Lighting Research Results

[matthewepler]

Thanks for the drawing @cpixip. I’m much better with visuals

Unfortunately I drew and posted my diagram before reading the article on the enlargers. I see what you mean.

Do you think this approach shown in your diagram will work for the Kinograph? I imagine we’d probably just print with white plastic or make the box out of white acrylic plastic which is glossy and reflective.

It seems like a good approach and one that we could easily test. I will have access to a laser cutter in addition to the 3D printer within a week or so.

[johnarthurkelly]

@cpixip @matthewepler - to the point about collimated light, it would definitely not be desirable at the image plane, but using a collimated beam to illuminate a diffuser at (or very near to) the image plane might be slightly advantageous in that we could afford to move the light source a little further away (potentially folding the light path with a mirror). This is why the bottom drawing would be better than the top drawing since you gain the ability to move most of the illuminant “stuff” away from the film path with roughly the same penalty in light loss.

Condensing the light and collimating it are not necessarily the same, and we would not necessarily have to collimate the beam to make it useful in a compact arrangement. Collimating the light (having the light rays all propagating in parallel) is most useful if we have to send the beam a long distance, but in our case always with a diffuser at the end

Re: the questions on the drawing-
A - The shape of the tube might not matter that much pre-condenser as the goal of the condenser would be to collect as much of the beam as possible and so it would likely be pretty close to the LED itself.
B - No, a mirror could be used in collimating a beam of light but in this case would be more complicated.
C - We can chose our distance without having to calculate it, to a degree, we can pick optics based on their focal length to suit the position of the light source and the diffuser. Basically we need to know what the shortest distance the beam would need to travel is, then we can pick condensing optics to suit. Of course based on what optics are available we may need to adjust that distance slightly. The second critical distance to consider is the distance between the diffuser and the film plane - we definitely don’t want to risk having the diffuser able to be in focus at the same time as the film. Since we’re mostly working under macro conditions I wouldn’t worry about this too much, as long as the film isn’t sitting directly against the diffuser.
D - No other special advantage!

[cpixip]

Ok, this starts to get interesting. Here are some additional, random thoughts I want to throw into the discussion:

• It is probably a different challenge if you work with white-light LEDs compared to a setup with separate LEDs for red, green and blue. In the former case, you primary task is to evenly spread a single light source over the area of a 35 mm frame. In the later case, you also have to assure that the light of every single LED (placed at different positions in the setup) is evenly distributed independently.

• LEDs differ somewhat in their radiation pattern. A narrow-beam LED can be moved much farther away from the scan area than a wide-angle LED. Provided an appropriate diffusor is available.

• Any optical element in the illumination path, like a collimator or condensor for example, will make the unit more expensive. And potentially more difficult to setup and source.

• From my experience it is difficult to get an even illumination with using a diffusor only, in a light path with more or less collimated rays. In case of light radiating from a single LED, one usually “sees” a blurred version of the LED if the diffusor not sufficiently strong (a “hotspot”). You need a thick enough diffusor (or several diffusors after each other) to get a sufficiently diffusing action. But a thick diffusor reduces the amount of light reaching the scan area.

• From a theoretical point of view, an integrating sphere with the appropriate surface coating is the perfect thing to create an evenly illuminated aperature - without any diffusor (such a thing could actually have some negative effects here). Well, as I am a theoretical physicist by training, this was what I opted for in my design. Only to find out that it a) works, b) is not so easy to manufacture and c) turns out to be a little bit to bulky to integrate it into the rest of the machine.

• From an engineering point of view, it seems that the design efforts of nearly a century of photographic enlargers could be an interesting resource for the Kinograph project. Some enlargers work with a condensor lens system, but try to find some replacement parts for them nowadays! Many enlargers have used diffusing lightboxes (mixing boxes). This is the cheapest solution I know of. An appropriately designed rectangular lightbox mimics the behaviour of an integrated sphere, ie multiple random reflections between source and exit port (= appropriate dull inner surface + large enough room/surface area for mixing).

@matthewepler: - yes, I certainly think that an approach patterned after some old enlargers for 35 mm stock could work for the Kinograph. The important thing is that the inner surfaces of the mixing box need to be as dull as possible ( Lambertian reflectance) - which rules out raw 3D printed surfaces; you have to coat such 3d-printed surfaces if they need to work properly. But what about making a box directly out of Styrofoam as “in the old days”? Styrofoam can be cut by hot-wire cutters and it is easily available in most places.

[junker]

I was just going to suggest what @cpixip mentioned that you would want a matte surface and not a glossy surface for the light mixing (whether it would be a box or sphere).

@johnarthurkelly, looking at the images of your Direct Drive tests I noted that there was a hotspot because the camera was looking directly at the LED. By putting a 45° bend at the end of the tube and mounting the LED from the side, we might be able to get better results by essentially making a light mixing box. And it wouldn’t need a 3D printer to make anything, you could cut a shape out of foamcore (or maybe even poster board), score the bends on the backside, and then “snap” the folds to make a box. The foamcore/poster board would be matte enough to run some simple tests for comparison and could be made in a matter of minutes.

My frugal nature (my wife says “cheap”) leads me to look for the most simple, inexpensive solution. I don’t always get there (I really love my 3D printer) but it keeps me thinking. I’m also guided by the engineering adage, “Never make what you can buy (or reuse/repurpose).”

• I’ve been looking on Craigslist for broken monitors/flat screen TV’s that people are giving away to harvest the diffuser panels out of (a 42" TV would be a lifetime supply for our purposes)
• Foamcore would be a quick way to mock up sample light mixing boxes
• I’ve looked at hollow balls to make simple integrating spheres from (maybe just cut a racquetball in half?).
• From the AV Home Theater world, it seems that Sherwin Williams Duration Extra White household matte paint could be a good substitute for Barium Sulfate for coating integrating spheres/mixing boxes.

[matthewepler]

Foam core is a genius idea @junker. I like how you think The first Kinograph used the diffusion from a computer monitor and it worked pretty darn well.

@johnarthurkelly thanks for addressing my original questions. Much appreciated.

Firstly, for myself and anyone following I’d just like to make sure we’re all using the same terms in the same way.

• condenser - converts a divergent beam from a point source into a parallel or converging beam.
• collimator - converts a divergent beam into a parallel beam, a type of condenser.

In other words, it’s a rectangle/square thing. All collimators are condensers, but not all condensers are collimators. Some focus the beam into a small point, etc.

Do I have that right? Not trying to split hairs, just making sure we don’t confuse one another.

I’m trying to condense (pen intended) what we have into a way forward. Here are some thoughts…

1. Sphere vs. direct.
   There actually isn’t a huge difference for me in terms of easy of build. they both require printing/making parts and assembling them with electronics and mounting them to Kinograph. I’d have to attempt to do both to see the real trade-offs. Which brings me to #2…

2. Design factors for each:
   a) direct

• light should bounce multiple times before hitting diffusion
• if we need to move the light source away from the diffusion plane, a collimator could help cut down light loss in the straight-aways.
• what type of condensor we choose (if any) is largely based on the beam spread of the LED and how far the light source is from the diffusion plane.

b) sphere

• maybe some a few tweaks for ease of assembly, mounting, and efficient use of material

c) both direct and sphere

• inner surface should be matte white
• both should be easy to assemble, and relatively easy to build/source in small batches

3. Price - as far as I can tell, there still isn’t a huge difference here either. Again, it would take more tests and hands-on experimenting before we lock down the design and therefore the costs.

In sum, the sphere is actually pretty straight-forward from what we’ve discussed. We agree it’s a good light mixer. It just remains to be seen how it would integrate with the current design. I can help there. I have John’s CAD files which I can toss into Fusion360 tonight and see how it looks.

In the direct version, it seems like taking the simplest approach and building in complexity as needed is a good approach.

SO NOW WHAT?!

I say we pool our resources and start prototyping. If we want to have John test it with the same rig for comparison, we can ship what we make to him.

If anyone wants to source the LED and driver for their own experiments, I can front the cost of a few.

I can 3D print stuff and so can @junker. I can make runs to hardware stores for other supplies as well.

What say you, friends? What do you want to build and what do you need to do it?

[matthewepler]

I know that was a lot of words, so I’m summarizing here for my own sanity. The answers I hope to answer are:

1. can the integrating sphere fit, and if so what other considerations are there in terms of physical design

2. what dimensions, shape, and components are required for a “pretty darned good” direct illumination approach?

3. which of the two options makes the most sense for v2 Kinograph (with cost, ease of assembly, and ease of production as key factors)

[matthewepler]

Well, good news on the CAD front. Both options fit just fine without any need for adjustments to the existing design. The sphere is just the right size and the tube is small enough to be flexible on positioning.

I sketched some quick thoughts here. A couple of notes:

1. @johnarthurkelly how big of a deal would it be to move the LED from the top of the sphere to its center? I’m pretty sure that is less than ideal but just checking. I ask because of the next note:

2. We would want to consolidate connections as much as possible. Ideally we can integrate the LED driver into the Arduino shield and then just solder some wires to the LED, glue it in place, and then connect it to the shield. Alternatively (and more modular) the LED is part of a standalone PCB that connects to the shield only for PWM control. Lastly and perhaps most annoying is the LED is separate from its driver which is in turn connected to the shield. .
   What do you all think? Original pic here for download if needed.

3. The tube design might be easier to repair/replace for people without easy access to a 3D printer. Even if they do, printing itself can become a bottleneck if they’re troubleshooting failed prints, sourcing filament, etc. In a pinch, the tube could be created from just about any material. Advantage tube. Agree?

I’m tempted to just move forward with both and see if we can’t figure out the LED/PCB/Shield connection so that it could support either direction. Then it’s as easy as choosing your preference and printing/building it.

Okay, enough out of me for one night.

[cpixip]

@matthewepler - that looks nice. One word of caution: a good working integrating sphere does not need a diffusor on its exit port.

[matthewepler]

Ah, good point! Thanks for reminding me @cpixip. Would you make any changes to the tube design?

[cpixip]

@matthewepler: I am afraid that I can not really comment on this. The geometry will be heavily influenced by the spread of the LED-light (angular distribution) and the choice of the material at the inner surface of the tube. Also, the diffusing properties of the diffusor will play a role in this.

[matthewepler]

All good. Guess we’ll just have to build it and see how it goes.

[johnarthurkelly]

This is very true - I tested this to get a consistent measure of the light loss penalty of having one in the chain (since it’s a requirement for the direct-drive model).

@junker - I know, the hot spot is egregious for that level of diffusion! I wanted to test this on a large aperture because I had a good experience using a similar design (although for an 8mm gate with the same diffusing material) and I suspected that it might not hold up for a 35mm gate, but I didn’t know how bad it might or might not be. Good thing for testing!

[johnarthurkelly]

1. It’s not a terribly big deal, although it would absolutely require baffles to be added to the interior of the sphere. This would precipitate some more testing depending on the model of LED being used, generally enlarging the sphere in this case would be advisable.

2. LED control can be done from the shield, but in the case of the Yuji LED that I tested, it really likes to be driven at around 19V, which would require a dedicated power supply. Other high-output white LEDs would be similar (i.e. they probably won’t run off the ~5V available). A confounding factor here is that most LEDs of sufficient brightness will likely require some kind of cooling as well, so having the LED separate from the shield (i.e. leaving room for a cooling apparatus) is beneficial.

3. Which leads to your second drawing which might be the ideal hybrid of both designs (the version on the left) - it positions the LED so that it can be cooled and accessed easily, and based on my experience with the integrating sphere model and the direct drive model I believe it would provide a sufficiently bright and even field. Absolutely worth testing! (I’d be very curious to do all three measurements- no diff/ground glass/white diff - not being a sphere I think it would still need some diffusion, but I suspect we would be able to get away with something like the ground glass (brighter!))

Foamcore would work for prototyping but I’d be hesitant to use it in any kind of production environment For testing I think we should keep up with the Barium Sulfate coating process - BaSO4 was picked for this kind of application specifically because of its history as THE reference standard material. The diffuse reflectance characteristics are well known and pure white latex paint is known to have lower reflectance. A mix of the two for durability and ease of application is still ideal, I think!

[matthewepler]

I’m also going to order a true white PLA filament for comparison. @johnarthurkelly I can ship the print to you if you DM me your address.

[Greg]

FWIW, there are two points I’d like to make regarding the research and the comments in this thread:

1. The research paper indicated that there is not much difference between a full spectrum white LED vs. individually controlled LEDs. While that may be true, the latter does offer the additional benefit of allowing for scene level color correction.

2. For diffusing material, I have successfully used the diffusion material from junked flat screen monitors and TV sets. The larger sets have thicker plastic that can be cut with a jigsaw (a trick is to cover both sides with clear plastic packing tape which is a way to lubricate the blade in order to make a nice cut). Smaller sets have thin flexible sheeting. Either way, it’s a great way to obtain true diffusion material. Recycling centers and curbside are convenient sources of flat screen TV’s and computer monitors. Caution should be used if the backlight is the older fluorescent type (as opposed to LED) because the thin tubes easily break and contain mercury (a toxic substance even in small amounts). So as a precaution, please pay attention to that.

Greg

[johnarthurkelly]

Hi Greg - It’s not that I found there to be much of a difference between the two, theoretically they can be perceptually identical to the human eye, but the downstream requirements of the system are dramatically different for taking best advantage of each design. Scene-level color correction is a good thing to be able to do, but it definitely adds significant complexity overall where we can cover most situations with the more simple option.

WRT the diffuser material, there is a lot of variation in the kinds used in flat panel displays (partially dependent on backlight/edge-light design). I would be cautious about their spectral properties and overall diffusing power - certainly something to evaluate. The options used in the test allowed us to specifically compare a neutral, diffuse surface (ground glass) vs a diffuse material and evaluate the brightness attenuation penalty in each.

[Greg]

Hi John - Thanks for your clarifications. No doubt there is a complexity trade-off between the two approaches and my point was probably an obvious one. Actually, I was unaware that full spectrum white LEDs exist and so that it an option that should be considered. I don’t recall that it has been mentioned before in the discussions here (unless I missed it), so I’m glad to see it being presented in your paper. In regards to the diffusion material, that was more in response to using poster board since it is specifically designed for light diffusion. Of course testing the properties would be the best way to discover the true properties. I thought it was worth suggesting since so many solutions I’ve come across use “found objects” such as ping pong balls and ordinary paper. Thanks again for your detailed research!

[johnarthurkelly]

Hi Everyone - With @matthewepler’s help on some quarantine 3D printing I tested a model based on the Right-Angle Diffuser Box design:

(shown here as two halves for printing purposes)

Two versions of this were made, one printed with White PLA and one with Black PLA. I tested the White PLA model without any extra coating, and then both models after applying the same Barium Sulfate/Latex paint diffuse reflective coating that I used in testing the earlier designs in this thread. I performed the same measurements on these variations using the same White LED and driver from the earlier tests. These are the results:

The RA Diffuser Box model generally underperforms the Integration Sphere in both measured falloff and illuminance, however it is notably better than the earlier Direct Drive model. The uncoated White PLA performs quite poorly compared to the coated PLA.

I believe the falloff observed in this model is mainly the result of the LED being aimed directly at the “45 degree” diffuse reflector and causing a hot spot on the part of the reflector closest to the LED. Also the length of the tube between the reflector and the aperture likely caused some of the measured falloff. The next iteration of this design will place the LED such that the light will have to perform at least one bounce before reaching the final reflective surface as well as a shorter distance from the final reflective surface and the aperture.

[matthewepler]

Thanks for publishing these results in such an easy to understand format @johnarthurkelly. Really useful. So the integrating sphere remains the design to beat!

If I’m reading it correctly, it seems like the coated white performs slightly better than the coated black. Is that right? I assume this would apply to the sphere as well?

[johnarthurkelly]

Perhaps ever so slightly, although that may be within the margin of error for the light meter. It could also be the result of variations in the thickness of the coating - I coated each version the same number of times but since I was applying the coating with a brush it may have some unevenness that I can’t see by eye. I’ve been considering using a small paint sprayer (hobbyist size for models and the like) that might help even out the coats.