Image Sensor / Optical Components


#21

Does anyone know if there are monochrome cameras that cover a wide spectrum of colors? Most cameras peak at certain wavelengths, and with film fading, those cameras might be blind to certain colors.


#22

Yes, the PGR cameras nearly all come in monochrome variants. You can then take 3 exposures with R, G and B light sources and combine the images to produce a full colour image.
This results in full resolution for each colour channel (unlike bayer mask colour cameras) and also you get some noise reduction through the combining of 3 exposures.


#23

Hi all, in this thread, @Gerhard34 mentions the relationship between LEDs and sensor type (CMOS vs. CCD). Thought I’d start a conversation surrounding those two types of sensors so we can track our options.

CMOS is quicker, has better low-light performance, and is more power efficient. For this reason, it is the favorite for consumer electronics and DSLRs that need lots of performance. Each pixel is its own circuit and outputs a digital signal.

CCD outputs an overall analog signal for the entire sensor and is therefore not as fast. It is, however, said to be better at capturing accurate color, although many articles I found suggest that due to the ubiquity of CMOS, it is now as good or better.

Really, it will come down to cost vs performance for the Kinograph, since our goal is to keep things affordable and to use available technology to keep things accessible. I thought we could start a conversation, though. So…go for it!


#24

Also, I have reserved a spot on the waiting list for an Axiom Beta camera. This project promises a very high performance camera for the price, and its technical specifications match our needs. Specifically:

  • global shutter
  • high speed capture
  • external trigger
  • build-in storage capabilities, also expandable to external drives, PC
  • totally customizable and open source

As their development progresses, I will be making payments to them. When complete, I will have a Beta to play with at roughly the same cost as a DSLR but with all the things Kinograph needs out of the box. Will post updates as they happen.


#25

Looks good, how fast is the external trigger? i.e. can it external trigger and sync at 24fps or faster?
The 10 stops in linear mode is okay.


#26

@Peter, not sure on the trigger time. I suspect because it is built to do high-rate 4K video that we should be able to get 24fps or very very close.

When you say 10-stops is “okay,” do you mean it’s acceptable, or do you think it’s actually not good enough? I’ve seen cameras like the Red Scarlet offer 16+ stops but it’s well out of reach financially for most people.

What’s a “good” or “great” expectation for dynamic range?

M


#27

Anything less than 12 will mean that it will struggle with dense film stocks.


#28

It appears that a machine camera or microscope cameras can be purchased for a very low price, but some questions are still outstanding.
Your comments Matt, or anyone else!

  1. It appears that the most microscope cameras have lens built on the units, some units costing less then $20. is this a viable solution for super 8 fiom transfer on par with filming movie projection on the wall, or are better results possible?
  2. It appears that capture software is included with these cameras. Is the software sufficient for the purpose intended. Do products like virtualdub, ffmpeg and other similar products provide all the features necessary. or is it necessary to pay the price for better software.
    I am interested in low cost, but reasonable quality, so any comments or suggestions are welcome. I have a Olympus E-PL1 camera with lens capable of full sensor overlap on 8mm super 8 film, and either continuous or “run - pause - run” transport control with arduino and stepper motor drive. However, the shutter wear possibility is not very encouraging.
    Douse the this issue also effect microscope or machine cameras?
    Tuneturkey

#29

Hard to tell what works without trying it. I usually buy what I can afford and try it out. I will start posting results as I get to some tests with optics. Right now I’m working on machine control code.

Capture software can be tricky. Often they are not open source and do not allow you to make adjustments. If the software does exactly what you want, then great! If not, then you’re stuck. I’m not exactly sure what cameras you’re referring to when you say “these cameras,” though so maybe my answer is not very helpful.

In a perfect world, we would find someone who can help us hack the firmware on an existing camera, something like Magic Lantern for Canon. In Live View (cinema mode), all moving parts inside the camera are non-operational and the sensor is still fully functional. I would love to be able to use that mode to get a series of high-res stills, even raw. I know Magic Lantern has a mode that allows for full raw video capture. Perhaps we can add spikes to the soundtrack to mark frames and extract them in post. I will try to do something like that in the coming months.

If, in the meantime, someone else would like to give it a test - you are most welcome. Report back!

M


#30

If you want to use a DSLR, get one that has HDMI output on the camera, then you can use a relatively cheap HDMI capture board (like the BMD Intensity) to capture a frame at a time from the HDMI feed.

The cheap microscope cameras are a waste of time and money, I have dozens of them here from old experiments. The most inexpensive camera that will give you HD resolution at decent quality (i.e. better than filming off the wall) at 24fps is the PGR Blackfly. $495 giving 1920 x 1200 at 41FPS
https://www.ptgrey.com/blackfly-23-mp-color-usb3-vision-sony-pregius-imx249

You also then do not require an interface board. The $295 Chameleon cameras could be used if you can live with less than HD resolution 1288x964. These give a stunning, true, uncompressed 12bit image with great dynamic range and low noise and can do 30FPS.

If you are looking for a standard definition capture method, then there are cheaper alternatives out there, but the really cheap stuff is noisy with terrible dynamic range, slow frame-rates and compressed images, and aren’t worth the time really.


#31

The $295 camera sounds interesting, especially for 8mm home movies. I always just wonder what the curve of diminishing returns looks like when it comes to capturing them. It’s one thing to want full hd, 2k or 4k when talking 35mm taken recently, but how much benefit would there be in those setups trying to capture 8mm film, likely shot in less than ideal conditions by complete amateurs 50 years ago?

Would a $295 camera look just as good as a $600 or $6000 camera given the source material?


#32

Thanks, Peter. One important advantage in going the monochrome route would also be the ability to have more than 3 channel, and the opportunity to change the colors of the light sources according to the film being scanned. That would be very helpful when scanning non-standard material (for example negatives, faded film, hand colored films, experimental films where the artist drew directly onto the film, tinted film, specialist film stocks, …).


#33

Yes, all of the professional level scanners we build are monochrome sensors, and we do a separate Red, Green, Blue and Infrared pass (to capture a damage-matte). You then have to be able to register and combine the images fo course, to make a single colour RGB image with the damage matte in the alpha channel.

If using a bayer sensor, then you can still use a tri-LED light source, and adjust the light-mix accordingly for different film stocks, or to adjust for fade etc. We only use tri-LED light sources whether for Bayer or Mono sensors, but with the mono sensor, you have to be able to have the film stationary for the 3 (or 4) captures, whereas with a bayer you can have continous motion film, and trigger the LEDs to flash at a fast enough rate to ‘freeze’ the single frame for capture.
Either way you get full control over the light-mix for capturing.


#34

All of the PGR cameras I mentioned are available in mono or bayer-colour.

To answer Digitap, the $295 camera does an excellent job, and is generally more than enough for Super8 home movies, I personally would move up to the $495 camera, but you are right, depending on the source you hit diminishing returns pretty quickly, and the $295 model gives extremely good pictures, and they scale up to HD resolution very nicely for viewing on current display devices.
I like to capture in higher resolutions so I can apply stabilisation and cropping and have it all happen at better than HD so that I can then scale down the result to HD, but realistically for most ‘family memories’ a 1288 x 964 capture scaled to 720P for final delivery looks amazing.


#35

You just blew my mind, @Peter. I’m thinking of doing tests with that and/or the Blackfly. Been getting some interesting results with my 8mm setup using Raspberry Pi, but the image quality is nothing to scream about.

In general, my only problem with non-line-scanner cameras is figuring out a way to rebuild intermittent motion in a way that is reliable and cost-effective enough to make sense if we have to make small batches of them. I’ve looked at geneva gears, tore apart 10+ projectors, and in every case we’re looking at high tolerance parts that are hard to make.

I did find one cheap option, but it only works on film that is in good shape.

Of course, this issue it moot if speed is not a concern. You can move the film slow enough that blur isn’t an issue and then there is no need for intermittent motion.

Does anyone have any bright ideas on how to solve the intermittent motion problem without gutting/using old projectors? Or is that really our best shot at an affordable machine??

M


#36

@Matthew “Does anyone have any bright ideas on how to solve the intermittent motion problem without gutting/using old projectors?”

There is, quite understandably, a lot of misconception surrounding the subject of home-made gate/claw mechanisms. It is instinctive to think that extremely accurate and high tolerance engineering is necessary (especially for the minuscule 8mm format), but that is not actually so. If it was, low-cost 8mm movie cameras and projectors of yesteryear could not have been produced.

Years ago, I was involved in the design of 35mm animation cameras for use on multi-plane rostrums, but have not considered this technology for many years - until recently when I inherited a large collection of films made by my father and his uncle, dating back to the early 1900’s. These comprise just about every format that exists up to 16mm, including the earliest that were hand-perforated.

I do not have projectors that will handle all these formats, and besides I know that many of the films are very fragile (not nitrate, fortunately) and may not survive the brutality of a projector. That is when I discovered the Kinograph website which I found very interesting. However, I can see that it needs a relatively expensive live-mode camera in order to capture the frame within a very few milliseconds, which I do not have. There are a few webcams out here that have full 1080 HD resolution (for stills), and excellent optics (which can be easily converted in seconds to macro), but their image chips are scanned (unlike DSLRs), which makes them unsuitable for moving subjects.

So, I revisited my past, and investigated if it was feasible to make a mechanism with my 3D printer (Ultimaker2) that would meet all my requirements. This time, I did a complete analysis of my earlier work, including a worst-case tolerance assessment. I was also able to adapt the design for 3D printing, which can produce parts that are difficult to make using traditional machining methods.

The result is illustrated in this animated image.

The dimensions are generic, and do not relate to any particular film gauge but, although some parts are quite small (especially for 8mm) they can easily be printed in ABS, nylon or carbon-fibre to the required tolerance.

For example, for Std8mm, the mechanism could have a pull-down of 4mm, with a claw-height of 0.85mm, given a tolerance of +/- 0.1mm. For larger gauges, this tolerance could be relaxed.

I have produced an Excel spreadsheet, which explains how claw mechanisms work, the relationship between pull-down and claw-height, and tables which show the trade-off between each of these parameters, for variable tolerances. There are pages that detail 9.5mm, Std8, Super8, and 16mm. I do not want to post that here, but if anyone has a genuine interest in it, then I am willing to email them a copy. My address is support@vitalsparks.com.

Jeff


#37

Amazing, @VitalSparks! Thanks so much for the handy Gif, too. Do you have any STL/OBJ files you can post for people to try out?

Sending you an email about the spreadsheet now…


#38

If you don’t want to the hassle of an intermittent drive, then just use an LED ‘flash’ light source. It can easily run at 24fps or faster with no blur, the Muller works exactly that way.
You need an array of Red, Green and Blue LEDs to get enough light, and some large-ish capacitors to make sure you have enough juice to flash them reliably, but it works perfectly and the film can remain in constant motion. The only tricky part is you need a sprocket sensor to reliably trigger the ‘flash’ at the correct time.
Frank Vine manufactures a triggered light source that works perfectly for this, (he also designed the one in the Muller scanner) and it isn’t all that difficult to make your own, though it would probably make more sense to pay Frank to design one for you.
The PGR cameras I have mentioned in the other threads, (that start at a few hundred dollars), are global shutter and work perfectly with this method. You are right that rolling shutter cameras ar not a good choice for this kind of work.

A big part of the problem of intermittent drives is that they can damage fragile film, and that the mechanism itself is prone to wear, it has to cycle 170,000 times for every two hours of film, and you also have to have loops of film to cater for the movement itself to stop the film getting damaged or putting pressure on the drive system.
The low cost projectors of yesteryear are notoriously hard on film. The cheap cameras only handled the film once, and only every brand-new perfect film, so they could be quite tough on film and get away with it.

A flashscan system avoids those issues, but introduces the need to trigger via sprockets.


#39

@matthewepler

Sorry, don’t have anything ready for publishing yet, because the final design is still being iterated, and I’m working with 9.5mm, which may not be of any use to many people.

However, if anyone with a 3D printer is used to designing their own things with CAD, then here is the most essential snippet of information that is the heart of the design - the rest is common sense. This relates to that weird-shaped cam. It took me many hours to work that out, but in the end I got it!. I have found out since that the shape has been used for centuries for countless other things. It is called a Reuleaux Triangle, a rounded triangle, or a ‘constant width triangle’ - the latter giving a clue to how I am using it.

I was a bit miffed when I found those darned clever Romans got it first, but I bet they didn’t use it for a telecine rig. Or did they…

Here’s how to draw it -

Draw an equal-sided triangle, with sides exactly equal to the pull-down dimension you want (not the spacing of the sprocket holes - my spreadsheet details why). At each point of the triangle, draw an arc that bisects the other two points - that’s it, there’s the shape. To use it, rotate it about one of it’s points as I show in the gif I posted earlier. The rest you can work out for yourself by watching the gif.

Hope that helps.
Jeff

@Peter
I hear what you say about intermittent gates, and badly designed ones are bad news. The one I am working on is just about as gentle as it gets. The claw does not touch the hole when it enters, but when it does, it has a ‘soft’ grab before it pulls down. Plastic parts are very durable, especially nylon which is very tough but ‘giving’ and is self-lubricating. Look inside most projectors and you will find loads of nylon parts.

Now, here’s a thing… Just today I started working on a new idea for intermittent frame transport, that does not use a claw, can tolerate broken socket holes, and theoretically can position the frame within 5microns. I have also worked out a method of triggering the capture device directly from the sprocket holes, even if they are damaged, or missing. The whole thing can be 3D printed in about 4 hours. I estimate that for 8mm, the frame rate would be about 3-5 fps, just right for my LifeCamStudio webcam.

When I have demonstrated a working model to myself, I will post details.

Jeff


#40

We have used a similar cam design in our own scanner designs in the past, we got the idea from the wankel engine utilised in Mazda rotary-engines, like the old RX-7, and it works well.

You absolutely can make fairly gentle pull-down mechanisms, but many film owners will not allow their films to be scanned on anything using one, so it is something to be aware of. All the stuff we have built in the last five years or so has utilised roller-driven, constant velocity systems to avoid the full down, with the exception of the mono sensor equipment, which operates at low-speed and the films are fully inspected and repaired if need be before scanning. Re reliability, yes nylon parts last quite a long time, it was more pointing out to any DIY people that anything you build has to be reliable and accurate over a huge amount of cycles, and wear can cause problems when the targets are as small as sprocket holes and you are running at speed, especially when film may have shrunk from its original spec.

The stuff you are working on sounds great, and it would be wonderful to have something that accurate that works on broken sprockets as well. In the past we have gotten around that by having large diameter wheels so that a missing sprocket or two is automatically corrected by the surrounding sprocket holes being engaged, I’m keen to see better solutions come out of the community.