Condenser lens retrofit for a lighting application

[PhilR]

TL;DR Summary

I'm looking for some advice on a hobby project which aims to retrofit LED emitters into theatrical lighting devices which previously used gas discharge light sources.

Hello!

I should preface this by saying that I work in the film industry where we use lots of different lighting devices, but this is not a commercial project. It's a hobby effort to help a bunch of slightly-used equipment avoid becoming e-waste. Not sure if this is high school, undergrad or advanced, really.

Some of our lighting devices gas discharge light sources (see "MSR," "HMI"). They form beams from these sources, often using a pair of plano-convex condenser lenses near the source and then two more lenses near the output, the final one being adjustable for focus. Between those two lens groups tends to be a selection of devices intended to create effects in the beam, particularly cut metal shapes which will be projected, colour filters, etc.

See image of a typical setup here; reflector is at right, the bulb itself is missing, the two condensers are visible. To the left is a wheel of circular glass colour filters which can be rotated into the beam. See image of the sort of results we are going for here.

There is a general desire to move away from using gas discharge lights for this. LED is lightweight, robust, somewhat more efficient, easier to control, and does not use mercury. The problem is that simply removing the arc light and replacing it with an LED does not work very well. The LED is a much physically larger light source. The whole thing ends up being very inefficient with a lot of light falling outside the originally designed optical path, and it projects a larger image than is really desired.

Is this a problem I can plausibly solve with by retrofitting a different optical arrangement? It ideally needs to be something I can do with parts I can purchase. Unfortunately I'm only vaguely aware of the function and behaviour of condenser lenses and I'm not sure if what I want to do is practical.

Not really sure where to start with this so any advice greatly appreciated. More than happy to do the hard work myself if this is something I could teach myself to do. I've looked at things like this, but it's all going a bit over my head in terms of how I should approach the problem I describe.

 

[PhilR]

Further to earlier (and I hope it's OK to document my thought process here) I've done a bit more reading.

From what I'm finding on places like Wikipedia, there is a fundamental problem with what I'm attempting. The article calls it étendue, from the French, but it's also called "acceptance" and "optical extent" which makes it impossible to design a system based around simple lenses which can take the output of a large-area LED light source and make it behave like a smaller-area gas discharge arc light source.

(The example given is that you can set fire to things with a magnifying glass and the sun, but you can't set fire to things using a magnifying glass and the moon, no matter how big the magnifying glass is.)

Having done a bit more reading, I discover the concept of fibre coupling. This seems to be mostly used to combine the output power of multiple solid state lasers. My layperson's understanding of this is that lots of lasers are fired into the end of a very narrow optical fibre, and by the time the light has made it all the way down the fibre, it's internally reflected off the walls enough times that all of the beam angles are effectively randomised, so that the end of the fibre becomes a point source of light.

As I understand it, it would be possible to fibre couple LEDs in the same way. I don't need my light source to be the size of a tiny hair-fine optical fibre. It can be a few millimetres across. I could conceivably fire a lot of LEDs into one end of, say, an acrylic rod, and so long as the acrylic rod had a few bends in it, the other end would become a small light source in its own right.

This is described as being more efficient than using diffusers to effectively randomise the light.

Do I have any of this anywhere near right?

 

[Dale]

Unfortunately, I don't know enough optics to help with your question. @Drakkith you did some optics, right?

 

[BvU]

Hi,

Liked the links in your post #1.

there is a fundamental problem

Extent is less relevant if your lightspot diameter is a lot bigger than the diameter of the light source. COB LED are stamp-size to 72mm for a 1500 Watt (at $ 1350 ).

Commercial solutions for 'the sort of results you are going for' aplenty, like here, or here. Looks like they can manage with condensor lenses, so renting one and reverse engineering would be my approach

I liked watching an intro (but not the commercial half way): current limiting and proper cooling are important. I've seen similar youtube videos where they went to much higher powers (can't find them easily now )
[edit] turns out to be the same bloke. But the video is only about powering and cooling, so not relevant here

$\$

 

[hutchphd]

there is a fundamental problem with what I'm attempting. The article calls it étendue, from the French, but it's also called "acceptance" and "optical extent" which makes it impossible to design a system based around simple lenses which can take the output of a large-area LED light source and make it behave like a smaller-area gas discharge arc light source

Congratulations. You have in fact correctly ascertained the fundamental limitation in your endeavor ( I have been down this road many times designing LED reflectometers for medical diagnostic equipment where more light give you better signal to noise). This limitation is dictated by the second law of thermodynamics and is very "slippery" to circumvent. Fundamental means just that. Combining multiple sources always creates geometrical impossibilities. Like most attempts to violate the second law, the hunt is often beguiling and tantalizing, but the results seldom satisdy.

 

[Andy Resnick]

Further to earlier (and I hope it's OK to document my thought process here) I've done a bit more reading.

From what I'm finding on places like Wikipedia, there is a fundamental problem with what I'm attempting. The article calls it étendue, from the French, but it's also called "acceptance" and "optical extent" which makes it impossible to design a system based around simple lenses which can take the output of a large-area LED light source and make it behave like a smaller-area gas discharge arc light source.

(The example given is that you can set fire to things with a magnifying glass and the sun, but you can't set fire to things using a magnifying glass and the moon, no matter how big the magnifying glass is.)

Having done a bit more reading, I discover the concept of fibre coupling. This seems to be mostly used to combine the output power of multiple solid state lasers. My layperson's understanding of this is that lots of lasers are fired into the end of a very narrow optical fibre, and by the time the light has made it all the way down the fibre, it's internally reflected off the walls enough times that all of the beam angles are effectively randomised, so that the end of the fibre becomes a point source of light.

As I understand it, it would be possible to fibre couple LEDs in the same way. I don't need my light source to be the size of a tiny hair-fine optical fibre. It can be a few millimetres across. I could conceivably fire a lot of LEDs into one end of, say, an acrylic rod, and so long as the acrylic rod had a few bends in it, the other end would become a small light source in its own right.

This is described as being more efficient than using diffusers to effectively randomise the light.

Do I have any of this anywhere near right?

I think you have done an excellent job educating yourself!

You are on the right track- put more simply, the basic challenge is that the arc and LED are (1) different sizes and (2) emit light into different 'cones' (the arc emits essentially emits into all directions, the LED into a restricted cone angle). There are also differences in the emitted spectra which will impact color rendering.
Replacing an arc with an LED is not trivial- my experience is trying to do this with a microscope- but I think you are on the right track with fiber-coupling. There are some companies that offer high-power LED retrofit modules for microscopes, for example https://www.bw-optik.de/, but I don't think the output power is sufficient for your needs. In fact, that's the basic challenge- LED light output simply can't match a high power arc.

My personal experience trying to do something like this (replacing the stock CSI 250W arc lamp with an LED) led to failure after failure for the reasons listed above (didn't try fiber coupling because the expected power losses made that approach unusable), and in the end I simply switched to a different arc- one that is currently being manufactured- and had an adapter made to allow me to mount the new arc in the correct location. I also needed to add some UV filters b/c the CSI bulb didn't emit much but pretty much every other arc in the world does.

Is your motivation simply to get away from mercury arcs? Could you work with a Xenon or metal-halide arc? High-power arc sources are very resistant to alternate technologies.

 

[PhilR]

Well, that's refreshing to know! Thanks, everyone.

I can't take much credit for figuring it out. Having had these parts in my hands there was a certain instinctiveness to the whole thing. It made sense to me that where there's a large area light source with light coming out of every point on its surface in every possible direction at once, there isn't going to be a single solution to make all of that light part of a parallel beam simultaneously. Not sure if I'm thinking about it in the right way.

I have some LED-powered theatrical beam lights which can project patterns of broadly the type BvU refers to (UK calls this a "profile," the USA calls it a "leko" or "ellipsoidal"). Most of them seem to have a double plano-convex condenser arrangement like an arc-powered light. These project patterns represents on cut metal shapes which are generally 60mm in diameter or more, which is a bit bigger than the LED chip-on-board module that drives them. In the intelligent moving-mirror lights which can point in different directions, the shapes are much smaller, often 25-30mm, which is smaller than the LED source I'd like to drive it with. The whole optical path is just much smaller on these moving lights. That appears to be the rub.

Anyway, my thoughts turn again to the idea of fibre coupling. I only persist in this line of thought only because it is absolutely being used in shipping products to combine the output of a large number of smaller laser diodes where a high power laser is needed. I wonder if this is potentially comparable to what I'm considering, on the basis that my 100-watt LEDs are really a ten-by-ten matrix of one-watt LEDs, or similar. I'm not sure if I'm missing anything here, other than that there will inevitably be some loss in doing this and I'm not sure how feasible it is to guide the output of that (say) 30x30mm LED array into a (say) 10mm light pipe. Am I recapitulating the same problem? As I say, laser cutters do it.

As to cooling and power levels, that's much more my area. I'm happy to design an appropriate power supply (actually, it's more or less off the shelf) and cooling solution. At these high power levels (hundreds of watts) it's mostly a challenge of cooling the thing in a workable amount of space and without making a lot of noise. Liquid cooling has been used. Often computer cooling parts can be used as the power levels are comparable.

LEDs are lighter, a bit more efficient, robust and controllable, and can offer RGB colour mixing. The replacement lamps for the old school lights are becoming rare and very, very expensive. Xenon is rather low efficiency.

PR

 

[Drakkith]

You are on the right track- put more simply, the basic challenge is that the arc and LED are (1) different sizes and (2) emit light into different 'cones' (the arc emits essentially emits into all directions, the LED into a restricted cone angle).

This was my first thought as well. LED's emit light into a cone and not isotropically like other light sources, so the existing optical elements might not work properly. You'd likely have to redesign the optical system to make it work.

I wonder if a simple diffuser would solve the issue. That would turn the light source back into something more like the arc lamp. The spacing would still need to be correct, so it might still be a problem if your new light source is physically larger than your old one.

 

[PhilR]

That is in fact the problem.

The old light source is an arc light so it mostly has a size of roughly the distance between the electrodes, on the order of 6-10mm by my eye.

The new light source is an array of LEDs perhaps 25mm square. some are round. It is already reasonably diffuse. White-emitting versions and certain colours are phosphor converted. The light is not entirely uniform (it hot-spots where the emitters are) but it is essentially a diffuse source. We might also use red, green and blue emitters and seek to combine them, for colour mixing.

I guess my interest at this point is whether we can do something like the following, just with LEDs. This seems to imply that the many parallel beams can be reduced into a fiber using practically-achievable optical parts (it shows green and blue being combined, but each colour is combined from many diodes).

Does it only work on solid state lasers, because of their ultra-small source size? Would this suffer horrible efficiency problems with LEDs? I mean, LEDs don't have enormous sources, individually. Each one-watt chip is perhaps a couple of millimetres square. I appreciate the source of a solid state laser is the size of a bacterium, or something, but am I making any sense?

 

[hutchphd]

As mentioned the chase is beguiling, but, in my experience, a fool's errand. The conservation of etendue is real. Were it not true, one could devise a passive optical method to increase the temperature of the hot side of the Carnot engine thereby violating the second law. This does not mean that LEDs cannot be used for a spotlight but it does mean that converting its output to look like an arclamp ins probably not a viable start. Therefore the "drop in" replacement seems unlikely to me.

 

[hutchphd]

green and blue being combined

A beam splitter cannot be run in reverse to make a "beamcombiner". It will again be a beam splitter."

 

[PhilR]

A beam splitter cannot be run in reverse to make a "beamcombiner". It will again be a beam splitter."

I'm aware, I just couldn't find a diagram with a single set of emitters on it! I've seen them in reality.

And yes, I'm getting the idea. I think what I'm after could probably be done with solid state lasers but it'd be impossibly expensive, and I'm not sure how efficient it would really be.

One manufacturer does make a highly collimated beam light with a laser light source, possibly because of exactly the issues we've been pondering. Inconveniently, it does start to engage laser safety regulations in some jurisdictions.

 

[Drakkith]

In the context of simple options that would work within the existing components, you could try using a stop somewhere in the optical path to block some of the unwanted light from spilling out.

 

[PhilR]

That's pretty much what we get now, with the naive substitution approach. There's enough internal structure on these things to block out the unwanted light. Problem is, the results are inefficient, dim, and generally provoke sadness and regret.

I think we've established why people generally aren't turning out retrofit kits.

Still, this thread, in conjunction with a chat with an AI, has given me a list of things that might at least be worth an afternoon playing around on the bench.

 

[berkeman]

this thread, in conjunction with a chat with an AI, has given me

Just a heads-up that we don't allow use of AI chatbots as technical references at PF. You are of course free to use them on your own, but please be aware that they are wrong a significant fraction of the time.

 

[PhilR]

Oh, I'm sorely aware. They're bad enough in areas I do know about that I'm not going to trust them in areas I don't. Still, it's a start, at least.

 

[Gleb1964]

..the basic challenge is that the arc and LED are (1) different sizes and (2) emit light into different 'cones'...

The main challenge is that LED has much less brightness* compare to arc.
(*I may be should use term luminance or luminous flux per area instead of term brightness, because the term brightness is often somewhat inaccurately used.)

 

[hutchphd]

may be should use term luminance

Perhaps you should use the appropriate term "etendue" as twice previously mentioned. Caution: this may require new knowledge.

 

[Gleb1964]

Perhaps.. the units of etendue is area*solid angle and it conserved in an optical tube, if the optical tube done properly. Like illumination system we are discussing.
Withing given etendue we need to maintain the light flux, which would drop proportional to the brightness of LED source relative to arc lamp.

 

[hutchphd]

Withing given etendue we need to mountain the light flux

The LED itself is already an optical system with a given etendue (emitter, reflector on chip perhaps lensing from the epoxy cap. There is no way tomake that number higher with lenses or light pipes. It is equivalent to adding more bicycle wheels to your perpetual motion machine. The details do not matter, however beguiling they may be.
Of course the optic can (and usually does) lower that number.

 

[Gleb1964]

Right, keeping the same light flux is not possible at given etendue of illumination system, if the LED brightness is less than the arc source. Normally, less LED brightness would be compensated by increasing the emitting area by placing more LEDs in array, but that would not be accepted by the limited etendue of projection system, which is designed to use an arc source.
One would need agree on a less light flux as a trade off.

 

[sophiecentaur]

Normally, less LED brightness would be compensated by increasing the emitting area by placing more LEDs in array,

I remember that old filament projector bulbs had a large rectangular array of filament coils the 'illuminant path' (condenser / mirror etc.)was arranged to get all the light through the projector lens (afair, an image of the rectangular matrix was focused on the projector lens to get the most out in the direction of the screen).The nearly point source from the new halogen bulbs required different optics so you couldn't just up-grade / swap those bulb types without compromising the illumination of the slide. Perhaps the better system for a large LED array would be more like the system for the old filament bulbs. Definitely not a plug in replacement for an arc.

Having looked inside several old fashioned projectors, as a lad, I was always struck by the poor quality of the condenser system - full of bubbles and streaks- but image quality doesn't matter in that application.

 

[Gleb1964]

..See image of a typical setup here;

There is a general desire to move away from using gas discharge lights for this. LED is lightweight, robust, somewhat more efficient, easier to control, and does not use mercury. The problem is that simply removing the arc light and replacing it with an LED does not work very well. The LED is a much physically larger light source. The whole thing ends up being very inefficient with a lot of light falling outside the originally designed optical path, and it projects a larger image than is really desired.
..

After looking closer at your problem it may be not look as desperate. The arc lamps are in an order or more brighter than the LEDs (brightness in the meaning of the flux density). But I would suggest that there may be a solution for your problem.
Here is the picture of your projection system, I tried to guess the scale, may be it is not correct:

There are LEDs lamp intended for halogen lamp replacement for the cars headlight, they are suit the best for replacement of your arc lamp. I would suggest that you may try something like this LED headlight bulbs https://www.auxito.com/products/9005-led-headlight-bulb-80w-16000-lumen:

LEDs are mounted on both sides and can provide about 8000 lm flux from every side, 16000 lm total, but it can be the problem that light from the back side would be blocked by the lamp itself.

Or, may be this is better, with double LED modules mounted very close on the one side, providing 16000 lm:

Using those type of LED modules hopefully the condenser system would fit the image of the LED to the exit pupil of the projection lens, utilising not all, but significant part of the LED output.

 

[sophiecentaur]

Just a heads-up that we don't allow use of AI chatbots as technical references at PF. You are of course free to use them on your own, but please be aware that they are wrong a significant fraction of the time.

Not half!! AI can be no help whatsoever for some things. Try selling something on eBay and the AI description of your aging telescope or electric mixer is positively embarrassing.

 

[Gleb1964]

Having looked inside several old fashioned projectors, as a lad, I was always struck by the poor quality of the condenser system - full of bubbles and streaks- but image quality doesn't matter in that application.

Until defects are not coming close to the projected image plane and affect its uniformity the spec on defects is pretty free. There are condensers lens with integrated ground surface to improve the uniformity. Of course, improving uniformity by using diffuse surfaces would sacrifice the source luminance , averaging it over larger area and solid angle.