Scanning film - How much resolution is too much?

[dkeyes]

I remember seeing info on this and other sites but can't seem to find anything about film resolution and maximum scan sizes.

I shoot 6x7cm transparency film and make digital inkjet prints up to 40x50".
What is the max. resolution one can get out of this size film.
For example. Will a 5000ppi scan of 6x7 trans. capture all the info that is in the film?

I usually scan my 6x7 transparencies at 5000ppi (8bit). This may be overkill for that film size, thus the question. I've been doing this because it gives me a 40x50 image at 270ppi and it's the max my scan provider can do. I've been happy with the scans but wonder if it's actually too much info/wasted pixels.

On the flip side, I could go to 16bit at the same ppi or smaller. (cost is an issue since it can double the price) Maybe smaller ppi but larger bit depth would be good?

Thanks for the input,
Doug

[michael]

Like most such subjects in photography there are many differing opinions. Also the answer will vary with the type of scanner, brand, and type of film.

Most knowledable observers that I've discussed this with seem to agree that beyond about 4000 ppi all you're doing is scanning film grain, not usable information.

Michael

[Dale_Cotton]

Doug: it's been a couple years since I was actively involved in scanning, so take the following with a generous helping of salt. Michael's 4000 ppi ceiling works for me ... assuming that the scanner's rated ppi is in the same ballpark as its actual ppi. The question in my mind is how much lower can you go? I suggest a test series along this line:

Find a really sharp slide with no DOF issues; scan it at 5000 and at whatever lower ppi options your scanner offers (real, not interpolated). No need to scan the whole slide if you can scan just a cropped area. If you didn't crop in the scanner, you can crop now to get a good part of the frame, maybe 2000x2000 pixels from the 5000 ppi scan. For the sake of the test, up-res each of the crops from the lower ppi scans to the same size as the crop from the 5000 ppi scan. Now give each as much USM as needed to regain the acutance you are used to seeing at actual pixels, then print. If any of the lower ppi crops shows no degradation compared to the 5000 ppi crop, then with some further testing you might be able to stick to that scan res for future work.

The main benefit of 16-bit is to minimize the damage from radical contrast and colour changes in post-processing. If you don't anticipate needing to do a radical edit on a given slide, you can forgo the extra bit depth; but of course this is also something you can put to the test.

Another perspective, however, is to make your best scan - 5000 ppi @ 16-bits/ch - of each slide you consider potentially worth printing, archive that as a master reference file, then do your post-processing on a down-sampled and bit-depth reduced copy should practicalities so dictate. The ideal here is that you would shed no tears if your dog chewed one of your slides so long as its master reference file remains safe.

[Ben Rubinstein]

I've been doing a lot of research on this recently, the answer is really dependant on the scanner. A drum scan at 4000DPI is going to have more information than an Imacon at that resolution and a whole heck of a lot more than a consumer flatbed.

[Doug Fisher]

What scanner are you using?  Very few scanners have a true optical resolution of 5000 ppi.  Find the true resolution of your scanner, not interpolated and not even the claimed ppi/dpi (e.g. a 4800 flatbed will usually give you a true resolution of less than 2800 ppi) and scan at that true resolution.  From there you will have to do the calculation to see what printing ppi will be yielded by that scan resolution.

Doug

[dkeyes]

What scanner are you using?  Very few scanners have a true optical resolution of 5000 ppi.  Find the true resolution of your scanner, not interpolated and not even the claimed ppi/dpi (e.g. a 4800 flatbed will usually give you a true resolution of less than 2800 ppi) and scan at that true resolution.  From there you will have to do the calculation to see what printing ppi will be yielded by that scan resolution.

Doug
[a href=\"index.php?act=findpost&pid=115343\"][{POST_SNAPBACK}][/a]

I should have mentioned that I am having the transparencies drum scanned. Don't know the brand of scannner but I'm sure it is true 5000ppi (not interpolated). Of course I could just scan everything at 16 bit 5000ppi but the cost is too high to do everything that way. I imagine the only way to know for sure is scan at a couple resolutions and bit depths. An expensive test but might save me money in long run.

[Dale_Cotton]

Senior moment on my part. I saw "it's the max my scan provider can do" and read "it's the max my scanner can do" and thereby got the idea you were using your own scanner.

Getting a second, lower-res scan of a slide you've already had scanned at max res might indeed be worthwhile. Before that you might also try a simple reality check. Take a copy of one of your unedited 5000 ppi scan files and down-sample it to the next best res the service offers. Edit and print that down-sampled version (or a crop thereof) to see if the results are up to snuff. My recollection from working with my own 4000 ppi dedicated film scanner is that the result of dropping down to a lower res in the scanner and down-sampling in Photoshop were actually quite comparable.

The difference, I find, is that a lower scanning res from the same scanner not only reduces the amount of finest detail but results in less acutance, which in turn requires more USM to restore, which in turn further attacks fine detail. However, 100 ISO slide film typically has such little acutance itself - they're not called dye clouds for nothing ;) - that lowering the scanner res might have little impact until a certain threshold is reached.

[pfigen]

The specific scanner make and model does make a difference in real world resolution for large prints. Drum scanners have many different hardware optical resolutions, but most are bunched at the lower end of the resolution scale. Every drum scanner I know of jumps from the highest res to half of that in one step. Many operators are unaware of that. My Howtek 8000 jumps from 8000 to 4000. My Howtek 7500 has 5000 and 2500 respectively.

The other side of the equation is whether or not your film has any usable detail over 4000 ppi. I have scanned really sharp 6X7 Provias at both 4000 and 8000 and seen an improvement, but it's very small. Whether it makes a huge difference is debatable. Only your images will know for sure.

[Mark D Segal]

I usually scan my 6x7 transparencies at 5000ppi (8bit). This may be overkill for that film size, thus the question. I've been doing this because it gives me a 40x50 image at 270ppi and it's the max my scan provider can do. I've been happy with the scans but wonder if it's actually too much info/wasted pixels.

On the flip side, I could go to 16bit at the same ppi or smaller. (cost is an issue since it can double the price) Maybe smaller ppi but larger bit depth would be good?

Thanks for the input,
Doug
[a href=\"index.php?act=findpost&pid=115295\"][{POST_SNAPBACK}][/a]

What you re doing is sensible if you are making prints that large. The final arbiter of scan resolution is the required PPI output resolution being sent to the printer. Prints that size should be in the range of 240 to 300 PPI, hence your scan resolution of 5000 giving you output PPI of 270 makes sense. You wouldn't want to print much below that resolution, and as long as you don't edit in a colour space larger than Adobe RGB(98) 8 bit depth should usually be very satisfactory. Generally speaking, the fall-off of image quality once you get below 240 output PPI would be more noticeable than the difference between 8 and 16 bit depth.

[Ray]

The problem with scanning is that the quoted ppi scanning resolutions are always diminished by the limits of the lens. Scanning film is just a way of photographing your film with a digital camera, a scanning camera.

A scanner with optical resolution of 4000 dpi should be able to cature resolutions up to 80 lp/mm, one might think from the maths. But this is far from the truth. In order to do this, the scanner's lens would have to be perfect. There are no perfect lenses.

My Nikon 8000ED scans at 4000 dpi, but I would be surprised if it could capture any detail above 50 lp/mm. With 35mm film, this is sometimes not good enough. I can see extra detail when I scan technically sharp slides with my Minolta 5400 II (at 5400 dpi). However, I think 6x7cm film would be slightly less demanding.

[Tim May]

Thanks!

[dkeyes]

Thanks for all the input everyone.
I am scanning my next batch of 6x7 trannies at 5000ppi 16bit to see if I see any difference from 8 to 16 bit. I imagine this will help with some of the images as they are night time shots, mostly dark shadows and a few highlights. I also want to make some big color moves on the rest of the images so again, having 16 bits is supposed to be ideal in this situation. Many of you have affirmed my assumption that having a 270ppi image at 40x50 (5000ppi scan) is better than a 220ppi image (4000ppi scan).

Of course there is always the issue of what's the best match for your printer. I'm using the z3100 and it seems to like the higher resolution images (300-360ppi). I've printed several at 300ppi and noticed a difference in smoothness in some areas compared to same print on Epson 9800 using 300 ppi.
- Doug

[Jonathan Wienke]

Here's a general rule of thumb for scan resolution:

Look at your scan at 100% in Photoshop. If you're devoting multiple pixels to individual film grain particles / dye clouds, you are committing overkill, and increasing scan resolution will not benefit you. Try downsizing your max-resolution scan to each of the resolutions offered. Look for the lowest resolution that still captures all true image detail (as opposed to detail of grain structure), and then go one step higher. That will give you some fudge room if you do noise reduction and things like that--the noise reduction will destroy less detail when run on a slightly-overscanned file. You may need to do equal-sized comparison prints at various resolutions to decide what scan PPI is the best.

I definitely recommend scanning in 16-bit mode; your noise reduction will work better and you'll be able to do much more aggressive curve and level adjustments before encountering posterization/banding.