And that is a positive contribution to this thread because...?
Yes, drum scanners are the best quality scanners currently available, and there are a number of fine drum scanner operators who take part in these forums. The reasons we are engaged in our project have been well-detailed in this thread. If you are interested in making a positive contribution to the project, then you are welcome to participate. If you're only interested in hawking your wares, then please do so elsewhere.
“You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
― Alexander Den Heijer, Nothing You Don't Already Know
I finished the negative stage today. I'll have to make some changes to the camera support structure before running any tests, and I'm trying to decide whether to use my 80mm Rodagon on a bellows or my Nikkor 55mm micro. Neither will be ideal, but I'd like to get to trying to stitch the frames fairly quickly, as that appears to be the biggest hurdle facing us.
On another note, the D800 announcement bodes well for the future. While I certainly won't be able to afford one, or something similar, for many years, this type of resolution could really cut down on the number of samples required for a scan. I expect that for most 35mm, no stitching would be needed. This would open up more time for hdr or focus stacking and still the process should be much faster than any current scanner.
“You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
― Alexander Den Heijer, Nothing You Don't Already Know
Hi Peter, how you doing?
I'm marco, I contacted you approximately 1 year ago about the Screen Cezanne...do you remember?
I still think there are better alternatives but if your going to DIY you should try using a black and white CCD sensor and RGB filters on a color wheel. Also you should be able to set the microns of your sample and the sample frequency independently.
If you sample a space smaller then your dye clouds then you get crazy color noise or "grain aliasing"... Its when you see the little red green and blue parts rather then what they look like as a whole.
The sample frequency or "resolution" can be increased past the sample size and will continue to improve the image. This is due to the stochastic nature of film.
Where does one find a black and white CCD sensor in packaging that can be replaced to overcome obsolescence in the future?Originally Posted by 8x10 user
Rick "who could design all sorts of stuff that would fail in the market just as current scanners are doing" Denney
Well, if you have a black and white ccd sensor that you'd like to lend me, then I'd be happy to pursue that route.
I do have a dslr, old as it is. And the tests run so far indicate that the quality minimum we are after is available in a single capture of a small area of a frame of film. The only important question left is whether stitching a bunch of samples to cover the negative will provide the required quality, and the only way to tell is to try it.
“You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
― Alexander Den Heijer, Nothing You Don't Already Know
Peter, as I understand it the quality minimum you are after is defined by the performance of a typical consumer flat bed scanner, such as the Epson. Let me try to be helpful in telling you what you will find.
On a single acquisition
1. The advantage of the area sensor (DSLR with an X x Y chip) over a linear sensor (Epson) is that you can magnify as much as you like in a device that images quickly. Very useful in imaging fine details of small targets while retaining MTF. However, larger targets (e.g. 4 x 5 film) require so many acquisitions that they become impractical. This is with a motor stage or rail system to move the sensor or target. Manual movement will fail in actual use, both because angular errors corrupt the alignment algorithms (your hand tilts the film a bit), and because it is just so tedious.
2. An area sensor will lose intensity resolution (precision, tonality, whatever you want to call it) in both dark and light areas. Your shadows will block up, and subtle intensity transitions in the light areas will be lost. These issues are generated both in software (correcting spatial intensity variation) and in the optical chain (flare). They are not solvable without altering the acquisition mode to one that is flare resistant (spot or line scanning). In other words, spatial resolution is the least of your problems. We shoot film for the tonality more than the detail and that is why scanners exist.
3. Color makes the problems much worse.
Stitching
1. Monochrome stitching suffers from an interaction between the correction field (your reference) and the intensities in the film. In other words, overall light levels during the scan affect the spatial distribution of the correction field. That results in noticeable intensity variation in the composite image (patches), with alignment errors also contributing to patchiness. Careful manual control/editing of the process helps, but it can take hours to do something like a 10 x 10 montage at high quality. Again, this is with a motor stage.
2. Color stitiching is a fearsomely difficult process for lots of reasons. Color images will be patchwork and muddy (because of loss of intensity resolution) unless you have discrete chips for each primary or a filter wheel. Even then, there will be a loss of color rendition and a subtle patch effect. Not better than Epson.
I will now shut up.
Keep on rockin, Peter! Don't pay attention to the naysayers. Especially the ones with lots of technical suggestions but no actual technical expertise.
It sounds like you're speaking from experience. Are you? Because doing four rows of six exposures (24 total shots) does not sound "impractical" to me.
You may be shocked to learn that there is a gigantic amount of flare in a flatbed scanner. There is flare from the scanner platen, as well as from the cheap plastic lens in front of the linear CCD.2. An area sensor will lose intensity resolution (precision, tonality, whatever you want to call it) in both dark and light areas. Your shadows will block up, and subtle intensity transitions in the light areas will be lost. These issues are generated both in software (correcting spatial intensity variation) and in the optical chain (flare). They are not solvable without altering the acquisition mode to one that is flare resistant (spot or line scanning). In other words, spatial resolution is the least of your problems. We shoot film for the tonality more than the detail and that is why scanners exist.
In contrast, a well-corrected macro lens at ƒ8 is incredibly flare resistant. I've verified this experimentally, but it could easily be guessed by the superiority of modern glass optics over the plastic scanner optics. Negatives are relatively low contrast, so you just don't excite much flaring.
Why? No explanation.3. Color makes the problems much worse.
I think you're referring to vignette, but it's not entirely clear. Vignette is easily removed mathematically by shooting a reference frame at the target distance and aperture. There is no technical reason why 2-3 dozen very carefully shot frames cannot be aligned and stitched with the proper software.1. Monochrome stitching suffers from an interaction between the correction field (your reference) and the intensities in the film.
I'm not sure what "loss of intensity resolution" is supposed to refer to, because you actually gain luminance resolution in the proposed system. "Patching" happens with vignette, which I've already addressed. You don't need a filter wheel or monochrome sensor. There's more than enough color sensitivity in the average DSLR at low ISO.2. Color stitiching is a fearsomely difficult process for lots of reasons. Color images will be patchwork and muddy (because of loss of intensity resolution) unless you have discrete chips for each primary or a filter wheel. Even then, there will be a loss of color rendition and a subtle patch effect. Not better than Epson.
I've shot my own tests, imaging color negatives at 1:1 with a 5D II, and saw immediately that this method produces sharper, clearer images than a flatbed. It sounds like you speak from experience, in which case I'd love to hear about it. Because virtually every point you make is contrary to my real-world findings.
I expect that Mr. Ramm knows very well what he is talking about, and this project may very well fail. I'm ok with that, but I'm not going to stop without giving it a try, especially since I know someone who has built such as system for a client, and the client is very happy with the results. And even if they results currently obtained aren't that good, there's been some major advances in all areas the project touches on, digital sensors, robotics, software development..., and I expect these advances will continue. Thus, what might not be practical now might very well become so over the next couple of years.
“You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
― Alexander Den Heijer, Nothing You Don't Already Know
Posting Permissions
Reply With Quote
Bookmarks