Nikon LS-4000: Unacceptable Depth of Focus

Discussion in 'Digital Darkroom' started by rishij, Feb 28, 2008.

  1. I have posted previously about the dismal depth of focus for the Nikon film scanners, especially the LS- 4000. On these fora, we discussed extensively the use of glass holders, anti-newton glass holders, etc. This is all documented here:
    However, one thing most of us generally agreed on was that, when using the SA-21 motorized strip film feeder, the middle frames of the strip (not the first or last frames) scanned sharply across the full frame. Assuming you had a strip of 6 frames (images), frame #1 and frame #6 ended up having significant softness on upwards of 30% of one side of the frame, due to the fact that film curls and due to the fact that the edge frames do not have excess film for the motorized portion of the SA-21 to latch onto and thereby flatten the film.
    Well, I'm overturning that theory that the mid frames (frames 2-5) are sharp across the frame. My scans of frames 2-5 are consistently unsharp on the left 15-20% of the frame. Focus point is set to somewhere a bit left of the center -- Vuescan's default. Mendel Leisk has previously confirmed that setting the focus point to a little off-center helps, since the scanner DOES have SOME depth of focus.
    In order to get the extreme left portion of the frame completely in focus, one must focus on the extreme left of the frame, thereby loosing sharpness in the center of the image.
    Please have a detailed look here:
    Please right-click and download the image, or view in a separate window at 100%, since the image has been downscaled above. Alternatively, click here:
    Nikon- LackOfDepthOfFocus
    You scanners out there -- are you finding this sort of softness even for frames 2-5 when using the SA- 21 adapter? It's really unacceptable. What's the point of even having L lenses when this happens when scanning (above image taken with a Canon 17-40 f/4L lens at f/11).
    Moreover, I don't even think this problem is addressable in a hardware-fix. At this point, after so much experimenting (and NOT wanting to get into wet mounting), I'm beginning to think that the simplest solution would be to scan images with different focus points, then combine them. I do this manually in Photoshop, but perhaps we could write software, or ask Ed Hamrick to modify Vuescan, to incorporate this feature?
    I am so close to perfecting the technique of scanning (finally, IT8 targeting with some new open-source profiling software based on little CMS has given me spot-on color accuracy in terms of matching what I see on my lightbox); but this final little detail is obnoxious. Surely, there must be a way for us to solve this problem.
    Let's engage in discussion!
    Many thanks in advance,
  2. I do not think bulk scanning will ever work with the precision you expect.

    Film is not flat unless you make it flat either with glass carriers or you put it under a book for a week in strip form.
  3. Hey Ho

    I always wondered why Nikon didn't put some slight 'curve' in the film path to keep it the film flat (across the width) when using the motorized strip feeder. Perhaps its something to do with imacon patents?

    dunno, but its certainly an issue and definitely inhibits getting as much from ones 35mm as one can see under a microscope exists on the frame.
  4. It's too bad they don't have a slower smaller apeture mode with somewhat more depth of field. There's no technical reason why you can't have a variable aperture scanner.
  5. "There's no technical reason why you can't have a variable aperture scanner."

    Diffraction limits if the CCD pixels are very small, perhaps?

    I still need to send you that target, Roger. I haven't forgotten, sorry!
  6. jtk


    LS4000 focus, as Minolta 5400 I & II, is inherently inferior to LS5000 and 50ED.

    I use a has plenty of depth-of-focus...running fully automatic, it scans grain-sharp from corner-to-corner except at ends of strips of especially curved film...which justifies FH-3 carrier.

    If film's especially curved one should focus half way from the center of the arc.. I've rarely found that useful for modern films... more often helpful with old-fashioned, thick emulsion films.
  7. "Diffraction limits if the CCD pixels are very small, perhaps? "

    Maybe, but there is a range of DOF in similar-type desktop scanners
    as far as I can tell with the Minolta 5400 and LS4000 being fairly poor, newer Nikons are said to be somewhat better, and my FS4000US has more generous DOF based on a few scans I've done on both.

    If you had variable aperture than users could choose if they want optimum sharpness at the point of focus, or if they can put up with some diffraction-induced sharpening and have an overall sharper image across the frame.
  8. Either way, this is ludicrous. How can Nikon put out a >$1000 product (when I bought it)
    that is largely non-functional to me because of this problem?!

    Anyone want to join me in a class-action suit against Nikon so we can upgrade our LS-
    4000's to something (LS-5000?) that can actually scan a full frame of film - a capability
    that SHOULD have been there in the first place with this thing?

    By the way, my film is already pretty darn flat. Except for some curl along the shorter axis
    of the film, which doesn't explain the above results, because the above results are
    resulting from lack of depth-of-field due to some tiny curvature along the long axis of the

    Also, I get better results with the SA-21 than the FH-3 carrier. The FH-3 carrier is
    seriously a piece of crap that can't even snap shut tightly.
  9. jtk


    Yes, FH-3 is crap. Necessary crap, sometimes.

    LS-4000 was the best there was, years ago. Still, people who work at it do get great scans, just as they do with Minoltas.

    fyi this issue doesn't relate to "depth of field," but to "depth of focus." Apples/oranges.

    Nikons (and the last Minolta) use a point light source (LED). Point light sources lose sharpness and actively create wild diffraction oddities with diaphragms. That's why point light source enlarger lenses are used wide open, not stopped down. The same El Nikor is SHOCKINGLY sharper wide-open with a point source than it is stopped down with an incandescent lamp in the same condenser enlarger. Nikon doesn't use a diaphragm because a) it wouldn't work and b) current Nikons have no need for greater depth of focus with modern film.
  10. "Nikons (and the last Minolta) use a point light source (LED). Point light sources lose sharpness and actively create wild diffraction oddities with diaphragms. That's why point light source enlarger lenses are used wide open, not stopped down. The same El Nikor is SHOCKINGLY sharper wide-open with a point source than it is stopped down with an incandescent lamp in the same condenser enlarger. "

    That's really interesting, John. I wonder if using a scanhancer-type diffusing plate would make a difference with this. Interesting that the Minolta with a fluorescent lamp had a scanhancer in the light path but not the II which used point source LEDs.

    Back to the original problem: using Vuescan you should be able to use manual focus to get numerical focus reads setting the focus cursor in different parts of the frame. I'd use that to see what the variation is and to try to pick a number beteween the center and the edge values. Looking at your sample image, I think a value that compromised between the two would yield acceptable results in combination with software sharpening.

    I also use folded over tape inside my film holder to keep it from flexing in places. Your holder is different than mine but it might help.

    Finally- for 35mm I find that the slide mounts hold the film flatter than my neg holder.
  11. jtk


    Roger, Scanhancer has trouble with Nikon, though supposedly can be made to work. Scanhancer wastes so much light that all of Nikon's brainwork gets whacked, needs to be heavily adjusted. Scans take far longer. Works with mounted slides, doesn't seem to with strips (I've not figured that out).

    FH-3 was exceptionally bad design and Nikon was stupid to actually go ahead and build it. Frustrating as hell, but it actually does work.

    The current Nikons are wonderful, don't need any improvement (except for FH-3). Problems come from failing to understand how they work and attempting to do things in the scan that should be done in Photoshop.
    I've heard they even work with Mac's latest animal, don't know about Vista.
  12. Rishi, I see 2 things you can do: 1. Follow Roger Smith's guidance in his 3rd paragraph posted just above (03:55 p.m). You have exacerbated matters by choosing to focus at the two extremes. The scanner has *some* depth of focus: the range within which sharpness is "acceptable". Take advantage of this. Assuming your film has a simple dome curvature, and the center is furthest from the scanner lens, you need to move your point of focus up that curve to a point where the center is still *just* acceptable sharp. I'll try to attach a jpeg sketch illustrating this. So to summarize, while in Vuescan, just make a sketch representing the image frame, a decent size. Then preview a frame with detail most everywhere, do manual focus at various points, and record your results. Divide up the frame kind of like a Union Jack flag, do points at all the extremes, the center, and in between. For starters, try splitting the difference between the 2 extremes, and setting that number manually. If that get's your whole frame passably focussed, your work is done. (Somehow I don't think this will be good enough for you, though, .) If this first try results in unacceptable softness at the centre and the edges, something has to give. I would adjust the focus point until you just get the center *reasonably* sharp, and then use that for future scans with the same film type, ie: similar curvature. It's worth noting that focus tends to fall-off more gradually on the far side than the near side. Also, Ed Hamrick has told me that the focus values -1 through +1 represent the full travel of the respective scanner's focus adjustment. 2. The heart of your problem is film curvature. If you can eliminate, or at least reduce the curvature, the problem is gone. Slide film has been problematic for me as well. By comparison, I have "one-hour photo" color negative film that is as flat as flap jacks. I wish I knew how this is accomplished: something to do with a rapid drying process? Or the film type? Anyway, one suggestion I can think of is to turn the strips into slides, and mount them, at least for the scanning process, in a tensioning mount like the Wess AHX500K. This was my process. What I found was the film *still* was not flat. But what did happen was the simple dome was tensioned enough that the center reversed, dimpling back until it was about the same level as the edges, and overall variation from highest to lowest was cut in half. I still chose to carefully manually focus with my scanner (Scan Elite 5400, using OEM software), focussing each frame. I would try to focus at the same "sweet-spot" on each frame, looking for an area of light grain, *without* a lot of detail, so that the scanner was focussing on grain alone. That was the theory, anyway. I also found with OEM software's focus mechanism that I would get *2* peaks, indicated point of best focus. I came to the conclusion that the first was a false peak, when the focus mechanism came to the film back. And the second was when the focus mechanism detected the emulsion. This may-or-may-not be a factor with the Nikons. Anyway, Rishi: I don't think you will get far trying to get Nikon to do anything. Whether or not this is right is a moot point, they're just too big, and pretty much the only company still making dedicated film scanners. I agree a little more depth of focus would be wonderful. But, to get things done, I would just put your efforts to improving what you have. I think using #1 and a bit of sharpening will likely do the trick.
  13. You know, looking at your centre focussed example, I don't see a lot of fall-off in sharpness. It does look like a slight adjustment of the focus value would allow you to keep center sharpness and recoup most or all at the edges.

    FWIW, I've exchanged slides with an LS5000 user and found the Nikon had more depth of focus than my Elite 5400, maybe by a factor of 50%.
  14. Thanks Mendel. The amount of work you've done re: this problem is incredible, and quite

    I've attempted your methods of using Post-It notes with Gepe slide mounts... worked well,
    but not well enough. I tried those Wess mounts that tension the film by latching on to the
    spokes... again, worked well, but not well enough. The tension introduced by stretching at
    the spokes introduced little 'humps' at every spoke, which resulted in a number of small
    'lines' of focus fall-off.

    More importantly, Vuescan ALREADY picks a point somewhere in between center and edge
    as the focus point. Hence I must correct myself: the 'center' focus above is actually
    somewhere ~20% left of the center point. And still I see this sharpness fall-off.

    I guess what it boils down to is this:

    I don't want ACCEPTABLE when it's possible to get much better. I want *as sharp* as one
    can get from film, because only then is it on par w/ even a 10 MP digital CMOS sensor.

    I guess that is too much to ask from this 'dedicated film scanner'? It truly irks me. I guess I
    want to try a LS-5000 now... but who's going to buy my LS-4000? Plus, should I really
    spend half the price of a Canon 5D for a scanner that *might* solve my problems.
    Sometimes the state of consumer products just appalls me. I work in a scientific
    laboratory and this sort of $hit would never pass in scientific instruments. And it's not like
    my LS-4000 was a $10 or even a couple hundred bucks. A big company like Nikon should
    be held responsible... but, then again, how many of us crazy film-scanning folk are out
    there for them to cater to?

    Again, Mendel, you've done an incredible job researching this problem. I'm just not
    satisfied with any of the solutions... I'll post some of my examples from using the mounts
    and from focusing somewhere between center and edge... still, I've already done these and
    just not found them reasonable given how sharp I can get the center *if* I just focus
    somewhere near the center... other focus points make it less sharp than it could have
    been, and who wants that when one shoots with top quality L-lenses to begin with?

    Rishi :)
  15. If you're "center" focus is at Vuescan's default (roughly 2/3 of the way up a portrait oriented image at center), then you're already compromising a little. That's evident in you "center" focussed sample, the edge doesn't look too bad. Try this:

    Assuming the same portrait oriented image (Vuescan's default), push the focus up a bit more, and over to the left, so that it's at about the center of a diagonal line from center of image to upper left corner. Do a few test scans, moving it forward and back along this line, noting the numbers as you go.

    Open pairs of scans focussed at different points, zoom to the same point at (say) 200%, toggle between. See how long the center stays sharp, and how much you can improve the corners.

    Bottom line: there are no absolutes. Best focus is a razor thin slice. Acceptable focus is much thicker, likely not as thick as the deflection in your film, but close. It seems to me you're leaning towards something like wet scanning, which can bring new issues. Still, I think you're chasing diminishing returns. Acceptable isn't as bad as it sounds. By that I mean almost unoticable softening at center, and slight softening around the very edges, which attract little notice, and often end up behind a mat.

    Also, watch out for gotchas. Like: is your film perpendicular to the scanners lens.
  16. Rishi, are you shooting dslr too? I picked up a very nice film slr soon after getting into scanning. About a year later I got a 20D, and have not shot any film since. I've since switched to 5D. I have decades of film to deal with, but quietly decided to quit beating my head against the scanning wall, for any future photography.

    Regarding the stopping down ideas posted, the Coolscan line would be good candidates, *if* Nikon was interested. Just one stop down, would double your scan times, I suppose, but the Coolscan's initial speed would mean the slowdown would not be that much of a burden. And the cool light source would be an important factor. With a warm light source, doubling your scan time increases film flex. Of course, all this is pie-in-the-sky.

    From what I've read, the latest generation of Coolscans, both the V and 5000, have somewhat improved depth of focus over the 4000.
  17. jtk


    Les, I think there actually IS a better FH-3 equivalent holder for Nikon... the guy that makes it (simple lucite sandwich with registration marks)..seemingly only for odd formats (110, 126 etc)...

    I've never seen him advertising a 35mm version, maybe because he makes his living scanning 35, not selling holders... advertises on 'Bay :)

    Any sensible machinist would design and execute better than Nikon did, not to mention Minolta and Microtek...would use aircraft aluminum with lateral positioning (not sprocket registration) pins, just as Omega enlarger does.
  18. jtk


    Les, Google "4000 Vs 5000" I happened to this'll find references to 5000 being much better than 4000 because of improved depth-of-focus. 5000's twice as fast, and Ice4 may have advantages over Ice3, but focus is the clincher.

    Right, the 'Bay carrier isn't "fancier." Nikon's FH-3 substitutes fancy for good. That's a typical weakness in Japanese industrial design, typically compensated-for by quest for perfection (Toyota)...but not in this case.

    A simple sandwich with something to position film (pins or even tape) would be better...a clicker for frame positioning could be mounted on the exterior of MA-21, but simple registration marks would do the job... MA-21 could easily be modified to use the simple sandwich...I think (don't know) that'd be necessary to support the sandwich as it passed through...don't know what the 'Bay guy does in that respect.
  19. Film scanning can be aggravating for these reasons- there's so much that can go wrong at every step. From flattening my negs/slide strips in heavy books to finding a lab that keeps water spots off of them to repeatedly servicing cameras that like to leave little horizontal scratches on the film, it really does take a special sort of fanatic to do this with a large amount of film.

    I have to say I'm enjoying my 20D for the times when I'm shooting a lot and my old film cameras for times when I am going to set up and only take a few shots.
  20. Les, I really do think that this is a blatant design issue. So many folk have had this
    problem. Re: servicing -- I don't particularly feel like shelling out hundreds of dollars.
    Would rather save up for a 5D.

    Mendel -- no, haven't gone digital yet. Decided to invest in L lenses first. Regardless, I
    have too much old valuable film lying around to just forego the scanning process. Much
    like yourself.

    I will try your method re: moving my focus point, and report back.

    John, it's hard to try tape or to modify the film itself at all since the SA-21 latches onto
    the 'spokes' of the film. A modified FH-3 might be the best option, but seems like a huge
    design project.

    Perhaps I need to try borrowing a LS-5000 to compare...
  21. By the way, out of curiosity, in the 100% crop above, is it me or is the bottom right sample
    (the center of the image, with focus point near center) amazingly sharp for an unsharpened
    film scan?

    This was shot with a 17-40L lens, and I don't remember any of my previous scans, when
    using the el cheapo 28-80 II and/or 75-300mm III, looking this sharp anywhere in the
    frame. To me, that unsharpened film scan looks about as sharp as an unsharpened RAW
    file coming out of a 5D (in my limited tests).

    Does anyone agree? Just curious. It IS pretty difficult to assay lens sharpness while being a
    film shooter, so just curious as to what others think... if I could consistently get this sort of
    sharp scans across the full frame from my film, I would be quite content holding off
    switching to digital...
  22. Les,

    Awesome, thanks for your help.

    WOW, you've only got loss of sharpening on the left/right of 10% of the frame at most...

    But for me, I swear, it starts getting soft 33% in from the left edge. Like 1/3 of the image
    is soft. I think I only really noticed it considerably now, and not on previous scans,
    because of how sharp these L series lenses are and because I'm not shooting slide and
    because this image had a lot of detail in the bricks in it.

    1/3 of the image! Wow! Dismal! I'm so appalled I can't begin to explain...

    By the way... I did take this thing apart once to clean the mirror inside. It was very dirty. I
    cleaned it up real nice with very soft paper. Put it all back together. But the clips holding
    the mirror -- it all seemed sorta haphazard. I'm pretty sure I slipped it back in place
    exactly as it was originally, but do you think there's any possibility that a tiny
    misalignment of this mirror might be rendering the left part of this frame out of focus?

    Hard to say, I guess... I'd probably at least have to do a lot of scans and confirm that it's
    really only the left portion that goes out of focus, not the right.

    K, until I report back... :)

  23. Sheesh, Les, your scans are so sharp, with only a tiny portion of the images OOF due to depth of focus.
    Is there really that much difference between the LS-4000 & LS-5000?? I really don't think is something Nikon as a company should be allowed to get away with.
    Maybe it IS my mirror having been slightly misplaced after I opened up my scanner? Does anyone know exactly what the purpose of that small mirror is, that people are told to clean every now and then?
    Here's the full size JPEG of the scan... see for yourself that some 35% of the image is soft on the left side:
    App allingDOF_fullsize.jpg
    What do you think, Les, and others? The center is tack sharp... the rest of the image, well, makes me want to puke.
  24. I've seen differing opinions on whether or not the depth of focus problem was taken care of
    in the LS-5000 & V. Some people say there was no improvement between the LS-4000 and
    LS-5000 in terms of edge to edge sharpness (due to depth of focus).

    Has anyone had experience using both, who can comment and definitively say that the LS-
    5000 has a greater depth of focus than the LS-4000?

    For that matter, anyone have a LS-5000 to sell? :)
  25. Les,
    I actually don't think that it's that mirror being misplaced, as last night I mounted a frame of Velvia in a Wess mount that, due its spokes that intercalate into the film spokes, stretches the film a bit to make it flatter. Using Mendel's method of selecting a focus point halfway between center and edge along a diagonal from center to corner, here's the result I got:
    Right-click image --> Open in New Window to view full size, or use the link below the image...
    It's pretty sharp, actually, across the frame, except for the top third of the image...
    I would now love to try this Wess mount on a LS-5000... I'm trying to get my hands on one at the University to do a comparison...
  26. Here's another idea: blend scans with different focus points:

    So, if you don't think scanning is a PITA, blend exposures to deal with hardware DR issues and now you can blend focus, too. Joy!
  27. Roger -- yeah that was my original idea in the post above, wondering if it could be
    implemented into Vuescan.

    But now you throw this software at me -- thanks! Very interesting! Have you tried it? I'm
    curious as to the algorithm used to detect which parts of the image are more in focus in
    which images... will give it a whir as soon as I get a chance!

    Thank you,

    P.S. Do you find the above WessMount.jpg acceptable in terms of across the frame
    sharpness? It's hard to see it, but there is a loss of grain structure in the top third of the
  28. I have not tried Heliconfocus but will keep it in mind for any scans where the shot I want is at the edge of a curved strip. My scanner's focus is generally quite good.

    I think the sharpness is good enough in the wessmount image you posted. I don't care if my sky isn't grain sharp. I'd say maybe only the top 1/8th or so is noticably soft. The rest is good enough. I do think it's odd that the bottom doesn't get nearly as soft as the top.

    If you're shooting with a zoom I guess give yourself a little room to crop- you'll both get rid of the parts where your lens is the least sharp but also where the scanner is unsharp.

    Here's a page on Nikon focus why may be helpful:
  29. OK, everyone.

    I am in a lab now scanning some of my slides with a Nikon LS-5000.

    I'm in utter SHOCK and DISBELIEF at how much incredible detail and film grain I'm seeing,
    edge to edge, in my slide scans now.

    I've seen this kind of focus & detail before with my LS-4000 -- IN PERHAPS 10% OF THE

    It is utterly RIDICULOUS how much more DOF the LS-5000 has in comparison to the LS-
    4000. It amazes me even more that no where on the internet can I even find this one,
    definitive statement:

    "The LS-4000 is unusable in terms of DOF when compared to the LS-5000, which scans
    consistently sharp from edge to edge if focused correctly."

    I'll be back to post more results, but, in the meantime, I'd like to know what people think
    about this...

    I mean, it's ENTIRELY unfair that I spent over $1000 on my LS-4000, albeit years ago, to
    now learn of this AFTER trying the LS-5000. IMHO, the LS-4000 DOES NOT DO WHAT IT
    WAS ADVERTISED TO DO! It can hardly keep focus on more than 10% of any given image in
    comparison now to the LS-5000! Just wait until you see my comparisons! I want my $1000
    back, or I want my LS-4000 replaced by a LS-5000! Is anyone else in the same boat?
    Anyone else want to launch a class-action lawsuit?
  30. I bet you could still sell the LS-4000 for a reasonable price and put the money towards a new LS-5000. Even if the focus of the 4000 is fairly poor, it's still better than what some others are using.

    Have you had any better luck with mounted slides vs strips?
  31. I've been a long time user. More than 10 years now I think. This is my first post in some time. This problem that we are dealing with has been a big issue for me. I've gone down all the paths you guys have. Flattening the film doesnt do it. The curvatures are so small and the depth of field/focus so thin you cannot fix it with hardware. I am convinced the solution is two scans then merged together. There were two software packages that offered this when I asked about it over a year ago. It seems these Helicon guys have updated theirs. When I tried it the first time it didnt work for me since my file sizes were big (Minolta 5400 16 bit files). Perhaps they've improved? What I'd really like to see is a Vue Scan or a SilverFast incorporate this into their scanning software.
    I've been holding off scanning until someone has solved this problem for us.
  32. "I don't want ACCEPTABLE when it's possible to get much better. I want *as sharp* as one can get from film, because only then is it on par w/ even a 10 MP digital CMOS sensor. "

    Right. And to get that you need fluid mounting - end of story. This can be done with the Nikon Coolscan LS9000ED and LS8000ED:
  33. Ellis,

    True, you make a good point. But I recently compared a Nikon LS-5000 scan with a
    Imacon Flextight scan, and the Nikon wasn't too far off from the Imacon in maintaining
    sharpness across the frame. Of course, the Imacon did a better job since it has the rolling

    Imacon's implementation is much easier then fluid mounting. Yes I really want sharp scans
    but no I'm not willing to put in the effort of fluid mounting because I haven't found any
    evidence that fluid mounting is better than Imacon's drum-like technique. Please correct
    me if I'm wrong.
  34. I tried the demo of Helicon last night. Seems to handle the large files and 16 bit now. But it will take more testing to see if the output of combining two scans(center and edge) is better than an off-center compromise scan.
  35. Nope. Helicon doesnt do the trick at high magnifications. By this I mean a full 72 megabyte
    5400 dpi scan of a b&w negative. The image doubles up. I tested it against about 6 negatives
    each with two scans(center and edge). It turns out you're better with the one off center
    compromise scan than using Helicon.
  36. Russell, could you post 100% crops to illustrate?

Share This Page