Scanner Settings? 8bit,16bit, 3600dpi or lower?

Discussion in 'Digital Darkroom' started by john_elder|1, Jan 16, 2007.

  1. Please tolerate this question. I just bought a a Primefilm Scanner Model
    3650U.I scanned my first 2 images ever last night. It seems to work great.
    After doing some research, I came to the conclusion that I should scan at 16
    bits rather than 8 bits. This advice was recieved at "Scantips" and a couple
    other sources. Independent of that research was the recomendations to scan at
    the highest Resolution the scanner offers which is 3600dpi. So, if I scan at
    3600 dpi and at 16bits, I get a huge file of app 100MB. I am looking for a
    file size of around 50MB. What do you folks reccomend: reducing the DPI to
    2700, OR scanning at 8 bits. What are the tradeoffs? Thanks from a newbie. By
    the way the CyberView X software has both histogram and curves, with before
    and after pictures above the graphs. Any suggestions will be well recieved.
     
  2. Further information i forgot to add. I am scanning exclusively B+W 400 film. No slides and no color negatives. I want to be able to print 35mm film 8x12 inch full frame images at 300DPI. I want files sufficiently large to be able to send to certain magazines. For example the submission guidlines for LENSWORK are no smaller than 2200 pixels or larger than 6000 pixels for either verticle or width.
     
  3. 8 bit will give you 256 different levels of intensity per color. 16 bit will give you 65,536 differnet levels of intensity per color. There are thre colors....Red, Green, Blue....I know.....you're not doing color.......but actually unless you scan in grayscale, you are scanning in color........well, at least all three colors (or channels). We'll leave that there for the minute.

    So, if you scan in 8 bit color, you have 3 channels of 256 variations. Or, 16.7 million combinations (256x256x256=16,700,000 apprx)

    If you scan in 16 bit color you have 3 channels of 65,536 variations. Or, 281 trillion combinations.

    If you scan in grayscale........you have one channel in each of the scenarios above. 256 for 8 bit.......or 65,536 for 16 bit.

    What all of that is, is essentially tonal gradations. The more variations you have available, the finer and more detailed the tonal gradation will be.

    I would NEVER scan in 8 bit grayscale (or color, even for an eventual black and white image). 16 bit grayscale will probably suffice....I use this. This decision to use 16 bit is especially important if you are going to do a lot of levels or curves editting in Photoshop later. The less variations you have, the more the possiblities of gaps in your tonal gradation. Plus you want to save the scan in TIFF. TIFF is a lossless image file and allows 16 bit images. JPEG is a lossy image file (meaning it's a compression of the actual information to reduce file size) and only allows 8 bit.

    So, for tonal variation granularity, and non destructive (well, near non desttuctive) editting, a 16 bit grayscale scan saved as a TIFF file is an extremely good compromise. Better yet is a 16 bit RGB (color....even though it's only B&W)) scan saved as a TIFF file is as good as you can get.

    Now, resolution. I own a scanner whose max rez is 2700 and 8x12s are no problem. But, if I owned a scanner that did 3600......I would personally scan at that. You may want to crop later. You may want to print larger later. Why scan twice? Scan as big as you can ONCE, and then you can do whatever you want later.

    Yep, that means more storage. Either external hard drives or optical (CD or DVD). Welcome to the memory hog world of digital imaging. i started with CD.......pain in the butt........went to DVD, not bad.......but since external USB (only) hard drives are down to below $150 for 250 Giga byte of storage, I now use them (with DVDs as a back up)............oh yeah, BACK UP EVERYTHING TWICE! Start now........*X&$#**X&$#**X&$#**X&$#* happens.

    But, yeah, 2700 will suffice for the prints you want. I didn't actually do the math for the publications, but it doesn't seem all that big, so I think it will suffice for that also.
     
  4. I am looking for a file size of around 50MB.
    Why ? Is storage space an issue?
     
  5. I want to be able to print 35mm film 8x12 inch full frame images at 300DPI
    then you need a file that is 2400 x 3600 pixels . At 8 bits per channel (bpc) that is a 24.7mb size file. At 16 bpc, it's 49.4mb. Either way, it is scanned at 2700ppi. As long as you aren't planning on cropping it seems like either way you are under what appears to me from the outside to be an arbitrary size limit. There are two very good reasons for scanning at a greater resolution. The first is obvious, you'll have a large enough image that you can crop if you decide to and still hit your H x W in inchesat 300dpi requirement. the other good reason to scan at a higher resolution is by reducing (downwards interpolation) the scan in Photoshop to the H x W x ppi you require, you'll go a long way towards knocking out the electronic noise pattern generated by your scanner.
     
  6. Thanks Thomas and Ellis. SO, your advice is to keep doing what I did on the first 2 scans:3600dpi and 16bits(rbg)yeilding about a 100MG tiff file. Thanks, thats what I will do. It just seemed to me after 3 or 4 days of research that that size file wasn't neccessary. Probably my conclusion was based on materials written before it was possible to buy a reasonably priced scanner yeilding that size file. I just wanted to confirm that what I was already doing wasn't stupid or an unreasonable waste of space. I already bought a 250GIG external hardrive and I have 190gig left on my internal drive. I only plan on scanning about 250 or 300 images. Thanks for the fast responses. I'll have more stupid questions later.
     
  7. I'm glad you like the 3650U. My only previous (and final) experience with a PrimeFilm product was the absolutely horrid 1800. Maybe their product line's much improved. However, before spending much time and effort on 3600dpi scans, I'd verify that the scanner optics is actually capable of resolving such. Try some sample scans of known sharp images at different scan resolution settings.

    16 bit scans are preferable to 8 bit if you'll be doing any digital darkroom work at all. You'll have much more headroom for tonal adjustments before posterization sets in.
     
  8. the other good reason to scan at a higher resolution is by reducing (downwards interpolation) the scan in Photoshop to the H x W x ppi you require, you'll go a long way towards knocking out the electronic noise pattern generated by your scanner.
    Does the Primefilm Scanner REALLY do 3600spi?
    Probably not. So why scan at that rez which introduces noise? I'd shoot for ~2400.
     
  9. "...Does the Primefilm Scanner REALLY do 3600spi?

    Probably not..."

    ....and you base that on...?.....what, exactly? did you look it up?
     
  10. John,

    A 2400x3600 pixel image should yield a good 300ppi print of 8x12". Now, for best detail, you should scan at your best scanner setting and then resize in Photoshop (or whatever image editor you use). The 16 bit scan is a necessity if you do any levels adjustment in postprocessing of the image. If you could manage to send the image directly from the scanner to the printer, 8 bits would suffice, but that's practically impossible; thus, you need to have more bits to allow the compression and expansion of the tonal zones of the image that levels adjustment will impose. Contrary to popular belief, if you bypass levels adjustment and go directly to curves adjustment, won't avoid this peoblem, although the histogram may not show it.

    Try this experiment: make a scan at 8 bits (any size), open the image in Photoshop and go to Adjustments>Levels. Move the midtone slider to the right about one quarter of the way, apply the changes and see what happens to the histogram. The continous peaks and valleys of it will acquire a comb-like pattern. The gaps are lost tones!

    Now, scanning at 3600dpi at 16 bits will yield an image (from a 35 neg) that is more than 130Mb in size, but that's a temp file. When you save it as TIFF or PSD, the image will be compressed and the resulting file will be much smaller.

    For printing, always try to scan at your scanner highest settings, even if you resize prior to printing. Bear in mind that many printers can resolve more than 300ppi. Good photo printers can resolve upwards of 480ppi. At a casual glance, this may not be noticeable but on close inspection, areas if fine detail, specially in the highlights, will clearly show the difference. OTOH, sometimes you can get by with a print resolution as low as 180ppi but a side by side comparison with a finer print will be evident.
     
  11. scan your BW neg at 16bit, that will help to get every piece of data out of it. But scan at the needed resolution, if you dont need a 100meg file, well scan at something smaller. It is *almost* as bad to reduce a big image to a smaller format that it is to upscale a small image to a bigger one (lost of sharpness and fine details in general)
     
  12. Pico makesagood point. Youreally wantto scan at no greater than the maximum optical resolution of the scanner --which should be stated clearly somewhere in the instructions and on the box and not at a software inflated resolution.
     
  13. Pico's "point" is yes, indeed well taken....but, the 3650U IS 3600dpi OPTICAL. I just wonder if he made an assumption, or actually found something to dispute the spec?
     
  14. Thanks everybody for the help. Yes, the scanner's optical resolution is 3600dpi or at least it is advertised to be. This scanner has film profiles as well including the film I have shot for the last 20 years:HP5+. On the first neg I scanned, I didn't make any corrections at all to the histogram or curves. The scanners automatic calibration nailed it. Of course I chose a full range negative that contact prints perfectly on a grade 2 gelatin silver paper. The 2nd neg, I deliberately chose a neg that I thought would need corrections prescan . The corrections were done very easily. The only thing holding me back right now is trying to decide my "workflow". I know to keep the"original tiff" unminipulated. I will copy the original and make a corected "master" 16 bit file. I haven't decided what file format for the "Master". Then I will copy the master, convert to 8 bits and make a jpeg for web use. Does that sound reasonable? Any suggestions? I am thinking about using the PSD format for my "Master" file, rather than a tiff. any thoughts? I am currently using Photoshop Elements 4 which came with the Scanner.
     
  15. I'm a little late to the party here... well OK, years late to the party, but my comment is still relevant to the topic for anyone who may have found and read this thread years later... me for example.



    The advice to always scan in 16 bits is fallacious for a reason I will explain shortly, but first let me say that there is nothing technically wrong with scanning in 16 bits, other than file size, which is for the most part not very important these day. It's just that in many (probably most, and possibly all) cases of practical interest 16 bit scans do not add any useful information to the digitized image.



    Usually the advice to scan in 16 bits comes down to the question of banding, as in "if I scan in 8 bits and later do a lot of digital manipulation to the image, will it result in banding?" The answer is a qualified no. Hold that thought, i.e. I included the word "qualified" ahead of the word "no".



    Now we come to the explanation. Film images contain graininess, which from the point of view of image capture and signal processing is a form of noise. In other words, the graininess in the film represents a form of noise superimposed on the image compared to an image acquired by a "perfect" image capture medium. If your scan exhibits noise (including a combination of film grain, scanner sensor noise, and any other form of noise) when scanning in 8 bit mode then it is not necessary to scan in 16 bit mode because those extra bits add no useful information to the data from a pictorial point of view. This is true, even if you subsequently do extreme image manipulation, provided you convert it to a 16 bit image before doing any manipulation of the image. In particular, you will not see banding in the gradients.



    This result is a specific example of a well-known principle in signal processing theory, and it is often taken advantage of. For example, some signal processing systems use a low-bit digitizer along with adding a small amount of noise before the digitization takes place. The reason for doing this is to get rid of digitization artifacts that would overlay steps onto a smoothly varying signal. The cost in doing this can be the introduction of a little bit of noise into the final digitized signal. However, if the signal already contains noise, such as film grain and/or sensor digitization noise, then it is not even necessary to artificially add noise to take advantage of the this effect.



    Now for a discussion of what I mean when I used the word "qualified" before the word "no" in an earlier paragraph: First, it is implicit in the discussion above that the graininess be captured in the scan. If the film grain is so fine as to be grainless in an 8 bit scan then all bets are off. Few if any films used for pictorial purposes would fall into that category. Note: it doesn't take much grain to meet the requirement. This can be framed in statistical language terms of standard deviations compared to the digitization step size, but I won't go into detail on that point.



    A second qualification is that the amount of grain captured depends on several factors, including the optical resolution of the scanner and the pixel size. The bottom line is that the lower the resolution (both optical resolution and pixel size matter) the less likely you are to capture the film grain. An implication is that you will want to use the highest resolution possible in your scanner. I'm talking about spatial resolution here, not digitization resolution, although if you can't capture the grain at the finest spatial resolution of the scanner then you would need to go to a higher digitization resolution to avoid the possibility of banding in a heavily manipulated image.



    A third qualification comes down to the question of "If I have to convert and image to 16 bits before doing extreme manipulation, then what difference does it make whether I scan in 8 bit mode or 16 bit mode?" There are a couple of answers to this question. The first is that it takes half as much storage space to save an 8 bit scan compared to a 16 bit scan. This is not too important these days because digital storage is not very expensive, but it can be a factor of concern, and it can affect other things, such as upload and download speed, limits on sizes of files during data transfer (e.g. email attachments), image sizes that can be handled within a computers memory, and so forth. Also, the intermediate 16 bit files used during image manipulation do not necessarily have to go into long-term storage. Usually an 8 bit file is fine for that purpose, so for long term storage the archival original scan can be 8 bits, and the final manipulate result can be 8 bits, and the intermediate 16 bit files do not need to be saved in long term storage. Furthermore, it is not always possible to acquire a scan at higher than 8 bits. For example, under certain hardware/software configurations the old Leaf scanners only allow 8 bit acquisition.



    A fourth qualification comes down to a very subtle point that will seldom if ever be relevant to scanning for pictorial purposes. If the purpose of the scan is to characterize the noisiness of the scanning process, including film grain and any other forms of noise in the process, then under certain conditions it may be necessary to scan at a higher digitization resolution. This is only likely to be a factor if the combined noise level is close to the digitization step size. At higher noise levels it won't really matter, and at noise levels much lower than the digitization step size one would need to shift to a higher resolution digitizer anyway.



    A fifth qualification comes down to whether the 8 bit scan is encoded using some non-linear protocol. This is complicated, and I won't discuss it here.
     
  16. digitaldog

    digitaldog Andrew Rodney

    A little more late to the party: In terms of scanning high bit (because few if ant scanners actually provide 16-bits of actual data) and WHY you absolutely do want to scan all the data you can, no matter the image, noise or possible future editing:
    http://digitaldog.net/files/TheHighBitdepthDebate.pdf
     
  17. Since we are apparently in a time warp in 2007, it may be possible to call up Cambridge Camera in NYC and order one of these from 1996:
    Forox-SD-Duplicator-C&D-95.jpg
     
    digitaldog likes this.
  18. digitaldog

    digitaldog Andrew Rodney

    I got far more done and earlier than 2007 scanning with those actual slide carousels using a desktop Kodak PIW (PhotoCD scanner, and writer) hooked up to a very old Mac.
    3 seconds a scan, albeit only 18mb per scan but just pop the carousel into the unit and walk away while it scans every slide.
    It was a rather rather expensive piece of hardware.... :eek:
     
  19. That's one of the secrets of capable, long-distance scanning -- use an older machine to work the scanner, which may even beep at you to tell you to load more slides. My old G4 Yikes (that was the code for it) runs an early OSX version and allows use of original NIkon CoolScan software as well.

    I used to be anti-revival on old posts, but it is difficult to notice the OP date, and I'm beginning to feel "what the hell", especially since the search engine here is one of the major not-so-good-as-it-might-be* aspects of P.net :rolleyes:

    _________
    * translation "it sucks majorly"
     
    Last edited: Jan 27, 2021

Share This Page