Overlaying light beams by one replacing the other where they intersect

[wosoka]

Need to overlay beams by having one replacing the other where they intersect rather than have the beams be superimposed.

this is what I want: https://i.imgur.com/F6HIYBl.png

this is what I do not want: https://i.imgur.com/ylhMeEa.png

This kind of beam combining can be either achieved by the "combiner" OR later in the setup by the "projection screen". Or in other words, having the original beam not reaching the projection surface is not the only solution as somehow not displaying the original beam on the projection surface even if it has reached it will also work.

The colors of the beams are for illustration purposes, so are their positions. Position of the smaller beam can change.

Is there any way at all to achieve this. Brightness not an issue so things such as polarizers can be used.

Any ideas?

 

[sophiecentaur]

That is very hard and impossible in a linear system. You need to ‘switch’ the large image with the light level in the small image. I would think a TV projector would be the best solution if it could be made to fit the overall idea.

 

[wosoka]

You need to ‘switch’ the large image with the light level in the small image.

Not sure I undertsand this

 

[sophiecentaur]

Not sure I undertsand this

It's how they inlay / overlay images onto a background scene. They use a 'green screen' behind an actor and the electronic switch turns on the background when the control image is green and replace it with the actor's image when the background is not green. Green screen or black screen (or even Blue, in the old days of TV) is an off the shelf effect which you could use if your light source could be a TV projector.
There are very few substances that can be switched between transmissive and opaque by having a light shone on the other side (which is what your first diagram is effectively showing.
If at all possible I think you should go for the TV based option.

 

[wosoka]

I don't know what a TV projector means but if it's a video projector then I can use that.

I've used green screens before but they were just a green cloth to be captured by a video camera digitally be replaced by something else in a program in the video file, not on the actual physical greenscreen. The kind of greenscreen you are talking about is new to me and interesting.

 

[sophiecentaur]

I don't know what a TV projector means but if it's a video projector then I can use that.

I've used green screens before but they were just a green cloth to be captured by a video camera digitally be replaced by something else in a program in the video file, not on the actual physical greenscreen. The kind of greenscreen you are talking about is new to me and interesting.

as far as I know, a "green screen" means the same to all of us and yes, I meant a video projector. So you know what I'm talking about and about superimposing one image on another. Would it be a solution for you? If you have electronically generated images then, of course, a physical green screen wouldn't be necessary.

 

[wosoka]

as far as I know, a "green screen" means the same to all of us and yes, I meant a video projector. So you know what I'm talking about and about superimposing one image on another. Would it be a solution for you? If you have electronically generated images then, of course, a physical green screen wouldn't be necessary.

I completely lost you here, you are talking about a video projector and green screen in the same paragraph.
green screen I know about requires software to generate a video and is useful for creating videos, not projecting them

 

[sophiecentaur]

I completely lost you here, you are talking about a video projector and green screen in the same paragraph.
green screen I know about requires software to generate a video and is useful for creating videos, not projecting them

I used the term “green screen” as a generic term for switching one image with another. From you replies, I gather that you have used a GS system . I therefore assumed that you would understand how the GS idea could apply to your requirement. If you don’t see the relevance then I would conclude it can’t be done using a TV technique and that’s the only way I can think of. Let’s see what anyone else has to say about the problem and if my idea makes any sense to them.

 

[Tom.G]

If the beams are always the same size and relative location, make your diagonal plain glass with a mirrored area where Beam 2 hits it.
To avoid fuzziness at the beam edges, you will need to add some optics to focus both beams on the diagonal and then more optics to do the final projection.

If Beam 2 has variable size or position that you can control or sense:

1. ) Use a Light Valve for the diagonal with the front surface silvered for 50% transmittance. A Light Valve is essentially an LCD display without a backlight. Project Beam 1 thru it but turn "off" the pixels where Beam 2 is to have priority. Drawback is you lose 50% of the light of each beam. Again, you will need to add some optics to focus both beams on the diagonal, etc.
   
2. ) If you have electronic control of Beam 1, instead of a Light Valve use a 50% mirror and just blank the region of Beam 1 where Beam 2 is to be displayed.

 

[wosoka]

If the beams are always the same size and relative location, make your diagonal plain glass with a mirrored area where Beam 2 hits it.
To avoid fuzziness at the beam edges, you will need to add some optics to focus both beams on the diagonal and then more optics to do the final projection.

If Beam 2 has variable size or position that you can control or sense:

1. ) Use a Light Valve for the diagonal with the front surface silvered for 50% transmittance. A Light Valve is essentially an LCD display without a backlight. Project Beam 1 thru it but turn "off" the pixels where Beam 2 is to have priority. Drawback is you lose 50% of the light of each beam. Again, you will need to add some optics to focus both beams on the diagonal, etc.
   
2. ) If you have electronic control of Beam 1, instead of a Light Valve use a 50% mirror and just blank the region of Beam 1 where Beam 2 is to be displayed.

See the issue is it's not easy to blank regions perfectly in software and make it all seamless. Limited resolutions of LCDs make it harder.

 

[wosoka]

How about a material that will replace a beamsplitter or partially silvered mirror, and can become opaque if is illuminated from one side?
A material that changes its transparency/opacity like this would be perfect

 

[Tom.G]

See the issue is it's not easy to blank regions perfectly in software and make it all seamless. Limited resolutions of LCDs make it harder.

Yeah. With the real world imperfect optics and the resolution of LCDs you will never reach "perfect." Even harder using only light waves, they don't interact with each other.

How about a material that will replace a beamsplitter or partially silvered mirror, and can become opaque if is illuminated from one side?

There are some materials that darken when exposed to UV light but the only ones I'm aware of take minutes to transition. Eyeglasses are available with that feature. You might try asking in the 'Chemistry' or the 'Materials and Chemical Engineering' forums to see what they can suggest.

A Google search https://www.google.com/search?&q=photoactive+light+valves
has over 20 000 hits, there may be something useful there.

https://www.google.com/search?&q=photochromic
has over 7 000 000 hits.

Just to satisfy my (and others) curiosity, I would like to know what you settle on though. Please post back here if you can.
Good Luck!
Tom

 

[wosoka]

Yeah, photochromic films are slow.

A Google search https://www.google.com/search?&q=photoactive+light+valves
has over 20 000 hits, there may be something useful there.

https://www.google.com/search?&q=photochromic
has over 7 000 000 hits.

Okay, give me sometime to go through the 7020000 links.

Okay, if being serious, I went through pages trying to find photochromic films that can by themselves or in certain chemical or physical conditions transition fast enough (8 ms is fast enough for most uses). Same with photoactive filters or light walves. Most results are just research articles, pretty much nothing when trying to find a ready product.
That said I did search like the first 30 pages of links and maybe didn't search the correct terms or phrases. I think 30 google pages is a good bet.

Just to satisfy my (and others) curiosity, I would like to know what you settle on though. Please post back here if you can.

sure

 

[Tom.G]

For light valves, I haven't looked, but the 23 inch LCD display on my wife's computer claims 5ms response time. Her tablet computer has a screen resolution of 225dpi -- and these are consumer products! Look at high end video projectors, the professional ones not consumer stuff, to see what's available. One possible place to start searching is: http://www.projectorcentral.com/projectors.cfm

Another possibility is a DLP (Digital Light Processor).
Try a google search: https://www.google.com/search?&q=digital+light+site:ti.com

 

[sophiecentaur]

How about a material that will replace a beamsplitter or partially silvered mirror, and can become opaque if is illuminated from one side?
A material that changes its transparency/opacity like this would be perfect

The edge effects would probably be a problem here. Even with high / perfect resolution, if the reflector is not in an image plane, there can be anything from blurring to complete crosstalk between the two images. That wouldn't be impossible to solve if your optical chain had an intermediate image plane. I have a feeling that may affect brightness though. Would that be a problem?
Also it would be useful to know that the 'inserted' image would be. Would it be a shape with uniform colour / luminance or would it be a normal scene? If you wanted to do the above with some photochromic material, the resulting transmission level would be dependent on that image.
But you don't seem to want to help with details of what you actually want to do and that limits what you can get out of this thread. You 'leaked' the term "foveal" in one post but that still doesn't help a lot.

 

[wosoka]

Video projection beam is a beam of pixels that can be red, green or blue with varying intensities each. If that means "normal scene" then yes. HOWEVER, you can also easily overlay an IR or UV beam in the projector optical path to go with the visible light projection beam, and we can use that beam instead of the visible light beam.

This requires surprisingly little modification to the original optical engine.

And no brightness loss is not a concern (around 75% light loss is okay).

Sorry if I appear as cryptic, I just don't have anything simple to show, there's too much technical lingo and 3d mapping techniques and tricks involved so its hard to explain in simpler terms.

 

[sophiecentaur]

If that means "normal scene" then yes.

If you are using the superimposed scene to define the switching points on the background then what information is there to determine the switching points? Or is the inset piece always the same size and shape? It is very difficult for me because I have to ask so many questions which may be wasted if you don't describe the actual problem properly. It may all be very obvious to you but I am not a mind reader.

 

[wosoka]

I know but I fear each new explanation will keep raising a new question of its own. Maybe wait till I prepare a proper document.

Or is the inset piece always the same size and shape

Yeah, it is.
The issue is since the background is much lower resolution, you will get jagged seams where the background and superimposed scene are to be blended.
Here's an illustration:

1) Take a lower resolution projection and try to hide the pixels digitally where the higher ones goes

https://i.imgur.com/9sVStQv.png

Now project the higher res beam in this hole. You will get a seam in form of holes.

https://i.imgur.com/zMI9yAm.png

What if you try making the middle beam here a bit bigger to cover those holes? Well you get seams in form of bright intersections because in those spots the light from the pixels is overlayed.

https://i.imgur.com/9ZuP3Ju.png

So digitally, there isn't much option in overlaying two projections, unless they will be very close in resolution (say 720p and WXGA) or if you just need to blend projections horizontally or vertically. But when one projection is bigger than the other in size then even if they are the same resolution you will have this problem.

If, however, there was a way to trim the pixels above which get overlayed in a way illustrated below, optically and not digitally,

https://i.imgur.com/zrSv7ZO.png

then you could easily place the other beam there with almost no seam.

https://i.imgur.com/yDaJd8N.png

One possible solution I proposed is having a transparent film or glass ( which I am not sure exists, at least one with 8ms transition time) which becomes opaque or any other state which can move beam 1 away from the optical path when bright IR or UV light is shone on that film/glass. Then all we would need would be to have that glass be half silvered and project the smaller inset beam also containing IR or UV. Setup would be like this:

https://i.imgur.com/5ku44hR.png

This would solve, or greatly decrease the seam between the two beams.

Any similar solution to this not onvolving pixels would work.
And as said in OP, this kind of beam combining can be also achieved later in the setup by the "projection screen". Or in other words, having the beam 1 section not reaching the projection surface is not the only solution as somehow not displaying the beam 1 section on the projection surface even if it has reached it will also work.

The other advantage is, you don't even have to worry which pixels to hide on beam 1, as they will get blocked anyway.

An LCD matrix or similar instead wouldn't work because there simply isn't one high resolution enough compared to Beam 2, not to mention it has to be bigger than beam 2 which can be anywhere on it, which means you would need an LCD panel with removed backing to be used as a shutter even higher res than beam2. This is what Tom.G. was suggesting, LCD is the most common Light Valve.
If such an LCD or other matrix would be usable we would just make beam 1 such a resolution. I can't as there isn't such a high res projection technology and even if I could the whole project would become pointless, as the purpose is saving pixel rendering time by only having many pixels where needed.

I hope I didn't fail at being clear enough this time.

 

[sophiecentaur]

I fear each new explanation will keep raising a new question of its own.

This is PF ! It's a price one has to pay but we often deliver satisfactorily.

I hope I didn't fail at being clear enough this time.

That is much better. Thank you.
How far short is the full screen resolution that's already available? You could use a matrix of very high res displays to make up your ultra high res. The 'joins' can be pretty good, I think. Sorry - that's a "new question" but are you sure it's not relevant?
I'm sure you could solve this with the appropriate interpolation for the pixellation of the background, added to the inserted bit (especially if the actual insert was rectangular. I can't see what you problem is with using a common display. It's very easy to up-convert the coarse pattern of the outer display and use a (one pixel) oversized inner image. Is there a fundamental reason why the outer image really has to be low res if you are doing this as a simulation? Is there a reason in the foveal system for a coarse display? I rather gathered the technique is due to a data rate rather than a display limitation. I can't imagine that the system would be processor limited (?).
I don't know enough about optical hardware implementations of this to exclude the idea but you really need a very good reason not to go for the DSP solution. It's cheap as chips and there's such a lot of expertise and software.

 

[wosoka]

Both the highres small beam and big beam resolution is 854x480 (well, 480x480 rectangle with middle circle used for the small one), but the small one is 1/20th the size of the bigger beam. 20 times more resolution projector than 480p simply doesn't exist. Nor an LCD that could be used as a shutter. You would need a 17080x9600 resolution projector or shutter. For context the most expensive 4K LCDs and projectors which are too big for this are 4096 × 2160.

Even if it did
1) I don't know of any signal or cable that could handle that much data
2) would fry the GPU even if it accepted such data
3) would cost me my kidney

I think these are good reasons.

 

[sophiecentaur]

OK. Again, the actual numbers would have helped from the start. You have been down these roads before and so does anyone else.
Bearing in mind that we / don't actually have a solution yet, then something different has to be considered. Why not have a rectangular insert with a border of a very few pixels, to allow for some good interpretation between the background and the hi res inner image? That actually doesn't call for much more processor power and the insert could be done with a rectangular (trapezoidal / rectangular?) mirror. Registration would be important but it would be anyway.
Ohh - does the high res section need to move about?

 

[wosoka]

OK. Again, the actual numbers would have helped from the start.

Okay. To be fair there's a fine line between not giving enough info and posting a wall of text which is obnoxious to go through. It's hard to decide what info is and isn't needed sometimes. Doesn't help that I draw diagrams and write text by hand on a notebook that I have to digitize later to share on the internet.
And to be fair I did mention such high res LCDs and projectors don't exist before.

Ohh - does the high res section need to move about?

Yeah. Don't say I didn't say this one from the start, its in the OP.

 

[sophiecentaur]

Yeah. Don't say I didn't say this one from the start, its in the OP.

Just found that bit.
So you need to be moving the high res image about. How did you intend to do that? Even with your original idea of LCD gating, the insert still needs to be moved. I suppose that can be done by a slow deflection mechanism. That could be synchronised with the electronic 'hole' in the big image so no (real) problem. Some tracking of the final result would be needed by a further camera, looking at the result.

Doesn't help that I draw diagrams and write text by hand on a notebook

There's plenty of software that would let you post notes and diagrams directly. Have you not presented this to any of the team that you must be working with? (You surely won't do it all on your own?)

 

[wosoka]

(You surely won't do it all on your own?)

Not to brag or anything, but yeah right now this is my side project.

Just found that bit.
So you need to be moving the high res image about. How did you intend to do that?

Galvo scanners with big mirrors and very slow rotation speeds by scanner standards. Maybe will add the off axis parabolic mirror after the scanner to fix the beam distortion more in the optics than in software by having the beam parallel to the big beam rather than at an angle.

Even with your original idea of LCD gating, the insert still needs to be moved.

It wasn't my idea, my idea was just hiding the pixels on the bigger/lower res projection, no LCD. The LCD panel shutter wouldn't be much more accurate than the DLP/LCD projection anyway since the technologies are close in pixel count (I can get a HD projection for the big one instead of a HD LCD shutter but the cost and more importantly size of the device will increase).

If some UV/IR activated shutter like I was thinking about existed it would not only create a near seamless overlay but we aslo wouldn't even have to care which pixels to hide since they would be just blocked.

 

[sophiecentaur]

But what would the transition form on to off be like? There would still be a need to match the edge of the LCD mask to the inserted HD image. Are you sure that a display plus a LCD mask would cost less than just using a display with a blanking signal applied to the original image? Is a higher definition LCD mask available than a HD display?
I asked before whether the equipment you plan to make will be experimental or for 'production'. What luminance and size will you want for the displayed image? If you are planning to study foveal display basics then 'any' way of producing what you want is acceptable, surely. Technology has a habit of catching up on research.

 

[wosoka]

LCD, DLP, LCoS are pretty much the same today in terms of resolution.
I did not mean that LCD would cost less, but the same. Increasing the DLP resolution or getting an LCD pixel shutter would both add to the cost and more importantly size of the device and still not give enough precision in the "mask" area.

As mentioned in the OP, "Brightness not an issue".

This is not to study foveal display but to research possible solutions to already known current issues pointed out by the research papers from Microsoft and NVidia from 2012-2017. The issues they found can only either be solved by a better seamless overlaying of displays or consumer displays which are not likely to be available in the next 10 years (>8K resolution).

 

[sophiecentaur]

Help me with this one. You have a display with standard resolution and you can put a 'hole' in it by video switching. You can physically deflect your VHD image to be at the same place on your final projection screen as the hole and your DSP can take care of any interpolation at the transition. Is there a need for an additional LCD shutter? I haven't understood the advantage the added complication. You have been at this somewhat longer than I have so you probably have a good reason. What is it?
Do you even need a mirror if the two projectors are near a common axis and keystone correction is applied?
None of the problems associated with this are a fraction of the problem of making the old shadow mask colour crt displays and they were happily mass produced in their thousands. Light behaves much better than beams of electrons.

 

[wosoka]

I simply wasn't the one who suggested to use an LCD, Tom did. I was just explaining its resolution isn't really better nor is its price so its not a better solution, I don't know why you still have the impression I'm the one who thinks LCD here is a good idea. But it definitely would be a good idea if it had a much higher resolution and didnt cost too much as the mask it created would be smoother (like in the comparison illustrations I posted). But like I said already, today LCD, LCoS or DLP are pretty much the same.

Do you even need a mirror if the two projectors are near a common axis and keystone correction is applied?

I need two mirrors or one 2d mirror to move the smaller beam anywhere as needed on the bigger projection. I need a parabolic mirror or similar to reduce the distortion caused by the axis of the steering mirrors.

The reason why keystone correction can't replace this is the same reason keystone correction can't always replace optical zoom or lens shift. Its a digital procedure that is lossy, it sacrifices pixels.
Besides, you need to properly undistort something when you don't necessarily need to if optical solutions will work.

 

[Tom.G]

Have you taken this into account from page 5 of:
http://www.physics.mcgill.ca/~moore/P101/Lectures/Lecture-16.pdf

There is much more detail in the article, but what it says on page 5 is that if the image plane is 250mm (10 inches, closest focus distance for young eyes) from the eye the maximum resolution is 850dpi based on the cone density in the fovea. And that is in a spot on the image of only 3mm dia. Of course unless you are doing eye tracking you may need a spot larger than this.

... each cone is only about 2
microns in diameter at their receptor end...
Since the fovea is 17 mm from the optic center of the eye lens, one cone will subtend an angle of

θ = (2*10^-6)/(17*10^-3) = 1.2*10^-4 rad = 0.12 mrad

A dot which subtends this angle at a distance of 250 mm has a diameter d given by

d/0.25 = 1.2*10^-4

d = 0.25*1.2*10^-4 = 3.5*10^-5 m = 0.03mm

To get an idea of this size, consider that there would be about 850 of these dots in one inch (25.4 mm). In the standard unit used for printing measure, i.e. the “dot” of 1/72th of an inch (about 0.3 mm) these dots would be "850 dpi". So a print at 850 dpi would have no discernable "graininess" when held at 250 mm from your eye. If you could hold it at about 180 mm from your eye, and still focus on it clearly as young people can, the print would have to have about 1200 dpi for the graininess to disappear. This indeed is regarded as the standard required by professionals for "photo-quality" print.

However, as I have already pointed out, the fovea only has a diameter of about 0.2 mm, or 200 µ. Thus the angle of clear vision is only
about 100 times that of the resolution, or about 12 mrad. At a reading distance of 250 mm this is a spot diameter of 3 mm.

It is hard to believe that we only have clear vision in a spot of 3 mm diameter at 250 mm. This is about the size of a capital "O" of "12-pt" font, as it would appear in a print-out of this text.

 

[wosoka]

I have read enough articles on fovea and human visual acuity to know what I'm doing. 5 degrees is a good rule of thumb, the resolution falloff is actually a curve and the perfect 20/20 vision is actually in less than 5 degrees but since we don't have such screens with random pixel density from center and don't have perfect eye tracking 5 degrees works.

 

[sophiecentaur]

I simply wasn't the one who suggested to use an LCD,

True but that would be an instantly workable solution. I'm not sure you have a workable alternative, which is the reason for your OP. Hence I was comparing two workable systems.
Keystone correction, whether optical or electronic (when properly done) will introduce the same spatial frequency response variation across the field. It's not just a matter of throwing away pixels.

 

[wosoka]

Keystone correction is a term describing digitally correcting an angled projection, I've never heard of the term used in optical solution to the keystone effect.
What I am thinking of is making the beam parallel again to the projection surface so there will be no need of keystone correction and I mention one possible way how that may be done, although I've never done them before and it is just an idea.

 

[sophiecentaur]

Keystone correction is a term describing digitally correcting an angled projection, I've never heard of the term used in optical solution to the keystone effect.
What I am thinking of is making the beam parallel again to the projection surface so there will be no need of keystone correction and I mention one possible way how that may be done, although I've never done them before and it is just an idea.

Keystone distortion has been with us for at least as long as photography. Any off-axis image will have some keystone distortion and architectural photographs have been corrected for by using 'shift' lenses. Expensive and well out of my league and Photoshop does well enough for me.
But your scanning system with a galvo-style mirror will still introduce some keystone errors. They can just be less when the tilted mirror allows you to start off on the axis.

 

[wosoka]

I was talking about keystone correction being a digital thing, not keystone distortion.

Yeah, for sure galvo scanner or 2d mirror will introduce distortion but like I mentioned I'm thinking of having a third, bigger off-axis parabolic mirror after the scanner to help with that distortion. Again, I don't know how much that will help, just an idea someone game me.
https://i.imgur.com/jkSsPBk.jpg

 

[sophiecentaur]

I was talking about keystone correction being a digital thing, not keystone distortion.

So you are telling me that a shift lens doesn't do keystone correction? People are wasting thousands of pounds on their photography, in that case. (Did you search "shift lens" before you came back with that remark?)