Sprocketless Design Thoughts

Musings on a Sprocketless Telecine Design

  • Overall Targets
  • 10 frames a second
  • Resolution — minimum 1920x1080 (aka TV 1080p)
  • Continuous film movement — no sprocket or pull-down claws
  • Transport Ideas
  • Film handling
  • Light exposure
  • Optics & Capstan
  • Synchronization
  • Laser sprocket detection
  • Single sprocket
  • Multiple sprocket
  • Laser frame detection
  • Single point
  • Line detection

OVERALL TARGETS
By starting with a few assumptions we can design to those choices and see if a practical design emerges. So 10 frames a second is not excessive and is well below easily obtained cameras with 1920 x 1080 (1080p) resolution at 30 frames/second.

THE FILM TRANSPORT
In the 1950’s there were punched paper tape readers (GNT 27, Decitek, and CINCINNATI MILACRON ACRAREAD LS TAPE READER) all handled paper tape without using tensioning arms and they could read up to several thousand characters per second without sweating. Just approximating the handling speed:
(3000cps x 0.1” tapePitch)/0.1667”per8mmFramePitch = 1800 fps equivlent
So moving film at 10 fps without tension arms should be easy. The systems worked by using one motor on each reel and the photoelectric reader sensed the sprocket holes with quadrature, so the direction of the tape was discernible. Some electronics then calculated the required currents and direction desired. So:

1] Direct reel-to-reel film handling without tension arms.

Next, we want smooth motion, so no stopping the film. How about a shutter or flashing the frame at the right time (timing will be discussed in depth later.) First lets discuss how brief the shutter or flash needs to be.

Given that we want 10 fps. That means each frame is 0.1 second long and since the camera is going to be 1080 pixels wide per frame then:
The aperture time needs to be 0.1sec/1080 pixels <= 92 µs. For safety margin’s sake lets shoot for 30 µs. This also points out that if we desire to double the resolution to 2160p (4K) we only need to reduce flash time to 15µs and as LEDs can work in the µs realm, it’s simple.

2] The aperture time will be 30 µs or less.

Obviously no mechanical shutter, but we can flash the light in these kind of time frames. Flash tubes are out because they don’t extinguish that quickly. That leaves LED flashing that can be done in the µs range. So we need to design an optical system with colored LEDs of the RBG persuasion with ufficient intensity.

3] LEDs provide 30 µs image exposure to camera (can be as low as a few µs )

If the LEDs that fit the well can’t make enough light then the LEDs can be accommodated above
the well of the capstan, with a combination of dichroic prism combining more powerful red,
green, blue LEDs and directing the combined light down into the well and a right angle mirror/
prism can direct the light source out through the capstan wall into the film image.

THE CAPSTAN
Imagine a miniature straight sided translucent coffee cup whose diameter is roughly that of an 8mm
film reel hub (2.5”) mounted on a good, low friction shaft/ball bearing combination like Fig.1

Now the capstan smoothes the motion of the film driving it and maintains the vertical position of film in relation to the camera. The laser/photo-detector combination will trigger the LED flash on the trailing edge of the sprocket hole. The wrap around the capstan maintains an exact relationship of film plane camera without any friction or complex film gate. Wet gate sponges are easily added to the film path just prior to the wrap. Due to the continuous film motion magnetic or optical sound pickup is much easier. A phase lock loop locked to the sprocket could provide a clock to the A/D system, effectively locking the film and sound track together simplifying later electronic editing and exact film speed would not be critical.

CURVED SURFACE IMAGE DISTORTION CORRECTION

So despite using a curved surface of capstan, the actual amount of film curve over one frame is
low and the plus/minus cylinder lens combination can be adjusted to remove even that small amount
of curvature focus distortion as shown by the extreme example of the photos. Film gates can let
film bow leading to major focus errors. Being wrapped around a cylinder will virtually eliminate
the ability of the film to warp.

SYNCHRONIZATION OF FLASH
This item is critical to the proper functioning of stable capture of the image.

Important points:
A) The laser should be reflecting off the film and detected, not shining through. The problem here is that film is generally transparent to the general laser colors so the on/off ratio is very low if the film is transparent. Whereas the reflected light, although less, is a value reflecting off the film, or almost zero when the hole appears. That ratio will be very high. It’s easy to set a comparator that has AGC and will track the small temperature variations and laser aging issues because of this large ratio. 45 degrees as shown is a good compromise for reflectively vs accuracy. From pg 14 of the “brown.edu” paper next is a discussion about another important consideration and that is the effect of polarization of the light on the film reflectivity. See:

This diagram is for the air-to-glass case, but is very similar for a air-to-plastic interface. Note that near the 45 degree point, if the polarization is parallel, then the reflection is decreasing. Perpendicular polarization is preferable. One way to ensure that would be to choose a green laser because of the optical components used to generate the green laser light is polarized. Experiments with cheap light pens confirms random polarization of the red and blue, but green is strongly polarized. So use of a green laser pen will allow adjustment of maximum reflectivity while keeping a 45 degree angle of incidence.

B) Next is the beam width. A cheap green pen laser beam is comparable in size to the 8mm sprocket hole. That will affect the ability to discern the edge. Therefore a small lens needs to be added whose focal length is the distance from the sprocket edge to the lens that will reduce the beam spot to a few mills in diameter. Another improvement would be to use a pair of lenses to reduce the spot to a line about the width of a sprocket hole, but very narrow in the film movement direction. This would retain the accuracy and enhance the ability to ignore edge quality of the sprocket hole.

C) The detection circuit should be set to trigger the LED flash on rising edge of light amplitude that corresponds to the trailing edge of the sprocket hole. The reason for this is that the leading edge of the sprocket is more likely to have suffered damage from the intermittent pawl of the projectors it has been used on.

E) Multiple sprocket holes could be electronically averaged together to get even better resolution of the edge point.

F) A method to supplement or even supplant the sprocket holes would to be to do the line generating techniques mentioned in B) above, except make the line length the full width of the image frame to enhance the on/off ratio of image to black line of the film frame.

      2/10/2021 rev. 1
     Charles R. Patton
2 Likes

Wow, some very clever info here. Some questions/suggestions.

  • If your capstan is a large diameter, say 8 inches or so, you could reduce the distortion quite a bit. This would be similar to the DFT Scanity. But they use a line scan imager. What method would you use to cure the distortion?

  • Your use of a prism on the light side is interesting. Would you use separate r/g/b leds and further filter them with addition r and g and b filters? I have several camera prism assemblies that could work for this. What about additional light processing, lenses, opal glass, etc?

  • My testing agrees with your shutter speed and light speed analysis. But also you will need quite a bit more light at those speeds to expose darker/thicker film. We ran out of light using a 100w led. If each lamp was 100 watts that could work. Our next tests will use a 200 watt white, 12000K, led lamp.

  • Have you tested the laser pickup sprocket detection? It sounds like it should work.

  • I think you could probably eliminate the tensions arms, but why do you want to? Cheap quadrature encoders are available, and do work well to control the reel motors. Hall sensors will work as well.

  • One thing that has not been explored here (to my knowledge) is the need for the transport to reserve easily. Will your system do this?

Great ideas! Hope to here more!

-Grace McKay

Oh, also, Depth Of Field. More light equals more precise focusing, more DOF is better in this use case

I’m not familiar with this term “reserve,” Grace. Can you explain it for me, please?

Great thoughts, @Charles! Thanks for sharing. I recall @VitalSparks bringing up the idea of a phase-locked loop for timing or the camera trigger some years back. I was never able to test it and would love to see someone implement it on a little film loop and share the results.

The clear capstan idea is very interesting and I like the visual aid you provided with the soda cans. I wonder, however, if proper diffusion of the light source can be achieved at that distance. Perhaps if you closed the top of the capstan and filled the middle with light from “behind,” in-line with the axle? It would be interesting to see a test of that too, perhaps printing the roller with white filament on a 3D printer. Do you have access to one? If not I could print a roller and send one to you.

I will order a green light laser and try your reflective idea. I’ve gotten some tips recently from other folks that confirm the need to focus the beam, either by lens or pinhole. I also did a test with an off-the-shelf component that did this very well, but whose beam was not small enough to work on 8mm film.

I hope to try a couple things to further this test:

  • put a pinhole cover on the off-the-shelf component
  • make a homemade version of the same setup, with a pinhole
  • make a homemade version with a focusing lens (another forums member is sending me one)

I did not know about the green led physics. I’ll have to order one of those.

that is my dyslexic way to say “reverse”

1 Like

In order to simplify my responses I’ll answer them inline

From matthewepler
February 11

Great thoughts, @Charles! Thanks for sharing. I recall @VitalSparks bringing up the idea of a phase-locked loop for timing or the camera trigger some years back. I was never able to test it and would love to see someone implement it on a little film loop and share the results.

It’d been years since I gave away my 16mm projector and all I have now is silent home movie 8mm. But I’m good at using LTspice so I might throw together a sim for the circuit that should include most of the required characteristics of a closed loop using the laser pen sprocket detection.

The other facet I definitely want to try is sim the whole physical reel-to-reel loop plus capstan. It’s a much faster breadboarding technique and I can check the servo loop stability requirements.

The clear capstan idea is very interesting and I like the visual aid you provided with the soda cans. I wonder, however, if proper diffusion of the light source can be achieved at that distance. Perhaps if you closed the top of the capstan and filled the middle with light from “behind,” in-line with the axle?

That’s an interesting thought. What led me to the cup design idea was the concern with two items:

Light loss through the opal/diffuse solid would become very high with the minimum 2.5’ dia. And I’m grateful for Grace’s comment of the required flash light level feeds into that decision, too. (Discussed more a bit further down.)

I didn’t like the thought of a clear capstan as I was worried about any dirt, hair, scratches, etc that would be on the surface would be imaged along with the desired film image since the whole idea is to control the image plane with the curved surface of the capstan.

It would be interesting to see a test of that too, perhaps printing the roller with white filament on a 3D printer. Do you have access to one?

I do not. I do have a lathe and mill coupled with minimal machinist experience as my main background is electronics design and software programming.

If not I could print a roller and send one to you.

I would be wary of a 3-D printer part for this particular application for several reasons:

As mentioned above the opacity of material. Are there any mildly or even transparent 3-D printing plastics? Too opaque will be a problem

And even if such plastics exist they will surely exhibit density/refractive index changes throughout the printed part that would lead to light focusing variations onto the film image/optic path into the camera.

I will order a green light laser and try your reflective idea. I’ve gotten some tips recently from other folks that confirm the need to focus the beam, either by lens or pinhole. I also did a test with an off-the-shelf component that did this very well, but whose beam was not small enough to work on 8mm film.

I just bought mine for what I thought was a good price from Ebay:

www.ebay.com/itm/3-Packs-900Mile-Strong-Laser-Pointer-Pen-For-Cat-Pet-Toy-Red-Green-Blue-LED-UV/174479778832

A red, blue and green laser pen combo pack for $9.99 with free shipping from NY. So delivery was quick. You’ll be able to immediately see the differences.

I wouldn’t recommend a pinhole for this, as using a lens to focus the beam will retain the beam power and will actually result in a much smaller point. The reality is that a pinhole is likely to be in the 10 mil diameter whereas a lens should get close to the diffraction limit of around 532nm (0.020 mills). Of course the homogeneity of the beam enters into this, but the advantage is clear. Also in my first posting I was suggesting the use of a line generating lens setup for some possible improvement in the detection process. The optics will probably cost much more than the laser pen.

I hope to try a couple things to further this test:

put a pinhole cover on the off-the-shelf component

make a homemade version of the same setup, with a pinhole

make a homemade version with a focusing lens (another forums member is sending me one)

I did not know about the green led physics. I’ll have to order one of those.


And then some responses to Grace’s post.

From : vintagefilm
February 11

Wow, some very clever info here. Some questions/suggestions.

If your capstan is a large diameter, say 8 inches or so, you could reduce the distortion quite a bit. This would be similar to the DFT Scanity. But they use a line scan imager. What method would you use to cure the distortion?

Please revisit my first post. The use of a pair of cylinder (like for astigmatism in glasses) lenses, one plus and the other minus, paired in line, by twisting one relative to the other you can go from 0 to 2x the diopter value. That is how the photos were done. One to show the loss of focus in depth of the field, and then using a pair of plus/minus lenses I adjusted to correct the field as shown in the second photo. Of course that correction is not perfect, but considering the correction here was for almost 1.25” depth change, whereas the amount of correction on a 8mm image frame of 0.130”x0.177” on a 2.5” diameter. capstan is only 0.002” depth of focus change from image center to the two sides so practically all distortion will be removed with an astigmatic lens correction as it will pretty much be 1st order correction.

Your use of a prism on the light side is interesting. Would you use separate r/g/b leds and further filter them with addition r and g and b filters?

Why would I need filters? The color linewidth of red, blue and green LEDs is already narrower than the ability of most filters other than expensive dielectric versions to filter. If you don’t like the rendition vs the dye color of the film, There are some other LEDs colors available but I would suggest that since the film is dye filters, they are much wider than the LEDs. Adjust the data to fix the colorimetry of the film vs the LEDs. That’s an area I haven’t explored yet, but the film curves are weird in that they don’t use 3 peaks as we think about it, but have some bad outlying filter characteristics the film folk had to deal with. I’m sure other folks have explored this avenue extensively as I’m sure many are using 3 color LEDs to capture images.

I have several camera prism assemblies that could work for this. What about additional light processing, lenses, opal glass, etc?

I’m not sure what you mean about “camera prisms.” If you’re describing the camera lens finder versions, then no. In my research into prisms there seems to be a glut of dichroric mixer prisms – perhaps as a leftover from the LED projector market. One of those prisms can easily merge the 3 separate LED colors into one beam. Using Matthew’s idea for the solid capstan would make a very simple optical path. But using my cup capstan the only addition is the mirror or prism to redirect a beam entering the top of cup and turning at right angles to exit through the side and through the film image.

My testing agrees with your shutter speed and light speed analysis. But also you will need quite a bit more light at those speeds to expose darker/thicker film. We ran out of light using a 100w led. If each lamp was 100 watts that could work. Our next tests will use a 200 watt white, 12000K, led lamp.

At this point this is a thought experiment. I appreciate the data about LED wattage’s needed. I don’t have a ‘feel’ for it but just a gut feel calculation says a 20mW consumption LED would expose a 8mm frame in sub 100ms. So 100ms/30us = 3.3k. So I would need a 20mW x 3.3k = 67W. Hmmm. Right in line with your numbers. One comment though, is that the LED itself and are you running them in a continuous mode, because they could be much smaller as they are being flashed so the LED may only need to be a 1 to 10 W unit depending on its saturation light output to large pulse current. Do you have data like that?

Have you tested the laser pickup sprocket detection? It sounds like it should work.

Only to the extent of the reflection technique that is clearly superior to trying to make out the sprocket holes with a beam through the film. Take any clear piece of plastic and observe the difference in trying to detect the very slight dimming by the plastic with the beam through the plastic as opposed to the total on/off quality of the reflected beam. And the polarization effect is also easily observed. Additionally the reflection method would be immune to the differences of film that has black around the sprocket hole and clear around the sprocket hole.

I think you could probably eliminate the tensions arms, but why do you want to? Cheap quadrature encoders are available, and do work well to control the reel motors. Hall sensors will work as well.

This question triggers a bit of a diatribe. I’ve done a lot of design over the years. A kind of mantra is the word elegant . Some equipment I’ve observed is elegant and some isn’t. For example. Years ago when I was much younger I was a partner in CAD/CAM business. We did a lot of punched paper tape format conversion work. The first tape reader we bought was a star-wheel reader – an abomination! Next we picked up a Ferranti 1000 cps stop-on-character reader. Oh, what a wonderous mechanism. Electromagnetic clutches engaging a grit wheel capstan pulling the tape. The noise of it stopping and starting would shake the rack cabinet it was mounted in. The tape reels were driven by multi-hundred watt pancake motors. You did not want your hands in the way if the tape broke as it used tension arms to feed back demand to those motors, I still remain in awe that we came away from that period without the loss of fingers. That unit was at least 3’ of vertical rack space and perhaps 100 pounds. Then went through a pair of upgrade improvements ending up with the GNT reader.

The purpose of a reader is to read the data from the tape and it did this exceedingly well and reliably. Small motors, one on each reel, gently moved the tape through the photocell head. No noise, gentle tape handling that never broke tape. No tension reels or devices to generate a complex path to thread. The tape just gently sped up and slowed down quietly. Total rack height about 9” and weight perhaps 5 to 10 pounds. Elegant. We also marveled at the beautiful, reliable performance of this device. I hadn’t formed the philosophy of elegant design yet, but the seeds were planted.

Subsequent years I had to work with equipment, systems, and software designed like the Ferranti. They worked after a fashion but looked like the idea started out simply, then the fixes for this and that get hung on and you have the Rube Goldberg design. I fully admit I’ve had my share of those designs, too. But I do my best to avoid that syndrome by better upfront planning, hence this design idea I posted. Discuss it and maybe it will be elegant.

To that end the only thing that tension arms bring to the design is servo stability for the reel/tape drives trying to bring up the tape start/stop speed. But that right off the top takes more power and three motors. The GNT reader only had two motors and was gentle by the elegant design step of putting in quadrature sense of the tape sprocket holes and using it to electronically clock the data into an IC FIFO. The FIFO had a digital output of half full that triggered the reel motors and so as data is pulled by the computer, the FIFO would start emptying out. When it passed the half full mark it would pull more tape. The computer stops taking data, the tape continues moving until the FIFO is more than half full. Reel motors stop and the FIFO waits for more data pull requests. Elegant. It was faster than the huge Ferranti.

So I argue that with modern small embedded computers like the Arduino or Raspberry PI, the computation required to do a similar task of gently moving film is no more difficult. The lack of need for tension sensing devices allows a very compact layout of image capture and film movement. Even the power requirements will be minimal. Talk of the 100W LEDs is not the actual average power requirement. Some 60 to 100 mW perhaps due to flashing. Even only a few W, perhaps 10-20 W total for the motors. Don’t try to accelerate instantly and the power required drops down to only the running friction of which this design has a very low amount.

This is why I don’t want to use tension arms. It’s one step away from the elegance of the main purpose, capture the film images with:

Maximum resolution

Gentle film handling

Minimum power requirement

Quiet operation.

Compact footprint

Lightweight

Simple, robust, low maintenance

So this is the way I think, obviously the goals often require tradeoffs. That’s what a designer is there for. But this is a foray into a telecine design just for the fun of it. Lord knows there’s not a large market out there! So no one is going to get rich.

One thing that has not been explored here (to my knowledge) is the need for the transport to reserve ( reverse ) easily. Will your system do this?

This is easily done but why? The reel design contemplated would not work well as a Moviola due to the already stated technique that it would not accelerate quickly – no stop on image. It is thought that the film would be started, left alone, and captured.

Having said that, it would be relatively easy to count frames as you move along and keep track of that. Stop accepting data to the computer, send a command to reverse the film to a frame number, wait for the data stream to come available again, while the transport pulls the film back past the desired point, then reverses to move forward again and start streaming data when it re-reaches the desired frame. Is this suitable?

Oh, also, Depth Of Field. More light equals more precise focusing, more DOF is better in this use case

Here I would disagree. Depth of Field is achieved by stopping down, creating the need for more light. But it is not sharper – the resolution is decreasing for an overall illusion of sharpness, but not really due to the Airy disc phenomenon. See: https://en.wikipedia.org/wiki/Airy_disk If you look at applications requiring good edge resolution like photomasks, printing plates, etc. the lenses are actually larger (and a lot more expensive) for just this reason. That is the problem they’re fighting constantly.

By using a curved capstan with a very controlled image plane, I argue that this design will give a more constant, better resolution taking full advantage of the better lenses that some builders will use on a larger format camera. If they use these larger format cameras for 8mm capture I’m not convinced that it’s not overkill as the image in the film probably doesn’t support it. But the only thing to change is the flash duration if there is a desire to increase to a larger resolution camera.

Perhaps the most critical aspect is perhaps the sprocket edge timing resolution. That goes something like this. 0.130”FrameSize / 1080pxlsPerFrame = 0.12 mil tape position needed for 1 pixel resolution. Already shown earlier is that w green laser with a lens spot will be in the class of 0.02 mills. And the electronic comparison of amplitude discerning as the light goes from 0 to 100% is easily at the 10% mark, so the actual possible resolution of tape position is 0.002 mils. Substantially better than our calculated need of 0.12 mils. So nothing jumps out about the non feasibility of this design.

Comments welcome. Charles R. Patton

Sitting here contemplating Matthew’s suggestion about the solid capstan. Yes indeed, I realized the method is to just cast some clear epoxy, with added diffusing powder to lightly cloud it. It would be easy to machine it, including the guiding channel for the film. Only light polishing should be necessary and it would be done. The clouding should suffice to diffuse the LED light sources behind it so nothing fancy would be needed for that. The only addition I can think of is it may need a cylinder lens to counteract the self focusing effects of the capstan cylinder.

Need to find a free optics design program to check that out.

Charles R. Patton

vintagefilm

"Our next tests will use a 200 watt white, 12000K, led lamp."

I assume you know, but just in case, saying “white” LED brings up the point that some “white” LEDs are 2-color sources, blue and yellow, that looks white to human eyes, but can create problems when trying to use to illuminate 3 color items such as film images. Another recent example of that happened with my daughter when she tried to use a green screen for a virtual Zoom call. The virtual feature was unusable. The problem was the LED lights in her chandelier caused the green screen to look dark grey, not green to her webcam. Changing the lights fixed the problem. So obviously there was no green in the light spectrum with the old LED lamps. Pretty much any single ‘white’ LED will exhibit this problem because the LED is actually blue and then the die is covered with a phosphor that exited with the blue emits yellow. The leakage blue plus yellow gives us the illusion of white. But it is not true full spectrum white. But for that matter, neither is the RGB, but it is more inline with our needs for 3 colors for film imaging.
Look at this colorimetry chart:


It shows the RBG LED color triangle and to imagine the blue/yellow, draw a line from the blue point straight across the chart going through the center point D65 (white) to the other side. Notice the end point is on yellow. The sum is the ‘white’ center point D65.

Another point to note is cheap green LEDs are around 560 to 570 nm much as shown by the white triangle on the color chart whereas the green laser is more around 532nm, much further into the green as shown by the black triangle. That argues a green laser instead of a green LED might be advantageous in use with film imaging. I’ll have to look into the film dye color transmission curves to see if that is a factor to consider.

Just some idle ramblings.
Regards,
Charles R. Patton

I don’t really follow, but I was thinking the “CRI” rating would now provide people with a quick number to compare how well the LED performs with colour.

Interesting idea about the light in the capstan roller! it seems quite complicated - I’m not sure of the advantages compared to a traditional gate and offset capstan? After making an integrating sphere for a light source to test, I wouldn’t even bother trying anything else. the light output is perfect for a film scanner. cracking and scratches are drastically reduced, and the light you input into the sphere is output just as bright. reducing the need for a high powered LED. The only reason I wouldn’t use one is if I were to sell the scanner. I’m pretty sure Blackmagic cintel has a patent registered for the use of an integrating sphere for film scanning.

Charles, thanks for the reply. You said

The use of a pair of cylinder (like for astigmatism in glasses) lenses, one plus and the other minus, paired in line, by twisting one relative to the other you can go from 0 to 2x the diopter value. That is how the photos were done. One to show the loss of focus in depth of the field, and then using a pair of plus/minus lenses I adjusted to correct the field as shown in the second photo. Of course that correction is not perfect, but considering the correction here was for almost 1.25” depth change, whereas the amount of correction on a 8mm image frame of 0.130”x0.177” on a 2.5” diameter. capstan is only 0.002” depth of focus change from image center to the two sides so practically all distortion will be removed with an astigmatic lens correction as it will pretty much be 1st order correction.

I did miss this on your post. I made the assumption that you were talking about digital correction. So this lens pair would be put between the camera lens and the gate? On my system, we only have an inch or so between the nose of the lens and the gate using a Componen enlarger lens.

Next, I said

Our next tests will use a 200 watt white, 12000K, led lamp.

And on this I was wrong. The latest lamp I bought is a 200w high CRI lamp, and I forget what the color temperature is. which I think was recommended on this forum. I have not yet tested it. And I do believe in using opal glass as a diffuser, which does require more light. Without it, it is possible to pickup the individual led elements from the lamp. Opal glass is a cheap fix for this.

One comment though, is that the LED itself and are you running them in a continuous mode, because they could be much smaller as they are being flashed so the LED may only need to be a 1 to 10 W unit depending on its saturation light output to large pulse current. Do you have data like that?

I have always flashed the LED’s. So no data on continuous running. We started with a 10w lamp, way back, and moved to a 100w right away, to get enough exposure. The camera I was using – a Prosilica/AlliedVision GT1920c needed more light to reduce dark current noise. With 100w we could get a decent picture a lot of the time, but some films needed more.

I should say, I appreciate the theoretical viewpoint, along with the references and calculations. It helps give perspective that I no longer have. My practical background on this began around 1999 with the first CCD telecine I purchased–a Bosch FDL60. I have been all over that system and all of the ones that followed, the Quadra, Shadow, Spirit, and now the Grass Valley/DFT-era Spirit 4K machines. Which brings me to “Elegant” I appreciate what you are saying about design and elegance. I consider the Spirit line of machines to be the most elegant in the field. At the retail price of $800,000 to $1,200,000, they had better be elegant.

One of the things that those machines have in common is the tension arm system. It has evolved over the years, but the design theory is much the same. They use a hall sensor at the center of travel to sense position, and modify the reel motor speed based on that. A secondary function is film tension, which is regulated by a spring whose length can be modified based on feedback, and by menu settings.

Here I would disagree. Depth of Field is achieved by stopping down, creating the need for more light.

I assume I will have enough light to be able to stop down a bit if I need to increase the depth of field, if not, get a bigger light. The lens choice should be a large enough diameter to cover the film. Schneider recommended to me to use the Componen 2.8/28mm to cover 16mm

Another quick comment that is unexplored. Machine tolerances play a big part in getting a stable sharp image. A little bit of play or run-out in the capstan can add distortion and jitter in the image. Avoiding the possibility can also be a design consideration.

Andy, thanks for the info. I will look up the patent and I need to learn about the CRI rating. I had been thinking about the integrating sphere already, and this confirms the idea. I need to understand Grace’s optic path. Her two points in her post bother me:

We ran out of light using a 100w led. If each lamp was 100 watts that could work. Our next tests will use a 200 watt white, 12000K, led lamp.

The color temperature is too blue it seems and and the wattage seems too high like perhaps using continuous lamps instead of flashing. I haven’t gone back yet to work out the light requirements vs time. I need to back into a number like lumen-seconds or mw-seconds per frame needed for the typical digital camera. ISO numbers converted to that number perhaps?

One other comment on the capstan idea. Matthew brought up the idea of a solid capstan. And another aspect came to mind. The cup cylinder I proposed has another very positive quality. If we start with a solid clear cylinder, then the ‘line’ focusing effect the cylinder would create would need some optic correction. A cup cylinder will not create focusing effects so it would be relatively easy to shoot the light beam across the empty ‘cup’ and through the film image without the lossy use of a ‘milky’ capstan. We can live with the multiple reflection losses of the 4 surfaces.

I also want to address the film gate discussion. Most film gates I’m familiar with use a constant pressure plate to try to keep the film flat in the image plane area hence are subject to friction effects scraping the film. That’s an effect I wanted to eliminate as the likelyhood the film is already old and damaged – why add more stress? The film will not be sliding on the capstan, so a gentle solution. Additionally film going through a typical film gate is flat, but does the film itself really want to lie flat – a much bigger problem in high intensity light situations of theater and drive-in projection where heat cause the film to expand unevenly? That in part is why some film gates were curved, too, for the same reason I want the curved cylinder of a capstan. The geometry makes it almost impossible for the film to buckle thus lending a very stable image plane position and the use of a capstan means the film is not sliding on a surface.

Regards,
Charles R. Patton

Andy,
I’ve started the sims on the transport idea along with thinking about the film path with the varying diameter of the reel on the side with the laser sprocket hole detector The angle variance may require a another capstan just to make a constant path. If so, then maybe driving it is easier. Not sure yet. As I mentioned, simple elegant design suffers when reality strikes and the fixes are added.

I went looking into US patents assigned to Blackmagic Design. It returns 29 patents, 22 of which are cosmetic design patents. However of interest is that this patent:
https://patents.google.com/patent/US10388306B2/en?oq=+10388306
is the use of speed detection of the film and using it to adjust the audio sampling rate like I mentioned earlier. So apparently not the first time it’s been thought of. So maybe they’ve done it?

I was looking for the integrating light sphere idea you thought they had used and can’t find a sign of it. So maybe another company/person has? That will take me a while to dig deeper. But maybe you could give it a bit more thought for an alternate?

More musings on this area of light sources in the reply to Grace.

Comments welcome.
Charles R. Patton

Grace,
A long reply, primarily on LED choices as a PDF that includes a lot of graphics so PDF is easier than trying to load it here a a straight inline response.

replies audio light-GM.pdf (422.9 KB)

Regards,
Charles R. Patton

Andy,
I think I found the patent you were talking about using an integration sphere.
Method and apparatus for digitizing films using a stroboscopic scanning system
https://patents.google.com/patent/US5771109A/en?oq=5%2C771%2C109
PIXAR – “Method and apparatus for digitizing films using a stroboscopic scanning system”
Interesting. Using an integrating sphere they would flash blue for 1 second into a hi-res CCD chip (3Kx2K), then use 3 seconds to pull out the data. This is repeated for red and green, for a total frame digitization time of 10 sec. Then the film is pulled to the next frame using as described: “In the preferred embodiment of the invention, film transport 206 is a modified wide-body version of the Albert Howell design circa 1911 which is a cam operated, pin-registered, film shuttle mechanism used in the optical camera manufacturing industry.”

The only part I don’t fully understand is the reason for a flashing frequency that is different for each strobe color. Red at 15 Hz. Green at 8Hz . Blue at 30 Hz. As joules per flash are mentioned it may be that a flash is a fixed quanta of light (like a flash tube) and the sensitivity and conversion to a color is being compensated for by the number of flashes total.

Just for interest, if you want to make your own integrating sphere, this data from some old research for a design project may be handy;
“A Mixture of Barium Sulfate and White Paint is a Low-Cost Substitute Reflectance Standard for Spectralon®”
https://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=1010&context=cpl_techniquesinstruments
I’ve made about a 3” integrating sphere with a plastic softball to compare LED outputs. Since I wasn’t after absolute, I never coated the sphere – but I do have the barium sulfate if needed. Simple and cheap and worked fine. As I recall I had to buy a bag of several balls at the toy store or some such for a few dollars to get one.

One point that became clear was just any old webcam won’t do. A critical spec needed is a “Global Shutter”. Those are a bit harder and more expensive to find, but searching found:

Standalone Camera
ArduCAM - RobotShop 5.3V/lux-sec responsivity / Resolution: 54fps 1.2MP (1280 x 960) CMOS color global shutter camera module AR0134 $69.99
The only possible gotcha that I can’t figure out is whether the color version is still 1280x960 as this apparently same chip is also offered in B&W.

Obviously good SLR like digital cameras are often global shutter sensors, but they’re also $1,000’s. I’m cheap, so I’d like to see an inexpensive solution.

So all the above bring up a couple of additional ideas:
1] If B&W cameras are the best form, 3 cameras and 3 pcs of 100W LEDs could be situated around the capstan and 3 separate sprocket detectors used that are spaced some n+frame-fraction angular distance apart. Now the individual 100W LEDs could be flashed separately and the recovered images combined in memory. That ups the equivalent flash power and retains the constant motion. It does make the optical alignment a bit more tricky to get the images aligned and exactly overlapping. Oh well, I never believed in the free lunch anyway!

I found a US supplier for the various 100W LEDs:
Ebay store = due20150801 all look to be Chanzon LEDs
$9.88 Blue 100W CH_ due20150801 | eBay Stores
$14.13 Green 100W due20150801 | eBay Stores
$17.99 Red 100w due20150801 | eBay Stores
$14.99 RGB 100W 100W watts High Power SMD LED Chip COB Lamp White Red Blue Green UV Lights Board | eBay

If the color camera version is still 1280x960 and a RGB 100W LED is sufficient the cost is rather low for this scheme. $14.99 + 69.99. The camera can capture at 54 fps, so depending on the flash power, running at the actual film frame rate may be in reach and certainly the integrating sphere would allow going up to 300W with separate RGB LEDs. The optic path is simple at that point.

While on the subject of frame rate, if all that is desired is to up the apparent LED power. The film could be moved at a slower rate, like 5 fps would allow doubling the on time of the LED flash so effectively doubling the flash power the sensor sees. A very simple solution and it only slows the film down and the original design goal didn’t specify running in real time.

Comments welcome.
Charles R. Patton

1 Like

It took me a while to find some mention of the integrating sphere patent (apart from on Blackmagic’s website) but here’s what I read - https://patents.justia.com/patent/8009190

I did see that link about mixing barium sulphate with paint, I ended up following a slightly different recipe I found here - http://pauluszak.com/photonic/technical/integration.files/reflectance-characterization.pdf
although I did still use barium sulphate instead of titanium dioxide, and here in Australia I used Cabot’s clear polyurethane. I 3D printed my sphere at 88mm diameter which is all I could squeeze onto the scanner. I did see some hollow foam balls sold for arts and craft which would be good!

Well interesting find - however, I think it’s only a US patent (not world-wide). Also, I would be surprised if no prior art could be found. For example, Fig. 5 of this expired Eastman Kodak Co patent is quite close to the integrating sphere concept. (It’s actually an integrating cylinder because it is used in the context of a line scanner.)

Seriously, the integrated sphere idea was published around 1900 by Richard Ulbricht (a professor at the “Königlich-Sächsischen Technischen Hochschule Dresden”, is a standard you will be treated to in any decent optics course at the university level (specifically, why the integrating sphere is a perfect mixer and under what circumstances that happens) and is widely used in a lot of photometric applications. Given, the Cintel patent application introduces some additional ideas, for example varying the reflectivity of the sphere locally to adapt to the task at hand. That is something different.

Hi all, just wanted to jump in here and remind you that we have some great research on this topic performed by one of the forum members. I’m following his recommendations for v2 of Kinograph. In fact, I’m working on it right now and hope to have a prototype running by the end of the week.

1 Like

The johnarthurkelly post on lighting is great. Makes me feel foolish for not researching the site more before I posted my musings.

I do admit a bit of a question though. I understand the collimated vs. diffuse (Callier effect) points.
But I was a bit lost at:


Why would wide spectrum light make any difference in the captured data as I thought the point is to capture the amount of dye per one of the three colors unless you’re saying the physical color of the dye varies across the exposure density. Is it the fact that the dyes overlap each other on the spectrum density curves and thus the correct amount of density of a particular color is incorrectly sensed by the image sensor? If that is so then wouldn’t using a B&W image sensor with sequential R,G, B flashes then combining them in the computer solve that problem? Although it would seem the problem would still exist for if the dye adsorption spectrum is overlapping, then the center narrow LED light spectrum will still be affected with the additional adsorption of the adjacent dye color. Does that effect somehow get ameliorated with the use of a uniform white light source? But I don’t understand how at this point.

Maybe it would make more sense if the sensor had the exact same spectral sensitivity as the human eye. Then it would be responding as the eye would and I assume color corrections could be made to correct the RGB channels on a pixel by pixel basis to make a color that the eye would see on the monitor that causes the eye to respond the way it would have to the film. But can’t that same correction be extracted from the narrow RGB images? It’s just a different transform matrix. Not necessarily simple, but doable with modern computers. Certainly not as difficult as the 3-D scene renderings being done today.

Please excuse me if these are simplistic questions - I’m still exploring these areas and trying to understand why people have done what they did.