PDA

View Full Version : 8K film recorders for repro vs. original film



bglick
20-Sep-2005, 21:48
I was considering duplicating drum scanned 6x7 film back to color film through a CCG film recorder. The Film recorder on the height writes 6144 pixles to the 56mm high film, or 55 lp/mm. I was curious, after drum scanning the film, sharpening it, adding contrast, etc....making the digital file just pristine, then going back to film at this high resolution, I would think the dupe would outshine the original? Has anyone ever tried such? I know for lower rez film recorders this does not happen as the in-camera film recorded more rez than the film recorder can output, but that is not the case here. TYIA.

bglick
20-Sep-2005, 21:49
One last bit.... the output film will be Fuji CDUII, high resolving and very tight grain film.

Will Strain
21-Sep-2005, 07:03
Resolution cannot be created. The digital output of the dupe (even if theoretically capable of more data than the original) will not have more data than the original. It may have been altered, and manipulated (contrast range, etc) But that is just interpolation of original data, not new.

That is not to say that it might not look better... but I don't know what you would gain in the dupe over simply printing from the digital file either. Tho it might be fun to crank out a cibachrome from such a tranny.

Neal Shields
21-Sep-2005, 08:57
55 lp/mm is theoretical. Actual is usually about .7 or less of theoretical.

So if you had 55 lp/mm on film and then scanned it and then wrote it back to film, you would have 55x.7x.7 or 27 lp/mm

One of these white papers get into that in detail but I can’t remember which one:

http://www.kodak.com/global/en/digital/ccd/publications/whitePapersArticles.jhtml

paulr
21-Sep-2005, 09:27
Based on what you discover the actual resolution to be, you'll need to consider the enlargement size you're trying to make (unless you're making enlarged film for contact printing; then it's completely irrelevent).

At print size, any resolution above 14 lp/mm is wasted. Only someone with perfect eyes, under ideal viewing conditions can even detect detail at that frequency. Most people in the real world can't see anything past 10 or 11 lp/mm. The only things to be gained from more resolution are the ability of the image to hold up when looked at through a loupe, and it's ability to render slightly diagonal lines as smoothely as possible.

The number one thing that makes a print look sharp and detailed to your eyes is the contrast (modulation) in the 4 to 7 lp/mm range. This you can control very easily with intelligent unsharp masking.

I have digital prints made at 720 ppi (about 14 lp/mm) that look more like contact prints than actual contact prints from the same neg. If that makes any sense. But I second Will's question ... why the intermediate step of making a dupe transparency? Is it for sending to pubs that don't take digital files?

bglick
21-Sep-2005, 10:38
> Resolution cannot be created.

Till about 3 years ago, I used to preach this same statement, as it makes so much sense in the real world...but...

After testing, one finds this all becomes semantics.... whether data is labeled, created, born, reborn, synthesized, interpolated, etc... in the end, good digital manipulation of the right image results in higher resolution and more appealing images. I know for sure, in prints, I have run experiments shooting subjects at different distances...the goal was to see how much farther away I could shoot a subject (less resolution at same final size) vs. the closer shot. Same lens, same camera, same film, etc. The results were pretty amazing....

The shots taken with digital camera benefited the most from the digital enhancements. I could easily add back 30% of sharpness which was lost from the further distance. This was mostly sharpening and contrast gains. Of course, if I enhanced the closer shot, it would have superceded the farther shot, but that was not the purpose.

Film, did not fare as well, maybe in the 15% - 20% range. Pixels born digital are more "enhancement friendly", as this makes sense, as the software is not being confused by grain.

Most film purist would find it ad nauseam to learn, that a "standard" digital file that was Born digital in a digital camera is output to a computer at x amount of recorded data. RAW processing of the file (mandatory) will make the file 4x in size (this is an average based on Bayer interpolation and color tonality interpration). So before any post processing such as Photoshop, the file only has 25% recorded (non-interplated) data. Now, add a small amount of up-rezing / sharpening, and the final file size (print size) now contains ~ 15% real, or "recorded" data. Seems absurd to most, including me...but in the end, the final output is truly amazing.

There has been several tests I have read through the years, that demonstrate how all this interpolation compares to cameras that recorded "all" the information at the moment of capture. (Pure RGB capture with full 16 bit tonal range) This is the purest method to compare interpolated data with pure recorded data. Without turning this into a digital thread, the overall concensus was, Bayer sensors can record about 1/2 the data vs. a camera which records 100% data which requires no interpolation...... after interpolation of the Bayer data, the files appear indistinguishable. This is the pure genius behind Bayer interpolation model. The Bayer system strategisticaly records key data to make interpolation 90% + effective. (this is a general statement, as there is all kinds of variables in digital capture / interpolation) The huge benefit is lower cost cameras, smaller size (due to recording less data), smaller storage devices in the field, etc.

I was actually informed a well known and highly successful photographer actually utilizes this method to push his final print sizes as he sells darkroom prints. The original film is drummed scaned, manipulated, output back to film, then the manipulated film is used in the darkroom to make prints. And based on the quality of the final prints, I have every reason to beleive this is the case, as I have never seen "pure film" enlargement looking so good at 20x - 25x enlargement sizes.

However, in the end, vs. film, the overall concensus is, the best 35mm color films, best scanners, etc. require about 9 - 12 Mega Pixel cameras to match the final print quality. Some of this is subjective as there is so many variables in qualtiy of digital cameras, quality of film, quality of scanner, human factor, etc. So following this general rule of thumb, and assuming LF shots are not too diffraction limited, even the forthcoming 39 MP sensors will only record the equiv. of 6x7 film, or possibly a bit more.

> 55 lp/mm is theoretical. Actual is usually about .7 or less of theoretical. So if you had 55 lp/mm on film and then scanned it and then wrote it back to film, you would have 55x.7x.7 or 27 lp/mm.

Neal..... here is what I anticipated.... at the point of exact focus, assuming I record 55 lp/mm, (2800 dpi) and overscan the film at about 5000 dpi, I should extract 90% of the recorded data, albeit with a penalty in file size. This is based on Seybold reports which took high end scanners and scanned test targets, at 70% overscan dpi, they resolved 95% of the target data.

Now, dealing with the rest of the efficiency chain...... I was hoping for an increase in perceived resolution of 30%, by shapening and increasing contrast. Then, efficiency losses going back to film, the 1/R nemesis, so up to a 30% loss (equal to your .7) So at the point of exact focus, I was hoping to record at least what was in the original. At the near far/point is where I felt I could do better, as sharpening / contrast improvements will have a greater % improvement as the recorded data is much less due to defocus, creating more space between "recorded film data" vs. max. possible data film will hold.

> At print size, any resolution above 14 lp/mm is wasted. Only someone with perfect eyes, under ideal viewing conditions can even detect detail at that frequency. Most people in the real world can't see anything past 10 or 11 lp/mm.

Paul, where did you get these numbers? I am really curious, as I have never seen numbers that exceed 5 - 6 lp/mm for the best human vision, I am really interested if you have any references. To date, the best reference I have read on this subject was Ctein book, which he states that although human vision can only resolve up to 5 lp/mm in test targets, (common acceptance value), but in photographic prints, he has demonstrated that humans can discern the differences in prints from 5 lp/mm through 30 lp/mm, consistently placing them in the proper order of resolution. I found this evidence quite compelling.... and btw, I do agree, contrast is just as important or more important for perceived sharpness vs. sharpening.

> I have digital prints made at 720 ppi (about 14 lp/mm) that look more like contact prints than actual contact prints from the same neg. If that makes any sense.

I beleive this... but have you ever pritned test targets on the paper to see just how much resolution the paper / printer / ink can hold? At this level, I am estimated from test reports, the printer / ink / paper efficiency drops 50%, so sending a clean 14 lp/mm file to the printer will only acheive 7 lp/mm on paper, which is still quite amazing, as most good prints are in the 2 lp/mm range.

> But I second Will's question ... why the intermediate step of making a dupe transparency? Is it for sending to pubs that don't take digital files?

No, the film is the final product, used in a stereo viewer.... I was researching what methods would produce the best possible dupes, in volume. The lenses in the stereo viewer will enlarge the film 4 - 5x, hence why I am trying to gain every ounce of rez into the dupes. At 15 lp/mm recorded at the near / far points, (15 / 5 = 3 lp/mm) I want to maxmize my gain in this area as improvements will surely be noticeable. In my opinion, back lit film is where human vision is most sensitive to resolution. And also, back lit film makes the eye ultra sensitive to contrast gains.