I'm making my own film scanner and will attempt to sell it

[MattKing]

Panasonic, Olympus, and others -- but M4/3 (AKA, MFT) is digital.

Yes, but it is essentially the same as 110 film. Not that I see a large market for scanning 110.

 

[MattKing]

This must be pretty close to the truth, if it's not actually more than that. Focusing on 135 at least initially is a very reasonable choice.

One comment about the continuous light source that @MattKing mentioned: please, look into this without relying on what "some guy said", even if that guy happens to be Matt whom I certainly respect and appreciate. It's doubtful if a continuous light source is necessary or even the best solution. In fact, especially with an eye on long-term performance, I doubt it is as it would involve durability issues that you don't have with discontinuous R, G and B illumination. Also in terms of color reproduction/fidelity, it's absolutely not evident that continuous light would be better. Just look at the crossover between spectral density curves of the dyes in color film as well as crossover between the sensing elements in your sensor array. I think in practice you would find yourself having to solve much less problems if you acquire as clean as possible an input signal in the first place. Otherwise you'll be confronted with the situation dSLR manufacturers had (and still do) to face in cleaning up an inherently very compromised signal.

I would agree that if you have relatively narrow band R G and B illumination that are each closely matched to the spectral sensitivity of your image sensors, you will have an excellent solution.
But how economical and easy is it to source such illumination?
A continuous spectrum source - incandescent/halogen or electronic flash avoids the problem with poorly matched to the sensor sources, is relatively easy and economic to source, and is proven technology.
If the choice is flash, you can also help prevent problems with vibration.

 

[koraks]

But how economical and easy is it to source such illumination?

LEDs are dirt cheap, omnipresent and as powerful as you'd want them to be. So it's much more economical, efficient and easier to implement than a flash tube, which would also involve high-voltage electronics and thus make matters unnecessarily complex and risky for those who decide to take the screwdriver to the device for repairs/ analysis.

 

[albireo]

I'm not sure if I'm missing something but I really don't understand the need for a fast scan mode if the full scan only takes a few minutes for 36 frames? I think it would take longer to do a 'fast' scan, select what you want to scan again. It's completely redundant and adds complexity to the software that's then completely unnecessary. I think this way of thinking is caused by the such long scan times people currently experience but this scanner will be very different in this regard.

I find I often use the 'preview' mode to make sure I'm getting everything 'right' before proceeding with the actual final scan.

One thing I like to do is to make sure I nail focus and I place it where I want it to be in the frame. This is particularly important if the film is unusually curly and the scanner holder is unable to guarantee planarity across the entirety of the frame. It does happen.

In that case, I will proceed with a fast preview, override the autofocus by placing the Vuescan focus crosshair where I want it to be, make sure I get the framing/crop/orientation right, and then proceed with the actual scan.

Of course, if your planned holder design will be such that it'll guarantee planarity every single time (there are some Coolscan holders using special glass to achieve this, but as I said my magnetic Scharf holders come awfully close), I might certainly do away with the focus override and change my workflow to scan absolutely everything only to then delete 85% of those huge tif files later.

 

[gswdh]

I find I often use the 'preview' mode to make sure I'm getting everything 'right' before proceeding with the actual final scan.

One thing I like to do is to make sure I nail focus and I place it where I want it to be in the frame. This is particularly important if the film is unusually curly and the scanner holder is unable to guarantee planarity across the entirety of the frame. It does happen.

In that case, I will proceed with a fast preview, override the autofocus by placing the Vuescan focus crosshair where I want it to be, make sure I get the framing/crop/orientation right, and then proceed with the actual scan.

Of course, if your planned holder design will be such that it'll guarantee planarity every single time (there are some Coolscan holders using special glass to achieve this, but as I said my magnetic Scharf holders come awfully close), I might certainly do away with the focus override and change my workflow to scan absolutely everything only to then delete 85% of those huge tif files later.

This will be a fed scanner meaning no carriers etc. and hopefully no focusing after the initial setup or none at all (done in the factory) ideally.

With proper profiles for the scanned film and good focusing, there shouldn't be a need for multiple passes. I would only foresee that for poorly or strangely exposed frames that could benefit from unique scanning settings. It might even be possible to do frame by frame preview settings but again this is a software thing.

 

[FotoD]

The properties I've found imortant in 35mm scanners are:
- ability to set and forget, to scan a whole roll with transport
- a good light source that doesn't accentuate grain in BW film
- ability to set exposure manually
- an IR channel for dust removal
- manual focus and autofocus
- 12bit color, 3000+ ppi
- lossless file format output
- Vuescan compatibility

 

[Chan Tran]

If I were to build a scanner I would work on a film transport that can take either the entire roll or strips of 4 to 6 frames and automaticall position the frame accurately and with good flatness. That's all. I would simply use a digital camera to digitize the film. Yes the price of good digital camera and the lens needed is expensive but I think few who scan film don't have a digital camera they use often. So like a computer there is a cost but most would already have it so perhaps don't count as cost.

 

[bfilm]

It seems to me that the purpose of this scanner might make it a prime candidate for a fixed focus build, which might also help with the lens cost. For example, Rodenstock offers high-quality Rodagon enlarging lenses from EUR ~400 retail price and Apo-Rodagon enlarging lenses from EUR ~530 retail price. Maybe if there is something suitable in Rodenstock's repertoire that would not entail significant custom development costs, it would be possible to use a higher-grade lens in a fixed focus build than could be obtained for comparable price in an auto- and/or manual-focus build.

 

[MattKing]

It seems to me that the purpose of this scanner might make it a prime candidate for a fixed focus build, which might also help with the lens cost. For example, Rodenstock offers high-quality Rodagon enlarging lenses from EUR ~400 retail price and Apo-Rodagon enlarging lenses from EUR ~530 retail price. Maybe if there is something suitable in Rodenstock's repertoire that would not entail significant custom development costs, it would be possible to use a higher-grade lens in a fixed focus build than could be obtained for comparable price in an auto- and/or manual-focus build.

How are you going to deal with film that isn't flat?
The lenses you reference may be an appropriate solution - depending on whether or not there are models optimized for the magnification that a scanner would require. That would, of course, depend on the sensor size you intend to employ.
On a more general note, I sense that your interest and focus is mostly in the electronics and software.
I would suggest though that it is the mechanical and engineering challenges, and the production costs relating to the possible solutions, that are the source of the market unavailability of reasonably priced, efficient and effective film scanners

 

[_T_]

Can you publish or point us to the code that implements ICE-like (possibly content aware) healing routines? I've read a bit on the subject (and I use IR channel data with great success in PS), but after all I've read implementing such functionality in a standalone one-man-band software seems like a non-trivial endeavour.

I don’t doubt that any kind of content aware implementation would be a great undertaking on a project like this but I personally haven’t had very good results with systems that attempt to clean up an entire image that way.

Ice in particular tends to leave artifacts on any dust of any shape more complex than a simple straight line or tiny dot. And I find that these artifacts are much more problematic than the dust itself because they’re much more difficult to spot when cleaning up the image after ice has been applied.

In my experience I have found that when attempting to clean up many disparate areas of an image at once a much simpler process like the “dust and scratches” filter is much more successful. The problem being that such filters when applied to an entire image tend to blur the image significantly. Hence I use the IR channel to mask the “dust and scratches” filter so that it only affects the areas which are absolutely certain to contain dust. With this process I find that there is a significant reduction in artifacts over ice.

In terms of healing and content aware editing of the image I would personally never apply such techniques in any way that is not manually, specifically targeted to a particular area.

 

[bfilm]

How are you going to deal with film that isn't flat?
The lenses you reference may be an appropriate solution - depending on whether or not there are models optimized for the magnification that a scanner would require. That would, of course, depend on the sensor size you intend to employ.
On a more general note, I sense that your interest and focus is mostly in the electronics and software.
I would suggest though that it is the mechanical and engineering challenges, and the production costs relating to the possible solutions, that are the source of the market unavailability of reasonably priced, efficient and effective film scanners

My understanding is that there are scanners that successfully use fixed focus. 35mm film is a fairly small piece of film, and it seems like the lens could potentially have enough depth of field to deal with some amount of curl without getting into too much loss of quality from diffraction.

And since this scanner is meant to scan an entire roll of film fairly quickly, it does not seem like the operator would often be focusing each frame individually. Now, as I mentioned earlier, hopefully it would also have the ability to scan shorter strips of film, and sometimes an operator might want to give individual frames more attention and then might appreciate more focus control. But it does seem like fixed focus can work for 35mm film.

I don't know exactly how enlarging lenses compare to the lens needed in a 35mm film scanner, but I used those simply as an example of the potentially lower cost of a simple lens from Rodenstock than the EUR 1000 that @gswdh mentioned. Although, maybe this has already been researched and the EUR 1000 accounts for custom development costs that might be necessary for an appropriate lens.

 

[bfilm]

I would suggest though that it is the mechanical and engineering challenges, and the production costs relating to the possible solutions, that are the source of the market unavailability of reasonably priced, efficient and effective film scanners

It seems to me that if good image quality can be obtained with fixed focus, that using this type of lens and build would significantly reduce the mechanical and engineering complexities and production costs. Although, you would have to be sure to build it very accurately, with close tolerances.

 

[MattKing]

But it does seem like fixed focus can work for 35mm film.

If you use a holder with the film between two pieces of glass.
The depth of field/focus at large magnifications is extremely small, and all film, including 35mm film tends to have a curl to it.
Enlarging lenses are a possibility, but they tend to designed for different magnification than special purpose process lenses.
What size sensor were you thinking would be employed? I'm not familiar with what sensors are easily available, but if you were going to employ a full frame sensor, the magnification that you need to have a lens optimized for would be 1:1 - with that information in mind, you can calculate the focal length of lens to employ, and from that calculate how much depth of field/focus is available. To perform that latter calculation, you have to decide how much out-of-focus blur (circle of confusion) you are willing to accept.

 

[gswdh]

If you use a holder with the film between two pieces of glass.
The depth of field/focus at large magnifications is extremely small, and all film, including 35mm film tends to have a curl to it.
Enlarging lenses are a possibility, but they tend to designed for different magnification than special purpose process lenses.
What size sensor were you thinking would be employed? I'm not familiar with what sensors are easily available, but if you were going to employ a full frame sensor, the magnification that you need to have a lens optimized for would be 1:1 - with that information in mind, you can calculate the focal length of lens to employ, and from that calculate how much depth of field/focus is available. To perform that latter calculation, you have to decide how much out-of-focus blur (circle of confusion) you are willing to accept.

So the curl in the film is easily solved by sending the film around a curve which will straighten it in the vertical direction.

The sensor I’m thinking of using is 28.2mm high so we’ll have a magnification of ~0.85x with the lens ‘backwards’.

From my rough calculations online etc. I think I’ll get around 1mm DOF at f8 with a 50mm lens. The track for the film is 0.5mm wide so should maintain the film in the right place well enough.

The focal length isn’t that important, it will most likely be determined by optimising the performance of the lens for the application. I would prefer as short as possible to keep the device compact.

We know the product is possible, optically, as the Pakon F135 used a fixed focus lens.

 

[koraks]

I think I’ll get around 1mm DOF at f8 with a 50mm lens

No way. You must be working with an unacceptable large CoF. It'll be an order of magnitude less than that, realistically.

I would prefer as short as possible

It'll be a tradeoff between compactness, cost and optical performance. Shorter focal length lenses are more challenging to correct.

I'd strongly advise getting in touch with someone who understands optics. You're facing decisions that have a very fundamental impact on image quality; it would be a shame to make these haphazardly on the basis of layman's knowledge.

 

[gswdh]

No way. You must be working with an unacceptable large CoF. It'll be an order of magnitude less than that, realistically.


It'll be a tradeoff between compactness, cost and optical performance. Shorter focal length lenses are more challenging to correct.

I'd strongly advise getting in touch with someone who understands optics. You're facing decisions that have a very fundamental impact on image quality; it would be a shame to make these haphazardly on the basis of layman's knowledge.

Sure, let’s see through some experimentation. I have sent enquires to about 10 optical manufacturers, hoping to hear back in the new year.

 

[MattKing]

So the curl in the film is easily solved by sending the film around a curve which will straighten it in the vertical direction.

35mm film curls in both directions.
And the curl varies with film and circumstances.

 

[MattKing]

For clarity, I wish you every success.
I'd never be in your target market at your proposed price. But there might be small labs around here who would be.

 

[mshchem]

Simple question. Do you intend to use injection molding for parts? I can't imagine how you can attain the volume required to make your product affordable and profitable.
Rather than all the technology discussions.
Ask, how much people are willing to pay for such a device.

 

[gswdh]

35mm film curls in both directions.
And the curl varies with film and circumstances.

It does but it can only do one at a time.

 

[gswdh]

For clarity, I wish you every success.
I'd never be in your target market at your proposed price. But there might be small labs around here who would be.

Thank you, I do appreciate it. You never know maybe it’s possible to bring the price down.

 

[gswdh]

Simple question. Do you intend to use injection molding for parts? I can't imagine how you can attain the volume required to make your product affordable and profitable.
Rather than all the technology discussions.
Ask, how much people are willing to pay for such a device.

Yes the product will use injection moulded parts. For shorted runs, lower cost aluminium moulds can be used. Injection moulding has a reputation for a crazy upfront cost however there’s many providers in China that have great support with very reasonable tooling costs nowadays. I have designed, sourced and had injection moulded parts made in 1000s for three products in my career now. Also, if you’re really smart with the design of the parts, you can really minimise the mould complexity and thus the tooling and per part cost.

 

[mshchem]

Yes the product will use injection moulded parts. For shorted runs, lower cost aluminium moulds can be used. Injection moulding has a reputation for a crazy upfront cost however there’s many providers in China that have great support with very reasonable tooling costs nowadays. I have designed, sourced and had injection moulded parts made in 1000s for three products in my career now. Also, if you’re really smart with the design of the parts, you can really minimise the mould complexity and thus the tooling and per part cost.

I've had some experience with molding too. Aluminum tooling is fine for prototype parts, my experience is with volume production, aluminum is problematic. Seems like all tooling comes from China these days, many times, maybe most times, these tools require local rework. Of course mostly the tooling stays in low cost countries for production.

Will this product require certification CSA, UL etc?

 

[wiltw]

LEDs are dirt cheap, omnipresent and as powerful as you'd want them to be. So it's much more economical, efficient and easier to implement than a flash tube, which would also involve high-voltage electronics and thus make matters unnecessarily complex and risky for those who decide to take the screwdriver to the device for repairs/ analysis.

But how can one ASSUME that the spectrum of light from any given LED source is truly as continuous as electronic flash or incandescent source?
If you start with Brand A model 1 LED, and then have to switch to Brand B model 3, will they MATCH in their spectral characteristics (in both continuity of spectrum, and in output at different points in the spectrum?

NOT saying it 'cannot' be done, but it certainly is a consideration that has to be factored into the design, unless one is intentionally offering a throw-away product that has zero repairability.

 

[albada]

So the curl in the film is easily solved by sending the film around a curve which will straighten it in the vertical direction.

Agreed!

There are two ways of scanning 35mm film:

1. Keep film stationary in a holder, and move the light+sensor combination.
2. Keep light+sensor stationary, and move the film.

You are correct that wrapping the film over two rollers removes curl. In fact, that removes curl in both directions.

In my prior job, I worked on the scanner team for HP all-in-one printers. We had a film-scanner project, which sadly was cancelled, but a prototype film-transport mechanism was made, which was given to me and I still have. Here it is:

It used method 2 above: moving the film. This mechanism could transport both film and slides. For film, two little spring-loaded rollers pressed the film onto the large rollers with orange rubber, wrapping it around the roller a little, removing curl from the film. Slides pushed those little spring rollers up, so slides passed straight through. I can post more photos of this mech from different angles if you're interested.

Mark