Led Backlight
Maybe a bit to the left of topic, but why not use a whole bunch of ccfl`s, say 10, with a fresnel lens at rear to spread the light source. Surely that would give an adequate ammount of light or am I missing something. Yes I know ccfl`s are about £10 ($15/16) but it would be low power and cool, maybe for smaller screen projectors?
Maybe a bit to the left of topic, but why not use a whole bunch of ccfl`s, say 10, with a fresnel lens at rear to spread the light source. Surely that would give an adequate ammount of light or am I missing something. Yes I know ccfl`s are about £10 ($15/16) but it would be low power and cool, maybe for smaller screen projectors?
Do you mean like those wrapped florescent tube lamps? I've tried that before, and the brightness is not that great, but mostly the color is terrible. It's hard to get much brightness because those lamps don't really have very high output to begin with (even 10 of them), and the light goes in every direction, so even with a fresnel you won't get much. That idea might work if you use a very small panel with one of those Fuji lenses that goes nearly in contact with the LCD, but you still have the color problem, and I still don't think you could get it as bright as even the most inefficient of metal halide light sources.
That said, the whole idea of this article is that you never know for sure until you try, so hopefully someone can prove me wrong! 🙂
That said, the whole idea of this article is that you never know for sure until you try, so hopefully someone can prove me wrong! 🙂
I am very interested in the possibility of using LED's, and have been for some time. I've crunched the numbers, and have tried some small scale experiments, but as of yet haven't managed to come up with anything worthwhile.
There is one thing that I wanted to mention that doesn't appear to have been brought up in this thread yet. Conventional white LED's will not work well, as they do not emit a true white light across the color spectrum. Instead, it is a combination of peaks and valleys that to the human eye looks white, but in reality is not. The problem with this is the color filters in the LCD's. If there is a valley in the color spectrum of the LED that is near the filter used in the LCD, it doesn't work too well. I believe there are some newer white LED's that do a better job of this, and may actually produce white light, but the cost would undoubtedly be more than for standard white LED's. The other option is combining red, green, and blue LED's, matched to the filters used in the LCD. This is more work (finding spec sheets for each color of LED and the LCD filters, computing how many of each color at what spacing at what power tp get, etc), but I believe this is more effective than the "white" LED's.
There is one thing that I wanted to mention that doesn't appear to have been brought up in this thread yet. Conventional white LED's will not work well, as they do not emit a true white light across the color spectrum. Instead, it is a combination of peaks and valleys that to the human eye looks white, but in reality is not. The problem with this is the color filters in the LCD's. If there is a valley in the color spectrum of the LED that is near the filter used in the LCD, it doesn't work too well. I believe there are some newer white LED's that do a better job of this, and may actually produce white light, but the cost would undoubtedly be more than for standard white LED's. The other option is combining red, green, and blue LED's, matched to the filters used in the LCD. This is more work (finding spec sheets for each color of LED and the LCD filters, computing how many of each color at what spacing at what power tp get, etc), but I believe this is more effective than the "white" LED's.
I agree that color is a big issue here, too. I actually think that using 3 seaparate LED's wouldn't improve anything because you'd still have peaks and valleys ... that leads to a very low CRI. We'll see how it works here - I keep talking about it, but haven't actually tested it yet.
About the color issue, if it turns out to be too blue, couldn't you put like red and green transparent pieces of glass in front of the light? Thus evening out the blue to a white light.
As a matter of technicality, yes you could make a filter for it, but the problem isn't necessarily the balance of the various colors. The problem, in my opinion, is white, in this case, is made up of green, red/orage, and blue peaks. There are a lot of colors missing. So even though our eyes see white, there are a lot of colors that the LCD can display that will effectively filter out a lot of the light coming in, since each pixel is a filter that lets only a very narrow range of colors through. Do you get what I'm saying? It's difficult for me to explain.
oh, I understand now. You might be better off combining christmas lights until you get a bright enough light to project an image.
ha, I'm kidding
Do they make like very large LEDs? If there are, you could get 1 green, 1 red and 1 blue large led and it would probably project enough light. If they don't have large LEDs, then why not?
ha, I'm kidding

Do they make like very large LEDs? If there are, you could get 1 green, 1 red and 1 blue large led and it would probably project enough light. If they don't have large LEDs, then why not?
They make large LED's, but they are still the same amount of light, just with a larger plastic mold around it. The way LED's work is to pass a current through some ionic material - you can actually see this stuff in a very transparent LED by turning it on and shining it into your eyes at a very close distance (with sunglasses on to avoid damage). I assume that the amount of "stuff" has to be very small to avoid it cracking under the heat and vibration.
It's ironic that you mention Christmas lights though - they are starting to make those from LED's as well since they don't burn out hardly ever and use VERY LITTLE energy compared to mini incandescent lamps. We should pool together some money and all invest in LED's! ... THEY'RE EVERYWHERE!!!
It's ironic that you mention Christmas lights though - they are starting to make those from LED's as well since they don't burn out hardly ever and use VERY LITTLE energy compared to mini incandescent lamps. We should pool together some money and all invest in LED's! ... THEY'RE EVERYWHERE!!!
I found a good place to buy LEDs for cheap!!! Check it out:
blue LEDs $2.99 for 50 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=66948&item=3828050318&tc=photo
green LEDs $10.00 for 10 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=26207&item=2586096304&tc=photo
Red LEDs $10.99 for 50 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=26207&item=3826957975&tc=photo
They are all different prices because they are all different brightnesses. Which could potentially cause a problem.
but, using those LEDs do you think it would be possible to make your own large LEDs, by combing a whole bunch of small ones, into a metal cylander. Then putting them in the right order to make white light. It's kind of like those really old projectors that have the three different colors of light that project into a picture.
This is what I mean:
blue LEDs $2.99 for 50 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=66948&item=3828050318&tc=photo
green LEDs $10.00 for 10 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=26207&item=2586096304&tc=photo
Red LEDs $10.99 for 50 pcs: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=26207&item=3826957975&tc=photo
They are all different prices because they are all different brightnesses. Which could potentially cause a problem.
but, using those LEDs do you think it would be possible to make your own large LEDs, by combing a whole bunch of small ones, into a metal cylander. Then putting them in the right order to make white light. It's kind of like those really old projectors that have the three different colors of light that project into a picture.
This is what I mean:
Attachments
I've been doing work with LEDs for quite a while, in quite a few applications. I'm coming up with projects to do with them all the time, and my recent idea was a projector. Little did I know that people have been trying to before me. However, I believe I can bring something new to the discussion.
There are a few things I have gathered from the posts on the board (I've been lurking for a couple weeks) .
1) LEDs won't work.
2) A generally good output of a bulb is around 4000 lumen (please correct me if I am wrong here)
Well, now to me, those 2 statememnts are contradictory.
First, let's start with an understanding of LED light measurement. LEDs are measured in candela (or rather, millicandela, mcd) and the bulbs people are using here are measured in lumen. These are in fact similar but fundamentally different measurements. Lumen is the flux of light, while candela is the brightness. So what we need is to get a conversion rate for candela to lumen before we can start doing the math. However, here is the tricky part. 1 candela = 1 lumen/steradian.
So, to get the lumen outout of an LED, we need to know not only the candela output but it's solid angle of output. This can be estimated by the following formula:
[pi] X (theta)^2
Where theta is the half viewing angle of the LED, in radians. Then you take that number and multiply it by your output in candela (careful, not mcd)
So let us take the blue LEDs from the previous post and calculate their lumen output. First, find the solid angle:
pi X (20 X pi /180)^2 = approximately 0.383
Now we multiply the solid angle by the candela(4.000), and get 1.5 lumen/LED.
Meaning we'd need some 2666 of the LEDs to produce 4000 lumen.
So, as you can see, it is possible to do it with LEDs, but it will take ALOT more than the 50 or so people have been trying it with.
Now, let's take at some white LEDs.
From that same seller, there are 11000mcd white LEDs with a viewing angle of 20 degrees again. So this time the output is almost 3 time as great, putting out about 4.3 lumen/LED. So we'd need a litttle under 1000 of those LEDs to produce 4000 lumen.
Now, onto another interesting fact. Bulbs are rated by lumen, but that lumen is for the entire light outputted by the bulb. As I'm sure most of you know, not all of that light actually ends up going through the panel. What I'd like to know is if someone has an estimate on how much of the light from a bulb actually ends up getting to the panel.
Finally, I'll finish up for tonight with some more stuff about LEDs. Pulsing. The peak current of those white LEDs (and most TI 3/4 LEDs) is approximately 20mA. Peak current is basically determined by how quickly the LED can dissipate heat. At 20mA, the LED can dissipate the heat created at the same rate it's being created, and so will not burn out. The theory behind pulsing is that you when you turn the LED off, it can dissipate more heat since it's not creating any new heat. Meaning you can increase the current a fair bit so long as you keep allowing it time to cool down. Normally this doesn't create any perceived increase in brightness, as the eyes register average brightness(for instance, twice the light running half the time ends up averaging to the same light all the time). However, I am unsure as to whether pulsing would be useful in this case. If, for instance, the LCD panel is running at 75 Hz, you could also pulse the LED at 75Hz, thus allowing you to increase the current, and in turn the brightness, but you since the screen is only displaying when the light is on anyway, it may actually be helpful.
Well, if you read all that, hopefully you have been enlightened (pun only partially intended) on LEDs, and now we can work together to see if we can mold this into a viable solution!
There are a few things I have gathered from the posts on the board (I've been lurking for a couple weeks) .
1) LEDs won't work.
2) A generally good output of a bulb is around 4000 lumen (please correct me if I am wrong here)
Well, now to me, those 2 statememnts are contradictory.
First, let's start with an understanding of LED light measurement. LEDs are measured in candela (or rather, millicandela, mcd) and the bulbs people are using here are measured in lumen. These are in fact similar but fundamentally different measurements. Lumen is the flux of light, while candela is the brightness. So what we need is to get a conversion rate for candela to lumen before we can start doing the math. However, here is the tricky part. 1 candela = 1 lumen/steradian.
So, to get the lumen outout of an LED, we need to know not only the candela output but it's solid angle of output. This can be estimated by the following formula:
[pi] X (theta)^2
Where theta is the half viewing angle of the LED, in radians. Then you take that number and multiply it by your output in candela (careful, not mcd)
So let us take the blue LEDs from the previous post and calculate their lumen output. First, find the solid angle:
pi X (20 X pi /180)^2 = approximately 0.383
Now we multiply the solid angle by the candela(4.000), and get 1.5 lumen/LED.
Meaning we'd need some 2666 of the LEDs to produce 4000 lumen.
So, as you can see, it is possible to do it with LEDs, but it will take ALOT more than the 50 or so people have been trying it with.
Now, let's take at some white LEDs.
From that same seller, there are 11000mcd white LEDs with a viewing angle of 20 degrees again. So this time the output is almost 3 time as great, putting out about 4.3 lumen/LED. So we'd need a litttle under 1000 of those LEDs to produce 4000 lumen.
Now, onto another interesting fact. Bulbs are rated by lumen, but that lumen is for the entire light outputted by the bulb. As I'm sure most of you know, not all of that light actually ends up going through the panel. What I'd like to know is if someone has an estimate on how much of the light from a bulb actually ends up getting to the panel.
Finally, I'll finish up for tonight with some more stuff about LEDs. Pulsing. The peak current of those white LEDs (and most TI 3/4 LEDs) is approximately 20mA. Peak current is basically determined by how quickly the LED can dissipate heat. At 20mA, the LED can dissipate the heat created at the same rate it's being created, and so will not burn out. The theory behind pulsing is that you when you turn the LED off, it can dissipate more heat since it's not creating any new heat. Meaning you can increase the current a fair bit so long as you keep allowing it time to cool down. Normally this doesn't create any perceived increase in brightness, as the eyes register average brightness(for instance, twice the light running half the time ends up averaging to the same light all the time). However, I am unsure as to whether pulsing would be useful in this case. If, for instance, the LCD panel is running at 75 Hz, you could also pulse the LED at 75Hz, thus allowing you to increase the current, and in turn the brightness, but you since the screen is only displaying when the light is on anyway, it may actually be helpful.
Well, if you read all that, hopefully you have been enlightened (pun only partially intended) on LEDs, and now we can work together to see if we can mold this into a viable solution!
MrMike said:As I'm sure most of you know, not all of that light actually ends up going through the panel. What I'd like to know is if someone has an estimate on how much of the light from a bulb actually ends up getting to the panel.
Well, I got pretty good view about that. My Sony VPL-SC50 is a 500 ANSI lumen projector with its standard 120W MH bulb (wíth kind of a perfect reflector and so on). That kind of bulb generates something like 8000-9000 lumens, and with the about perfect reflector, we get the 500 ANSI lumens out of the thing. Well... now I've retrofitted it and it has 150w MH (11000 lumens) with a condenser lens, and I first had a soup ladle as a reflector but I took it away because taking it away made no visible difference on the screen, but it did warm up the lamp and the projector and so on.... Well, with this setup I get roughly 200-300 ANSI lumens (not measured with any tools, only my own estimation) out of the projector. So... with perfect reflector/optics setup (like the case might be with leds used properly) we get 5,6-6,25% of the light out of the projector, and with less then perfect diy setup we get something like 2,3%... that is about 38% difference in effiency.
Maybe someone might be able to use this information someway, I dunno, I just answered the question :B
Regards
HB
Sounds like a reasonable estimation... (darn inefficient LCD panels ...)
One of these days I might have to break down and buy a digital light meter so that we can start making lumens calculations instead of lumens approximations.
One of these days I might have to break down and buy a digital light meter so that we can start making lumens calculations instead of lumens approximations.
Um... I've been thinking of making my own... it shouldn't be very hard... just a small solar cell or LDR resistor and something like voltage meter, then just go and measure voltages with few commercial projectors so you know how much is it with 1000 ansi lumens and with 500 ansi lumens just to see how the ldr/solar cell reacts, or something like that and then just calculate the scale...diylabs said:One of these days I might have to break down and buy a digital light meter so that we can start making lumens calculations instead of lumens approximations.
Maybe some electronics dude here would like to design a simple one and then we'd all build our own according to that? Maybe based on a 1k-5kohm LDR, they cost 0,5 euros here in finland 🙂
Actually.... it could be that it doesnt have to be anything but a 1k-5k LDR and a voltage meter, we just should agree on how to do the measurement... maybe 1 or 2 meters from the projector? And the LDR in such position that it gives highest value?
Then we could all post results here and all that have the enthusiasm or friends with commercial projectors or a commercial projector at work could measure those too?
Regards
HB
Does the light from an LED have a singular polarity?
If not, 50% of the light is unusable (the LCD panel is singularly polarized). Has anyone tried light recycling (polarizing beam splitter--- half-lambda plate--- second reflector--- collimator) with LEDs?
This might explain better.
As to not being able to integrate two sources of light, as previously mentioned, it was my impression that an x-prism integrates three. My understanding isn't complete on this, so please let me know if this is not the case.
To get the most out of any lamp, check out this system (reflectors are key).
And for the LED-based projector detractors out there, the technology is getting there (albeit small for now)...
Link 1
Link 2
Link 3
If not, 50% of the light is unusable (the LCD panel is singularly polarized). Has anyone tried light recycling (polarizing beam splitter--- half-lambda plate--- second reflector--- collimator) with LEDs?
This might explain better.
As to not being able to integrate two sources of light, as previously mentioned, it was my impression that an x-prism integrates three. My understanding isn't complete on this, so please let me know if this is not the case.
To get the most out of any lamp, check out this system (reflectors are key).
And for the LED-based projector detractors out there, the technology is getting there (albeit small for now)...
Link 1
Link 2
Link 3
The dual paraboloid demonstration is particularly interesting. Using a glass pipe or fiberoptic cluster seems to be a very neat way to channel the light where it needs to go, but in that case it seems useful only if you're using a very small LCD since you'd need to have a very short FL lens that is nearly as wide as the LCD itself. Still very cool.
I don't see any mention of Luxeons here. I saw a link to Lumileds, but the Luxeon IIIs and the 5Ws are really very bright, and the LuxIIIs are also favored by people who want a point source - the 5W are four 1W dies in one package.
Anyway, there's a ton of information at www.candlepowerforums.com - Luxeon IIIs can go up to 200 lumens iirc.
Anyway, there's a ton of information at www.candlepowerforums.com - Luxeon IIIs can go up to 200 lumens iirc.
I have to admit, I'm partial to a smaller transmissive panel. I think it's the way to an LED solution. But I'm also operating under the possibly false assumption that reduced surface requires less illumination.
Folks dismiss the smaller panels because of the resolution, price, or because they're monocular. But I think the new 5mm VGAs--- like those that will be incorporated into cell phones next winter--- hold promise.
I don't mean to divert this into a small panel discussion. However, has anyone tried using a transflective LCOS panel? My understanding is that because the panel is reflective (in addition to being transmissive) it allows more light to transverse. I don't know the workings of it... maybe the backlight isn't removable? or the contrast suffers?
Folks dismiss the smaller panels because of the resolution, price, or because they're monocular. But I think the new 5mm VGAs--- like those that will be incorporated into cell phones next winter--- hold promise.
I don't mean to divert this into a small panel discussion. However, has anyone tried using a transflective LCOS panel? My understanding is that because the panel is reflective (in addition to being transmissive) it allows more light to transverse. I don't know the workings of it... maybe the backlight isn't removable? or the contrast suffers?
Figured I should add my 2c...
First off glad the LED's are ressurected, I still want to see it work!
I did a load of calculations a while back and documented them HERE - my conclusion was that you'd need about 1000 (10,000mcd) LED's prior to any pulsing techniques (this was to generate roughly 1,000 ANSI based on the magical 10% efficiency figure that floated around here in the early days).
On to pulsing... Most white LED specs have a peak current that is only marginally over continuous current rating - that is they can't dissipate the heat quick enough to put a larger amount of current through at a lower duty cycle. Probably the only worthwhile duty cycle to look at is 50% (spending half the time on at double the current - then half the time turned off). Anything over this and you'll be cutting into the life of the white LED's quite considerably, although with individual colours you can probably take this further.
Assuming 50% DC - This gives an approximate brightness of 1.5x, the persistence of the human eye actually allows us to percieve the brightness pretty much in the middle of average and peak brightness. From what I've read it's slightly over the middle value which I'm going to take to cancel out the LED's lower efficiency rating when pulsed (makes the maths simpler that way!)
Therefore at this duty cycle with this type of LED (10,000 mcd) you'll need approx. 700 LED's. Of course you could buy the new 20,000 mcd versions but they're still very pricey, last time I checked 12,000 mcd were best on a cost to mcd ratio - but that's still 560 ish that you need.
I don't mean to be negative with this paragraph, but personally I can't afford to buy that many LED's to experiment with, let alone be bothered to solder them all to a bread board with the hope that they all line up. This task is made even harder if you're trying to pulse them as you need the circuitry to switch them at a predefined current. As far as experimenting goes you're probably best avoiding the pulsing option but that = more cost and more soldering.
Another thing to consider is that for best light uniformity you want to pack the LED's as close to each other as possible - how many can you fit behind your screen? Even if you place them side by side (a hexagon pattern would be best but this is a simpler guide) filling a 15" screen is going to require 60 x 45 = 2700 LED's behind it...
Rest assured I'll be popping in on a semi-regular basis to check up on how you're getting on!
Best of luck, live the dream etc.
Steve
First off glad the LED's are ressurected, I still want to see it work!
I did a load of calculations a while back and documented them HERE - my conclusion was that you'd need about 1000 (10,000mcd) LED's prior to any pulsing techniques (this was to generate roughly 1,000 ANSI based on the magical 10% efficiency figure that floated around here in the early days).
On to pulsing... Most white LED specs have a peak current that is only marginally over continuous current rating - that is they can't dissipate the heat quick enough to put a larger amount of current through at a lower duty cycle. Probably the only worthwhile duty cycle to look at is 50% (spending half the time on at double the current - then half the time turned off). Anything over this and you'll be cutting into the life of the white LED's quite considerably, although with individual colours you can probably take this further.
Assuming 50% DC - This gives an approximate brightness of 1.5x, the persistence of the human eye actually allows us to percieve the brightness pretty much in the middle of average and peak brightness. From what I've read it's slightly over the middle value which I'm going to take to cancel out the LED's lower efficiency rating when pulsed (makes the maths simpler that way!)
Therefore at this duty cycle with this type of LED (10,000 mcd) you'll need approx. 700 LED's. Of course you could buy the new 20,000 mcd versions but they're still very pricey, last time I checked 12,000 mcd were best on a cost to mcd ratio - but that's still 560 ish that you need.
I don't mean to be negative with this paragraph, but personally I can't afford to buy that many LED's to experiment with, let alone be bothered to solder them all to a bread board with the hope that they all line up. This task is made even harder if you're trying to pulse them as you need the circuitry to switch them at a predefined current. As far as experimenting goes you're probably best avoiding the pulsing option but that = more cost and more soldering.
Another thing to consider is that for best light uniformity you want to pack the LED's as close to each other as possible - how many can you fit behind your screen? Even if you place them side by side (a hexagon pattern would be best but this is a simpler guide) filling a 15" screen is going to require 60 x 45 = 2700 LED's behind it...
Rest assured I'll be popping in on a semi-regular basis to check up on how you're getting on!
Best of luck, live the dream etc.
Steve
I finished my first tests using the LED's. First, I tried just replacing the lamp in my old OHP with a cluster of LED's. The brightness (with no LCD panel installed) was actually impressive, but the area illuminated was very narrow. Then I tried putting the LCD in. I couldn't make out an image very well. It was pretty dim, but I could see where the image was on the wall. Next, I put the LED's immediately behind the fresnel lens. That seemed to work better (brighter), but now there was a double image. Next I tried it without a fresnel... the fresnel needs to be in there apparently because I couldn't make out anything at all without it.
My conclusions from the first test are that the LED's I have aren't quite bright enough, but since they are 16,000 mcd and new ones can go up to 20,000 or even 25,000 I think, there is some potential here to be bright enough. Secondly, the LED's have to be pretty well aligned in order for this to work - having a few stray LED's will distort your image (that's why I think I had a double image).
My conclusions from the first test are that the LED's I have aren't quite bright enough, but since they are 16,000 mcd and new ones can go up to 20,000 or even 25,000 I think, there is some potential here to be bright enough. Secondly, the LED's have to be pretty well aligned in order for this to work - having a few stray LED's will distort your image (that's why I think I had a double image).
- Status
- Not open for further replies.
- Home
- General Interest
- Everything Else
- The Moving Image
- Lighting and OHP
- Collaborative LED Light Source Experiment