Scanning LF Negative 4"x5"

Discussion in 'Large Format' started by stephen_curran|1, Aug 20, 2010.

  1. I am just starting in with LF and am curious; does any one know how large a file would be of a scanned 4"x5" negative to make a large print say 44" x 54" ?
     
  2. The file would be about 600Mb at 300dpi.
     
  3. There is not 600 megs "worth" of info in a 4x5 negative; more like a few hundred.
    One can make 44x54" prints from any size file you want to. If that is just part of a hockey dasher board; it is sub VGA; ie Barbie cam.
    Practically; much 4x5 is really just scanned in the 2000 dpi region; this is a 230 meg file.
     
  4. Work it out:
    44x55" is an 11x enlargement, so assuming a 300ppi print, you'll need a 3300ppi scan. Assuming 3 bytes per pixel (8-bit RGB),
    In [4]: 4*5*3300*3300*3/1024/1024.
    Out[4]: 623.1298828125​
    So about 600Mb it is.
    Notably 3300ppi is beyond the optical resolution of flatbed/desktop scanners, so you can either
    • take it for a drum-scan, or
    • settle for less than 300ppi (reasonable: if you're pixel-peeping a large print that close, get a life), or
    • investigate alternative tricks (super-resolution/enfuse from several scans offset by a few pixels)
     
  5. I measured a 4x5 exposure. It is not quite 4"x5". Mine was 94mm x 118mm. Your results may vary. However with those dimensions in mind...

    44" @ 300PPI = 13,200
    45" @ 300PPI = 16,500

    13200 x 16500 = 217,800, times three for three colors equals 653.4MP for the file size. However it would be referred to as a 217.8MB scan.

    13200/94 = 140.43 pixels per mm or 70.2 lp/mm in a scan.
    16500/118 = 139.85 pixels per mm or 62.92 lp/mm in a scan.

    Those numbers are assuming that you use every last bit of the exposure. It would require a scanner capable of giving 3550+ real PPI, not the advertised specification. There are a few, but very few 4x5 lens/film combinations that will provide those numbers in the center of the exposure. Plus 70 lp/mm is quite possible in the center of the film, but at the edges, 50 lp/mm would be a very high number.

    Sounds like you need 8x10 film. Even though lenses that cover 8x10 don't get any where near 70 lp/mm, you'd be dealing with a lot more real pixels of information on the film as there are some 8x10 lenses that exceed 35 lp/mm at the edges (1/2 of the lp/mm needed on 4x5). 2000 PPI is quite doable with high quality flatbed scanners and would give you a little working leeway on the edges.

    On the other hand you can compromise on a little less PPI in the print. That is not something I like to do but have to sometimes.

    A. T. Burke
     
  6. As Beavis and Trumann said; one cannot polish a turd.
    You can do all the wacky math you want; most folks do not drum scan 4x5 at high numbers; it is often just wastefull
    A huge common average is just 2000 dpi drum scans for 4x5. It really is a newcomers thing to scan 4x5 at absurdly dpi's; or you are not paying the bill; or you have a 1 in 100 killer negative
    In all my scanning since 1989 for the public; all I can say is 600 megabytes is total bull dung for 4x5; you drank the Koolaid. It is turd in a punchbowl number. It is like saying Home depot has 80 Horsepower lawn powers; you have a factor of 3 to 4 error best case.
    You can get a 4000 dpi drum scan; or 3200, 2800, 2400, 2000; or even 1600 and 1200 too.
    The few clients that I have had that I farmed out 4x5 for 4000 dpi scans did it for ego reasons; I can downsize the damn things and loose NO details; in about all cases. As a printer; it gets old to see turd polishing still after 21 years of scanning.
    What a 4000 or 3200 dpi drum scan sometimes does is grab some actual details on axis; in rare cases. It is real typical that 1/3 way out one really only needed 2400 to 2000; and the corners maybe 1600.
    If you really want 600 megabytes of usefull real pixel; use a bigger format; ie 5x7 or 8x10 or 11x14.
    If you want real data recorded; strive to get some design margin. Use more film area, scan at lower dpi numbers. Pictorial images is what most folks print; BUT for who do scan tests like 1:1000 test target numbers; and using film curves where the response is low.
    If one scans 4x5 at 4000 dpi and 8x10 at 2000 dpi; or 35mm at a zillion dpi; 8x10 wins.
     
  7. If the question is how large of a file do you need for a great looking 44x55 inch print that answer is less then 10MBytes.
    9.8 MByte Image
    That would me a 44x55 inch print at 150 ppi. The pixels are very sharp and unless you have your nose in the print you would not be able to tell the difference between a print at 150 and 300 ppi. So the short answer is you need around 10 Mbytes for a great looking 44x55 inch print. However that just says what you could get away with, that stores the image in jpeg mode with a fair bit of compression and whereas you would never see the difference in a print that large disk space is cheap so keeping a tiff would make sense. And since disk space is cheap you might as well scan at 2000 ppi, which gets you up to 240 Mbytes.
     
  8. In making BIG wall maps that are say 42" by 72" or 84"; one can have files that are 1 gig in size. One has actual details; street names; tiny fonts. 4x5 is too small for this, it is created with either mapping programs, CAD, combos of many detailed aerials. If this extreme; one can be printing at a real 300 ppi or 360 or 400. That is required when roads are close and one has names. 150 ppi will just not cut it then.
    If it is a 72x72" image on a dentist office wall while you are in a chair; it might be 10 feet away. Thus instead of 300 ppi at 1 foot; one only needs 30 ppi at 10 feet. This means one only needs a 6x6cm negative scanned at 2000 dpi.
    The issue /trap about everybody falls into; is not defining the VIEWING DISTANCE. If one wants a great 72x72" print with gobs of details so one can be 1 foot way; go shoot 8x10 film or 11x14 film.
    In old process camera work; our old rule was never above 4x; thus for 72x72" one needed negative of 18x18".
    The concept of being conservative to get actual repeatable results is often not one on photo.net; thus one has endless talk of 4000 dpi scans; like they always somehow work all the time.
     
  9. Kelly, You've added a whole bunch of parameters and suppositions about the author's question. Unfortunately, no details were given as to what the intention was for the image.
    As someone whose primary photo income is from large mural size prints for designers, I typically have MF and LF images drum scanned to file sizes around 400 to 500 mb. Of course you don't need 300 ppi prints for some viewing distances, but my colleagues (who shoot mainly MF panos) and I are always working with files of that size.
    While you can print at 150 ppi, I personally don't care for the loss of resolution. You are right, in that the negative must be very good to get a high res image, and the workflow in PS to ensure the image is prepared for that size printing can be tedious, but that's what I trade in. Printing sub-par images at that size simply magnifies the problems. But the OP didn't specify much information about the intentions for his image, nor the workflow to get there. So forgive me for polishing the turd... I didn't know that's what I was working with.
     
  10. Regarding 400 to 500 mb scans, without seeing the pixels it is really hard to say if all those extra pixels are buying you anything. That is why in discussions like this I think it is a real good idea to at least post a few 100% crops. I have seen a lot of 4000 ppi scans that when down sized to 2000 ppi loose very little detail and what detail is lost would not be seen in a print.
    As for how many pixels are needed for a large print, without seeing the pixels it is hard to judge, if the pixels are sharp enough then you really had to get you nose right up to a photo to see the difference between 150 and 300. Where a lot of people think 300 is needed comes from having pretty soft pixels to start with. A print make with soft pixels at 300 ppi will look lots sharper then a 150 ppi print with the same softness of pixels. But that speaks more to the quality of the pixels then how many pre inch are needed.
     
  11. Michael, I think Kelly's main point, with which I agree, is that although many common printers natively print at 300 ppi, and this is about an 11.6x enlargement (at least from my 4x5 film holders, 96x121mm), that does not mean it makes sense to simply crunch the numbers and say you need a 3493 ppi / 1.26 GB (at 16 bits per channel, or 630 MB at 8 bits per channel) scan.
    In a recent photo.net thread (http://www.photo.net/film-and-processing-forum/00X4Nd), a know-nothing know-it-all posted a chart asserting that 8x10 film contains over 1 gigapixel of information. Basically, he derived this figure by saying if you scan 8x10 film at 4000 ppi, that's what you get. Of course, 4000 ppi is about 79 lp/mm, and very few (if any?) 8x10 lens-and-film combinations will capture anything approaching that. Indeed, at typical 8x10 apertures, diffraction will limit you to far less resolution. Recall that one of Ansel Adams' early photo circles was called Group f/64. Well an 8x10 at f/64 has the same depth of field as a 35mm with a roughly equivalent field of view at f/8. Anyway, even at f/32 (DoF equivalent to 35mm at f/4), diffraction is going to put a hard limit around 50 lp/mm, with some effect starting around 25 lp/mm. So assuming you want more than really shallow depth of field, the 8x10 is going to be limited to around 500 MP, assuming everything else (including both lens and film giving 100% MTF at 50 lp/mm--ha!) is perfect. So for almost all real-world 8x10 frames, a 1200 to 2000 ppi scan will get essentially all the real picture detail that is there.
    With 4x5 the case is not quite so dire because, all else being equal, you shoot with the lens two stops wider open and diffraction becomes the problem at twice as high a resolution. But at a common taking aperture of f/22, the combined effects of diffraction, garden-variety lens performance limits, and film MTF roll-off mean that substantial detail finer than about 35-40 lp/mm (about 1800 to 2000 ppi) is pretty unlikely. Scanning at a higher resolution than appropriate for the limits of the real detail in the film makes no sense.
    The upshot of all this is that the OP probably should get the 4x5 film scanned at about 2000 ppi, which would capture essentially all the real detail in the shot and give him about 172 ppi at the intended print size, which he could then up-interpolate to make the print (or perhaps better, if a pro is printing it, let the RIP do the up-interpolation). In 16 bits per channel RGB, such a scan would result in about a 400 MB uncompressed TIFF, and at 8 bits per channel, about 200 MB.
     
  12. I had a picture I wanted to do at 30x40 inches which is the largest that Chrome in San Diego could put on silver halide paper with their printer, a Chromira ProLab I believe. They scanned my 4x5 Astia. I had tested several lenses on Astia in a studio with a 1951 USAF chart to find one that would give 50 lp/mm, even near the edges at F: 5.6 and/or F:8. With the chosen lens I took the picture and had it scanned on Chrome's Scitex EverSmart Pro scanner at 2400PPI, necessary for 300PPI printing. It was very sharp right up to as close as my eyes could focus.
    Anything larger would have had to entail some compromise be it less PPI in the print or "over-scanning (scanning beyond the detail available in the original).
    A. T. Burke
     
  13. Thomas,
    When you say give you 50 lp/mm that has little meaning, what was the contrast at 50 pl/mm and what was the contrast of the target? It is all well and good to talk about number like 50 ln/mm but unless we see the pixels it mean very little. As I have said before a print at 300 ppi when the pixel as soft is really not worth any more then a print at 150 ppi when the pixels are sharp. how about showing us how sharp the pixels are that you are talking about, and not with a test chart but the real scene.
     
  14. Scott, when you say it would be helpful to see a 100% crop: how would that be meaningful if it is at 300 dpi, and monitors are 72 dpi? You'd lose a lot of sharpness in that exercise, wouldn't you? On 4x5 images, I believe I am getting something by scanning files at that size and printing the equivalent of 300 ppi. I have tried smaller sizes of equivalent negatives to see if I could save money, and I wasn't happy with the output I was getting from the smaller files.
    Now if you scan a 4x5 at 500 mb, and print 16x20" prints, I totally agree; you're wasting time and processing power.
     
  15. Michael, a 100% crop helps me what kind of sharpness you are talking about at 300 ppi. It is somewhat meaningless to talk about what ppi is needed for a print if we donโ€™t know how sharp the pixels are.
    For example if we print this image at 300 ppi the print would be fairly sharp, plenty good enough for a large print.
    [​IMG]
    If your pixels look about that sharp at 300ppi I would say you will get a sharp print.
    But that image is upsized from an image at 200ppi to 300ppi, the 200ppi image is here
    [​IMG]
    Since the top image came from the bottom one it is clear that if you can get a sharp print from the top image at 300ppi then you can get a sharp print from the bottom one at 200ppi.
    I have seen people argue for the need of 300 ppi and when we see just how soft the pixel are it is clear why they believe that they need 300 ppi.
    I hope this makes sense.
     
  16. It all depends on the printer, paper and tolerance for quality compromises.
    Specifically, 24"x30" test prints from a 5DII (150dpi) look awful next to 4000dpi scans of 6x7 (360dpi) printed on my Epson 7880. The printer's native resolutions are 360 and 720.
    It is not a minor difference. It is night and day. The 5DII is unacceptable and 6x7 is plain perfect.
    I am not sure I understand how some people find 150dpi acceptable for a print that is to be displayed. I regretfully printed at 24x36 for photographers commissioned work they took with 5DIIs to later find out they framed them for their clients and ended up on a wall. Not good.
     
  17. What I find in printing for the public is most all folks are *FIXATED* on the 300 ppi number; thus folks are always upsizing to this number.
    Their ego makes them BELIEVE that upsizing and scanning at absurd dpi levels creates more details
    Thus if their image sharp input for a poster is un-upsized image is 36x54 inch image at 50 ppi; I might as a printer upsize it to 100 to 150 to smooth it out.
    ****Most all folks believe if one upsizes more to 300 or 360 ppi; it will look better and the man on the grassy knoll will appear.
    Folks watch CSI New York on TV and see details come out as they process a crime scene image off a security camera or cell phone; thus 99 percent of folks think that if one scans at 5400 dpi or upsizes 10x; one will see details come out.
    Turd polishing is common. You would not believe what the general public does.
    Folks will take an image from a 3 megapixel digital; and upsize it so on image requires a DVD; ie one gets 1 to 2 gig images.
    If anything the public is dumber; folks FIXATE on 300 ppi and upsize it; even if the image is a dumb VGA screen capture for a giant poster.
    Thus there is a greater Goober factor in dealing with the public today in printing than in past eras. It is not uncommon to have folks upsize a file 100 times to hit 300 ppi; because always believe 300 is better. Thus if Kilroy's poster image is really just 15 ppi; they will upsize to 300. One can print a sample section at 100, 150 and 300 and see no difference; but they usually will not believe a sample either.

    It is like a whole mess of folks equate more pixels is better; even if one sees no difference at all. Thus on has a dumbing down of inputs and much hand holding.
     
  18. It was far easier 10 to 20 years ago to deal with the public about digital printing. Often in that era one got a digital file or a transparency to scan and the public allowed printers to do their jobs.
    Today everybody is now an expert;
    Thus if a client brings in a crappy Kodachrome shot with an Instamatic; I might look at it on a light box and see it is not so sharp.
    Today the client will often want a 4000 dpi film scan; or farmed out drum scan for an crappy original that really only holds say 1200 dpi worth of info; ie a dumb flatbed is overkill.
    One has the delicate situation where here I have scanned 35mm since 1989; and the know it all newbie client has a giant ego; and wants an expensive drum scan that is wastefull.
    One can show many folks on a light box that the original is just fair; and many folks still want a high end scan. If one just does it without mentioning it; one can get labeled as ripping folks off. If one brings it up; you deal with their fragile egos.
    The bulk of folks read that giant prints need 300 ppi and that 35mm holds gobs of info. Then they bring in a poor original; and want it scanned or printed at high levels.
    10 to 20 years ago the average Joe would look at their beloved slide on a light table and believe you that it only needed a moderate scan. Today many folks are stupider; yes dumber.
    A huge number of folks today seem "not to get it" that there are good pixels and useless ones; ie bloated.
    In a way a printer today is like a chap that works on automatic transmissions; and the customer now dictates all sorts of goofy things that drive up costs; add no value.
     
  19. Michael;
    Re "Kelly, You've added a whole bunch of parameters and suppositions about the author's question. Unfortunately, no details were given as to what the intention was for the image."
    One has the CORE tenet on photo.net that about 99 4/10 percent of all folks who ask these daily questions about :
    (1)"how big can I enlarge",
    (2)"what file size;
    (3) "what ppi"
    have no details.
    You see with an actual professional job; one has an actual client. One has a purpose for an image. Thus one can box in the distance of the print.
    One has on Photo.net many many thousands of the same question; that get repeated every day.
    It is like asking on Shipping.net how much string or tape is required to ship a box; but nobody provides any details as the boxes size or weight.
    Call up Home Depot and ask them what wood costs to build your gizmo. When they ask what the gizmo is; say what does this matter; just answer the simple question!
    Folks want an exact answer; but cannot provide any details; since they have zero goals.
    The basis of every one of these daily questions on photo.net is to ask a fuzzy question; and provide no viewing distance; no details; not purpose. It is in the DNA of most folks to want exact a black and white answer to a fuzzy question
    Most all folks will yak; walk over hot coals, debate; than actual provide any details.
    On has on Photo.net often the same type of question being asked and answered on a dozen active threads. A whole mess of folks give answers; when their really are no details. About 99 percent of folks are assumers
     
  20. Mauro,

    It would help if you could post a sample of the image that you found unacceptable at 150pp.

    Overall I find that many photographers look at prints in ways that the public does not. We tend to stick our noses right into the photo and are unhappy if we see softness from 10 inches away, which is find for a 4x6 inch print and ok for a 8x12 inch print. When you start getting to prints up in the 20x30 inch range people no longer view from 10 inches away, even 20 inches is fairly close for a print that size.

    So I see two different cases regarding 150 ppi, in one the pixels are soft to start with and the viewing distance is much closer than normal. In the other the pixels are sharp and the viewing distance is reasonable.
     
  21. Mauro; in order for 300 to look better than a 150 ppi image one has to have two things:
    (1) One has to be closer that 2 feet away
    (2) One has to actually have a usefull image at 300 ppi; not the common upsized stuff folks worship here
    I base this just on printing over 300k worth of images over the last 20 years.
    Typically customers who worry about 150 versus 300 creates the worst images; ie no soul. So much of their brainpower is wasted on what doesnt matter; that they often do not use an image with impact.
    These do not matter "It all depends on the printer, paper and tolerance for quality compromises."
    You really need a non bull-dung image that is really 300 ppi and not the typical upsized one.
    Almost NEVER is the printer the real limit; it is the lay publics images; often upsized.
    A core tenet on photo.net is endless talk of what ppi to send to printers; with little if any focus about delivering images that really hold details at 300 ppi.
    If the image really is only 100 ppi; upsizing to 300 does not make details appear.
    A core tenet on photo.net is the slacker ways of believing upsizing and 300 ppi brings more details. As Beavis and Truman said; one cannot polish a turd.
    The sad fact is that the bulk of the lay publics inputs for poster today can be printed on my brand new Canons; or 1994 Novajet and they limit of quality is NOT the 300 dpi 1994 Novajet that just supports about a 165 ppi image and NO more
    The limit is almost never the printer at all; it is the publics dumb images that are not the best.
     
  22. Kelly, you are correct generating a 360 dpi print master through scanning won't give a better output than 150 dpi master if the picture was out of focus or taken with a plastic lens.
    Conversely, an 8000 dpi scan from a good Velvia shot will contain more detail than a 4000 dpi scan.
     
  23. Scott, here is a print test you can run at home:
    00X7WU-271385584.jpg
     
  24. Mauro;
    The 35mm stuff I have farmed out that is for scanning above what a common 4000 dpi Nikon or Canon film scanner "will do" has at best been nothing to rave about.
    In a few few *rare cases* more details are pulled out
    Thus often a 5400 ,6400, or 8000 dpi high end scan really does nothing at all but waste gobs of money.
    I farm out stuff like this for lay customers to please their big egos. They are BS'er s at heart; thus their makeup craves BS scans too ! :)
    You really with a scan like this trying to pull out info past about 75 line pairs per mm on film ; which is a pipe dream one most originals. With a high end scan; details at the 75 line pair number might gain tad more contrast; if one looks hard.
    It is more of a pipe dream since folks who get 75 line pairs per mm do it with fine grain B&W and shoot 1:1000 test targets; with camera on granite blocks.





    ***Going past 4000 dpi is like vacuuming the kitchen floor eight times; instead of four times; in theory it is better.
    You pick a tad more crud; that one really would have to hunt for to tell any difference.
    A math major likes this type of stuff; you double or triple costs and gain 1/2 percent improvement; that in an actual giant print 1 out of 1000 folks might notice' maybe 1 in 100 if you SHOW them were to look.





    Scanning at absurd dpi numbers does feed many folks egos.
    One does an in house scan at 4000 dpi on a 3 grand Nikon film scanner; and the customer BELIEVES that a higher end scan is warranted.
    One has to tactfully try to steer Mr Ego from wasting money; or saying they are wrong. Thus I farm their "stuff" out for a higher end scan; and almost always not more details are captured, Having done this for along time; the trend is getting worse.
    A typical higher end like your 8000 dpi of Velvia from amateurs or pros *about never* contains any more details than a 4000 dpi one. This happens maybe in one out of 100 to 1000 cases.
    Having farmed out thousands of bucks worth of stuff for Mr Ego; his ego just flushes cash down the toilet. It is usually just turd polishing; folks want to believe that their beloved original has 160 line pairs per mm on film and a high contast; but the features are really at 50 line pairs per mm with some contast.





    To say that a 8000 dpi scan pulls out more info than a 4000 dpi one is stretch; a lie to a customer. It does sometimes in rare cases; in a few best case originals.
    If BS'ed like this to customers; I would not be in the scanning business since the 1980's.
    In Farming out stuff above 4000 dpi; the print shop here has to deal with Mr Egos deflated ego when a scan come back; since the more expensive scan of their original is about the same as the in house 4000 dpi one. Sometimes it is worse too. Some want me to farm it to another place; or eat the scan cost. It sort is like is an insane customer wants their lawn mower blade to be as sharp as sissors or a razor blade. They pay gobs extra in cost; and results are practically exactly the same.





    TRANSFER FUNCTIONS:
    *****The camera's Lens; FILM, and scanner have a response curve of resolution versus contast.
    Beyond 4000 dpi both film and scanner are down in response (contrast/ modulation) ; and now one has the product of THREE hokey things down in the weeds.
    ***A typical person who wants a scan beyond 4000 dpi believes in their mind that then will get 4 times the details.
    They ignore the response rolloff; in their mind the 3 bogeys have high contrast at hundreds of line pairs per mm; thus getting details past 4000 dpi is a slam dunk.
    It really is more like completing a long bomb in football; or a grand slam in baseball; or a hole in one in golf; or a hat trick in hockey; it is not something that happens with ease all the time for the average pro player.





    *****Thus with the public at least one has to say that:
    scanning beyond 4000 dpi; or
    polishing ones riding lawn mower; or
    removing a 1 oz letter from ones car
    ***will probably not noticeablly improve performance in a radical way; ie you probably will not notice.
     
  25. Mauro; your left print is already blurred
    Here is your captured image downsampled by 150/360; so it is now NOT UP SIZED.
    you have stacked the deck for you argument; since the un-upsized image is really not sharp at all:
    You started off with a blurred image:
    The text in the screen capture is sharper than the image itself
    you stacked the deck!
    You lens faults show
    [​IMG]
    Here is a direct 100 percent pixels non upsized image; see how it is sharper
    you can see fine cracks in the fireplace and you have fuzzy flowers,
    [​IMG]
    non cropped 35 megapixels
    \[​IMG]
     
  26. Here is the crop without upsampling:
    00X7Xl-271409584.jpg
     
  27. I agree most times 4000 dpi captures at least 80 percent of the detail available on my film. That's why I scan almost everything with a Coolscan 9000.
    Regarding the exercise for comparing prints; there is a large difference that I see on the 24x30 print between the 5DII at 150 dpi and 6x7 film at 360 dpi (native scan).
    The 5DII shot above is not mine (or my lens). It was provided to me for use in the forum as a good sample of sharpness from the 5DII. Not sure what you mean by stack the deck but please provide your best sample of a 5DII upsampled from 150 dpi to 360 dpi. I will gladly print it to evaluate it.
     
  28. What camera did you use for the 35mp shot?
     
  29. Just for clarification Kelly, is it your opinion that a 24x30 print from a 5DII (150dpi) has the same detail and quality than a print from a 4000dpi scan from 6x7 film (360dpi)?
    With large responses I sometimes miss the overall statement.
     
  30. Mauro; Here is my debate or issue:
    ***The 5DII shot of yours (left) just does not look that sharp;
    that is why it can lead somebody in the mire for comparing to another image.
    Maybe the lens was not the best; or not at a good f-stop. I just mention all this because the fuzzy white hairs or whatever look not that well resolved. ie one could downsample and not loose details. Thus it appears up up-sampled an image that is already not sharp.
    Maybe you can find another 5DII image that is sharp to use as a comparison. This would be better.
    At full resolution the Canon 5DII is a 5616 x 3744 pixel rig. That means one would get a 156 ppi image at about 24x36mm; close to you example at 150 ppi. The pixel pitch on the sensor would be 36mm divided by 5616 pixels; ie 36000 microns/5616 pixels= 6.41 microns. I got 156 ppi from 5616 pixels/36 inches=156 pixels per inch. ie one has a 24x36 inch image at 156 pixels per inch without up sizing.
    My shot is with a 36 megapixel scan back from circa 1996; an old Phase One 4x5 scan back that really scans a 7x10cm area. It makes a 5000 x 7000 pixel image. The pixel density is way larger than a 5dII; it is 100mm/7000 pixels; ie 100,000 microns/7000 pixels = 14 microns. This shot is with a 1947 4x5 Speed Graphic with a 150mm F9 Apo Ronar about at F16.
    ***I TOO see a big difference between your left and right shots; the left one even when I reduced it to a non upsized version looks not that sharp.
    Thus in summary from here it looks like you have a crummy canon 5DII image; ie one not so sharp.
     
  31. Here is another thought I had on a walk this morning: What does it mean when we scan at 4000dpi vs. 8000dpi? To me, we're not looking for more image resolution, lens resolution, or film resolution. We're looking for silver resolution. Are the edges of the grain sharp, or not?
    A 4000dpi scan is not going to give you more image quality than a 2000dpi scan, but it may give you more grain resolution. It depends largely on the scanner quality as a major factor. So while we're talking about MTF charts and the like, that doesn't seem terribly relevant to what I look for in my hi-rez scans. If I took the image with a Kraponar lens, then the image is never going to get better with a great scan.
     
  32. Mr. Wilson...

    1. 50 lp/mm was rounding down. With the 1951 USAF target there is not a converging curve of lines. There are lines printed at specific distances apart. As one is able to distinguish each pair of lines the the value jumps up more than an integer. The difference in value is relevant to and changeable by the distance from the nodal point of the lens and the lens "length", in this case 210mm. In preparation for writing this response I looked up my lens here:

    http://www.hevanet.com/cperez/testing.html

    I think the numbers I got for F:8 were about the same as he got for F:16... 67, 67, 54, which gave me better than 50 at the edge. I am not sure about my numbers this long after the fact.

    2. Contrast of the target? I do not know the number that Edmund Scientific assigns to their paper 24" x 36" (nominal) target, do you? Giving the publisher the benefit of the doubt, I would think that their target is a lot closet to the 1000:1 standard rather than the 6:1 standard. It looks very black and very white to me. That last sentence is the only honest one I can make without assuming that which I do not know.

    More significant is the person's judgment and the equipment he/she uses to view the results on film. Scanners can give a lower value by being out of registration with their scanned material. Their software can also help the results look more definitive than they really are. I use a 60x, 100x and 300x microscope to view the film. After that it is a value judgment as to what constitutes a definable line. Also unless your eyes are older than 90 years, you may get even better numbers than I did. Younger eyes will see a greater difference.

    Mr. Kelley....

    "What I find in printing for the public is most all folks are *FIXATED* on the 300 ppi number; thus folks are always upsizing to this number."

    1. Not me... I like to use what the printer uses. I also use scan values and crop so there will not be any rescaling or resizing as the data enters the printer. Many printers are set up for even multiples/divisors of 300 PPI (NOT dpi). They do the least interpolation (a chance to reduce sharpness) at 75, 150, 300 & 600 input. Canon and HP inkjets use that standard. My best ink jet is an Epson which used the 90, 180, 360 and 720 chain. I also have a Fuji Pictography 4500N in my basement. It uses 200 and 400 PPI on silver halide paper. A 12 x 18 sure looks good at 400, IF I did my job with the input.

    2. Yes, upsizing or up scaling or whatever does not add to the original information base. High end programs like Genuine Fractals can make an upsized scan look better on paper but cannot add to resolution.

    3. IF..one starts with high definition film, having shot higher contrast views, used a tripod, used an excellent lens and gotten the exposure right, an 8000 PPI scan can get more real data than a 4000 PPI scan. Even if there is the equivalent of only 4000 PPI on the film source, an 8000 PPI scan can do better because it reduces registration errors. But how often does one find input with 4000 PPI (78.7 lp/mm) worth of definition on film? Looking at the hevanet reference above you will see many lenses touted as good to great by "professionals" only having 20 lp/mm or less at the edges. At those values even the 2000 PPI scan is overkill.

    In general...

    PPI only came upon us with the digital age. I like prints through glass and good glass at that. The square boxes of scans and pixels cannot really represent the irregular shape of the dye clumps or silver grains. Square peg, round hole?

    Furthermore, how much sharpness is required? How mush unsharpness or fuzz is acceptable. The last time I was at a big city photographic exhibit was about five years ago in San Diego. THE artistically "in" photographer of the big city of San Diego, CA had an exhibition/sale in the theater building where I was to attend a play. Most of the stuff was between two feet and four feet on the longest side. A few of the best reached my sense of poor quality. The rest, even worse, looked like way over enlarged Lomography. Was that the in look? If so, cutting off heads, poor composition and nothing subjects must be in also. I would not hang a picture of my ex mother in law, if taken and printed that poorly, over my toilet.

    The prices seemed awfully high, but then there weren't empty spaces on the walls either. They were not exactly blowing out the door. Is this level of professional work typical? Are we so dumbed down as a society or as photographers? Hopefully his reason for the show was that he was the panty-boy for one San Diego's more prominent decadent dowagers.

    Real photographers still use film and print through glass.

    A. T. Burke
     
  33. In all my scanning since 1989 for the public; all I can say is 600 megabytes is total bull dung for 4x5; you drank the Koolaid. It is turd in a punchbowl number. It is like saying Home depot has 80 Horsepower lawn (m)owers
    I bet you $100 that it would sell.
    And I don't mean something out of a movie or a redneck tailgater party; I mean a regular push-mower with 80hp printed on it.
    Lawnmower World Land Speed Record
     
  34. Kelly, the 150 dpi image from the 5dII needs to be upsized (by PS or the printer) to 360dpi for the Epson 7880.
    There is no way around it.
     
  35. Regarding printing at 360 dpi vs 150 dpi, you can pick the sharpest 5DII picture available on-line and print it at both 150 dpi and 360 dpi.
    If you don't think there is a loss of quality on the print, then 150 dpi is fine for you.
     
  36. I don't know if anyone has already posted this, but just in case here is a link to some large format lens test results. I think these are 4x5 results.
    http://www.hevanet.com/cperez/testing.html
    A few of the lenses resolve 80 lpm or almost 80 lpm in the center. Most are around 60 lpm, more or less.
    Of course the edges are worse. However, I think that ideally one would want to come as close to preserving the center detail as possible rather than just the worst part of the image.
    Life isn't always ideal, especially if one has financial limitations (as most of us do), but at least here are some real numbers from actual lens tests that can help frame the discussion.
     
  37. Mauro,
    I took the DPReview 5D mkII ISO 100 sample (E5D2hSLI00100_NR_STD.JPG), scaled it in Photoshop to 24x36" at 360 ppi to match your 24x30" size for 6x7, and then took a 342 x 1064 crop from one of the bottles. Unfortunately I can't post it directly to photo.net (I got in trouble for that once before), but I believe I can post a link to it. Here it is next to your 5D mkII and MF crops, both unsharpened and with Smart Sharpen 1.5 x 100%:
    https://docs.google.com/leaf?id=0Bz1sfHfXHVDCNWIxOGRmNjgtOWQ0MS00YmRkLTg2MGMtOTM0NzFkMzQyOTQw&hl=en&authkey=CJHd3tMN
    The thing I abhor about this is that it's not the same target under conditions which are equal to whatever degree is possible. But it's clear that the 5D mkII is capable of recording more detail and providing greater sharpness than is observed in the 5D mkII sample you presented next to your MF scan.
    At their best, a 6x7 film scan file will offer higher IQ than a 5D mkII file. And this higher IQ will be quite noticeable given challenging subject matter and very large prints. I find it takes three 7D files stitched to clearly match or exceed a top notch MF scan of a challenging subject, which gives some indication of where MF sits.
    That said, I've printed crops from the various Rand McNally tests we've shot and presented in past threads, including my 7D tests, Les Sarile's MF scans, and one of your MF scan crops. To my surprise at the time, there was very little to discern between them at 24" and even 30" equivalent size. The MF samples were better, but that does not become clearly evident or important until one prints at sizes larger than about 30". This is consistent with my observations of prints of real world subject matter. (Note that I am assuming optimum processing when enlarging the 7D file, particularly sharpening and local contrast enhancement. Also note that RAW files yield higher resolution and my comparison involved a 7D RAW file.)
    Regarding the larger issue of dpi/ppi in this thread: 300 ppi is not a magical number. Neither is 360 ppi. Humans can discern much finer details given monochromatic line art, and generally struggle to discern that much detail given lower contrast colors. Optimum ppi depends greatly on subject matter and viewing distance. And sharpness of coarse details can matter more than resolution or sharpness of finer details, again depending on subject matter, viewing distance, audience, etc.
    Stephen gave us nothing to go on in terms of subject matter, audience, quality of the film, etc. But odds are he can get away with a roughly 200 ppi print, or 2000 ppi scan.
     
  38. That's true Daniel, unless one is close (a couple feet) to the print the difference may not be noticeable between 150 dpi and 360 dpi. That's why I point out that it is relevant for artistic display but it may not be for commercial or casual applications.
    I will go to the link and print it.
     
  39. I don't view this as a film vs digital thing, in fact digital has nothing to do with this. What I see is people who believe that you have to have 300 ppi no matter what and then show samples that are very soft at the pixel level, yeah if you are using soft pixels you better have 300 ppi.
    Mauro your image of the leaf that is scanned at 4000 ppi is very soft so much so that when resized to half resolution and back up it is very hard to see a difference. In the link below you can do a mouse roll over to flip between the original and an image that has been down sized to half resolution and back up, the text show a change but hardly any in the image. Clearly there would be next to nothing lost using a 2000 ppi scan in this case.
    Link to comparison

    Now if we do the same test using a sharp image the difference is huge
    Link to comparision using a sharp image
    When we talk about how many pixels are needed we really need to talk about how sharp the image is at the pixel level.
     
  40. Sending an image to a printer as a 8x10 inch image at 300 ppi goes back before photo.net; before most all here even saw a digital camera. It is a 1980's number. Think MTV time frame.
    If one takes a National Geographic printed cover; it measures about 160 to 175 line screen. You can measure this with a C-Thru Ruler line screen device. You rotated the template until one gets a star pattern; then you read the line screen number.
    A higher line screen is used for a higher quality printing; one done with dots; ie offset lithography.
    A Line Screen is the measure of how many halftone lines are printed in a linear inch. The value is expressed as Lines Per Inch ie LPI. You use CMYK inks. In Black and White; ie where only a black dot pattern is used on white paper; this process goes back 200 years.
    National Geographic is about 160 to 175 line screen; a regular magazine might be only 120 to 150. Newspapers are in the 85 region. A cheap flyer might be only 65 line screen.
    With papers that absorb a lot of ink; the paper limits the line screen level because the dots will merge to a blob
    In Inkjet 15 to say 22 years ago if one had a "600 dpi class printer"; a rough rule of thumb is it would only support half this number as the ppi of the full size image/ print sent to it. Thus a 600 dpi printer would only support a 300 ppi image. My 36" wide color inkjet printers from 1991 and 1994 are just like this ; both are 300 dpi devices. If I send images 24x36" SHARP IMAGES WITH GOBS OF DETAILS at 100, 120, 140, 150, 160, 170. 180 ppi that are NOT UPSIZED; about 160 is about all it will support. Higher numbers buy one nothing at all; no more details appear.
    The 300 ppi of the prints image number at actual target size comes about because that is what is required to be sharp when about 1 foot away; when a printer of at least 600 dpi is used. *IF* one makes a mess of sharp images say at 600 pixels per inch, 550, 500, 450, 350, 300, 250, 200, 150, 100; and one has a mess of folks examine them; you well get a mess of answers. NONE of this is anything really new; it was done when Reagan was in office. It was done before digital cameras existed. It goes back to the DOS era. In Fax machines it goes back into the 1960s and 1970's!
    You will get "different answers" depending on the subject matter. With our 1990 DOS engineering copier scanner; it was a 400 dpi scan; and about a 400 ppi print. When one had fine features; a scan at 400 and print at 400 reproduced more details than scan at 300 dpi and print at 300 ppi. This whole jargon probably bothers newcomers. The reason 400 is better than 300 its that one has a high contast line work one is dealing with; ie non pictoral. IF one made a greyscale image with a buckshot pattern and sent it the 1990 machine; its real resolution is far less; about only 150 to 200 since one is making the image with a dot pattern.
    In maps that I make and print on a 11x17 20k buck color copier; the device is a 600 dpi class printer. With a combo map that is part photo, part topo map and part my custom text; the fine text and fine lines on the map govern the "ppi " of the image. A 11x17 inch image at 300 ppi looks great; *BUT* 350 to 360 does allow some more details; 400 is to the point where practically only a few might tell if one looked at page for 5 minutes; and compared back and forth the 360 versus the 400 one. About 1/2 of the folks see no difference.
    With EVERY printer that I buy; I just run my own tests to see what the printer will support; and use others comments as only a rough starting point. Having done this now since the 1970's 36" pen plotter era; I find it interesting how folks debate printer numbers without an tests at all.
    Without any mention of the paper used; you all are are really hip shooting at the higher ppi settings to a printer.
    The paper has a strong influence of the resolution supported in lith printing and inkjet printing; it is often the gating item today; more important than the printer itself.
    With a big new 54" printer here; it is called a 1200 dpi device for the nozzle pitch; and the Print resolution spec is hawked as "up to" 2400 x 1200 dpi max. If one sends a USAF target so it is like dime size; glossy high buck paper here with about no bleed shows it has details in the 600 to 800 ppi range; only if one gooses the test to lighten up the black a tad to reduce micro bleed. With a pictorial image; the support practically is way less; about 300 to 600; depending on the paper and subject matter.
    About 90 percent of the giant posters I do are from files that when not upsized are in the 40 to 120 ppi region; for color posters/pictures that are 24x36", 30x42", 36x48" sizes.
    Most folks do not own 20 plus megapixel cameras; most folks 35mm slides hold no more info past a 4000 dpi scans
    EXAMPLE : When somebodies 24x36mm slide is scanned at 4000 dpi for a 24x36" poster; it works out to about a 158 ppi image at 24x36 inches. Whether if it is the 158 ppi image; or upsized to 250 or 360 ppi ; here I never see any real difference; even after 20 years of using Novajets; Lasermasters, Canons, HP's , Epsons, Oce's etc. One is just "upsizing" an image that typically really not even a 4000 dpi anyway. You are just polishing a turd that as already polished. One really does it to please a customer. Folks watch CSI and believe that making pixels up helps make the license plate appear. In practice one just upsizes smaller sub 150 ppi images to about 150 to 200 for smoothness; not details.
    If that 35mm slide scanned at 4000 dpi is for a 36x54 inch poster; it is about a 105 ppi image at 36x54 inches. If the viewer is at 3 feet; 105 ppi is fine. If it is some artsy print one might upsize it to 150 to 200 to smooth it out; 300 will look exactly the same; it just makes a lay useless boated file.
    The average person believes that if that 36x54 inch print is printed with an image upsized to 360 ppi at 36x54"; that it is ALWAYS better. They believe this no matter if the image is from a Hit camera; cell phone; or common 4000 dpi 35mm film scan. Most all the time it really is turd polishing. *IF* one really has a naturally killer sharp non upisized image that is 200 ppi at 36x54 inches; a mild upsize and mild sharpening and a 300 ppi or 360 ppi image may look a tad sharper.
    The real problem is folks "BEST CASE" the entire chain of events
    As a practical a matter few in any folks really even get 50 line pairs per mm on film in pictoral images.
    The average person on the dpi Koolaid agenda quotes a test report showing their 50mm F1.4 lens recorded 78 line pairs per mm back in 1983 at F8; shot with Panatomic-X and 1:1000 USAF test targets. They then bring in a iso 800 Kodak Max zoom color negative of their kid at soccer. It is underexposed; the DOF is shallow; it was probably shot at F2.8; one has some micro subject motion too. With a loupe one can tell it doesnt not hold gobs of info. They want a farmed out drum scan; but you know that a 2700 dpi film scanner would be overkill. You scan at 4000 dpi on the Nikon Coolscan please their ego.
    Now then it gets to the giant poster; they bring up upsizing and what dpi/ppi to print at. One cannot tell then that they are stupid. Their 35mm negative really only holds what a flatbed will capture; ie about 1800 to 2000. Thus one really might have only about 40 ppi worth of OK pixels at the target giant posters size of 36x54"; but they want it printed at 360 or 400 ppi "since it is better".
    The bulk of folks who want posters want to dictate to upsize; "because it is better". One has a dumber customer than 15 to 20 years ago; folks are in to BS more; ie most all think more pixels are better. No wonder the world is in a slump; folks lack brains. Customers who Know nothing or know alot and have been around awhile are the easiest to deal with. Folks who know some often want to dictate how to use a printer; that they have never use before. It is like if Goober shoots a duck and brings it to a 5 star eating place and dictates how to cook it; in an unknown kitchen.
    There is a point with each scanner and printer that higher levels buy one nothing.
     
  41. To all responses, thank you, I'll have a lot to absorb.
     
  42. Please allow me to add the following insights from practical experience.
    - The labs where I have had large prints made print at 200 DPI, not 300 DPI, when dealing with images larger than 30 inches on the long side. WCI claims that their large Chromira prints actually look BETTER when printed at 200 DPI. The printer's internal algorithms seem to like 200 DPI better than 300.
    - I've seen 15 foot wide prints made from 35mm film. Yes, I said 15 FEET. They look amazing from the other side of the room. From three feet away, they look like unintelligible globs, i.e. like magnified newsprint.
    - Viewing distance matters, there's no reason why a sharp 4x5 exposure cannot be printed to 55 inches when scanned optimally. A 300 MB 8-bit scan or a 600 MB 16-bit scan should be just about right. It won't look as sharp as a 30-inch print when view from six inches away, but it will look just fine from two or three feet away. How close do you have to stand to a 55-inch print, anyway? You can't even see the whole thing from a viewing distance of two feet.
     
  43. I found a web page that discussed some of the issues being discussed here, and it uses some experimental results. Here is the link.
    http://www.clarkvision.com/imagedetail/eye-resolution.html
    A couple or relevant points from the web page:
    Based on the experimentally measured acuity of the human eye the author of the web page states "Consider a 20 x 13.3-inch print viewed at 20 inches. The Print subtends an angle of 53 x 35.3 degrees, thus requiring 53*60/.3 = 10600 x 35*60/.3 = 7000 pixels, for a total of ~74 megapixels to show detail at the limits of human visual acuity." To apply some interpretation to this, if you have three colors and 16 bits per pixel per color you have an estimate of 444 megabytes.
    The author also comments on lines per inch:
    "in a recent printer test I showed a 600 ppi print had more detail than a 300 ppi print on an HP1220C printer (1200x2400 print dots). I've conducted some blind tests where a viewer had to sort 4 photos (150, 300, 600 and 600 ppi prints). The two 600 ppi were printed at 1200x1200 and 1200x2400 dpi. So far all have gotten the correct order of highest to lowest ppi (includes people up to age 50)."
    What he doesn't say here is the viewing distance or the print size. However, he links to one of his own web pages that gives a little more information:
    "In dim lighting, such as a 60 watt light bulb at 3 feet (~1 meter) I can't tell any difference between the 300 and 600 ppi prints. But in good lighting (daytime near a window, or typical office lighting) the 600 ppi print is noticeably sharper."
     
  44. The mouse over is very cool Scott.
    I see a difference on the Ektar scan but not on the 5DII. This is expected since the 5DII does not have that level of detail to begin with since it is already upsized.
    Even with the Ektar sample, I think the subject on film was probably lower frequency than 4000 dpi.
     
  45. Scott, here is an example with TMAX 400 you can print to evaluate.
    Also consider than an original capture at 2000 dpi won't be as good as a 4000 dpi capture downsampled.
    Can you do the mouse over thing with the two crops of TMAX 400?
    00X7xc-271823584.jpg
     
  46. Mauro,
    Feel free to our the mouse over page, you just give it the url to two image at the bottom and the text for each and hit Submit. This will make a rather large url but one that can be used as a link in Photo.net forums. To get the page without image already on it just go to http://sewcon.com/compare/
    This page only work well if the two images have the same dimentions.
    The cool part is I don't load the photos onto my server, I just past the links on and so the bandwidth for me is very low.
    As for the Ektar scan, this is a very small difference between them but not much, the main point is that 300 ppi of a fairly soft scan like that are not the same as much sharper pixels. For example a print made from a 4000 ppi scan of 35mm compare to a 1000 ppi scan from 4x5 will clearly show a much sharper print with the LF. In fact I believe that 170 ppi print from a 1000 ppi scan will look sharper then a 300 ppi print from a 4000 ppi scan. The point is the sharper the pixels the fewer of them you need per inch.
     
  47. Mauro,
    I am out of sync with you, if you can post the two crops as separate photos then I can easily to the mouse over thing.
    For that matter you can also do the mouse over thing as long as you have urls to the two crops. BTW I assume you know you can right click on the image to get its url.
     
  48. I did not know how to get the url or to do the mouse thing. Trying it now...
     
  49. 1st image
    00X82c-271883784.jpg
     
  50. 2nd image
    00X82f-271883884.jpg
     
  51. Here is the overlap (i missed by a pixel or two). It is easy to tell how some of the eyelashes or facial hair merges or gets distorted with the transition. More obviously the eyelashes are thicker than in the original.
    http://sewcon.com/compare/index.php?photo1=+http%3A%2F%2Fstatic.photo.net%2Fattachments%2Fbboard%2F00X%2F00X82c-271883784.jpg&text1=1st&photo2=http%3A%2F%2Fstatic.photo.net%2Fattachments%2Fbboard%2F00X%2F00X82f-271883884.jpg&text2=2nd&B1=Submit
     
  52. You can see a difference which shows that there is some more detail in the 4000 ppi scan then in the 2000 ppi, which is not a surprise for BW film. But the 4000 ppi scan is not making as good use of its pixels as the 2000 ppi image is. A good test for this is to take an image scanned at 2000 ppi and down size to 1000 ppi and back up and then to a compare.
    As an example, and trying to keep this all about film, here is a film scan I down sized until it was very sharp, then compared the original to a copy that was down size to 50% and back up. The Difference is pretty striking.
    Link
    The best was to really see how much of the possible detail a film scan is capturing would be to shot the best shot you can with a good sharp 50mm lens, then shoot the exact same scene with a 200mm lens. Scan both at 4000 ppi but down size the shot with the 200mm lens down to 25%. The shot with the 200mm lens will be dramatically sharper, what this shows again is that not all pixels are the same, a 300 ppi print from a 4000 ppi scan is not worth nearly as much as a 300 ppi print from a 2000 ppi scan. And of course if we go up to 300 ppi print from a 8000 ppi scan, well it is worth less than a 100 ppi print from a 2000 ppi scan.
     
  53. This is really easy. A straight 16 bit scan at 2400 ppi scans below grain thresholds, and provides for an approximate 240 output to print at 40x50. With film flatness and lens diffraction, you rarely obtain much more than this.
    For a 16 bit file at 2400 ppi from 4x5, you get about 690mb....or 345 in 8 bit. Obviously, for the best tonality and to avoid scanner noise, it's best to scan in 16 bit....especially with consumer scanners.
    An output to 300 or 360 ppi on print isn't necessary at that size....and the data is rarely real detail anyway....unles your shooting with B&W, not stopped all the way down, and using glossy papers.
     
  54. For 6x7 and 35mm there is a clear advantage to 4000 dpi. Here the bottleneck is the scanner, then the film then the lens.
    As Dave said for 4x5 probably not. Here, at 4000 dpi the bottle neck is likely the lens with its limitations, aberrations and diffraction, then the scanner and lastly the film. At 2000 dpi the scanner is probably back to being the bottleneck.
     
  55. Agreed Mauro,
    I do not recommend scanning below 2400 ppi. At that size, it's easy to go over 600mb in file size. The beauty of a 2400 ppi scan from 4x5 is that it's grainless with modern day films like Astia, Velvia, TMax 100, Delta 100, Neopan 100, etc.
    For MF though, you need 3200ppi and up to get the details off the film as MF lenses typically are sharper, film is held flatter, larger apertures can be used while still maintaining DOF.
     
  56. Mauro RE:
    "For 6x7 and 35mm there is a clear advantage to 4000 dpi. Here the bottleneck is the scanner, then the film then the lens."
    Scanning 6x7 and 35mm at a 4000 dpi real film scanner pulls out about all the practical info these sizes hold.
    There *IS* an advantage going to this 4000 dpi level with a tack sharp original; but really not much more as a practical matter.
    The "bottleneck" is about never the scanner at this resolution. It might be in one out of 100 to 1000 scans for high end pros; or one out of a million to 1000 scans with an amateur. For a typical amateur who uses 800 iso color print films; I have yet to see where beyond 4000 dpi matters; most only needs about 2700 dpi film scan or way less.
    When I look at actual many many many thousands of 35mm and MF I have done for pros and amateurs since 1989; here I can just say the "bottleneck" is almost NEVER the scanner at the 4000 dpi level ; it the the shooter's non perfect input.
    A *real* dedicated film scanner is almost never the bottleneck at 4000 dpi; the bottleneck is the shooters sub perfect piece of film.
    A film scanner at 4000 dpi is almost never the weak link.
    Very little if any even pro stuff needs or warrants scans above 4000 dpi; ie one in 100 or 1 in 1000 inputs really warrants me to farm them out for a drum scan. One typically does this to please the customers ego; often there is really little or NONE extra captured; thus Mr Ego's bubble often gets popped when reality sets in; ie their masterpiece is not chock full of info past what a Nikon coolscan 4000 dpi device can capture.
    With 35m and MF the most common bottleneck in the resolution chains is the shooters beloved original does not tax the limits of the 4000 dpi class real film scanner
    Nikons 35mm film scanner of 1988 was about 4300 dpi device; later they DROPPED the dpi levels newer scanners to be 4000 and 2700 regions.
    5400 dpi film scanners were a marketing flop; only a few makers ever made them.
    With a Nikon 4000 dpi film scanner; it is going to capture about all the usefull details of folks:
    *"culled out super sharp/great original pile"* ; maybe just maybe one can farm the few percent reculling best of the *"best of the best"* out for drum scans at 5400 to 8000 dpi. Then one has to really search most of the time really to see any practical benefit. You have to ask yourself is if really work cutting the yard twice; or polishing the riding mower to make if faster? To a math major; if the mower burns 1 millionth less gas it matters. To a resolution freak with scanning; if one out of 1000 folks can see the difference in the final print; it matters.
    At the 4000 dpi scan setting level with a dedicated film scanner; one is already practically at most all folks limits for film already.
    A 4000 dpi real film scanner is the bottleneck in only a few shots in folks lifetime; these can be farmed out for a high end drum scan if warranted.
     
  57. In drum scans of 4x5 stuff; a 3000 or 3200 dpi level often pulls out no more info than a 2000 one; except for the on axis stuff. Drum scans of 4x5 at the 2000 level are common; so are 2400 and 2800. drum scanning 4x5 at 3200 or 4000 is done way less; ones costs are skyrocketing and one picks up little extra info
    ***EXAMPLE relating to ones pocketbook:

    You have to ask yourself does one want to have five 4x5's drum scanned at 2000 dpi for 200 bucks; or have them all scanned at 4000 dpi for 1000 bucks???

    What if one wanted 100 4x5's scanned? a 2000 dpi set is thus 4 grand; a 4000 dpi set just 20 grand
    Is 16 grand chump change to you or your client?

    The "question" about whether to waste money sinks in more if that 1000 buck bill is revolving at 29.99 percent on ones own credit card; or you client or wife questions if the extra 800 bucks was needed.
    As scan prices have dropped; folks ponder costs less. All photographers are loaded; and client pays one in gold bars too.
    In commercial work; costs matter. There is a live real client that cares about costs. They do not want to pay extra for things that adds no value, Thus if an 80 buck 2800 dpi 4x5 scan is totally ok for all applications; they do not want 120 bucks wasted to make it a 4000 dpi scan. Folks on photo.net seem not to understand costs; or "what is good enough" so it does not matter.
    If you have an unlimited budget; one can just have all ones 35mm and MF films high end drum scanned.
    Since decent film scanners came out 21 + years ago; less stuff is high end drum scanned. This is because 4000 dpi class real film scanners practically pull out all the info in a 35mm original.
    None of this is anything new; Nikon found this out with their first 35mm scanner when Reagan was President; this is older than Photo.net. It goes back to Barney Scan and Photoshops birth
    Twenty years from now folks will be asking how much info is in a 35mm film original; the question will be asked forever
     
  58. Kelly,
    What you are basically saying is that very good is good enough for practical purposes.
    However, it is an experimental fact that in some careful tests 100 lines per mm has been demonstrated for some camera-lens-film combinations, at least in 35mm. One of the photo magazines showed this some years ago. To faithfully digitize an image such as this would require a theoretical minimum of ~5000 true pixels per inch, and to avoid Moire patterns in some images, as well as to deal with the fact that technical limitations would limit the true resolution of an instrument to somewhat less than that which would theoretically be achieved at 5000 pixels per inch, one would have to scan at somewhat higher resolution than this.
    While it is no doubt true that few practical images would reach this level of resolution, it is nevertheless true that it is possible to achieve this level, and it will be achieved in some images.
    Therefore, to settle for an instrument with lower resolution is to accept the fact that you are settling for something less than maximum possible performance. Hence, your strategy is equivalent to saying that very good performance is good enough for most practical purposes. It is not equivalent to saying that very good performance is equivalent to good enough for all cases.
    Then there is the rather complex issue of grain aliasing, which is a topic I admit I am not completely comfortable with and don't fully understand, but which seems to have something to do with the interaction between grain structure and sampling rate.
     
  59. Alan, in some case there is a benefit to higher resolution scanning. I have had some photographs that were printed at 40x50 that I printed to a fibre based Baryta paper. It was printed at 360 as opposed to 240 or less that would be normal for most uses. The film stock was Ilford Delta 100, processed in PMK Pyro. There was pretty much no grain. The scan at 3600 ppi should more detail at that print size than did the 2400 ppi setting. So yes, in some cases good isn't good enough.
    There can be a greater perception to depth with a higher rez scan.
     
  60. I have no issue with people scanning at 4000ppi to get a bit more detail. What I have an issue with is those people trying to claim that a large print needs 300ppi, based on a film scan at 4000ppi. Putting it very simply going from a 2000 ppi scan to a 4000 ppi scan does not allow you to print twice as large, in many of the scan I have seen you might be able to print 5% to 10% larger, in some cases a bit more but never close to 100% larger.
     
  61. Alan;

    It is true that with a lab run/special test one can get high numbers with miles per gallon and what 35 mm film resolves.
    I here have done these "what will it do" with about every scanner I buy. I have some ultra high glass etched test targets I had made over 2 decades ago
    I too have gotten 52 to 102 miles per gallon in a 1980's Camaro. You just pump up the tires to 50 psi so they are about to pop. You have everything all warmed up for the test. You use a 4 cylinder model with a 5 speed. You find a flat road on in the California Desert in the middle of nowhere. You start the car and lug it an get it up to 45 MPH and cut the engine off and coast. You restart it about 20 mph and do it again.
    In tests of Normal lenses for 35mm film cameras that I have done personaly; I have gotten a few with high numbers if I look back through 38 + years of tests over many dozens of lenses. I used Panatomic-X in microdol and a solid tripod; the lights are turned on and off for the exposure; thus there is no camera shake. I got 85 line pairs per mm at F5.6 with a Konica Auto S2; 93 line pair per mm with a 50mm F2 Summicron at F5.6. In the common 50mm F1.4 Nikkors for the Nikon F; I have had the peak data be as low as 58; and the high be 95 in one case; with many in the 65 to 78 region as the peak
    In the camera magazines; I remember there was on case where a lens topped 100; the 90mm F2.5 Series I got 105 line pairs per mm at a close test distance. With my own three tests of the same lens; NONE breached 100 at close of far distances; the best of three peaked at 85 line pairs per mm.
    Thus as a practical matter; I could buy 1000 lenses on ebay or new from B&H and test them all; and maybe find one that really hits 100 line pairs per mm on fine film.
    The customers who have their lenses that they claim hit 100 on film; that I have tested just clock in at about 50 to 75 line pairs per mm on film at their best fstops. Folks either uses the wrong distance or use the wrong values for the group pairs.
    I use to work on 35mm microfilm cameras for check sorters. With microfilm and a 1:1000 test target and if one varies the exposure by 1/3 stops; I have gotten over 100 line pairs per mm on film; ie 105 to 110 numbers. As a practical matter; with a colored check; with tolerances of the checks positions; the issue of exposure not being within 1/3 stop one usually gets about 1/3 the resolution numbers on a practical basis.
    It is possible for folks to get 100 line pairs per mm on film. It is also possible that they have the same chance of winning a lottery; or that gold bars will arrive in the mail too.
    Getting 100 line pairs per mm on pictorial film; of a pictorial object is as easy as getting 100 miles per gallon in the typical USA car or truck. IT CAN BE DONE! If one waks up to another at a gas pump and you tell another that you get 100 MPG with one's Ford Ranger; must folks will think you are nuts. One photo.net many folks preach " I saw where once somebody got 100 line pairs per mm on film" ; and most folks think it is easy; or practical; when it is about impossible.
    Alan; it *IS* possible that tomorrow you will win the lottery; or get 100 line pairs per mm on pictorial film; of a pictorial object.
    In is also possible for Los Angeles to get 5 inches of rain in July too. But just think of what the odds are; one has 1/100 of an inch of rain as the 100 year average for JULYS for Los Angeles downtown. A Bookie in Vegas would think there are better odds for Brett Farve to play pro football for another 20 years.
    Thus the issue is not whether one can get 100 line pairs per mm on 35mm film; or get 100 MPG in a 1980's 3rd generation Camaro. IT IS HOW OFTEN DOES IT HAPPEN IN THE real world.
    Getting 100 line pairs per mm on film for a pictorial image is like getting a hole in one; it happens at times
    Look at golf; even Tiger Woods "settles" that few are every going to get a mess of holes in one in one 9 or 18 hole game.
    Look at Baseball; it is possible that games go into overtime; it happens a lot. The media does not plan that every game will go into overtime with scheduling
    Kodak in San Diego did tests of how much info 35mm microfilm will record back in the later 1970's and early 1980's for digital storage usage. Kodak and the military did in too during World War II for microfilming of mail for GI's; ie VMAIL. Microfilm got mailed. Even in the 1920's and 1930's 35mm microfilm like Kodaks later Microfile microfilm would record fantasy high numbers under best case test conditions
    Alan; a 4000 dpi real film scanner works well with most all folks super sharp originals; the best of the best can be thus farmed out if they want to chase more details. The quest is like not always worth it; ie vacuuming the floor 3 times instead of 2 times. ie one has a high cost versus what extra is captured. It is like mowing the lawn twice; taking a bath 3 times a day. About every single customer who wants stuff farmed out repeats the 100 line per mm tales; ie one camera out of 100 ever tested; Panatomic-X film shot of 1:1000 test objects.
    It really is not a complex subject or even technical one either. It is one where lay folks believe their fairly sharp 35mm original hols more info than a common 4000 dpi Nikon film scanner; when it about always does not. It really one is of the customer has this insane belief that based on pure faith that a better drum scanner will MAGICALLY make details appear.
    Thus by using just a 4000 dpi film scanner; there is really NOT just settling for results. The average Joe's sharp original often never breaches the scanners limit. Thus one just farms out their gems of gems for higher end drum scans. If one told them that many of their better stuff benefits from drum scanning; one has the issue that one is a darn liar; folks will resent being lied to; being ripped off; ie paying for higher costs services that buy one nothing at all.
    Saying that going above 4000 dpi is warranted for a lot of stuff makes one look like a BSer; it really is consumer fraud from a customer relations standpoint. In the long run the smarter customers will resent being ripped off; ie paying for scans that add little if any extra details. After one DOES farm out Mr Egos gems of gems for high end drum scanning; the majority often go through this crisis. If it is an account customer they may balk at paying for their invoice; since the farmed out drum scans did not make the grassy knoll have gobs more details/info.
    A typical customer who has it in their heads that a 4000 dpi film scanner is not good enough does have som stuff farmed out for drum scans. After awhile the numbers of slides/negatives that get farmed out drops off. ie they "grew up" ; ie the Honeymoon/Newbie phase is over. They weight the extra costs with not much if any gained.
    It is as simple as paying to have ones lawn cut twice in one day; in theory it looks a bit better. After awhile the benefit to cost ratio looks poorer.
    It is most common that folks with 35mm get fewer and fewer stuff framed out for drum scans; as they mature. 4000 dpi film scans are usually good enough. The bottleneck is their original is not some 100 line pairs per mm test negative; but more just a common sharp pictorial image
     
  62. ... which is a long way of saying that I am right when I say that you are of the opinion that very good (but not quite the best possible) performance is good enough most of the time.
    The following is an excerpt from a post I made a couple of months ago on some experimental results of camera/film resolution:
    Modern Photography, October, 1978.
    Here are some of the results:
    Summicron 50mm f/2 on Pan-X film: 88 lines/mm at f/4
    Nikkor 50mm f/1.8 on Pan-X film: 88 lines/mm at f/4
    Canon 50mm f/1.8 on Pan-X film: 86 lines/mm at f/4
    Summicron 50mm f/2 on Tech Pan: 96 lines/mm at f/4
    Nikkor 50mm f/1.8 on Tech Pan: 96 lines/mm at f/4
    Canon 50mm f/1.8 on Tech Pan: 92 lines/mm at f/4
    Summicron 50mm f/2 on High Contrast Copy: 105 lines/mm at f/4
    Nikkor 50mm f/1.8 on High Contrast Copy: 105 lines/mm at f/4
    Canon 50mm f/1.8 on High Contrast Copy: 102 lines/mm at f/4
    Summicron on Kodachrome II: 86 lines/mm at f/4
    Nikkor on Kodachrome II: 82 lines/mm at f/4
    Canon on Kodachrome II: 80 lines/mm at f/4
    Summicron on Micro-Ektachrome: 102 lines/mm at f/4
    Nikkor on Micro-Ektachrome: 100 lines/mm at f/4
    Canon on Micro-Ektachrome: 100 lines/mm at f/4
    So, they got around 100 lines/mm using several different lens/film combinations, though with normal pictorial quality films they were not quite getting that high, more in the high 80's lines/mm.
    Clearly, under optimal conditions there are certain film/camera combinations that can yield a resolution greater than that which can be resolved by a 4000 pixel per inch scanner, even for some pictorial films.
    However, I am sure you are correct when you say that this is not needed in most cases.
    On the other hand, to shift gears a little bit, a flatbed scanner capable of resolution equivalent to 2000 pixels per inch is definitely not good enough to pull all the detail out of a 4x5 negative. A lot of lenses for 4x5 cameras resolve 60 lines/mm or better (some up around 80 lines/mm), and these cameras are usually used on a tripod using good all-around technique.
     
  63. One other comment, more of a philosophical comment than anything else, posed in the form of a question: What is better, to use equipment with performance reserve capable of handling all situations or equipment which is incapable of providing full performance under certain conditions? Answer that question and it may determine whether one believes it would be good to have a scanner with more than 4000 pixel per inch.
     
  64. Alan,
    If you truly find you need more resolution then what 4000 ppi gives you then you really need a larger piece of film not a higher resolution scanner. The extra detail going from 2000 ppi to 4000 ppi is most often very small, going past 4000 ppi means that you have push the size film you have much too far, IMHO.
     
  65. Alan; I have that issue too . The one test that got 100 line pair per mm (not lines like your text) was with High Contast copy film
    Thus you used a film about none every use for pictorial usage. I still have my 150 foot roll of Kodak High Conrtast Copy film frozen; I got in in 1976. It is an ASA 6 film; a LITH film,
    One cannot get micro-ektachrome anymore. Panatomic-x is gone too.,
    Again; you are confusing high contrast best case conditions; with pictorial ones; dead wrong assumption most all make.
    You are making the basic lay begineers assumption,
    You want to believe that the best case test data with high contrast film and targets; matches pictorial subjects and films. It is just a small error of about 1.5 to TWO TO ONE!
    You quote test data shot with high contrast films of high contrast 1:1000 USAF test targets.
    In the real world of non bull dung; folks shoot images of people; sports, cars; children; buildings.
    *******None of this stuff has 1:1000 contrast bars unless it has a referees jersey. This fact alone hacks the super best
    case numbers down to the 70's or 60's.
    *******THEN one has the fact that folks are using pictorial films; NOT high contrast stuff. This hacks the numbers typically down into the 50's.
    Folks who want scans above 4000 dpi are obsessive compulsive folks; it truely is an insane behavour at times dealing with them as customers. They have a mindset that more is better; they are a dangerous lot in an uncontrolled project. They are the ones who waste gobs of money on things that usually not matter
    One can re-shoot those best case 88 line pair per mm tests with special USAF targets with pictorial contrasts of 1:4, 1:10 and the numbers drop like a rock. ie sometimes in the 30's to 50's; and that is with Panatomic-X or Micro File microfilm.
    My own Konica Auto S2 that bests at about the 85 line pair one mm with Panatomic-x at 1:1000 targets drops down to the 50's when MY special grey type USAF charts I used for check sorters. If I try 1/500 at F5.6 shooting grey targets by hand; and shoot many and cherry pick the best ones; you get 40's to 50. ALL the numbers get hammered down to reality. If I reshoot those 1:1000 targets with iso 800 superia; one gets not 85 but numbers in the 40s to low 50's. With grey targets ; tripod and iso 800 superia one gets 30's to low 40's
    Thus you are barking up the wrong tree to think folks get even close to the lab data under pictorial conditions; about nobody gets into that region at all.
    Thus from a customer relations standpoint; the lay want high end scans that usually gather no extra info at all. One cannot come out and tell them directly they are dumb/stupid/dolts/; you just have to advise that the higher end scan may not be warranted.; and it COSTS MORE.
    It really is just thus an irrational thing; folks have stuck in their brains that 1978 best case 100 line pair per mm number with B&W film; timed lights; tripod; ASA 32 Panatomic-X. I have several hundred lens tests from magizines. Folks quote the best tests of the century.
    The irrational part is their lens is worse; it was not at the best fstop, The whole autofocus thing with many cameras is called locked when one is good enough (about 40 line pairs); they use faster films; they use a kit zoom
    One can say why not just drum scan everything; or wash ones hands 10 times; or have the yard cut 4 times in one day.
    I did all this 35 years with microfilm cameras for checks. If a check has a micro USAF 1:1000 target and one has perfect exposure and perfect focus; you would get sky high numbers. The with just lower contrast you got reality.
    The main point is most lay folks cannot understand transfer functions.
    They believe that 1:1000 test data is about what they will get with pictorial objects.
    Their brains are not wired to understand MTF's; or combos of MTF's.
    Thus one has to let lay beginners learn by failure; you advise them to cover liability. You watch newcomers have stuff farmed out; that really does not tax the 4000 dpi film scanner. You have to let the unbeliever who is obsessive compulsive waste enough cash to where it hurts their pocketbook,You basically are like a parent with a know it all 18 year old who knows everything.
    One can often just use a lightbox and tell in a second what scan level is required; after 2 decades it is easy. Folks really want that drum scan to polish their beloved image; one has to deal with the disappointment.
    One customer the other day said that the 150 bucks worth of stuff I farmed out were not even drums scans; "BECAUSE THEY ARE NOT BETTER" thus one has dumbing down of folks understanding.
    The philosophical part today is the public is dumber; you as a service bureau know nothing; they as a customer know everything. It is real common that folks today cannot grasp that if an original looks just fair on a lightbox; they might not even breach a flatbeds limits.
    maybe it is because few project slides anymore; or few use an enlarger anymore.
    From a philosophical standpoint having done 21 years worth of 35mm film scanning; a 4000 dpi real film scanner pulls about about all there is with about all folks stuff.
    The average persons "sharp original" often really only needs a 2700 dpi scan level. There are gob of folks with Kodachromes that really only need a flatbed; or 2000 dpi film scanner.
    Look at it this way; it 1 in a million slides needed a 8400 dpi drum scan instead of a 7000 dpi one; would you say that 8400 is required for all?
    You have the situation where already the 4000 dpi film scanner has a VAST RESERVE; most folks do not tax these at all.
    4000 dpi film scanners already have big reserve already.
    One can use an micro etched glass 35mm test slide I had made and breach any scanners limit.
    If folks believe that film scanners should be 5000 dpi; maybe 1 in 10 million will breach its limits
    You have the situation where file size grows; and one picks on slide in the scanners MTBF will breach its limits.
    The common 4000 dpi film scanner that is a couple decades old already has a lot of reserve; since little breaches it limits. Nikon learned this 25 years ago; that is why early units back in 1988 were 4300. Later coolscan units got down to 4000 and 2700.
    As a Philosophical comment; If you want the ultimate; just pay the shot and have all ones 35mm stuff scanned at 8400 dpi on drum scans; since 1 out of 1000 to 5000 great sharp shots might be noticeable.
    It is really a money issue; do you want to pay double to ten times more for something that can be noticed after a lot of hard looking; in one out of thousand scans
    To an person that is irrational; that one crumb they find after 2 hours of vaccuming the Kitchen is "worth it"
    Maybe the impass is folks do not consider costs. With 4x5 stuff drum scanning at 2000 dpi or even flatbed scanning at 2400dpi is very common. In pro work the costs matter. In amateur work costs do not matter; there is no client.
    The most odd thing is many folks who want drums scans with 35mm have these slides of sunsets. There are NO sharp details at all. A 2000 dpi film scan is often overkill for details; one uses 2700 or 4000 and see which looks better of the three in the final print.
     
  66. Kelley,
    A few corrections and clarifications. First, most of the test results quoted above were 80 lines per mm (OK, "line pairs" smart aleck) or higher. There is not a 4000 pixel per inch scanner of the face of the earth that can resolve 80 line pairs per mm, let alone higher than that. Why? Because the Nyquist limit for a perfect sampling device with 4000 pixels per inch is 78.4 line pairs per mm, and a real scanner cannot reach perfection. (If a "4000" pixel per inch does reach 80 line pairs per mm then it is not truly a 4000 pixel per inch scanner but something higher than that.)
    Second, it does not require non-pictorial film to exceed the limits of a 4000 line pair per mm scanner. Note that of the five films listed above that achieved >80 line pairs per mm two and one half of them are pictorial films. (I am counting Tech Pan as half-pictorial, since it has sometimes been used as a pictorial film.)
    Third, the results were not limited to monochrome film. Some results for Kodachrome exceeded 80 line pairs per mm, and I believe some of the more modern films exceed Kodachrome in resolution.
    Fourth, I have the article in my hand and it says nothing about a 1:1000 contrast USAF target. They used a USAF target, but the only special qualities mentioned in the test report is that it was large. Presumably it was more like 1:100 contrast, since I believe that is about what can be achieved with a reflective target, but I am prepared to be corrected on that point if wrong.
    Fifth, I believe there are comparison scans on the web between Minolta (5400 pixels per inch sensor) and Nikon (4000 pixels per inch sensor) where the resolution of the Minolta was slightly better than the Nikon. Unfortunately, I can't prove this assertion right now because I don't have the links handy.
    Sixth, I am not disputing that it is often (even usually) sufficient to scan at less than the maximum possible resolution. What I do dispute is your assertion that less than maximum possible resolution is virtually never useful.
    Seventh, most of your discussion is based on an assumption that maximum technique is virtually never used. I think this is somewhat of an exaggeration.
    Eighth, what kind of lenses do you use on your camera? Do you realize that almost all the arguments you have presented for less-than-maximum scanner specification are also arguments that there is almost no reason to buy the sharpest lenses available?
     
  67. Correction, 4000 pixels per mm can theoretically resolve up to 78.74, not 78.4. I read my calculator wrong.
     
  68. Alan, for fine grained films, 4000 ppi scans are indeed the bottleneck. For 4x5 sheet film, it's pretty rare to achieve numbers much beyond 80 lp/mm. For some 35mm and MF, you can indeed go well beyond that. It doesn't need to be a high contrast copy film. I know when Roger Smith tested the Voigtlander Bessa R3a with Leica glass and Ilford Delta 100, he achieved 125 lp/mm. A 4000 ppi scan won't come near that.
    It is also not necessary to photograph a test chart to achieve these resolutions....just take a landscape where trees cover the sky. Like I said, for 4x5 it's not really necessary. For 35mm and MF, I've done enough scans over the years to see that scanning higher than 3200 ppi is very often warranted.
     
  69. Alan;
    FIRST; the lens test data I quoted were with formal USAF 1:1000 charts; you shoot a negative; you read the pairs and groups with a microscope. There is NO scanner in the loop. This type of lens testing has been done for decades. Since there is NO scanner in the loop; the scanner does not matter.
    SECOND; The lens test you quoted are of 1:1000 test charts; nobody's facial features has a TEN F stop range between details. The actual contrast is little. One has a 2 or 4 or 8 ratio between details; instead of 1000.
    THIRD; again these Kodachrome tests are not of real world subjects; but of a high contrast 1:1000 charts. Actual pictorial details do not have 1:1000 contrasts; thus pictorial data is not in this fairy tale region.
    FOURTH; ALL these tests are with charts are 1:1000 contrast; unless otherwise noted. It is as fundamental as knowing baseball has 9 innings. Thus all this data of high numbers 100 line pairs per mm resolution is with this lab test super high contrasts; that NOBODY shoots at.
    ***This is the point I have been trying to get across; that those HIGH 85 to 105 line pair per mm test numbers are of super super high contrast test targets; like about all published tests are. The test is run to see the resolution one gets with these targets; which are not what pictorial stuff is.
    ***When one repeats the same type of test with a *grey* set of bars; the data drops like a stone/ rock. One does not get these high numbers anymore; and this is BEFORE one even has a scanner in the loop. ALL those high numbers get hammered when a typical details object is shot; ie the resolution numbers on film of the test chart are way way lower.
    As the contrast of the target drops; does the ultimate resolution.
    FIFTH; if some Minoltas at 5400 dpi tested higher than a 4000 dpi Nikon ask yourself why was it a marketing dud? It is because on a practical basis it did not really matter. It is sort of like buying a faster dialup modem of 200k; up the phone system has it all capped at about 53K anyway
    SIXTH; one can use above 4000 dpi sometimes in rare cases. If you want to you can have all your own stuff drum scanned at 12000 dpi; there really is no law against it. There really is no black and white answer.
    You may ponder why if super end 8400 to 12000 dpi scans have been around for a long time; why do very few use them? Why in heck do folks as you say "scan at less than the maximum possible resolution." Is it that a 11,000 dpi scan might be 250 bucks each?
    Most all folks on the planet "Scan at less than the maximum possible resolution" it has been that way for decades. Some folks actually look at costs and weigh the tradeoffs. One could have stuff high end drummed scanned 5,10, 15 or evey 20 years ago.
    You bring up an interesting point; why do not all photographers not use Summicrons at F4 on tripods and have all drum scanned at 11,000 dpi?
    SEVENTH; I am saying that the average persons super great shot with great technique often makes images that do not tax a 4000 dpi film scanner. The stuff that is done well still gets boxed in by the the laws of optics; one has little contrast at all way out there with a high number of cycles; thus there is little if any benefit needed of a scanner beyond 4000 dpi.
    There can be stuff out there beyond 4000 dpi; this well understood 20 years ago
    The best of the best can be farmed out and then folks will be either impressed at that 200 bucks spent per frame; or they will cry due their stupidity; they flushed cash down the toilet.
    There really is a growing group of lay folks; it is a darn good thing the economy is poor; it often wakes up folks with blinders on.
    From a customer relations standpoint it is really about education. Morally one needs to say to folks that you probably will not notice. It is sort of like telling somebody they really do not need new tires if they have a lot of tread.
    Formal lens tests in those old magazines are with USAF test charts with 1:1000 contrast. Nobody shoots objects like that in normal life.; unless it is for microfilming engineering drawings; that was done eons ago.
    Modern Photography sold the same charts they used as a "KIT" for about 15 years. You used Panatomic-X, you had two lights 45 degrees to the charts on a wall. You expose by turning the lights on and off to quash all vibrations.




    The 1:1000 is the contrast of the USAF charts; the white reflects 1000 times more than the black. In f stops; that is 10 f stops; ie 2 to the 10 th power is 1024.
    The grave error is assuming that 1:1000 test data matches folks stuff shot with pictorial contrasts.
    It is a grave gross error of assuming that those great HIGH test numbers done with 1:1000 test charts are what folks get with regular contrast subjects. An analogy is like assuming a cars best MPG test data for a steady 40MPH matches a persons normal commute with stop and go and varied speeds. One gets numbers one half or less.
    Pictorial images are not of the same contrasts as those lens tests; they are of actual objects that only have a stop of two; or a fraction of stop difference.
    The "Crime" is preaching using data where is does not apply. When Test data for one high contrast lens test is shot with lower contrast test targets; the test data drops like a rock. The same goes with car MPG tests; if one wants to be a con-artist; you quote data that is high over one speed of 40 MPH ; and let others assume it applies to their case with stop and go; picking up kids.
    Thus the actual real life pictorial negatives when sharp are down in the 45 to 60 line pair per mm and never 100. At the limit of the lens tests 100 data; the resultant image on film at 100 line pairs per mm is not 1:1000 in contrast; more like about none. One is just barely resolving bars at the limit of performance. The transfer function is such that the 1:1000 test targets image on film is now with little contrast. When the same lens shoot grey scale USAF charts with a low pictorial contrast; one gets number no where near 100; more like 40 to 70's numbers. One gets numbers less than a 4000 dpi scanner. One really has a poor situation; the film/lens is down and one is near the scanners limit too
    At some point the fibbing about getting 100 lines per mm in pictorial has to be pointed out. These are pure fibbing manure numbers. . This manure is spread around as gospel; when it really is just done to stack the deck; ie a con artists gambit.
    Having worked in optical engineering; it is easy to spot BS numbers.




    Most sharp pictorial shot are really in the 50 line pair per mm range in resolution; with stuff that was done with good technique and a lot of care. As a practical basis; nobody ever gets 100 line pair per mm on film with normal pictorial subjects. The ultimate great stuff might be 60 to 70; in rare few cases. If on has actual details on this range; a higher end drum scan at 8400 dpi can have a better transfer function than a 4000 dpi box. It just means that one might get some higher contrast at the higher resolutions; compared to a 4000 dpi box. As another said; "it often is just not worth it"




    Saying one gets negatives in the 100 line per mm range with pictorial is really NUTS. In effect the dogma is like a used car salesmen that cons folks. The salesmens goal is to sell cars and use BS data. He has no morals and doe not care if his dogma is wrong or misleading. Thus he finds test data that that the used 3400 Lb Buick gets 45 MPG at 40 MPH and says the car gets 45 MPG to sell the car. The end user then finds out he actually gets about 18 MPG in the city and 26 MPG on highway trips. The end user drives 55 to 80 MPH on the highway; and not a steady 40 MPH. The used car guy cherry picks this single high number data point to be a devil; to purposely mislead folks. He implies that one gets 45 MPG with normal driving. The error is about the same as with lenses; just a TWO to one error




    It is really not that the 100 number shot with 1:1000 test targets is wrong; it is just used by folks where the data is invalid; ie with murky grey objects.




    The numbers I quoted for my formal USAF tests of lenses where with no scanner; you use a microscope and quote the lessor of the tangential and radial data for each zone.
    You have to realize what the immoral thing is about preaching numbers that are not valid for other cases. It really is fraud in a legal sense
    Tests with high contrast test targets go back 100 years now; the USAF chart is over 50 years old. High contrast test targets give one data point; and not the entire ball of wax.
    By using the best case 1:1000 test data and assuming folks get this type of resolution numbers on a practical basis with pictorial shots; you are off base by about a factor of TWO.
    It is as bad as the used car chap that misleads folks; legally it is a dis-service to the public to use data the wrong way.




    It was well known that a 4000 dpi perfect scanner might resolve about 78 line pairs per mm with a perfect target 2 decades ago. I am not sure if you just learned this; or got a calculator or just think I did. All it is 4000/2/25.4= 78.74
    The problem is that folks do not shoot 1:1000 shots of people; cars etc. There are not 10 f stops difference on eyelashes; glasses; hair etc. It is only a fraction of stop or two. Thus when one repeats the typical great best case lens tests of 80 to 100 line pairs per mm with 1:1000 targets; one is really only getting 35 to say 65 recorded on film.




    Thus 4000 dpi film scanners are about never breached; since folks sharp stuff really is in the 35 to 65 line pair range.
    By using the best case 1:1000 test data and assuming folks get this type of resolution numbers on a practical basis with pictorial shots; you are off base by about a factor of TWO.
    Folks have always been able to drum scan 35mm stuff; one could do it in the 1980's.
    When 4000 dpi film scanners became widely used; 35mm stuff farmed out for drum scans tanked/dropped; since little really warrants a higher end scan.




    There really has been NO barriers for folks scanning a lot of there stuff above 4000 dpi except RAW CASH. Some folks do not have unlimited cash to waste. They just farm out the higher end stuff that needs it the blue moon cases.







    Farming out stuff for scans above 4000 dpi for higher end drum scans is really about religion. One has customers that are often like brainwashed and have joined a cult; they believe that it matters a lot. On has to do handholding; they often go thru a detox; debrainwashed phase when they find out the farmed out stuff is about no better. Some will not want to pay; others want it sent back to be drum scanned again; a rare few will say you are right. There is this fallout phase for them; the sacred dogma got shattered.
     
  70. Hmmm,
    I've seen the difference in prints....so the numbers aren't manure. I guess some of us print and scan better than others.
     
  71. All one has to do is farm out that rare stuff that needs a scan beyond 4000 dpi; if one runs 4000 dpi class scanners.
    If one farms it out for another; one most often has to deal with their emotional crash; after the canned dogma cracks. For some it is a PITA; they do not want to pay. They want you to try another drum scan outfit. You are seeking holy grail scanner that will make their original have more info.
    The folk who crash the worse look at the whole issue as a black and white one; they want a simpleton black and white answer to a complex issue. Their brains cannot handle any ratios; contrasts; MTF curves; reality . If one shows a Gaussian plot and says most pros only need this tiny tail of the Bell Curve scanned beyond 4000 dpi; they cannot understand it; or do not want to. It is more like trying to get across stuff to somebody who has been into Scientology too deep; and has been brain washed.
    If one looks at each customers bell curve of slides; some could shoot 2 million and non need above a 4000 dpi scan; it is all low end stuff; most doed not even need a 2000 dpi scan.
    Maybe M perfect might want 5 percent farmed out; and he is a 3 sigma customer that is rich; uses a tripod; shoots at F5.6; sweats the details. Another pro might have 1 out of a box of slides scanned above 400 dpi and he gets ticked off since the Nikon 4000 dpi film scan is like better.
    farming out drumscan 35mm stuff is about managing the disappointed customer; they paid all this extra cash for the better Gizmo; and now they are disappointed.
     
  72. Kelley,
    The lens tests I referred to were in Modern Photography in 1978. Where do you get the information that they used a resolution target with a 1:1000 contrast ratio in that test? The only information I find in the article is that it was a large custom made USAF target. The contrast ratio is not mentioned in the article.
    Maybe all 1951 USAF targets have a 1:1000 contrast ratio, but so far I can't find that specification anywhere.
    If it really is a 1:1000 contrast ratio it would be interesting to know how they accomplished that challenging technical feat given the fact that it is not easy to get more than a 1:100 contrast ratio on an evenly illuminated reflective target. Perhaps you know how they did it.
     
  73. All; I have about at least FIVE of the Modern Photography test kits they sold over more than a decade. Herbert Keppler of then Modern Photography started the magazine selling them so one could do ones own tests. You got the test booklet; the charts; and a dinky Edmund magnifer to try to read ones negatives.
    The USAF chart is described here in MIL standard 150A; ie MIL-STD-150A; it is about a 4.3 meg PDF
    http://www.everyspec.com/MIL-STD/MIL-STD+%280100+-+0299%29/MIL-STD-150A_16197/
    Page 21 (31 on the PDF file) has the description.
    There are 3 types of charts; High, Medium and Low Contrast
    High is has a density difference of at least 2.0
    Medium has a density difference of 0.8
    Low has a density difference of 0.20
    High is the one that Modern Photography use to use and sell. They uses the same charts in house as the ones they sold in the kits
    A density difference of at least 2.0 is a logarithmic scale. 10 to the 2 power is 100. It means the target has to have at least 100 times difference in reflectance between the white and black.
    Since most of the time these are contact printed; the difference is closer to 1000 than 100; thus often called the 1:1000 target in slang.
    When one does a test also with the medium and low USAF charts; the resolution drops. On gets poor numbers with the LOW contrast chart.
    A density difference of 0.3 is a two to one difference in reflectance; actually 0.30103
    If one had charts at exactly 1:100 versus 1:1000 it would not change test result numbers much; it would be less than variabilty of the tests
    The working test charts themselves are just high quality glossy photos; contact prints; the ones I have are not RC prints but the old ferrotype stuff where one use a chrome plate.
    One can get about 5 to maybe 8 stops range with photopapers. It varies by the paper; developer etc. Some papers have a reflection density of about 2.3; that means about a 200 to 1 range; this is pushing it. The eye is suppose to be able to see 100 to 200 different level of grey in a great photo/chemical based print. To get dark blacks one need fresh paper; leaving it in the developer long enough and not a bad safelight.
    I hope some of this helps; regards.
    DOWNFALL/shortcoming of just the high contrast Target:
    The down fall of just using the high contrast charts was found out in the 1950's when TV video cameras emerged. TV engineers found out that lenses that had high numbers on those high contrast charts for the ultimate resolution sometimes were WORSE than lessor lenses; when used on LOW res TV taking tubes. The crafty Electrical engineers got involved and the MTF for lenses got started; it gives the entire ball of wax.
    A test of a lens A versus lens B with all three types of charts can have both lenses giving high numbers with the high contrast charts; and B might be far better than A with the medium and low contrast chart.
    Most folk never even know that there are three variants of the USA charts; i used the medium and low ones too with microfilm check sorters too.
     
  74. Two orders of magnitude: 1:100 contrast ratio, just the figure I suggested in an earlier post.
    By the way, I bought one of Modern's lens test kits back in 1971 for a class I was taking. I have no idea what happened to the kit.
     
  75. "...if some Minoltas at 5400 dpi tested higher than a 4000 dpi Nikon ask yourself why was it a marketing dud?"
    I dunno, maybe for the same reason that essentially all vendors (except Plustek and arguably Nikon) have dropped out of the 35mm film scanner market.
     
  76. The resolution difference between the Nikon and the Minolta had nothing to do with the Minolta being a dud. The Minolta easily outresolved the Nikon. In fine grained B&W work, I was always able to produce more fine detail in a large print from the Minolta.
    It came down to marking. Why did Beta fail against VHS when it resolved more? Marketing. The real dud was the use of this as a comparison. Folks, we don't need to read a post of War and Peace every page to see the difference. Pull a high rez scan vs a lower rez scan and make a large print from it. Some films will benefit, some won't. Some optics will, some won't. I've compared the differences for many years....and have never resorted to test charts.
     
  77. Must admit it's great to see Kelly get fired up over this topic.....many will get caught up with the idea that more is always better. I've found that high dpi's much resemble high megapixel sensors...more is not necessarily better. I have scanned 4x5's and 6x17's on both a drum scanner (with oil attach), and flatbeds but found little if any gain past 1200 dpi. The 4800 and 9600 levels are ludicrous, increasing file size no end, but offering very little (if any) increase in resolution. I think that hi res begins and ends with the camera and its lens and your ability to hit sharp focus. I now use an old Arcus 1200 (beastly thing) that at its maximum stretch of 1200dpi renders large images that would make your eyes pop.
     
  78. Dave;
    All of us have different results with these scanners. I drank the Koolaid too and I believed the Minolta 5400 units were better. I got excited and then got disappointed.
    ****My own experience is different than yours.
    When the first 5400 unit was no better; I waited and later got another and it too was no better than my Nikon or Canon 4000 dpi units.
    I gave up after two Minoltas; they got sold off on Ebay. I suppose I could have bought a dozen and found a gem; but then again one has no global data base to tell what the odd are of cherry picking a great 5400 unit.
    Thus at least here I compared gem of gems sharp stuff shot with Kodak Ektar 25 with super high end drum scans versus my Canon and Nikon 4000 dpi film scans versus the two Minolta 5400 units. I purposely used the rare example cases that breached the 4000 dpi units; ie where the super high end drums can pickup up a tad more details. Then with the two 5400 Minolta units I had; plus a buddies I never saw where the 5400 units really helped.
    I am not saying you that you do not have a hot 5400 unit that betters the 4000 ones; but here I just tested only a sample of three units; thus I declared my own tests as strike out.
    The real question is how much raw cash does one spend buying "5400 dpi/ppi" class units and testing them to get one that betters what I already have?
    My own experience is the 5400 units were not clearly better; it was a wash.
    Ok so YOURS "easily outresolved the Nikon"
    I have an Epson 2450 flatbed here that was one of the first units; it is excellent for a flatbed; I really have not found any newer units that better it; ie 3200; 4800; the V500; V755 etc.
    With another 2450 model; it is not as good as the first one. One can mess around with z heights on both to get the best focus; and the earlier one is noticeably better.
    Here based on 3 samples; I would call myself a liar if I said that the Minolta 5400 was much better; or even measureable any better than a common 4000 dpi unit. Your experience is different; based I guess on as sample one unit.
    I am not saying that others too might get better results with a 5400 minolta; if they do that is cool.
    In a way it is like chasing a better 50mm lens; these scanners DO VARY thus we all can be correct with what we have used; but differ since we have different serial number units.
    With one data point one can prove any cause.
    Here at least none of the Minolta 5400 units really scnned any more details than my 4000 dpi units; and no more details showed up in any big prints. And that was with using the ultra rare cases where an original breaches a 4000 dpi unit
    ******The whole reason as a service bureau I got the 5400 calls units is to reduce the few/little stuft that was farm out for high end drum scans.
    Thus the 5400 units were tested on the select ultra rare stuff that was sent out to be drum scanned. The rare stuff that got drum scanned out also got a 5400 dpi scan on the two Minoltas; the results really were no better than the 4000 dpi stuff; thus no sane customer would go for the lack of value; but the drum scans pulled out a tad more.
    Thus the 5400 dpi minolta project was failure. It is sort of like buying a new 7200 dpi flatbed and finding out gee it is the same as my old great Epson 2450.
    When I dumped the 5400 Minoltas On ebay; I said that they were really about a 4000 dpi unit; for fear of getting negative feedback due to fibbing. I really did not want some poor soul to find out these two were no better than 4000 dpi ones; then blast me with negative feedback for claiming they were.
    I am not saying you that you do not have a hot 5400 unit that betters the 4000 ones; but here I just tested only a sample of three units; thus I declared my own tests as strike out.
    The real question is how many units does one have to go through; to get a good one. Mine two we brand new units; my buddies was a used Ebay one.

    maybe these vary all over the place like Russian Zorkis?
     
  79. Go to this link and read the test results on Velvia scanned at 4000 and 8000 dpi:
    http://www.normankoren.com/Tutorials/Scan8000.html
     

Share This Page