Nikon Super CoolScan 4000 ED - some questions

Discussion in 'Digital Photography' started by Hans-Georg Michna, Sep 12, 2005.

  1. Does anybody here have some experience with this scanner? I have
    a few questions.

    I got this scanner to scan all my slides and am not quite sure
    about the optimal settings.

    1. I find that this scanner, like practically every other, isn't
    as sharp as the manufacturer states. It cannot nearly reproduce
    a good slide in all of its resolution. Sharp, fine text, for
    example becomes mushy and, in extreme cases, unreadable.

    I conclude therefore that it serves no purpose to set the
    scanner to its full native resolution. The pictures would only
    be bigger and unsharp. So I reduced the resolution from 4,000 to
    2,400 dpi. Even at that resolution the results are clearly very
    unsharp.

    Does this make sense?

    2. Has anybody experimented with the sharpening function, the
    one called unsharp masking? Does this in fact compensate for a
    scanner shortcoming or does it only exaggerate edges? What did
    you find to be the best settings?

    3. When scanning with 8 bits per color, rather than the full
    color depth, does it still make sense to do multiple passes per
    slide? Or is the sensor noise lower than what can be represented
    in 8 bits per color?

    4. While scanning, the scan software uses 100% of an Athlon 64
    3000+ processor. Is this normal? Is the processor really the
    bottleneck? Anything I could set up differently to make it scan
    faster?

    5. Any other hints?

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 12, 2005
    #1
    1. Advertising

  2. Hans-Georg Michna

    Guest

    Regarding 'sharpness'. I have used a Nikon 5000ED (same res as your
    4000), and have found that it easily resolves grain. If your scanner is
    not dirty or malfunctioning, or just not focused, it should be quite
    sharp. Can you see the grain when view at 'actual pixels'? Also, if
    your slides are heavily curved, that may be the problem. There can be
    depth of field issues.

    You may want to post your questions to 'comps.periphs.scanners' as that
    is where the scanner community mostly is.

    W
    , Sep 12, 2005
    #2
    1. Advertising

  3. Hans-Georg Michna

    Stacey Guest

    Hans-Georg Michna wrote:

    > Does anybody here have some experience with this scanner? I have
    > a few questions.
    >
    > I got this scanner to scan all my slides and am not quite sure
    > about the optimal settings.
    >
    > 1. I find that this scanner, like practically every other, isn't
    > as sharp as the manufacturer states. It cannot nearly reproduce
    > a good slide in all of its resolution. Sharp, fine text, for
    > example becomes mushy and, in extreme cases, unreadable.
    >



    Sounds like you have focus problems or the film isn't flat. Read the manual
    on the software, use the "ctrl" while clicking the focus button and you can
    select different spots to focus on. If they vary a bunch (over 20-30
    points), you'll never get the whole thing sharp.

    >
    > 3. When scanning with 8 bits per color, rather than the full
    > color depth, does it still make sense to do multiple passes per
    > slide?


    I'd only do this if you have REALLY dense area's that show noise.

    >
    > 4. While scanning, the scan software uses 100% of an Athlon 64
    > 3000+ processor. Is this normal? Is the processor really the
    > bottleneck? Anything I could set up differently to make it scan
    > faster?


    Do you have AV software running? That can cause all sorts of problems from
    what I've read with scanning.

    I have the big brother LS 8000 and it scans everything I can see of the
    film..

    --

    Stacey
    Stacey, Sep 13, 2005
    #3
  4. On 12 Sep 2005 06:41:18 -0700, ""
    <> wrote:

    >Regarding 'sharpness'. I have used a Nikon 5000ED (same res as your
    >4000), and have found that it easily resolves grain. If your scanner is
    >not dirty or malfunctioning, or just not focused, it should be quite
    >sharp. Can you see the grain when view at 'actual pixels'? Also, if
    >your slides are heavily curved, that may be the problem. There can be
    >depth of field issues.


    It does show grain, but there is a lot of softness. An edge that
    is fairly sharp on the slide covers several pixels, something
    like 4, on the scan. Slides are in glass, but taking them out
    doesn't give any better results. They should at least be sharp
    in the center, but they never quite are.

    I must admit that I have yet to see any scanner that reproduces
    a one pixel wide line as a one pixel line. Gues I need 8,000 dpi
    to get a sharp 4,000 dpi picture.

    >You may want to post your questions to 'comps.periphs.scanners' as that
    >is where the scanner community mostly is.


    Thanks! That's a useful hint. Will do this.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 13, 2005
    #4
  5. On Tue, 13 Sep 2005 01:07:34 -0400, Stacey <>
    wrote:

    >Hans-Georg Michna wrote:


    >> 1. I find that this scanner, like practically every other, isn't
    >> as sharp as the manufacturer states. It cannot nearly reproduce
    >> a good slide in all of its resolution. Sharp, fine text, for
    >> example becomes mushy and, in extreme cases, unreadable.


    >Sounds like you have focus problems or the film isn't flat. Read the manual
    >on the software, use the "ctrl" while clicking the focus button and you can
    >select different spots to focus on. If they vary a bunch (over 20-30
    >points), you'll never get the whole thing sharp.


    Stacey,

    thanks for your reply! I had already tried that, but except for
    a few far out of focus scans this is not the case. The autofocus
    works well and the results are repeatable.

    I think that this scanner is no better than any other and
    exaggerates its resolution by a factor of 2. Scanning at 2,000
    dpi, rather than its rated 4,000, yields reasonably, though
    still not perfectly sharp pictures.

    Try this. Put a piece of paper into a slide and scan the edge.
    Then check how many pixels it covers. It's at least 4 to 4 grey
    pixels between the black and the white area, rather than 1 or
    perhaps 2.

    >> 3. When scanning with 8 bits per color, rather than the full
    >> color depth, does it still make sense to do multiple passes per
    >> slide?


    >I'd only do this if you have REALLY dense area's that show noise.


    Thanks, thought so. I don't do it. It seems that the noise from
    the film overwhelms the noise from the sensor, at least when I
    scan at 2,000 dpi, half the rated resolution. Thus four pixels
    become one and a lot of noise is cancelled out.

    >> 4. While scanning, the scan software uses 100% of an Athlon 64
    >> 3000+ processor. Is this normal? Is the processor really the
    >> bottleneck? Anything I could set up differently to make it scan
    >> faster?


    >Do you have AV software running? That can cause all sorts of problems from
    >what I've read with scanning.


    No AV running in the background.

    >I have the big brother LS 8000 and it scans everything I can see of the
    >film..


    Nice! Ever looked at the processor load? I have this impression
    that the processor load is genuine, because any further load
    actually slows the scan. Those scanners do make use of the
    processor.

    I have now, after a few tests, switched on the Digital ICE4 and
    am deeply impressed. Never switched it off again.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 13, 2005
    #5
  6. In article <>, Hans-Georg
    Michna <> writes
    >Does anybody here have some experience with this scanner? I have
    >a few questions.
    >

    Yup, years of experience with it. ;-)

    >I got this scanner to scan all my slides and am not quite sure
    >about the optimal settings.
    >
    >1. I find that this scanner, like practically every other, isn't
    >as sharp as the manufacturer states.


    Well that raises an interesting question: Do you know what "as sharp as
    the manufacturer states" should look like? If you don't (and too many
    people think it means you should get adjacent black and white pixels -
    it doesn't!) how can you determine that it isn't as sharp as it should
    be?

    >It cannot nearly reproduce
    >a good slide in all of its resolution. Sharp, fine text, for
    >example becomes mushy and, in extreme cases, unreadable.
    >

    OK, you clearly have a problem with focus or dirt. The LS-4000 is
    particularly fussy about film flatness, so if your film is curved by a
    couple of millimetres then it will not maintain focus across the entire
    frame.

    Assuming your film is flat, is the scanner set to focus automatically
    before each scan or preview is made? If you have inadvertently switched
    this option off in the preferences section then the scanner will just
    hold the last focus position it had - which could be miles (well, not
    quite that far) from the correct focus position.

    You can set the auto-focus on any particular part of the image using the
    focus position icon in the scan software, or manually adjust the focus
    through a range of values - though I have never found a real use for
    this latter control other than assessing the amount of film curvature.

    >I conclude therefore that it serves no purpose to set the
    >scanner to its full native resolution.


    I conclude the opposite - that you have a problem, not that the scanner
    is incapable of delivering its native resolution.

    >The pictures would only
    >be bigger and unsharp. So I reduced the resolution from 4,000 to
    >2,400 dpi. Even at that resolution the results are clearly very
    >unsharp.
    >
    >Does this make sense?
    >

    Yes - it suggests you either aren't focussing or your scanner has
    dirt/dust on the mirrors.

    The easiest diagnosis of dirt/dust is to scan a frame edge against a
    white background on slide film. eg. the black gap between two lighter
    frames or the edge of the slide mount with light subject matter at the
    edge of the frame. If you have dirty mirrors or optics then you will
    notice a smearing of the lighter portions of the image into the black
    frame. If this is the problem then there is no alternative to having
    the scanner stripped and cleaned - something you can do yourself if you
    are confident of working with optical instruments, but otherwise it is
    better to pay someone who is.

    >2. Has anybody experimented with the sharpening function, the
    >one called unsharp masking? Does this in fact compensate for a
    >scanner shortcoming or does it only exaggerate edges? What did
    >you find to be the best settings?
    >

    All scanners have shortcomings in terms of their image sharpness, it is
    a fact of life and USM is one method of compensating for this. I find
    that something around 5% intensity, 5% halo width and 5-10 levels of
    threshold fairly acceptable, but often scan without any USM.

    >3. When scanning with 8 bits per color, rather than the full
    >color depth, does it still make sense to do multiple passes per
    >slide? Or is the sensor noise lower than what can be represented
    >in 8 bits per color?
    >

    Yes, it still makes sense to use multisampling - this scanner does not
    do multiple passes, it multisamples in a single pass. The reason why
    this is still worth doing is because the noise is, of course, more than
    the quantisation noise of the 14-bit ADC and this is for a linear scan.
    Gamma compensation for normal use results in the noise from all sources,
    including this quantisation, being amplified in the shadows. However
    you will need a fairly dense film to be able to show the difference -
    and you haven't a hope if you are getting scattered light from dirty
    optics. Also, measurements show that for near saturated scans,
    corresponding to shadows on negatives, the photon noise itself is
    equivalent to just over 8-bits for the CCD. This can be improved by
    multisampling, to over 10-bits at 16x. You will probably see more gain
    from multisampling negatives than slides, because the dynamic range of
    the scanner is very good.

    >4. While scanning, the scan software uses 100% of an Athlon 64
    >3000+ processor. Is this normal? Is the processor really the
    >bottleneck? Anything I could set up differently to make it scan
    >faster?
    >

    No idea, I don't use Athlons.

    >5. Any other hints?


    RTFMs - if you don't have copies then download them from the Nikon
    website. There are two - one for the scanner itself and one for the
    Nikonscan software.

    Obtain and install the latest firmware update (v1.10)

    Obtain and install the latest version of NikonScan (v4.02)
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
    Kennedy McEwen, Sep 13, 2005
    #6
  7. In article <>, Hans-Georg
    Michna <> writes
    >
    >Try this. Put a piece of paper into a slide and scan the edge.
    >Then check how many pixels it covers. It's at least 4 to 4 grey
    >pixels between the black and the white area, rather than 1 or
    >perhaps 2.


    Precisely the point I was making in my other post - you *don't* know
    what the manufacturer's specification should look like!

    Resolving something does *NOT* mean reproducing it at 100% contrast -
    which would be the case of having the single pixel transition between a
    black and white edge you think you should be getting!

    No visible optic in itself can achieve that, let alone one of limited
    size in a scanner at 4000ppi (2000cy/in) resolution! The lens has a
    finite MTF as does the CCD - all of which reduce the rate of change of
    edges. Just for reference, a *perfect* f/4 lens will only be capable of
    resolving 2000lppi at around 75% of the contrast of the full black to
    white transition - and that is peak green light, it will be worse at the
    red end. I don't know what the f/# is of the lens in the LS-4000, but I
    doubt it is much faster than f/4.

    A quick look at some published measurements from a number of LS-4000
    scanners shows them to be among the best resolving units available, with
    MTF-50 figures around 22-25cy/mm. That means that the scanner will
    reproduce a bar pattern (more accurately a sine wave) of 560-640lppi at
    50% of the contrast of the original. Since the scanner is capable of
    resolving 2000lppi, then 4 pixels to achieve full contrast transition is
    a reasonable comparison. However, the scanner could still be capable of
    *resolving* 2000lppi even if it took 10 or 20 pixels to transition peak
    white to black - unlikely, but certainly possible.

    Your general dissatisfaction with scanners appears to be down to a
    misunderstanding of what the manufacturers are offering in their
    specifications.

    What I find more surprising is that you have somehow convinced yourself
    that so much more information is available on the film to begin with!
    Try this for comparison: view your scanned edge transition or text on
    your screen at 100% magnification, so that you can see the softness you
    are concerned about. Typical displays these days are around 96ppi - so
    the image size from a full scanned 35mm frame would be around
    40x60inches. Now, go into a projection room and project your slide at
    that same size and view it from the same distance as you do the monitor
    - about 10-12". The original will, of course, look a little sharper
    but, if your scanner is functioning correctly, it won't be that much
    sharper and it will surprise you just how little information that the
    scanner is not capable of lifting off the film.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
    Kennedy McEwen, Sep 14, 2005
    #7
  8. Hans-Georg Michna

    Stacey Guest

    Hans-Georg Michna wrote:

    > I have now, after a few tests, switched on the Digital ICE4 and
    > am deeply impressed. Never switched it off again.
    >


    Yep, it's amazing how well it works! I suppose since I'm scanning mostly
    6X4.5 and 6X9 film the rez issues don't seem bad? I have done some 8X10
    prints from 35mm film scans and they look really good, about as far as I'd
    ever go with 35mm film anyway..
    --

    Stacey
    Stacey, Sep 14, 2005
    #8
  9. Hans-Georg Michna

    Stacey Guest

    Hans-Georg Michna wrote:

    > On 12 Sep 2005 06:41:18 -0700, ""
    >
    > I must admit that I have yet to see any scanner that reproduces
    > a one pixel wide line as a one pixel line.


    Stop pixel peeping and look at the prints..

    --

    Stacey
    Stacey, Sep 14, 2005
    #9
  10. On Wed, 14 Sep 2005 01:13:36 -0400, Stacey <>
    wrote:

    >Hans-Georg Michna wrote:


    >> I must admit that I have yet to see any scanner that reproduces
    >> a one pixel wide line as a one pixel line.


    >Stop pixel peeping and look at the prints..


    Stacey,

    no need to. I can compare the original slide under a magnifying
    glass with the scan. The results from the Nikon 4000 ED don't
    even come close to reproducing the sharpness of the slide.

    But then this is still a relatively cheap scanner, probably
    meant for the ambitious amateur, and for that it may be good
    enough.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 15, 2005
    #10
  11. On Tue, 13 Sep 2005 23:27:45 +0100, Kennedy McEwen
    <> wrote:

    >In article <>, Hans-Georg
    >Michna <> writes


    >>Does anybody here have some experience with this scanner? I have
    >>a few questions.


    >Yup, years of experience with it. ;-)


    Kennedy,

    thanks for your very interesting replies! I'll try to comment on
    both in this message.

    >Well that raises an interesting question: Do you know what "as sharp as
    >the manufacturer states" should look like? If you don't (and too many
    >people think it means you should get adjacent black and white pixels -
    >it doesn't!) how can you determine that it isn't as sharp as it should
    >be?


    I think we can lay this question to rest. What I understand when
    I read 4,000 dpi is that those pixels represent the scanned
    picture optimally. For example, a perfect black-white edge
    should have only one row of grey pixels. But in the real world
    they don't do that.

    No sense discussing this much, things are as they are. I already
    wrote in my first message that all scanners I know exaggerate
    their resolution this way.

    >>It cannot nearly reproduce
    >>a good slide in all of its resolution. Sharp, fine text, for
    >>example becomes mushy and, in extreme cases, unreadable.


    >OK, you clearly have a problem with focus or dirt. The LS-4000 is
    >particularly fussy about film flatness, so if your film is curved by a
    >couple of millimetres then it will not maintain focus across the entire
    >frame.


    Glass frames, but when I take the film out or scan non-glass
    frames, the results are the same. I tested that, of course.

    >Assuming your film is flat, is the scanner set to focus automatically
    >before each scan or preview is made? If you have inadvertently switched
    >this option off in the preferences section then the scanner will just
    >hold the last focus position it had - which could be miles (well, not
    >quite that far) from the correct focus position.


    I'm fully aware of when the scanner performs its autofocus
    procedure. I keep finding some slides that are out of focus and
    have to rescan them. Some came out that way because there is no
    contrast in the center, and I have to choose a different focus
    point, but for most the reason is inexplicable, since a
    subsequent scan under identical conditions produces a normally
    sharp picture. No big deal.

    >The easiest diagnosis of dirt/dust is to scan a frame edge against a
    >white background on slide film. eg. the black gap between two lighter
    >frames or the edge of the slide mount with light subject matter at the
    >edge of the frame. If you have dirty mirrors or optics then you will
    >notice a smearing of the lighter portions of the image into the black
    >frame. If this is the problem then there is no alternative to having
    >the scanner stripped and cleaned - something you can do yourself if you
    >are confident of working with optical instruments, but otherwise it is
    >better to pay someone who is.


    Thanks for this information! It is very useful and interesing!

    In my case the scanner seems to be fine and not too dirty.

    >[...] Also, measurements show that for near saturated scans,
    >corresponding to shadows on negatives, the photon noise itself is
    >equivalent to just over 8-bits for the CCD. This can be improved by
    >multisampling, to over 10-bits at 16x. You will probably see more gain
    >from multisampling negatives than slides, because the dynamic range of
    >the scanner is very good.


    Interesting again. I wasn't sure about how many photons each
    sensor pixel gets, but we're apparently near the limit.

    >RTFMs - if you don't have copies then download them from the Nikon
    >website. There are two - one for the scanner itself and one for the
    >Nikonscan software.


    Did that already. Needed the manual for some information.

    >Obtain and install the latest firmware update (v1.10)


    v1.10 was already loaded.

    >Obtain and install the latest version of NikonScan (v4.02)


    Had to do that anyway, because v3 doesn't run on Windows XP.


    On Wed, 14 Sep 2005 00:12:36 +0100, Kennedy McEwen
    <> wrote:

    >In article <>, Hans-Georg
    >Michna <> writes


    >[Precise, interesting and useful information snipped]


    >Your general dissatisfaction with scanners appears to be down to a
    >misunderstanding of what the manufacturers are offering in their
    >specifications.


    Let's say an unfulfilled hope that those 4,000 pixels per inch
    should reproduce the information on the slide as well as
    possible. Guess I'd have to buy an 8,000 or 16,000 dpi scanner
    to get a real 4,000 dpi resolution.

    But I understand your points and have to accept that the scanner
    manufacturers do it the way we see here.

    >What I find more surprising is that you have somehow convinced yourself
    >that so much more information is available on the film to begin with!
    >Try this for comparison: view your scanned edge transition or text on
    >your screen at 100% magnification, so that you can see the softness you
    >are concerned about. Typical displays these days are around 96ppi - so
    >the image size from a full scanned 35mm frame would be around
    >40x60inches. Now, go into a projection room and project your slide at
    >that same size and view it from the same distance as you do the monitor
    >- about 10-12". The original will, of course, look a little sharper
    >but, if your scanner is functioning correctly, it won't be that much
    >sharper and it will surprise you just how little information that the
    >scanner is not capable of lifting off the film.


    Here the truth is obvious. I look at a slide that contains some
    small black text on a white background, and I see that the slide
    reproduces that text impressively sharply.

    Then I scan the slide at 4,000 dpi, and the text is barely
    readable or not readable any more, even though the number of
    pixels per letter would easily suffice to have it readable. In
    fact, there is a lot of cross-feed from one pixel to the next,
    either because the optics are not sharp enough, could also be
    out of focus, or because there is some cross-feed on the sensor.

    But as I wrote, effectively a 4,000 dpi scanner like this is
    unable to reproduce the full sharpness of a slide, and I have to
    accept that grudgingly.

    I'm still under the impression that this scanner produces
    reasonably sharp pictures at 2,000 dpi. Of course raising the
    resolution always adds some information to the result, but
    raising it to the native 4,000 dpi yields rather little. Egdes,
    for example, don't become any sharper.

    Let me post an example. Please have a look at

    http://www.michna.com/temp/images/india0203_detail_2000dpi.png
    http://www.michna.com/temp/images/india0203_detail_4000dpi.png

    The first, 2,000 dpi scan was scanned earlier and taken from
    http://www.michna.com/photos/ . It was an automatic scan through
    the slide feeder. The second, 4,000 dpi scan I have taken just
    now for comparison.

    The samples show that the autofocus seems to be imperfect,
    because the higher resolution scan is actually less sharp than
    the 2,000 dpi scan. That's another factor one has to put up
    with.

    The text on the bus is readable, but when you look at the
    original slide, there is no comparison. Unfortunately I can't
    show it to you, but the text is pitch-black and perfectly sharp,
    totally incomparable to the scans. The scanner does not
    reproduce the quality of the slide by far. The film is
    Ektachrome 100, by the way. If the camera can do it, why not the
    slide scanner?

    Anyway, what you can see is that it serves no purpose to scan at
    4,000 dpi. The pictures do not contain significantly more
    information than 2,000 dpi scans.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 15, 2005
    #11
  12. In article <>,
    Hans-Georg Michna <> wrote:
    >Anyway, what you can see is that it serves no purpose to scan at
    >4,000 dpi. The pictures do not contain significantly more
    >information than 2,000 dpi scans.


    For comparison, here are a couple of scans on my LS-4000:

    <http://misc.hq.phicoh.net/edge/Image1a.tif>
    and
    <http://misc.hq.phicoh.net/edge/Image1b.tif>
    are 4000 ppi crops
    <http://misc.hq.phicoh.net/edge/Image1f.tif>
    gives and overview of the entire frame.

    <http://misc.hq.phicoh.net/edge/Image2a.tif>
    is a crop of a scan of a razor blade. See
    <http://misc.hq.phicoh.net/edge/notes.txt>
    for more details


    --
    That was it. Done. The faulty Monk was turned out into the desert where it
    could believe what it liked, including the idea that it had been hard done
    by. It was allowed to keep its horse, since horses were so cheap to make.
    -- Douglas Adams in Dirk Gently's Holistic Detective Agency
    Philip Homburg, Sep 15, 2005
    #12
  13. In article <>, Hans-Georg
    Michna <> writes
    >
    >Kennedy,
    >
    >thanks for your very interesting replies! I'll try to comment on
    >both in this message.
    >
    >>Well that raises an interesting question: Do you know what "as sharp as
    >>the manufacturer states" should look like? If you don't (and too many
    >>people think it means you should get adjacent black and white pixels -
    >>it doesn't!) how can you determine that it isn't as sharp as it should
    >>be?

    >
    >I think we can lay this question to rest. What I understand when
    >I read 4,000 dpi is that those pixels represent the scanned
    >picture optimally. For example, a perfect black-white edge
    >should have only one row of grey pixels. But in the real world
    >they don't do that.
    >
    >No sense discussing this much, things are as they are. I already
    >wrote in my first message that all scanners I know exaggerate
    >their resolution this way.


    It is worth noting that this interpretation of resolution, whilst not
    unique to yourself, is incorrect. Two forms of resolution take place in
    scanners - sampling resolution (which I prefer to call sampling density
    to avoid confusion) and optical resolution. Sampling resolution or
    density is the headline number that the manufacturers hit you with and
    is basically the number of pixels that the scanner samples per inch. It
    doesn't say anything about the actual performance of the scanner or how
    distinct any pixel is from adjacent pixels. That comes down to optical
    resolution and, as I am sure you have guessed, manufacturers are less
    than vociferous in making customers, let alone potential customers,
    aware of that. There are good reasons for this - people are stupid and
    will buy a scanner from a manufacturer that claims 4,000ppi resolution
    than one who offers only 1500cy/in resolution.

    However even optical resolution isn't anything close to your estimate.
    Optical resolution is based on the century old Rayleigh criterion - how
    far apart can two point sources be when the system just manages to
    distinguish them as separate objects rather than a single coalesced
    mass. So, your single grey pixel across an edge isn't realistic at all.
    You might have two black dots the size of several pixel size on a white
    background and move them closer together scanning at each separation.
    When you get each dot with a few light pixels in between then you will
    probably get a good contrasty signal. However that isn't the optical
    resolution - you have to move the dots closer still until they can just
    be identified as separate instead of a single rectangular blob. At this
    point, the signal between them will be just marginally lighter than the
    dots themselves. That distance, or rather the number of such distances
    in a reference unit, is the optical resolution.

    Bear in mind that even a perfect lens won't image an edge as perfect as
    you describe - it will always blur due to diffraction. A real lens will
    add blur from aberrations, but overcoming diffraction means using as big
    an aperture as possible. However, even an infinite aperture will only
    resolve (and I mean the Rayleigh version here) about half a wavelength
    of light - about 50,000ppi.

    So, as you see, reading 4000ppi or 4000dpi on a scanner data sheet and
    thinking you will get what you are looking for is just wrong on (at
    least) two counts. They are quoting sampling density as resolution, not
    optical resolution. It is industry standard, whether we like it or find
    it useful or not. But, even if the manufacturers came out and said what
    the optical resolution of their scanners were - and the Coolscan 4000 is
    pretty close to the equivalent of 4000ppi optical resolution - it still
    wouldn't be what you are expecting, and nor would 10,000ppi; perhaps
    50,000ppi might be close but there isn't that much information on the
    film so don't hold your breath or your wallet to pay for that humungous
    lens that it would need! ;-)

    >
    >>>It cannot nearly reproduce
    >>>a good slide in all of its resolution. Sharp, fine text, for
    >>>example becomes mushy and, in extreme cases, unreadable.

    >
    >>OK, you clearly have a problem with focus or dirt. The LS-4000 is
    >>particularly fussy about film flatness, so if your film is curved by a
    >>couple of millimetres then it will not maintain focus across the entire
    >>frame.

    >
    >Glass frames, but when I take the film out or scan non-glass
    >frames, the results are the same. I tested that, of course.
    >

    OK, worth bearing in mind for the future though - if you get this focus
    problem fixed then the next one you will hit is film flatness. :-(

    >>Assuming your film is flat, is the scanner set to focus automatically
    >>before each scan or preview is made? If you have inadvertently switched
    >>this option off in the preferences section then the scanner will just
    >>hold the last focus position it had - which could be miles (well, not
    >>quite that far) from the correct focus position.

    >
    >I'm fully aware of when the scanner performs its autofocus
    >procedure. I keep finding some slides that are out of focus and
    >have to rescan them. Some came out that way because there is no
    >contrast in the center, and I have to choose a different focus
    >point, but for most the reason is inexplicable, since a
    >subsequent scan under identical conditions produces a normally
    >sharp picture. No big deal.
    >

    OK, this is a bit worrying - especially the comment about "contrast in
    the center" causing focus problems. The autofocus works on the film
    grain, not on the image itself. The only time it should fail is if
    there is no film grain at the autofocus position, eg. a completely
    transparent emulsion. Even then, minor dust and scratches on the
    emulsion surface should be enough to focus on. This suggests that the
    optical system is not focussing well enough at its optimum position for
    the grain to be resolved very well at all - which confirms your
    complaint.

    BTW - are you aware than the autofocus point can be moved to any point
    in the image? This is particularly useful once you resolve your focus
    problem and only have to deal with film flatness - it lets you get the
    optimum focus on the part of the frame you think needs it, not just in
    the middle.

    >>The easiest diagnosis of dirt/dust is to scan a frame edge against a
    >>white background on slide film. eg. the black gap between two lighter
    >>frames or the edge of the slide mount with light subject matter at the
    >>edge of the frame. If you have dirty mirrors or optics then you will
    >>notice a smearing of the lighter portions of the image into the black
    >>frame. If this is the problem then there is no alternative to having
    >>the scanner stripped and cleaned - something you can do yourself if you
    >>are confident of working with optical instruments, but otherwise it is
    >>better to pay someone who is.

    >
    >Thanks for this information! It is very useful and interesing!
    >
    >In my case the scanner seems to be fine and not too dirty.
    >

    There are a couple of examples of dirty optics scans on the net if you
    search google for Coolscan dirty optics including:
    http://www.vad1.com/photo/dirty-scanner/
    which has links to cleaning these if you find this becomes an issue.

    >>[...] Also, measurements show that for near saturated scans,
    >>corresponding to shadows on negatives, the photon noise itself is
    >>equivalent to just over 8-bits for the CCD. This can be improved by
    >>multisampling, to over 10-bits at 16x. You will probably see more gain
    >>from multisampling negatives than slides, because the dynamic range of
    >>the scanner is very good.

    >
    >Interesting again. I wasn't sure about how many photons each
    >sensor pixel gets, but we're apparently near the limit.


    Not what it means. Photon noise is just the random variation on the
    arrival of photons at the sensor. This is just like tossing coins - if
    you only have one coin then there is just as much chance of getting a
    tail (0) as a head (1) - so the noise is the same as the signal. If you
    have 100 coins then the chances are the same, but for each time you toss
    100 coins the number of heads will vary by less than the 50 heads you
    get on average - it should average around 10, giving a signal to noise
    of 5. 10,000 coins should give 5,000 heads, but the variation on each
    run should average about 100, giving a signal to noise of about 50 and
    so on.

    So the fact that the photon noise is quite so low is a consequence of
    the limitation of the CCD, not an indication that you are close to the
    limits of what is achievable. Basically, it means that the CCD can only
    detect and store the signal from a limited number of photons before it
    saturates - around 100,000. If you could get a better CCD with more
    storage capacity in each cell then you could expose for a longer time,
    detect more photons and get better signal to noise. Since you are stuck
    with whatever Nikon felt gave them the best trade-off between
    performance and cost, the best you can do is to simulate a bigger
    capacity CCD. That is all that multiscanning does.

    >
    >>RTFMs - if you don't have copies then download them from the Nikon
    >>website. There are two - one for the scanner itself and one for the
    >>Nikonscan software.

    >
    >Did that already. Needed the manual for some information.
    >
    >>Obtain and install the latest firmware update (v1.10)

    >
    >v1.10 was already loaded.
    >
    >>Obtain and install the latest version of NikonScan (v4.02)

    >
    >Had to do that anyway, because v3 doesn't run on Windows XP.
    >
    >
    >>Your general dissatisfaction with scanners appears to be down to a
    >>misunderstanding of what the manufacturers are offering in their
    >>specifications.

    >
    >Let's say an unfulfilled hope that those 4,000 pixels per inch
    >should reproduce the information on the slide as well as
    >possible. Guess I'd have to buy an 8,000 or 16,000 dpi scanner
    >to get a real 4,000 dpi resolution.
    >
    >But I understand your points and have to accept that the scanner
    >manufacturers do it the way we see here.
    >

    From the discussion above, I hope you now realise that this is not only
    unfulfilled, but unrealistic in your terms, but I doubt there is much
    more information actually on the film in the first place.

    >>What I find more surprising is that you have somehow convinced yourself
    >>that so much more information is available on the film to begin with!
    >>Try this for comparison: view your scanned edge transition or text on
    >>your screen at 100% magnification, so that you can see the softness you
    >>are concerned about. Typical displays these days are around 96ppi - so
    >>the image size from a full scanned 35mm frame would be around
    >>40x60inches. Now, go into a projection room and project your slide at
    >>that same size and view it from the same distance as you do the monitor
    >>- about 10-12". The original will, of course, look a little sharper
    >>but, if your scanner is functioning correctly, it won't be that much
    >>sharper and it will surprise you just how little information that the
    >>scanner is not capable of lifting off the film.

    >
    >Here the truth is obvious. I look at a slide that contains some
    >small black text on a white background, and I see that the slide
    >reproduces that text impressively sharply.
    >
    >Then I scan the slide at 4,000 dpi, and the text is barely
    >readable or not readable any more, even though the number of
    >pixels per letter would easily suffice to have it readable. In
    >fact, there is a lot of cross-feed from one pixel to the next,
    >either because the optics are not sharp enough, could also be
    >out of focus, or because there is some cross-feed on the sensor.
    >
    >But as I wrote, effectively a 4,000 dpi scanner like this is
    >unable to reproduce the full sharpness of a slide, and I have to
    >accept that grudgingly.
    >

    Fuji Velvia, for example, goes out to 160lp/mm if the original contrast
    is about 1000:1 - but even then, the contrast on the film will only be
    about 15% - lower contrast detail will be lost. 160lp/mm is equivalent
    to a little over 8000ppi. At a more typical 1.6:1 contrast, Velvia can
    only resolve 4000ppi, again reproducing around 15% contrast on the film
    (1.15:1). There are marginally higher resolution colour films around,
    but not generally available, but this is just the film - the camera lens
    and any defocus or shake during exposure will limit the image resolution
    further.

    So, unless your images often contain very fine detail at a very high
    contrast and you shot them with top of the range lenses at their optimum
    aperture and used a tripod for everything then it is unlikely there is
    much more than 4000ppi on the film itself.

    Of course whatever is on the film will be compromised by using a similar
    performance scanner and some examples suggest that 5400ppi available
    from the Konica Minolta DSE-5400 can pull a bit more off the film than
    the 4000ppi Nikons when it is present.
    See http://www.xs4all.nl/~bvdwolf/main/foto/scan/se5400/se5400-5.htm


    >I'm still under the impression that this scanner produces
    >reasonably sharp pictures at 2,000 dpi. Of course raising the
    >resolution always adds some information to the result, but
    >raising it to the native 4,000 dpi yields rather little. Egdes,
    >for example, don't become any sharper.
    >

    They may not look like it to you, but if you run some measurements, they
    certainly are.


    >Let me post an example. Please have a look at
    >
    >http://www.michna.com/temp/images/india0203_detail_2000dpi.png
    >http://www.michna.com/temp/images/india0203_detail_4000dpi.png
    >

    Well, just as an instance, in the 4000ppi scan I can easily read
    "MYSORE" on the back of the bus and the text below that is almost
    legible. It certainly isn't as legible in the 2000ppi scan even after
    scaling it back up to the same size, although it is discernible, but the
    text below that certainly isn't. In particular, look at the definition
    of the R in "MYSORE". Information is present in the 4000ppi scan that
    is not there on the 2000ppi scan, though I have to say, it is a lot less
    than I have seen on my LS-4000 with similar exercises.

    Incidentally, these images show some evidence of the dirty optics I was
    referring to earlier - look at how the white of the bus advert has
    bloomed out into adjacent areas, such as the black hair of the guy with
    the light shirt. Its possible that the original slide is like that
    given the type of shot, but I would not be surprised at the scanner
    being the source.

    Also, what scanner settings did you use for these because both show
    rather a lot of posterisation and, given the format, I don't think that
    is a compression artefact.

    >
    >The samples show that the autofocus seems to be imperfect,
    >because the higher resolution scan is actually less sharp than
    >the 2,000 dpi scan. That's another factor one has to put up
    >with.
    >

    Autofocus is undertaken at the same resolution irrespective of the
    resolution of the final scan. If you are getting inconsistent focus it
    isn't because you are changing resolution.

    >The film is
    >Ektachrome 100, by the way. If the camera can do it, why not the
    >slide scanner?
    >

    Well, I bet if you compared the camera to the actual scene (impractical
    in all but set tests of course) then you would notice the same
    shortfalls on the camera. Any instrument will only make the best
    interpretation of the information it has and no instrument is perfect.
    Your camera is interpreting the scene and produces what you see on the
    film - a lot less than what is in the scene though. However you are
    then presenting that to the scanner - it doesn't have the original scene
    to work with, so its interpretation of the image will be further
    degraded. That will always happen whether your scanner is 4000ppi or
    40000ppi.

    >Anyway, what you can see is that it serves no purpose to scan at
    >4,000 dpi. The pictures do not contain significantly more
    >information than 2,000 dpi scans.
    >

    I disagree, and your examples have not changed my view.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
    Kennedy McEwen, Sep 16, 2005
    #13
  14. "Kennedy McEwen" <> wrote:
    >>

    > BTW - are you aware than the autofocus point can be moved to any point in
    > the image? This is particularly useful once you resolve your focus
    > problem and only have to deal with film flatness - it lets you get the
    > optimum focus on the part of the frame you think needs it, not just in the
    > middle.


    It's even better than that. You can use the AF system to measure the
    position of the film at enough points on the film to determine the range of
    distances, and then set the focus manually to the center of that range and
    get everything on the frame in focus.

    >>Anyway, what you can see is that it serves no purpose to scan at
    >>4,000 dpi. The pictures do not contain significantly more
    >>information than 2,000 dpi scans.
    >>

    > I disagree, and your examples have not changed my view.


    The salient term in there is "significant", and we may have different
    definitions. I _think_ that scanning at 4000 ppi, applying noise reduction
    and a tad of sharpening, and then downsampling to 2200 ppi or so will
    produce better pixels* than just scanning at 2000 ppi. I think. But in
    general, most scans can be bicubic downsampled to 1/2 the scanning density
    and then bicubic upsampled to the original scanning density without
    _significant_ loss of information.

    *: The US$3500 question I'm agonizing with this month is whether 12MP of 5D
    pixels are (a) really "worth more" than 16MP (645) of these fairly nice 2200
    ppi pixels (I'm pretty sure 12MP wins hands down here), and (b) whether I
    will be _un_happy enough with those 12MP pixels printed at 224 ppi (13x19)
    compared to 28MP (6x7) of those fairly nice pixels printed at 312 ppi (also
    13x19). This (b) part is hard. If the flipping 5D had 16.7MP (250 ppi at
    13x19) I wouldn't be worrying this. Aaaaaaaaaaaarg.

    David J. Littleboy

    Tokyo, Japan
    David J. Littleboy, Sep 16, 2005
    #14
  15. In article <dgdv02$koh$>, David J. Littleboy
    <> writes
    >
    >*: The US$3500 question I'm agonizing with this month is whether 12MP of 5D
    >pixels are (a) really "worth more" than 16MP (645) of these fairly nice 2200
    >ppi pixels (I'm pretty sure 12MP wins hands down here), and (b) whether I
    >will be _un_happy enough with those 12MP pixels printed at 224 ppi (13x19)
    >compared to 28MP (6x7) of those fairly nice pixels printed at 312 ppi (also
    >13x19). This (b) part is hard.
    >

    Whilst not agonizing (other things have been causing that much more
    literally recently) I am going through a similar debate myself, having
    rejected the digital camera route for anything more than toys thus far
    due to combinations of cost and performance. The 5D is likely to break
    the rule I had previously set myself of waiting till the 1Ds
    became affordable within my priorities.

    I am fairly confident that 5D output printed at 13x19" will be perfectly
    acceptable even though the source resolution will be around 220ppi - the
    sample images available at 100 & 200ASA look so noise free that
    resampling sensibly to the printer native resolution will pretty much
    artefact free. 220ppi is only just below the resolution of an unaided
    human eye, so nobody will notice the pixel deficiency from the sensor
    unless comparing it directly to something else with aided vision. It is
    more a question of whether optics are really up to that sort of
    enlargement - 14x scaling (excluding crops) is really asking a lot.

    >If the flipping 5D had 16.7MP (250 ppi at
    >13x19) I wouldn't be worrying this. Aaaaaaaaaaaarg.


    Yes you would - you'd just be at the next layer of the onion and
    worrying about the optical scaling, wishing the 5D was a 645 format
    camera instead of 35mm.

    The question I have to ask myself is more along the lines of does the
    quality of the 5D justify investing 7-10k of UKP on a complete new set
    of optics and accessories. The answer is probably yes, but its a hard
    one to implement.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
    Kennedy McEwen, Sep 16, 2005
    #15
  16. "Kennedy McEwen" <> wrote:
    > David J. Littleboy <> writes
    >>
    >>*: The US$3500 question I'm agonizing with this month is whether 12MP of
    >>5D
    >>pixels are (a) really "worth more" than 16MP (645) of these fairly nice
    >>2200
    >>ppi pixels (I'm pretty sure 12MP wins hands down here), and (b) whether I
    >>will be _un_happy enough with those 12MP pixels printed at 224 ppi (13x19)
    >>compared to 28MP (6x7) of those fairly nice pixels printed at 312 ppi
    >>(also
    >>13x19). This (b) part is hard.
    >>

    > Whilst not agonizing (other things have been causing that much more
    > literally recently) I am going through a similar debate myself, having
    > rejected the digital camera route for anything more than toys thus far due
    > to combinations of cost and performance.


    We're more on the same page than I expected<g>. (Although I purchased a 300D
    for family snaps and low-light work.)

    > I am fairly confident that 5D output printed at 13x19" will be perfectly
    > acceptable even though the source resolution will be around 220ppi - the
    > sample images available at 100 & 200ASA look so noise free that resampling
    > sensibly to the printer native resolution will pretty much artefact free.
    > 220ppi is only just below the resolution of an unaided human eye, so
    > nobody will notice the pixel deficiency from the sensor unless comparing
    > it directly to something else with aided vision.


    I hope you're right...

    > It is more a question of whether optics are really up to that sort of
    > enlargement - 14x scaling (excluding crops) is really asking a lot.


    Here, though, I'm more sanguine. The resolution (not the "limiting
    resolution" but the "apparent resolution when test charts are viewed at 100%
    on the screen") is about 40 lp/mm, and that's not too painful. The 17-40 is
    going to be a tad soft in the corners at 17, but I'm expecting it to be fine
    in the 20 to 24mm range. And any prime 35mm or longer will be fine.

    >>If the flipping 5D had 16.7MP (250 ppi at
    >>13x19) I wouldn't be worrying this. Aaaaaaaaaaaarg.

    >
    > Yes you would - you'd just be at the next layer of the onion and worrying
    > about the optical scaling, wishing the 5D was a 645 format camera instead
    > of 35mm.


    Well, actually, no. I really don't need anything more than 13x19, so 16.7MP
    would be perfect. (But lens infelicities would raise their ugly head more
    than with 12MP.)

    > The question I have to ask myself is more along the lines of does the
    > quality of the 5D justify investing 7-10k of UKP on a complete new set of
    > optics and accessories. The answer is probably yes, but its a hard one to
    > implement.


    I planned ahead. When I bought my 300D last year, I bought the lenses I'd
    need for the 5D. (Well, all except the 85/1.8: the Tamron 28-75/2.8 will
    have to do in the short term, but my wife wants the 300D, and that's the
    lens she'll want on it.)

    David J. Littleboy
    Tokyo, Japan
    David J. Littleboy, Sep 16, 2005
    #16
  17. On Fri, 16 Sep 2005 00:43:38 +0200,
    (Philip Homburg) wrote:

    ><http://misc.hq.phicoh.net/edge/Image2a.tif>
    >is a crop of a scan of a razor blade.


    Philip,

    good example! Instead of the ideal one grey pixel between the
    black and white areas it has six. In other words, the scan is
    very decidedly unsharp. You can reduce the resolution by a
    factor of 2, 4, or even 6 without losing much information.
    That's also what I observe.

    The scanner is incomparably worse than the camera and the film I
    used to take photos.

    I guess that's just how it is. I don't quite understand why a
    scanner of a quality comparable to a camera has to be many times
    more expensive than a camera, but I can't change it. Perhaps it
    is because many more cameras are made than film scanners.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 16, 2005
    #17
  18. On Fri, 16 Sep 2005 05:42:24 +0100, Kennedy McEwen
    <> wrote:

    >It is worth noting that this interpretation of resolution, whilst not
    >unique to yourself, is incorrect. Two forms of resolution take place in
    >scanners - sampling resolution (which I prefer to call sampling density
    >to avoid confusion) and optical resolution. Sampling resolution or
    >density is the headline number that the manufacturers hit you with and
    >is basically the number of pixels that the scanner samples per inch. It
    >doesn't say anything about the actual performance of the scanner or how
    >distinct any pixel is from adjacent pixels. That comes down to optical
    >resolution and, as I am sure you have guessed, manufacturers are less
    >than vociferous in making customers, let alone potential customers,
    >aware of that. There are good reasons for this - people are stupid and
    >will buy a scanner from a manufacturer that claims 4,000ppi resolution
    >than one who offers only 1500cy/in resolution.


    Kennedy,

    yes, I'm fully aware of this. That's why I reduced my scan
    resolution from 4,000 ppi first to 2,400 ppi, then to 2,000 ppi.
    Of course I lose some information that way, I'm fully aware of
    that too, but not enough to warrant the much larger files.

    >The autofocus works on the film
    >grain, not on the image itself. The only time it should fail is if
    >there is no film grain at the autofocus position, eg. a completely
    >transparent emulsion. Even then, minor dust and scratches on the
    >emulsion surface should be enough to focus on. This suggests that the
    >optical system is not focussing well enough at its optimum position for
    >the grain to be resolved very well at all - which confirms your
    >complaint.


    Interesting! I guess I will have to have the scanner cleaned and
    checked. However, perhaps the errors happened because the
    scanner focussed on the glass surface, rather than on the film.
    I use GePe anti-Newton glass frames.

    >BTW - are you aware than the autofocus point can be moved to any point
    >in the image? This is particularly useful once you resolve your focus
    >problem and only have to deal with film flatness - it lets you get the
    >optimum focus on the part of the frame you think needs it, not just in
    >the middle.


    Yes, I used that function in a few cases. Do you happen to know
    under which circumstances the scanner puts the autofocus point
    back in the center? I'd like it back when I scan the next
    picture, but I couldn't find out, so I always moved it back by
    hand, which may be unnecessary.

    >Not what it means. Photon noise is just the random variation on the
    >arrival of photons at the sensor. [...]


    >So the fact that the photon noise is quite so low is a consequence of
    >the limitation of the CCD, not an indication that you are close to the
    >limits of what is achievable. Basically, it means that the CCD can only
    >detect and store the signal from a limited number of photons before it
    >saturates - around 100,000. If you could get a better CCD with more
    >storage capacity in each cell then you could expose for a longer time,
    >detect more photons and get better signal to noise. Since you are stuck
    >with whatever Nikon felt gave them the best trade-off between
    >performance and cost, the best you can do is to simulate a bigger
    >capacity CCD. That is all that multiscanning does.


    Thanks for the very good explanation! I'm learning.

    > From the discussion above, I hope you now realise that this is not only
    >unfulfilled, but unrealistic in your terms, but I doubt there is much
    >more information actually on the film in the first place.


    Oh, but there is! When I look at the photo, from which I took
    the two samples, through a magnifying glass, the text on the bus
    is absolutely black and sharp, incomparably better than the
    light grey fuzzy stuff in the scans. I can only surmise that the
    optical quality of the scanner optics is way below that of the
    camera lens through which I took the photo.

    >Fuji Velvia, for example, goes out to 160lp/mm if the original contrast
    >is about 1000:1 - but even then, the contrast on the film will only be
    >about 15% - lower contrast detail will be lost. 160lp/mm is equivalent
    >to a little over 8000ppi. At a more typical 1.6:1 contrast, Velvia can
    >only resolve 4000ppi, again reproducing around 15% contrast on the film
    >(1.15:1). There are marginally higher resolution colour films around,
    >but not generally available, but this is just the film - the camera lens
    >and any defocus or shake during exposure will limit the image resolution
    >further.
    >
    >So, unless your images often contain very fine detail at a very high
    >contrast and you shot them with top of the range lenses at their optimum
    >aperture and used a tripod for everything then it is unlikely there is
    >much more than 4000ppi on the film itself.


    Wish I could send you the original slide! This was Ektachrome
    100 shot through a good Canon lens. Too bad I have no way of
    scanning through a microscope to show you what's actually on the
    slide.

    But assuming the film contains good 4,000 ppi information, but
    the scanner's optics are obviously way below that, the
    phenomenon is already explained.

    >In article <>, Hans-Georg
    >Michna <> writes


    >>I'm still under the impression that this scanner produces
    >>reasonably sharp pictures at 2,000 dpi. Of course raising the
    >>resolution always adds some information to the result, but
    >>raising it to the native 4,000 dpi yields rather little. Egdes,
    >>for example, don't become any sharper.


    >They may not look like it to you, but if you run some measurements, they
    >certainly are.


    Of course the 4,000 ppi scans contain more information than the
    2,000 ppi scans, but the difference appears small to me,
    particularly when compared to the roughly 4 times larger file
    size.

    >>Let me post an example. Please have a look at
    >>
    >>http://www.michna.com/temp/images/india0203_detail_2000dpi.png
    >>http://www.michna.com/temp/images/india0203_detail_4000dpi.png


    >Also, what scanner settings did you use for these because both show
    >rather a lot of posterisation and, given the format, I don't think that
    >is a compression artefact.


    The 2,000 ppi scan was stored in JPEG format at the highest
    quality setting. The 4,000 ppi scan was never compressed, only
    converted from TIFF to PNG, so it does not contain any
    compression artefacts. I converted both to PNG to avoid losing
    quality again after cropping.

    I guess that there are much better scanners that can easily
    resolve 4,000 ppi to the pixel. I'd guess they'd be advertised
    as 8,000 ppi at least, probably even 16,000 ppi. Do you happen
    to know such scanners? What do they cost?

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Sep 16, 2005
    #18
  19. In article <>,
    Hans-Georg Michna <> wrote:
    >On Fri, 16 Sep 2005 00:43:38 +0200,
    >(Philip Homburg) wrote:
    >good example! Instead of the ideal one grey pixel between the
    >black and white areas it has six. In other words, the scan is
    >very decidedly unsharp. You can reduce the resolution by a
    >factor of 2, 4, or even 6 without losing much information.
    >That's also what I observe.


    Check out Image3a.tif and Image1a.tif as well. I accidentally left ICE
    on (which of course doesn't work with a razer blade).

    Image1a suggests that manually focussing the scanner is not really an
    option.

    >The scanner is incomparably worse than the camera and the film I
    >used to take photos.


    No, the scanner is good enough for color film.


    --
    That was it. Done. The faulty Monk was turned out into the desert where it
    could believe what it liked, including the idea that it had been hard done
    by. It was allowed to keep its horse, since horses were so cheap to make.
    -- Douglas Adams in Dirk Gently's Holistic Detective Agency
    Philip Homburg, Sep 16, 2005
    #19
  20. Hans-Georg Michna

    Guest

    I don't think you have to worry about the 5D at 13"x19". I have prints
    made from the 20D. 12"x18" no sweat. Even 20"x30" are impressive. I
    realize the ppi number is something like 115 here. But practically (and
    I am sure there is some image dependance) the end result is what
    counts. (I upsample to 300ppi then USM) I think the 20D can create
    images in many ways superior to 35mm film. The 5D should be quite
    impressive!!


    David J. Littleboy wrote:
    > "Kennedy McEwen" <> wrote:
    > > David J. Littleboy <> writes
    > >>
    > >>*: The US$3500 question I'm agonizing with this month is whether 12MP of
    > >>5D
    > >>pixels are (a) really "worth more" than 16MP (645) of these fairly nice
    > >>2200
    > >>ppi pixels (I'm pretty sure 12MP wins hands down here), and (b) whether I
    > >>will be _un_happy enough with those 12MP pixels printed at 224 ppi (13x19)
    > >>compared to 28MP (6x7) of those fairly nice pixels printed at 312 ppi
    > >>(also
    > >>13x19). This (b) part is hard.
    > >>

    > > Whilst not agonizing (other things have been causing that much more
    > > literally recently) I am going through a similar debate myself, having
    > > rejected the digital camera route for anything more than toys thus far due
    > > to combinations of cost and performance.

    >
    > We're more on the same page than I expected<g>. (Although I purchased a 300D
    > for family snaps and low-light work.)
    >
    > > I am fairly confident that 5D output printed at 13x19" will be perfectly
    > > acceptable even though the source resolution will be around 220ppi - the
    > > sample images available at 100 & 200ASA look so noise free that resampling
    > > sensibly to the printer native resolution will pretty much artefact free.
    > > 220ppi is only just below the resolution of an unaided human eye, so
    > > nobody will notice the pixel deficiency from the sensor unless comparing
    > > it directly to something else with aided vision.

    >
    > I hope you're right...
    >
    > > It is more a question of whether optics are really up to that sort of
    > > enlargement - 14x scaling (excluding crops) is really asking a lot.

    >
    > Here, though, I'm more sanguine. The resolution (not the "limiting
    > resolution" but the "apparent resolution when test charts are viewed at 100%
    > on the screen") is about 40 lp/mm, and that's not too painful. The 17-40 is
    > going to be a tad soft in the corners at 17, but I'm expecting it to be fine
    > in the 20 to 24mm range. And any prime 35mm or longer will be fine.
    >
    > >>If the flipping 5D had 16.7MP (250 ppi at
    > >>13x19) I wouldn't be worrying this. Aaaaaaaaaaaarg.

    > >
    > > Yes you would - you'd just be at the next layer of the onion and worrying
    > > about the optical scaling, wishing the 5D was a 645 format camera instead
    > > of 35mm.

    >
    > Well, actually, no. I really don't need anything more than 13x19, so 16.7MP
    > would be perfect. (But lens infelicities would raise their ugly head more
    > than with 12MP.)
    >
    > > The question I have to ask myself is more along the lines of does the
    > > quality of the 5D justify investing 7-10k of UKP on a complete new set of
    > > optics and accessories. The answer is probably yes, but its a hard one to
    > > implement.

    >
    > I planned ahead. When I bought my 300D last year, I bought the lenses I'd
    > need for the 5D. (Well, all except the 85/1.8: the Tamron 28-75/2.8 will
    > have to do in the short term, but my wife wants the 300D, and that's the
    > lens she'll want on it.)
    >
    > David J. Littleboy
    > Tokyo, Japan
    , Sep 18, 2005
    #20
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Eric Crowder

    Banding problem with Nikon coolscan 4000....Please help....

    Eric Crowder, Oct 4, 2003, in forum: Digital Photography
    Replies:
    0
    Views:
    439
    Eric Crowder
    Oct 4, 2003
  2. TonyOfoto

    Nikon Coolscan V or 4000?

    TonyOfoto, Jan 9, 2004, in forum: Digital Photography
    Replies:
    0
    Views:
    430
    TonyOfoto
    Jan 9, 2004
  3. Mike
    Replies:
    6
    Views:
    446
  4. David Gintz

    Slide Scanning Help - Nikon Coolscan 4000 ED

    David Gintz, Apr 27, 2005, in forum: Digital Photography
    Replies:
    5
    Views:
    830
  5. David Gintz

    Increasing Crop Size on Nikon Super Coolscan 4000 ED

    David Gintz, Apr 30, 2005, in forum: Digital Photography
    Replies:
    4
    Views:
    429
Loading...

Share This Page