Designing a new film scanner; need your help

[jkjod]

Pakon 135+ with bigger files and 120 capability.

I agree with this - speed is pretty important to me. If there was a newer Pakon 135+ that allowed for slightly larger files and continued service, I'd happily buy one.

 

[Roger Hicks]

The distinction was one that was very consistently enforced in all of Kodak's publications and product descriptions. They even rejected "slide film" in favor of the more proper "film for slides" for their 35mm reversal films as the film itself didn't produce "slide", only "transparencies", though it would typically be mounted as slides.

One manufacturer's house style is worthless. "Native English speaker" is not the same as "native Kodak speaker" or even "native American speaker".

Cheers,

R.

 

[Roger Hicks]

There's a big conflict here between speed and convenience on the one hand, and quality on the other. My own vote is for maximum quality from the few trannies I choose to scan, rather than maximum speed: there are already plenty of ways of getting quick'n'dirty low-res scans from flatbeds, magazine-fed scanners, etc.

To the OP: decide whether you want the low-res market or printable pics.

Cheers,

R.

 

[x-ray]

Isn't it better to design a high precision x-y movement rig which can mount a camera and macro lens and take multiple pictures to stitch them together?

Plus software that would process the resulting images and create "scans" out of them.

Please excuse me if this sounds naive, I'm just curious, and not technologically sophisticated.

I use a Pakon scanner today, and the software's ability to nail white balance in any lighting situation more than makes up for the lower resolution scans.

I think I suggested something like this several posts back.

I'm with you, why reinvent the wheel. For $1000-3000 you're not going to get anything close to what a medium priced or even intro level DSLR will do in resolution, dynamic range and optical quality.

I owned and regularly used a Fuji Lanovia Quattro pre press scanner and currently own an Imacon 848. Both are comparable in output quality. The Fuji was a much more sophisticated scanner and designed to scan film or reflective up to 13x18 inches and the 848 will do both up to 5x7.

Professional scanners that deliver top quality are super expensive for a reason. The fuji weighed 150 pounds for a reason. It was to reduce detail killing vibration. the Imacon weighs about 50 - 60 pounds. The earlier version of the Fuji scanner I had weighed nearly 350 pounds and had a cast iron frame inside. Optics on both the Fuji and Imagon are top notch. Imacon I believe uses an apo Rodenstocl lens and the Fuji had 4 of Fuji's finest apo process lenses unlike consumer grade scanners like the v750 with plastic lenses. The Fuji auto focused up to 10mm depth on the fly and the Imacon pre focuses on every image. Epsons don't focus. Image noise levels are very low with both machines compared to the Epson.

The Fuji was over $30,000 with no accessories and the new Imacon I believe is $16,000 up depending on model and accessories. There's a reason for this price. For $1000-2000 you can't get good lenses and can't get an image system even close to what your medium level DSLR will do. I copied B&W negs with my Hasselblad digital and Macro Planar for several years with results comparable to the Fuji scanner.

Spend your time on developing a system to use existing cameras and lenses. This way when technology changes it's easy to update. You aren't redesigning imaging hardware only a mechanical transport and alignment system.

 

[Bille]

Spend your time on developing a system to use existing cameras and lenses.

You mean a light table and a repro stand?

 

[x-ray]

That's basically it. Simple enough. When I've done this I use black tape or paper to mask out stray light around the film.

The one change that might be incorporated is to make a jig where the light table could be moved into different positions to allow stitching of segments. Easy enough to do and would be cheap.

 

[Ranchu]

So how do you do C41 with this kind of setup, X ray? Do you get good skintones?

😉

 

[quejai]

Thanks for all the ideas, everyone - lots of things to think about.

Although I'd like to respond to each question/suggestion individually, I just haven't got enough time. But there are some more popular questions that should be clarified:
- This scanner prioritises results to convenience, but honestly this doesn't mean that it will be slow or hard to use.
- XY rig: That was actually the setup for the first prototype. To save on space, and to experiment with an autoload feature, the X-dimension was actually a conveyor belt made of transparrent plastic, with the film in a wet-mounted sandwich between. Even with continual focusing, it just didn't work out. Since that idea has been tested and crossed off, I'm comfortable mentioning it - but the same doesn't go for the current design. You'll see how that works when the kickstarter begins.

So long story short: The current design will have a prototype, and there are exciting times ahead.

 

[x-ray]

So how do you do C41 with this kind of setup, X ray? Do you get good skintones?

😉

I rarely shoot C-41 and when I have it's usually in my Holga. Most of what I've done this way is B&W large format.

I wouldn't think it would be too difficult to create a photoshop action to invert and neutralize the mask in color neg. A plugin for lightroom or stand alone app seems easier than designing an entire system including software.

I use a software to create profiles from a Macbeth chart for each of my cameras. When I edit in Lightroom I apply that profile. It seems reasonable that such a program could be written to analyze one frame of a roll with a Macbeth chart and create a profile to neutralize the Orange mask and set a relative color balance. Sound reasonable? Each emulsion type would be different but you can do it once and pretty much forget it. Of course different lighting will require adjustments for correct color balance just as it does on a scanner. That's basically what's been done with scanner profiles.

 

[kiemchacsu]

That's basically it. Simple enough. When I've done this I use black tape or paper to mask out stray light around the film.

The one change that might be incorporated is to make a jig where the light table could be moved into different positions to allow stitching of segments. Easy enough to do and would be cheap.

Would you mind to share your DSLR scanning set up in this thread

I am still struggling between the ideas to build a DSLR scanning system or buy a Flatbed scanner to digitalise 120 films.

 

[Ranchu]

I use a software to create profiles from a Macbeth chart for each of my cameras. When I edit in Lightroom I apply that profile. It seems reasonable that such a program could be written to analyze one frame of a roll with a Macbeth chart and create a profile to neutralize the Orange mask and set a relative color balance. Sound reasonable? Each emulsion type would be different but you can do it once and pretty much forget it. Of course different lighting will require adjustments for correct color balance just as it does on a scanner. That's basically what's been done with scanner profiles.

Sounds reasonable, except Adobe ony uses matrix profiles in their converters, the 'profiling' they do with that macbeth chart is some hsl edits rather than a proper LUT profile. Adobe sells their theory, but they don't really hold up their part of the deal imo. Your point about other stand alone software is good though, but even then won't the software be more accurate, and easier to use if it's made for a specific device? Of course it doesn't really matter with b+w 8x10s, does it?

 

[x-ray]

Would you mind to share your DSLR scanning set up in this thread

I am still struggling between the ideas to build a DSLR scanning system or buy a Flatbed scanner to digitalise 120 films.

It's pretty simple shat I've used. I have a 10' studio stand in my studio and position a Fuji light box on a table or the floor. I simply koi t the camera with a macro lens and use levels to insure the sensor of the camera is parallel to the plane of the film. Most of what Ive done was with my Hasselblad and digital back with a 120 micro planar but have done it with my D800 and other DSLRs. This could be done with a tripod as well. I do use black masking tape and black paper to mask off stray light that can cause flare. You'll also probably want to do this in a darkened room or one with subdued light to insure the ambient light doesn't influence the image. You may even need a black card like black foamcore above the camera to keep reflections down if there are any lights on in the room. It depends on your setup.

 

[x-ray]

Sounds reasonable, except Adobe ony uses matrix profiles in their converters, the 'profiling' they do with that macbeth chart is some hsl edits rather than a proper LUT profile. Adobe sells their theory, but they don't really hold up their part of the deal imo. Your point about other stand alone software is good though, but even then won't the software be more accurate, and easier to use if it's made for a specific device? Of course it doesn't really matter with b+w 8x10s, does it?

I'm just speculating here because I'm not a software writer. I use it but couldn't bdgin to write an app.

It may be that I simply don't know enough but would imagine a single program could work for generating profiles for most any capture device. It really an analytical program comparing data from standard colors and tonal values against what you capture with your device. I just don't see it as being that difficult.

I guess in reality this could be done by doing measurementsxfrom the film with a densatometer too and entering that I go I to the program. Basically the application I use does that from a digital capture of the Macbeth chart.

 

[brbo]

I use a software to create profiles from a Macbeth chart for each of my cameras. When I edit in Lightroom I apply that profile. It seems reasonable that such a program could be written to analyze one frame of a roll with a Macbeth chart and create a profile to neutralize the Orange mask and set a relative color balance. Sound reasonable? Each emulsion type would be different but you can do it once and pretty much forget it. Of course different lighting will require adjustments for correct color balance just as it does on a scanner. That's basically what's been done with scanner profiles.

If only it was that simple. We would have such a system in place 50 years ago. Unfortunately...

 

[tsiklonaut]

Just some random thoughts:

Interestingly, the ugliest scans I've ever seen are done through single-shot sensors, i.e. dSLR scanning. I've been to this route myself as well.

The demosaicing, noise shaping, LUTs etc etc a row of built-in manipulations happening make it look rather artificial - pixelized, over-edgy, lifeless or just sterile.

If you want speed and quantity you should bin analog and shoot digital anway.

In scanning IMHO every pixel deserves a full RGB info from a sample-spot on the film, not a combined info or a manipulated info. Personally I don't care about the speed, I care about the result. For me analog photography has always been a quality over quantity. I'm fine with half-an-hour per frame or even more really, as long as I know it's not some hype job manipulating scanner.

Market is full of über-fast scanners, cheated-up resolutions, demosaiced and microcontrast-enhanced scans that marketers brag about how incredibly sharp or good their scans look, but to me their sesults all tend to look like another digital photo from a digital camera making all this analog-shooting a very questionable pursuit if you end up like 99.99% of those digital photographers.

A non-interpolating, non-cheating scanner to "see" film as it is - now this is very hard or expensive to build and automate, something that also keeps film perfectly flat and has a working AF (mighty important I might add!). Ideally a drum scanner with proper PhotoMultiplier Tubes as sensors and variable scanning aperture to optically alter the image rendering, or go a grade lower compromise, i.e. virtual-drum holder like Imacon/Hasselblad scanners or even cheaper, such as a line- or row-sensor, i.e. Nikon 9000ED-ish.

Or alternatively to fill a spot between drum scanner and dedicated prosumer scanner: a stiching X-Y flatbed scanner with high end mechanics and sensor. I.e. Creo IQsmart as a platform to improve on. This platform also lets easily you to wet-mount, a feature that many dedicated prosumers scanners lack, i.e. Nikon 9K or Imacon/Hasselblad.

If you really want to go the cut-corners route and build a camera-type scanner, well there are non-interpolating CCDs available, i.e. Foveon X3 or Hasselblad shifting-sensor, but they still mess with microcontrast and noise shaping. I guess then you're still back to square one on this post.

Just me 2c.

Good luck with the project!

 

[Ranchu]

A LUT profile just allows you to build color correction into the characterization of the scanner, so neutral can be 'maintained' from 0 to 255 at the beginning. It's a good thing. I don't think scanners even have demosaicing (??), just one of each rg and b.

 

[lrochfort]

Just some random thoughts:

Interestingly, the ugliest scans I've ever seen are done through single-shot sensors, i.e. dSLR scanning. I've been to this route myself as well.

The demosaicing, noise shaping, LUTs etc etc a row of built-in manipulations happening make it look rather artificial - pixelized, over-edgy, lifeless or just sterile.

If you want speed and quantity you should bin analog and shoot digital anway.

In scanning IMHO every pixel deserves a full RGB info from a sample-spot on the film, not a combined info or a manipulated info. Personally I don't care about the speed, I care about the result. For me analog photography has always been a quality over quantity. I'm fine with half-an-hour per frame or even more really, as long as I know it's not some hype job manipulating scanner.

Market is full of über-fast scanners, cheated-up resolutions, demosaiced and microcontrast-enhanced scans that marketers brag about how incredibly sharp or good their scans look, but to me their sesults all tend to look like another digital photo from a digital camera making all this analog-shooting a very questionable pursuit if you end up like 99.99% of those digital photographers.

A non-interpolating, non-cheating scanner to "see" film as it is - now this is very hard or expensive to build and automate, something that also keeps film perfectly flat and has a working AF (mighty important I might add!). Ideally a drum scanner with proper PhotoMultiplier Tubes as sensors and variable scanning aperture to optically alter the image rendering, or go a grade lower compromise, i.e. virtual-drum holder like Imacon/Hasselblad scanners or even cheaper, such as a line- or row-sensor, i.e. Nikon 9000ED-ish.

Or alternatively to fill a spot between drum scanner and dedicated prosumer scanner: a stiching X-Y flatbed scanner with high end mechanics and sensor. I.e. Creo IQsmart as a platform to improve on. This platform also lets easily you to wet-mount, a feature that many dedicated prosumers scanners lack, i.e. Nikon 9K or Imacon/Hasselblad.

If you really want to go the cut-corners route and build a camera-type scanner, well there are non-interpolating CCDs available, i.e. Foveon X3 or Hasselblad shifting-sensor, but they still mess with microcontrast and noise shaping. I guess then you're still back to square one on this post.

Just me 2c.

Good luck with the project!

I couldn't have put it better myself.

 

[ChrisPlatt]

At $600 or more (perhaps far more) only very serious hobbyists will be interested.
At higher cost only institutions or professionals still using film might be customers.

Chris

 

[Scapevision]

Why haven't I already bought the readily available multiformat dedicated film scanner? Because they cost around 2,000 and higher. That should answer your question.

 

[jzagaja]

Try make something like Kinograph with monochrome sensor and XY for higher res for stills and without for telecine. You will be haveing cinematographers and photographers. Even a drum scanners (mine) has troubles with Super8 format. For example Ochoypico gets nice results.