Jump to content

DIY Film Scanner (With Samples)


Josh Gladstone

Recommended Posts

Why go through all that trouble and then use an 8bit camera? I'd look at something like this instead: $700 and it's a 10 or 12 bit 5.5k camera using a pretty nice Sony sensor: https://www.ptgrey.com/blackfly-s-color-200-mp-usb3-vision-sony-imx183 you can get a 2k version for about $500, but honestly, I'd go for the full 5.5k if you're using the color bayer version, and then downsample to 2k. You'll basically get true RGB color sampling that way, negating the (arguably inconsequential) effects of the bayer filter. If your design uses the intermittent motion of the projector, the rolling shutter shouldn't matter.

 

What are you doing for a light source? You'll want to run something cooler than a traditional bulb, at those slow frame rates, or you'll burn up the film.

 

Thanks Perry,

 

Well, that's why I was asking, as I am just learning about this and am unfamiliar with bits and color depth and such, but I was up until 1AM reading up on it and discovered I should at least use 10 bit, maybe more. I think I will still pick up this camera to give me something to play with as it's cheap and comes with a Pentax lens. Don't have much money to spend right now, lots of time though, but am searching for a better camera. And funny thing, Point Grey is only about 300km from here, but is now owned by another company. I called them asking if they had any "old" or returned inventory, still waiting for a call back. :unsure:

 

I am planning on slow continuous motion, maybe 2-4 fps, I won't know until I get it assembled, but I have a plethora of goodies on the way with the Arduino, optical sensors, hall effect sensors, PWM motor drivers, several different RGB square LED arrays, some single white LEDs with different color temp, and a few other things. The first thing I removed was that 500W lamp, holy crap that thing is a beast. I also removed the condenser lens which focused the light on the film gate, but left in the heat shield glass hoping it may work as an IR filter. In the slot of that removed lens, I placed several stacked sheets of diffusion plastic sheets obtained from an old LCD monitor, cool stuff, it really makes my temporary 1W LED light source pretty decent.

 

I could go on and on, but will save it for the next post.

Thanks for the input, really having fun with this and looking forward to learning Arduino code, I ordered 3 boards as I know I will find uses for them around our property.

 

Cheers

post-76079-0-68102200-1548174094_thumb.jpg

Link to comment
Share on other sites

The Sony sensors are definitely worth looking at - Point Grey is owned by Flir now, but there are other manufacturers using these sensors as well. Sony is killing it when it comes to sensor tech right now: these have crazy dynamic range and almost no noise, and the range of models is really impressive. The IMX183 in the 5k camera I linked to above uses a rolling shutter, which might work for your continuous motion scanner, if the speed is slow enough. But if you're not using the intermittent motion of the projector, you may need to look into a global shutter camera, which is often a bit more expensive. Without that you might get weird motion artifacts because of the motion of the film: https://www.google.com/search?q=rolling+shutter+artifacts&newwindow=1&safe=off&source=lnms&tbm=isch&sa=X&ved=0ahUKEwjS_73k7YHgAhWHm-AKHTW4DHMQ_AUIDigB&biw=1550&bih=1252

Link to comment
Share on other sites

Okay thank you again, really good info. I am going to look into the difference between rolling and global shutter as I know nothing about this yet. As for my projector method, I can certainly make it stop-start, it's easy with the Arduino and a sensor on the wheel, if that is what I need to do to make it work. Currently I have about 15 or so 400' reels to record of standard 8, then I will be getting my Beaulieu 4008ZM4 and 5008 out of the closet, rebuild my darkroom and start shooting and developing my own film again...it's been over 30 years. Will be fun, but that is another subject.

 

Cheers

Link to comment
Share on other sites

Well shoot, I may be stuck with frame by frame, slow mode, as I just discovered the 3 hole shutter is an integral part of the projector, it is also the drive pulley for the main shaft, so can't be removed. It will make capturing a frame a bit of a challenge for a moving film stock.

 

More later.

Link to comment
Share on other sites

I would leave the intermittent motion in place. You could probably use the shutter blade and a sensor to trigger an exposure on the camera. At 2-3fps, you're leaving the film in place long enough to get a good shot with a rolling shutter that doesn't result in artifacts. It seems easier to me to do that than to try to time it with the film in constant motion. Not that it cant' be done, but I'd let the existing mechanics work for you!

Link to comment
Share on other sites

I would leave the intermittent motion in place. You could probably use the shutter blade and a sensor to trigger an exposure on the camera. At 2-3fps, you're leaving the film in place long enough to get a good shot with a rolling shutter that doesn't result in artifacts. It seems easier to me to do that than to try to time it with the film in constant motion. Not that it cant' be done, but I'd let the existing mechanics work for you!

 

Mmm, I think I may have lead you astray when I said continuous motion. What I meant was the motor would be running continuously to operate at about 2-3 fps, but with all the original mechanics in place, so yes the sprockets still pulling the film through, the pin moving the film through the gate with it's 3 hole shutter and inherit pause, so yes, I will leave it all intact and continue on with my original plans. I was planning on either hall sensor with small magnet on the shutter wheel, or light sensor shining thru holes in the shutter wheel, either way I have both parts on order, be here Thursday.

 

Cheers

Link to comment
Share on other sites

If you can fit one into the projector somewhere, take a look at these: http://www.ia.omron.com/products/category/sensors/photomicro-sensors/slot-type/index.html

 

They're very inexpensive and simple to use. Just put it in a place where the shutter blade just triggers it (or stops triggering it, however you want to set it up), when the shutter is open at the gate. They're really inexpensive too - like $3-$4 USD. It's basically just an optical switch. With something like this you don't need to modify the shutter to make holes, you can just let the leading or trailing edge of the shutter blade be the trigger.

Link to comment
Share on other sites

Woo Hooo! Look what I found.

 

http://www.adept.net.au/cameras/avt/pdf/OSCAR_F_510C.pdf

 

TONS of features, many aren't listed on the poop sheet. It's a bit dated, discontinued 2014, but I think it will work quite well for my application. I downloaded all the software for it and user manuals, should be here in a few weeks. Now I need to find a C mount lens that will work, but first some research to determine what focal length etc I need.

 

What do you think?

 

Cheers

Link to comment
Share on other sites

For me DIY film scanning is about ease, the convenience and getting the most out of the Super 8 or 16mm frame. A while back I decided to scan my Super 8 and 16mm films; at first I did what everyone else seems to do, I adapted a conventional projector, but I quickly realised the limitations of projectors, I realised that the best way to get the most out of new film was to design a new and simple film transport system to advance either Super 8 and 16mm. Mine is a simple standalone machine, I don’t use complex electronics or software, I have used a bright white ‘even’ LED and a normal consumer mirror-less digital camera, which saves each frame as RAW or JPEG, by using a ‘normal’ camera I don’t need to run the computer at the same time. I am in the process of re-working the system to include a filter tray for scanning negative film and am building a new gate and a claw which is kinder to film. It’s a frame by frame system where the claw triggers the camera via a micro-switch and the cameras remote. Currently I am using a Samsung NX Mini, the interesting this is that these little cameras have lens inter-changeability and they have a simple remote which can be easily linked to the micro-switch, I am using a 50mm Schneider enlarger lens. In the next few weeks I shall be scanning some new Kodak 250D which I have used in a Super 16 camera.

 

http://www.lightbreeze.co.uk/Film%20Technology.htm#telecine

 

Pav

Link to comment
Share on other sites

Hello Pav,

That it an interesting system you have built, very different from what I am doing, but achieves the same goal. Cool design.

 

So I need some high level help here, understanding the RGB lighting concept. As mentioned I am going to use the Arduino controller, and am planning on using an RGB LED square array behind some diffusion material, but what I need some help in understanding, is what is the controlling sensor that says hey we need more green, or you reds, sit down a bit. Is it the capture software that sees the image being captured? If so, how does that control the Arduino to dim the reds or bring up the green? I know the Arduino uses USB for programming but is it programmable on the fly? I don't think it works that way does it? I have read thru this thread multiple times and I know Josh shared his code, but I couldn't see where there was any feedback to control the brightness of each color channel. This is my first exposure to programming a ucontroller, but I am an electronics eng of 40 years but it's been all analog and RF. My bits n bytes training was back in 1981-83.

 

So far, my plan is to have the DC motor PWM controlled by Arduino and operating the projector slowly, about 2-3 fps with all original shutters, sprockets and pull down pin, so there should be sufficient time for a good capture at that rate. Planning on triggering the machine vision camera with either Hall sensor (magnet on shutter wheel) or light sensor (LDR or Perry's Omron recommendation) shining thru the rectangular holes in the shutter wheel. So in this design, the Arduino is just supplying a fixed PWM signal for the motor controller and lighting the LED array, the capture is strictly electro-mechanical to the camera which triggers the software to capture a frame.

 

Maybe I could just use a nice daylight LED and let the cameras' auto WB do the work, same for the exposure, would be a lot blinkin' easier. Lot's of experimenting to do when the camera arrives.

 

Anyway, thanks for helping me to understand this lighting feedback loop thing.

 

Cheers

Edited by Paul Robertson
Link to comment
Share on other sites

 

I think what you'll likely find with the Samsung is that it doesn't have the lifespan you need. Even with an electronic shutter, it's a consumer grade sensor in that camera, and it will likely fail after a few hundred thousand frames. Considering a 50' reel of Super 8 is 3600 frames, you're only talking 50-60 reels before you start looking at the end of the camera's usable lifespan.

 

Based on your link, are you making the decision not to use the machine vision camera because of the image quality of the scanned file? Because what I'm seeing from the Point Grey camera is essentially what we see from our ScanStation with a flat scan. It will require a bit of color correction, but generally has more to work with. If you're scanning to JPG on the Samsung, you're only getting an 8 bit image as well. Not sure what the sensor is in that, and what's actually recording with RAW. Do those RAW files open up easily in something like Resolve?

Link to comment
Share on other sites

So I need some high level help here, understanding the RGB lighting concept. As mentioned I am going to use the Arduino controller, and am planning on using an RGB LED square array behind some diffusion material, but what I need some help in understanding, is what is the controlling sensor that says hey we need more green, or you reds, sit down a bit. Is it the capture software that sees the image being captured? If so, how does that control the Arduino to dim the reds or bring up the green?

 

You might be able to do this in hardware with something custom, but the easier way is in software. Assuming you have individual control over each color in the LED, you'd want to calibrate the scanner before scanning: with no film loaded, you'd take an image with only red light, sample some pixels and average their color value, then adjust to get it to where you need it. Then do the same with Green and Blue. This will require knowing what the RGB values need to be, and writing some software to do the analysis. Adjusting the amount of each color channel is a matter of changing the PWM frequency, assuming you're using PWM for that.

 

Take a look at OpenCV. If you're writing your control software (on the computer, not the arduino) in most common programming languages, there's a way to integrate OpenCV, and it will handle most of the basic image manipulation stuff you'd need to do. That said, for something like calibration, you could just do it in your control software, by analyzing captured images.

 

So once you've done this, you know the light coming through is white. Honestly, though, if you're using white light, you might be better off just using a very powerful white LED, doing the same measurements to figure out the exact color balance of the image, and then adjusting that color with a LUT that gets applied to all the captured frames before they're written to disk.

Edited by Perry Paolantonio
Link to comment
Share on other sites

Thanks Perry,

 

That's the sort of info I needed to get my head around lighting. I have OpenCV, but haven't played with it yet. I have written some computer code, years ago, actually ladder logic for PLCs, RS Logix I think, and I have read in this thread what has been done with OpenCV. I was following the recommendation to get the levels close to max for each color so that the CCD sensor is happy and dynamic range is good, before the image is written. I think to start, to at least get moving with this, I will use the white LED source and balance that, then once the system is working continue playing with RGB. I probably shouldn't be jumping in too deep to learn code for too many things without a digital life jacket. :D In the package that is due to arrive today, there are white LED arrays, square, and some single high power LEDs I can mount multiples of to a cct brd. I have a 120vac single high power LED in there now for experimenting with lenses, and produces a nice even illumination with all the diffusion material I have installed.

 

What are your thoughts re: auto white balance in the camera? What about auto exposure? This Allied Vision camera is loaded with many options and variables including LUT, binning, frame buffer etc.

 

Thanks again.

 

Cheers

Edited by Paul Robertson
Link to comment
Share on other sites

One thing to look at is that there are some super bright LEDs you can get now that are about a half inch in diameter - they're a lot simpler than an array. With most machine vision cameras, there's a trigger interface, so you can send a pulse to the camera and have it trigger a preset exposure setting. You can also tie this into your LED, so that when you pulse the LED it sends a signal to the camera, and the camera snaps the image.

 

Also, with a machine vision camera, you should have some software that lets you create a profile that can be uploaded to the camera. So you could do your basic calibration that way, in that software, then load that profile to the camera. So when you snap an image it uses those settings by default unless you override.

 

The camera you chose has an old firewire interface, so it may be a bit funky to work with, but they were pretty common 8-10 years ago. Not as fast as modern interfaces (CameraLink, 10GbE, or even USB3), but it should work. AlliedVision is still around, though you may have a hard time getting support for one that old.

 

As for the settings, you really just need to experiment, because it'll depend on the intensity and color balance of your light source, as well as how much light is impeded by the film once you have that in place. You really do need a crazy bright source of light to get the most out of the film. The f/stop on the lens will also come into play. The camera maker may recommend a specific one, but you may also need to play with that to see what you can do.

Link to comment
Share on other sites

  • Site Sponsor

Cameras are moving pretty fast and Sony has some new ones that are really great, the Pregius line has a new 4K (IMX253) and 6.5K (IMX342) which have incredible DR (71Db) and are fast and inexpensive global shutter cameras.

 

https://www.imperx.com/cmos-cameras/c4120/

 

The 253 runs about $3300 and is being used by LaserGrpahics to replace the noisy 5K CMOSIS camera and others we just got one for one of the DCS Xena machines.

Link to comment
Share on other sites

Take a look at this sort of thing for your light source: http://www.cob-led.com/What-is-cob-LED-chips-on-board.html

 

Well that is very cool. I just looked at Digikey and they have about 500 different ones in stock, various shades of white. So with your experience, what color temp should I pick, an ideal, and then a range in case the ideal isn't in stock? Then what size? Enough to cover the entire gate opening and then some I would imagine? And a CRI minimum of xx?

 

Box of Arduino stuff and sensors and LEDs are out for delivery, be here soon, and Digikey can deliver for tomorrow if I get my order in soon. I can play all weekend with this.

 

Cheers :D :lol:

Edited by Paul Robertson
Link to comment
Share on other sites

I haven't looked too closely at these. High CRI (high 90s if you can find them) is best, because you know you'll get more consistent results. You can also get RGB models that let you control the channels separately, but in one neat package. You can then mix your own white from that.

 

As for color temp, it probably doesn't matter a ton as long as you're calibrating your camera to it and there's enough light to get through the densest film.

 

With white light and enough diffusion you can just point the light at the film and stick the diffusion between, then with the camera look for hot spots (adjust the exposure time way down and stop the lens down as well. This will reveal hot spots. If you see these, then maybe use an integrating sphere of some kind - could be the mirror that was used int he projector, or something similar - A box with mirrored surfaces on the inside works - the light bounces around in there and you get better diffusion. Also look into holographic diffusers (which may be easier) - you can get a small piece of it online and use it in line with your regular photo diffusion material.

 

Link to comment
Share on other sites

OKay cool, well there are about 20 with a CRI of 97, and anywhere from 500mA to 4A and 1000-7000Lm, nice neat rectangular packages, easy to mount. Some of them are labeled as "3 level Mac Adam Eclipse", no idea what that means. I have 2 different frosted diffuser sheets, and one sheet, which I think is holographic as when I hold it up to a light, I get a funky 3D prismatic effect, neat stuff. I also have the original little 1" condenser lens from the projector which I could re-install if need be.

 

So you have given me lots to work with, thank you. I'll get them ordered and start playing after work today.

 

Cheers

Link to comment
Share on other sites

  • Premium Member

Macadam ellipses are shapes on a CIE 1931 diagram which demonstrate the size of the area in which the average human being can't tell the difference between colours.

 

Two things usually strike people about them: first, they're quite big, and a lot of the colour we regard as hugely important in film and TV work is represented by differences nobody could ever actually see. Second, they're very, very elliptical. You'd sort of expect them to be circular, given the amount of colour change required to become visible ought to be the same whether you drift towards red, green or blue. But it isn't. That's mainly because the CIE 1931 diagram is really nonlinear with respect to the human perception of colour. Small distances in one part of the chart look, to us, like bigger colour differences than similar distances in other parts of the chart.

 

This also leads to the realisation that the huge amount of greens left out by colourspaces like Rec. 709 actually aren't as huge as they look.

 

The chart looks like this:

 

325px-CIExy1931_MacAdam.png

Link to comment
Share on other sites

WoW! Phil, that is amazing, and quite amazing how such variance can be detected, or not detected by human vision. It's like audio volume, takes almost 3dB before we can tell there is a change, even though that is a doubling, or halving of power.

 

This is such a fun learning experience for me, thanks everyone for the help.

 

Arduino package arrived today...will be playing this weekend...after I finish rebuilding a computer for a friend's business..can you believe I am reinstalling WIN98 on a computer...yup still have the install disks, yikes! never throw anything away. :rolleyes:

 

Cheers

Edited by Paul Robertson
Link to comment
Share on other sites

 

I think what you'll likely find with the Samsung is that it doesn't have the lifespan you need. Even with an electronic shutter, it's a consumer grade sensor in that camera, and it will likely fail after a few hundred thousand frames. Considering a 50' reel of Super 8 is 3600 frames, you're only talking 50-60 reels before you start looking at the end of the camera's usable lifespan.

 

Based on your link, are you making the decision not to use the machine vision camera because of the image quality of the scanned file? Because what I'm seeing from the Point Grey camera is essentially what we see from our ScanStation with a flat scan. It will require a bit of color correction, but generally has more to work with. If you're scanning to JPG on the Samsung, you're only getting an 8 bit image as well. Not sure what the sensor is in that, and what's actually recording with RAW. Do those RAW files open up easily in something like Resolve?

 

Firstly, I must point out that for me the DIY method is not a substitute for professional transfer facilities and their scanning systems. I have created a something that produces great results, it’s easy to use where I can quickly check my films after processing. I totally agree with everything you say. I have scanned thousands of feet of 16mm and Super 8 and have used about three to four cameras over the past five years, this is mainly because each successive camera is better. I think DIY set ups should offer a certain ease and with the right care acceptable quality can be achieved, I have found the quality from consumer digital cameras pretty decent, most importantly they are cheap, quite disposable and quite easy for this work and best of all they allow us to have a standalone system. I have used machine vision cameras in the past, but I couldn’t get quality images with ease.

 

Pav

 

Edited by Pavan Deep
Link to comment
Share on other sites

Well the COB LED is amazing! Crap is it bright. 10W of 4000K light, and it is installed in the projector, complete with large heatsink and cooling fan, should have left the big blower in it. :rolleyes: With 4 layers of diffusion material in front of it, there is a nice even illuminated panel behind the film gate.

 

Going to be playing with Arduino and some software to see how I can control the brightness with feedback from the capture software, should be fun.

post-76079-0-86686700-1548609516_thumb.jpg

Edited by Paul Robertson
  • Like 1
  • Upvote 1
Link to comment
Share on other sites

One more pic. The light and fan are currently wired to the original 3 position switch, which in the end will be for setting up the camera and exposure etc., then it will switch over to PWM computer control for brightness adjustment. That's the theory anyway.

 

Cheers

post-76079-0-30569800-1548609722_thumb.jpg

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...