the lighting amount, maybe a bank of 4 to 6 devices in order to keep them in their safe operating area.

The way to individually regulate each string of 3 LEDs is with a resistor, so we need 45 closely matched resistors. 1% tolerance resistors will be perfect.

What's the current ratings for those LEDs?

Essentially, though, what we want to do is feed a fixed and regulated voltage to the LEDs. The LEDs can be in parallel strings, that's not a problem. A resistor in each series string will ensure that current gets shared equally, even if the LED devices themselves are not identical.

Typical Vf for Cree XLamp XR-Es...
@350 mA: 3.3
@700 mA: 3.5
@1000 mA: 3.7

So, just so I know I've got this straight... The resistors are based on the input voltage and the desired current, so it wont matter that the LEDs will likely have different Vfs, right?
I do think it would be cool to be able to switch the projector to some kind of "high mode" when I decide I need to! Hopefully "low" will be good enough, 'specially at night.

Hmmm... LED tint test: failure.

A quick note about Cree LEDs...
Cree LEDs have anode and cathode connection points on both the top and bottom of it's little rectangle base. The little tiny contact pads are very difficult to solder on. A good size reference is 0.7 mm is a standard size for mechanical pencil lead.


When I first plugged in the LEDs after hooking them up for one reason or another 2 of them did not light up. What I find especially odd is that they were numbers 6 and 7 in the series string of 8. As a non-electrician the only thing I could come up with to troubleshoot the problem was to unplug and re-plug the AC adapter. When I plugged it back in only 3 of the LEDs lit up (numbers 1, 2, and 8 IIRC). I did the plug cycle again, and the original set lit up again (the 2 that were off the first time were still off). Discouraging for sure, but not the end of the world, I figured since the 2 'duds' were from different tints I could still do the test and figure out the problem later.

So, I turned them on and messed with the fresnel array for a while (if you don't know what I'm talking about stay tuned, I'll put up some pics soon...). I was trying to see if I could get rid of the seams between the lenslets. After a while that proved futile so I turned my attention to the actual differences between the tints.

With the LEDs running at about 350 mA and without the LCD in place I'd say the WH tint LEDs looked the most "white" to me. Oddly enough, the WG looked very yellow. The WC and WD were noticeably blue. With the LCD in the light was too dim to judge which looked best, but if I had to choose I'd say the WC looked the most white. So I decided to turn up the juice to the LEDs. I turned them up a couple different times, but when I plugged them in after turning them up to what should have been ~875 mA they got pretty bright for a second or 2 and went off. I've heard something like that happens when you drive them with too much current, but that shouldn't have been the case. Now when I plug them in nothing happens.

I figure it's possible the LEDs were shorted out on the heatsink (through those bottom connection points). My previous test heatsink was anodized, this one was not. I anticipated it could be a problem, but after thermal-epoxying them all to the heatsink I did a continuity test in every way I could think of without finding any errant connections so I thought I was OK. Also, the thermal epoxy I was using, Arctic Silver, is not technically non-electrically conductive. The manufacturer says the electrical conductivity of Arctic Silver Thermal Epoxy is "negligible". Maybe after turning up the current the negligible conductivity turned egligible? I'll have to order some Artcic Alumina for the final build (and maybe the $60 to have the heatsinks anodized would be a good idea... or kill 2 birds with one stone and use the LEDs mounted to "stars").

I haven't given-up on these LEDs yet, I'll test them with some double-A batteries to see if there's any life left in 'em.

On the plus side this test setup looks pretty darn cool!

Dead LED. Yep, that's the result of overheating, or overcurrent. They're not coming back.

There is no visual distinction between a live LED and a dead one, just one works, and one doesn't.

I know very little about led but I bought a 3 led light a while back that operates on 3 aaa batteries from the $ store. The 3 led are each sitting in a chrome cone shaped reflector. These look very bright and are impossible to look at directly without creating semi permanent spots before the eyes that last close to a 1/2 hour.

I pointed it close to a white melamine shelf until I could see the 3 distinct shapes of the led reflected on the board. The reflection of the led itself at the center about 3/16 of an inch wide was super white and then the ring got paler/dimmer as it got away from the center. It was still pretty bright but the difference was visible. If an led array has that same separation of brightness, wouldn't that create a hot spotted look in the final projection?

Is this what you hope to correct with a fresnel mosaic or are you just worried about filling the gaps in the array?

Darn! I wonder if there's something wrong with this LED driver... It's controlled by potentiometer, by changing the resistance on the pot I should be able to change the output current (much like those transistors I think).
Click to view attachment
Resistance to output

The time that the LEDs stopped working I had the driver set to <12 K ohm.

Maybe since they weren't electrically isolated they shorted out and were essentially connected backward (reversed polarity I think it's called). Would an LED that was hooked up backward have any physical clues that I could look for?

I knew I should have waited until I had better heatsink epoxy!

Since that light was from the dollar store I have to assume it was made with 5 or 10 mm LEDs. Those LEDs usually have their optics setup so that they create a very tight spot (say 15º or even as low as 6º). The manufacturers cannot control the light perfectly, so some spills out the side. I think what you saw was the center spot created by the optics and the dim spillage that is impossible (or at least very difficult) to avoid.

The LEDs I'm using have a much broader beam angle (around 120º I think).

The light experts here have said that light outside of a 6º spot wont make it through the LCD and triplet, so for my LEDs I'm loosing something like 114º of my 120º beam. My plan is to use some kind of lens to collect as much of the light as I can and bend it down to as small of an angle as I can. There seems to be 2 realistic options: a fresnel array or using individual LED optics. The fresnel array will create a perfectly collimated beam (0º beam angle) but don't collect as much light, the LED optics collect almost all of the light, but don't make a perfectly collimated beam. Each lens type has it's pluses and minuses, but right now the complete unavailability of suitable fresnel arrays makes the individual LED optics the clear winner.

So the short answer to your question: I was essentially trying to create that hot spot by using the fresnel array. The idea was to create a square hotspot with almost no side spill.

About the power supply for leds, i thinking in use 2 or 3 old AT Power Supply, that is less then USD$5,00 for me, i will make some modifications on it, and will got about 14,4~14,6v @ ~15A from it. (how i can make it?? use the translator then make easy for u: http://babelfish.altavista.com/ and translate this link: http://paginas.terra.com.br/lazer/py4bl/fonte.htm for ur language)

Probably i will use this array




So for a best configuration with 88 leds, i will use the tension on 14,4 ~ 14,6v for 4 leds in serie (3,6 ~ 3,65v for each led). At total i will have 88x3,7w (~325w) with a good amount of lumens (~14,000lm) i hope i have good results with it.

For me was to much expensite USD370 in leds, so i be a little limited in quantity of it.
Im buying the power supplies and when i got the y i will start make the modifications and post they (i have to make a plog).

Good lucky.

PS: if u have msn add me (murilo.sonegatti@gmail.com)

Interesting! I haven't looked at that link yet, but I'm curious: what is the advantage of 14.4 V over 12 V? What I mean is: isn't it possible to use 12 V and just divide your array into sets of 3 (lets see, 88 / 3 = 29 sets of 3 and 1 by itself OR 28 sets of 3 and 2 sets of 2).

[edit]I looked at that page, it's very confusing after being translated... I can't be sure I'd understand it even if it was written in English though.

I technically have an MSN account, but I rarely use it.

I need get or ~14,5v(for 4leds) or ~11v(for 3 leds) to make it drain about 900mA ~ 1000mA from each set then i dont need control the current from power suplly , if i use 12v in 3 leds i will have 4v in each led, when the max voltage is 3,7v.

I thought the voltage rating of LEDs was different than most devices. I thought it was OK to supply a higher voltage so long as the current was regulated.

For instance, imagine a set of 4 series connected LEDs on a 14.6 V line where each LED has a Vf of 3.7 V. Wouldn't the first LED get the full 14.6 V?

Also, wouldn't the second get 10.9 V, third: 7.2 V, and the fourth would be left with only 3.5 V?

I can be wrong, but in a circuit in serie, all components has the same current, so, if one led have 3,7v@1000mA, the others will have 1000mA too, but in my case i dont have 3,7v for each led but at max 3,65v.

I know, I think it also, but I couldn't find the delete button

No the voltage will be divided in 4, (not perfect in 4, because the voltage of each led will be different, not a lot but a little bit) But the total will always be 14,6V.

For example: 3,7 + 3,6 + 3,8 + 3,5 = 14,6V





You are not wrong, but correct In serie all the components have the same current

Grtz Keires

If I can help you Hirundin, let me know

Um... so will the voltage be divided in 4 or not?

You start by saying that it'll be divided in 4, then you say it wont be perfect, then you show that it's not even close.

So, what if you had 4 LEDs that all needed exactly 3.7 volts, how would 14.6 V be distributed between the LEDs?

I will go with the 135 Q5 (more efficient and not a big number of LED) and the power supply I will recommend Notebook-power-supply
this is 90W 18.5V 4.9A and if you use 2 in serial connexion you will get 37V and use 11 LED / string 37/11=3.36 and will demand about 400mA at 3.36V and 11LED x 12 row =132 LED and 12 row x400mA is 4.8A
The advantage of using notebook power supply is size and no noise.

Okay, the way that I've always driven LEDs, and it's worked for me...

V = I * R + Vf

V is the circuit voltage. We'll call it 12V for the purposes of this circuit.

I is the amount of current. Let's just say 500mA or 0.500A.

R is the resistance, in ohms of the current limiting resistor

Vf is the total forward voltage of the LED(s)

So, for a series of 3 of your diodes, we'll assume a nominal Vf of 3.6V works out well, and call it 10.8V

So we get:

12 = 0.500 * R + 10.8

Subtract 10.8V from both sides...

1.2 = 0.500 * R

2.4 = R

This means that you need 2.4 Ohm resistors in series with each 3 element array. ([url=http://led.linear1.org/led.wiz]This[url] design tool which I found on the web says 2.7 ohms.)

To drive them to 1000mA (1A) you would need 1.2 ohm resistors.

I always design with a safety factor involved with LEDs, particularly in large arrays. Replacing single elements is always a pain. Figure the safety factor as follows:

Resistors have a tolerance. Always assume that they'll be to the bad side of the tolerance. I typically use 1% tolerance resistors, but for high power resistors (1A X 1.2 Ohms = 1.2W -- more than a 1/4W device is going to take. You need a 2W resistor to be able to do this job.) 5% is typically the best tolerance that you will get. Assume that it'll be 5% low.

The LEDs will have a tolerance as well. I've seen some as much as 10% out on their measured vs. nominal Vf. Fortunately you can assume a slightly closer tolerance, since you're dealing with 3 in series. This statistically works in your favor, but again, assume a 5% tolerance, and low. Your regulated voltage may be off, as well, by as much as 5%. Add 'em all up, and you could easily be 15% off in your calculated values. Being a bit low on current will cause you to lose a little bit of light, but being a little high will cause you to lose a lot of light. Like all of it. The 2.7 (or 2.8 is I think in the E12 series which is easier to find) for 500mA of current is a safer bet. 1.5 Ohm resistors would be a safe value for higher current, provided that you have adequate heat sinking. Again, 2W devices are a minimum.

[I got a very long phone call while I was writing this... I figured I may as well post it, I haven't read SupraGuy's post above yet...]

Yeah, the more I think about it the more the 135 Q5 method makes sense.

Using laptop power supplies might work, but I would be nervous about them overheating. Running them at full load for hours on end doesn't sound like a good idea to me. Taking them out of their plastic housing and sticking a fan on them might solve that concern, but that means they wont be silent any longer.

I'm really trying to do this "right". From everything I've read from people who know high power LEDs slapping a power brick on them and plugging it in isn't the "right way".

A quote from an "instructable" discussing the different methods to drive LEDs...
From Step 3 of this: http://www.instructables.com/id/Circuits-f...gh-Power-LED_s/

According to step 4 of that instructable "the humble resistor" might actually be an adequate solution (this is of course what SupraGuy suggested as well). But the circuit in step 6 does sound better, and supposidly it can be made for $1. And done to it in step 9 sound good too (for a little extra cost). If I can manage to get it to run a string of 9 LEDs in series (or is the correct term "serie") $15 sounds like money well spent.

I've been wondering: how do you heatsink a resistor? Do I just cover them with a blob of heatsink grease or something?

Lets say I wanted to be on the very safe side, could I just throw 2 W resistors on there (if they exist), is there any reason that would be bad?

2W resistors do indeed exist, as do 5W ones.

The 5W ones are typically wirewound resistors which are in a block of ceramic. Those are heatsinked by sandwiching the ceramic between a plate and the heatsink.

2W ones are typically just physically larger than their small 1/4W cousins, and are often carbon or wirewound resistors, which means that you don't get that fine 1% tolerance of the metal film resistors, but they do have the abilioty to disspiate more power. You can often buy a lot more than you need, and get the tolerance by individually checking resistors and grouping them by actual measured resistance. Generally a batch will all be fairly close, but I've found some where I had trouble with a batch of 50 finding 20 that were within 1% of each other, let alone their rated resistance.

In this application, it's more important that the resistors match each other than that they match the rated value, seeing as how what we're going for is consistency, but given that we've no guarantee that the LEDs are going to match each other, it's probably all moot. Besides, it's already been adequately proven that the light levels can fary signifigantly without the human eye catching the variation. In other words, just get the 5% tolerance resistors with an adequate safety margin, and it'll be fine.

Hirundin,



The thing is that not each led needs exactly 3,7V. (due to the manufacturing variances of the LED). Some leds will work with 3,6V, others with 3,5V etc... I just wanted to say, it will be divided in 4, but the forward voltage of all the leds will not be the same

Btw you can also use the LM317 with a resistor, it's easier and it works very good! I also use it to drive my leds... You only need a resistor and the LM317. And if you want to regulate your current, you can use an extra regulatable resistor.