Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: syncrasy on March 15, 2012, 09:50:37 pm
-
I have Silverfast 8 scanning software for my Nikon Coolscan 5000 scanner hooked up to my Mac Pro [OS 10.6.8]. I am scanning Kodachrome slides @ 4000 ppi, 48 bit, tiff output. The colors look great but when I zoom in to 100% in Photoshop or any other image viewer, I can see very narrow, but distinct, scan lines/banding throughout the image. The lines are most noticeable in darker areas, and create jagged edges where dark and light areas meet. I have spent literally dozens of hours trying to troubleshoot this, including corresponding with Silverfast tech support in Germany. They can't reproduce the problem on their test machines and so they deny there is a problem with the software. I have had to go back to Silverfast 6.6, which produces far superior results but only when I engage multi-scanning (set to at least 8x). Silverfast 8 dropped the multi-scanning feature, so I'm not sure if that's the cause of the problem, or if there is some other problem. The software appears to be designed to create scans at only one speed: very fast, with no options to improve scan quality. I wish I could go back to Nikon Scan, but as you know, it doesn't work on Snow Leopard.
Does anyone have any ideas on how to eliminate this problem? Am I not using the software properly?
-
I have Silverfast 8 scanning software for my Nikon Coolscan 5000 scanner hooked up to my Mac Pro [OS 10.6.8]. I am scanning Kodachrome slides @ 4000 ppi, 48 bit, tiff output. The colors look great but when I zoom in to 100% in Photoshop or any other image viewer, I can see distinct scan lines/banding throughout the image. The lines are most noticeable in darker areas, and create jagged edges where dark and light areas meet. I have spent literally dozens of hours trying to troubleshoot this, including corresponding with Silverfast tech support in Germany. They can't reproduce the problem on their test machines and so they deny there is a problem with the software. I have had to go back to Silverfast 6.6, which produces far superior results but only when I engage multi-scanning (set to at least 8x). Silverfast 8 dropped the multi-scanning feature, so I'm not sure if that's the cause of the problem, or if there is some other problem. The software appears to be designed to create scans at only one speed: very fast, with no options to improve scan quality.
Does anyone have any ideas on how to eliminate this problem? Am I not using it properly?
I have EXACTLY the same set-up that you have, I have done tons of work with it, and I have not noticed this problem. I am not in a position to do so now, but come the end of the month (I know, not instant gratification) I shall be, and I would be pleased to run some tests to see whether I can reproduce the observations you've made here. My preliminary assessment based on what I've and seen before is that this should not happen.
Lasersoft Imaging eliminated multi-sampling because they introduced multi-exposure, as their testing indicated to them that multi-exposure achieves everything that multi-sampling did, plus the specific intent of multi-exposure which is to open up shadow detail in slides and highlight detail in negatives. I have no doubt that if they thought there would be value-added to preserving multi-sampling as a separate tool they would have done so.
Scanning speed depends on the scanner and the resolution you set in the software. The higher the resolution the slower the scan, so in your case, it would be scanning at the slowest speed. I don't think that's the problem.
-
Thanks for the reply, Mark. I appreciate your offer to run a test. And no worries about the delay -- I'm using Silverfast 6.6 for now. Let me know if I can provide you with anything for your test. Would it be useful if I posted some samples here? Or should I wait until the end of the month?
Lasersoft Imaging eliminated multi-sampling because they introduced multi-exposure, as their testing indicated to them that multi-exposure achieves everything that multi-sampling did, plus the specific intent of multi-exposure which is to open up shadow detail in slides and highlight detail in negatives. I have no doubt that if they thought there would be value-added to preserving multi-sampling as a separate tool they would have done so.
I wish I had as much confidence in Lasersoft's tech team as you do. Upon launch of Silverfast 8, the tech support crew didn't believe me when I told them the software didn't work with the Nikon slide feeder. (Some time later they quietly released a bug fix.) And they tightly control their support forum so as to not allow any posts that might imply there is trouble with the software. I can't get even the most basic questions past the forum gate keeper. Maybe the developers are more on the ball than tech support, but the Silverfast 8 UI, while generally better than SF 6.6, still has some not-so-friendly elements (from a usability perspective).
Regarding multi-exposure. . . Yes, they also told me multi-exposure replaced multi-sampling. I tried it, but I don't see any significant improvement (perhaps 10% at best). The lines persist. I thought the problem was due to the age of the slides (1930s and 40s), but then I scanned some 1970s slides in great condition and still noticed the lines. A casual observer might not notice them, but once I saw them, they were hard to ignore.
Regarding speed/quality: My 4000 ppi scans take only about 10 seconds. Does that sound normal?
Thanks,
-- (another) Mark
-
Hi Mark,
It would be best to receive your images when I am in front of a decent computer screen allowing me to evaluate them properly. When the time comes, I'll send you a PM with arrangements for uploading the full raw scan files so I can see the issues in full detail.
While the Nikon 5000 is a fast scanner, ten seconds for a 4000 PPI scan does seem a tad fast, but I have never really timed it, so that would also be something to verify when I have access again to my scanner.
Cheers,
Mark
-
Thanks, Mark. I'll look for your PM in a couple weeks.
I should have mentioned that I also tried the VueScan demo and had similar problems (scan lines/banding).
-
Oops - that's a very important observation. If two different applications produce the same results it points to the issue being lodged somewhere outside the software. But we'll get into that anon.
-
hi Mark (syncrasy) --
I have seen this happen with black/white film.. makes sense that kodachrome could exhibit it. I think it might only apply to the 8000/9000 but I recall an option to scan with a single line from the CCD on the Nikon software.. it cleared it up as I remember. It was a rare anomaly in my experience. If Silverfast has that option you could try it there...
As to the Nikon software not working on snow leopard, have you tried? I'm not advocating you switch, just curious.
The time you mention I would guess is way too fast, I don't think you are getting true 4000ppi scans in 10 seconds or less.
Last thoughts. Have you tried making a scan with the controls zeroed? Wondering if potential contrast adjust exaggerating edges.
And finally, have you checked the tiffs in another program or other display? I've seen screen previews and redraws do wacky stuff..
I know I'm reaching, sorry to hear of your troubles and good look sir.
all best -- sean
-
You could try Super Fine Scan - it takes more time, but I heared, that banding problems which ocurred with the LS8000 were fixed by some people using SFS mode.
I never had banding problems with my LS9000 so far.
Concerning Silverfast - The new version is a mixed bag.
What is really good, is, that now the IR extraction for the Nikon scanners (finally!) works on windows, but they omitted the analog lamp control, which is bad.
There are also some bugs with the MF film holder for the LS9000 - I'm sure they'll fix it soonish.
-
Sean, yes, when I first got my new Mac Pro last year, I tried Nikon Scan immediately because I had read that it wouldn't work on Leopard or Snow Leopard and I wanted to see for myself whether it was true. The software would not launch. I tried for about 30 minutes (fixing permissions, restarting the Mac, relaunching, etc.). It still would not launch. Officially, Nikon says it should not work and is not supported. So I purchased SilverFast.
Today, when I read your post and Christoph's post, I was going to reply, "tried it and it didn't work" and "please see my post above". But I thought, "let me try it again, just for kicks." I tried launching Nikon Scan and, to my disbelief, it launched successfully. Then I scanned two test scans (1x and Super-Fine 8x). They worked! Nikon Scan 4 now works on my Snow Leopard Mac! I don't understand why it now works, but this is very good and unexpected news. The only problem is that Nikon Scan 4 never played well with my slide feeder (it always crashed during batch scans). SilverFast generally plays well with the slide feeder. I'll test Nikon Scan again to see if (somehow) that feature works better.
As for Silverfast 8, my time estimate of "10 seconds" was incorrect. It was a subjective and emotional impression, but I had failed to actually time it. So here are some scan timings and results (all at 4,000 ppi):
- Nikon Scan 4 (16 bit, multiscan Super-Fine 8x): 2 minutes, 15 seconds
- SilverFast 6.6 (48 bit, multiscan 1x): 30 seconds
- SilverFast 6.6 (48 bit, multiscan 8x): 2 minutes, 45 seconds
- SilverFast 8 (48 bit, multi-exposure off): 1 minute
- SilverFast 8 (48 bit, multi-exposure on): 2 minutes
All other settings are "zeroed out" (I always scan "raw" and process in Photoshop later). With multi-scanning, both NikonScan and SilverFast 6.6 produced acceptable images with no apparent scan lines. But SilverFast 8 creates scan lines. The problem is not a matter of previews or redraws. The scan lines are in the image (confirmed by opening in Photoshop and viewing at 100%).
See the attached JPEG. In the SilverFast 8 image on the right, you can see tiny scan lines along the edge of the roof and the roof beam. They make the roof line look jagged.
-
I hope you labelled the images correctly, because the Nikon Scan image looks the worst of the lot - the sky looks like coarse sandpaper, not the case for the two to the right. As for the two to the right, it made no difference with versus without multi-exposure - which can indeed happen when the deep quarter-tones are just too dense. I wonder about the colour balance of the SilverFast scans. Were you using a scanner profile?
-
You could try Super Fine Scan - it takes more time, but I heared, that banding problems which ocurred with the LS8000 were fixed by some people using SFS mode.
I never had banding problems with my LS9000 so far.
Concerning Silverfast - The new version is a mixed bag.
What is really good, is, that now the IR extraction for the Nikon scanners (finally!) works on windows, but they omitted the analog lamp control, which is bad.
There are also some bugs with the MF film holder for the LS9000 - I'm sure they'll fix it soonish.
Christoph - they didn't "omit" lamp control - but it isn't present yet. :-)
-
I hope you labelled the images correctly, because the Nikon Scan image looks the worst of the lot - the sky looks like coarse sandpaper, not the case for the two to the right.
Yes, I agree. Is that because Nikon Scan can only scan up to 16 bit (versus 48 bit for SilverFast)? Presumably Nikon would recommend fixing this using their grain smoothing tool during scan (e.g., GEM, or one of those acronyms that I can never remember).
As for the two to the right, it made no difference with versus without multi-exposure - which can indeed happen when the deep quarter-tones are just too dense. I wonder about the colour balance of the SilverFast scans. Were you using a scanner profile?
Sorry, but are you referring to sky courseness/fineness (which is the same in the two SilverFast scans)? Or are you referring to my claim of jaggedy scan lines (which are apparent in the right image, but not in the center image)? I'm not so worried about the color balance as I can fix that in Photoshop. And I'm assuming that color balance is a separate issue from the scan line issue. Are you suggesting the two issues might be related?
Regarding profiles, the initial SF8 scan had no profile (I guess it got deselected after a recent software update), but then I re-selected the SF Kodachrome for 5000 profile and got the same results (scan lines).
Attached are two color management preferences: (1) after reselecting the Kodachrome profile and (2) after resetting the SilverFast 8 software (which deletes all preferences). Jagged scan lines persist.
-
…Nikon Scan can only scan up to 16 bit (versus 48 bit for SilverFast)?
Mark
16 bit and 48 bit are the same! Nikon are quoting the per channel bit depth 16 red bits + 16 green bits + 16 blue bits = 48 total bits, whereas SilverFast are simply quoting the total bits. (The same occurs with 8 bits per channel versus 24 bits total). Note that some scanner files may include an extra channel for an infra-red channel that is used in some scanners for dust reduction giving either 32, 56 or 64 total bits per channel (note: 56 bits is a 16 bits per channel colour image with an 8 bit IR channel).
Regards
Nigel
Modified to correct typo.
-
Thank you for the clarification, Nigel.
-
Mark
16 bit and 48 bit are the same! Nikon are quoting the per channel bit depth 16 red bits + 16 green bits + 16 blue bits = 48 total bits, whereas SilverFast are simply quoting the total bits. (The same occurs with 8 bits per channel versus 24 bits total). Note that some scanner files may include an extra channel for an infra-red channel that is used in some scanners for dust reduction giving either 32, 56 or 6 total bits per channel (note: 56 bits is a 16 bits per channel colour image with an 8 bit IR channel).
Regards
Nigel
Nigel,
Yes it occurred to me too that it is just another numeraire for the same file format. In the case of SilverFast, if you select a 16 bit format preserving the infrared channel (such as HDRi it becomes a "64-bit scan" - 16 added for the IR channel.
Mark Segal
-
Christoph - they didn't "omit" lamp control - but it isn't present yet. :-)
Well - in version 6.6 it was there .... whatever it means.
I still like Silverfast, but they have to do some homework.
-
Yes, I agree. Is that because Nikon Scan can only scan up to 16 bit (versus 48 bit for SilverFast)? Presumably Nikon would recommend fixing this using their grain smoothing tool during scan (e.g., GEM, or one of those acronyms that I can never remember).
Sorry, but are you referring to sky courseness/fineness (which is the same in the two SilverFast scans)? Or are you referring to my claim of jaggedy scan lines (which are apparent in the right image, but not in the center image)? I'm not so worried about the color balance as I can fix that in Photoshop. And I'm assuming that color balance is a separate issue from the scan line issue. Are you suggesting the two issues might be related?
Regarding profiles, the initial SF8 scan had no profile (I guess it got deselected after a recent software update), but then I re-selected the SF Kodachrome for 5000 profile and got the same results (scan lines).
Attached are two color management preferences: (1) after reselecting the Kodachrome profile and (2) after resetting the SilverFast 8 software (which deletes all preferences). Jagged scan lines persist.
I regret that I am working from a MacBook Pro laptop for the time being - perhaps that combined with the JPEG low resolution makes it impossible for me to see the problem of jagged scan lines. I don't see a problem with the sky in the two SF scans, but the sky in the NikonScan version is unacceptable. While you can fix colour balance in Photoshop, it is far easier to do this with SilverFast's neutral pipette tool, or with Lightroom's white balance eyedropper tool, provided there is the slightest bit of data in the scene that should be neutral.
The lower profile set-up looks fine to me - except that you may wish to consider using ProPhoto as the internal colour space, because the most recent crop of Epson printers exceed ARGB(98) in some areas of the spectrum.
None of this however addresses the issue of the jagged scan lines you are seeing, but unfortunately in light of my current set-up I cannot. As I said, beginning of next month I shall be able to do some testing for this condition.
-
Well - in version 6.6 it was there .... whatever it means.
I still like Silverfast, but they have to do some homework.
What it means is that in those particular scanner models (and few others), you could control the exposure of each channel through software controls over the actual amount of light hitting the CCD in any channel. One possible advantage of this is that if you notice most of the scanner noise embedded say in the blue channel, you could increase blue channel exposure to increase the signal to noise ratio in that channel. One can also do some colour balancing with that tool, but I don't recommend it over other purely software based remapping approaches that are easier to manage.
My understanding is that the "homework" is being done.
-
I regret that I am working from a MacBook Pro laptop for the time being - perhaps that combined with the JPEG low resolution makes it impossible for me to see the problem of jagged scan lines.
Mark, just to prove that I'm not hallucinating, I've attached another comparison (SF6.6 vs SF8), blown up to 300%, so you can see the difference on your laptop. The scan lines, while exaggerated, are obvious on the right (SF8) image. The SF6.6 is holding up well even at 300%.
-
Mark, just to prove that I'm not hallucinating, I've attached another comparison (SF6.6 vs SF8), blown up to 300%, so you can see the difference on your laptop. The scan lines, while exaggerated, are obvious on the right (SF8) image. The SF6.6 is holding up well even at 300%.
Hi Mark - I didn't think you were hallucinating - I ascribed all the blame to my temporary viewing conditions. Thanks for the blow-ups - by comparing the two scans it's now clear what the phenomenon looks like. I am intrigued that the left image name suggests no multi-exposure and the right image name suggests with multi-exposure. If this is the only difference between the two scans, it does raise a question about whether there could be a slight registration problem between the two scan passes that multi-exposure creates, or an inordinate exaggeration of the dark pixels relative to the lighter edges - I'm purely drawing inferences or hypotheses here, and even if either were true, not clear whether it is software or hardware-related, but perhaps interesting avenues of further investigation - provided the image names correctly identify the differences of scan treatment.
-
Hi Mark - I didn't think you were hallucinating - I ascribed all the blame to my temporary viewing conditions. Thanks for the blow-ups - by comparing the two scans it's now clear what the phenomenon looks like. I am intrigued that the left image name suggests no multi-exposure and the right image name suggests with multi-exposure. If this is the only difference between the two scans, it does raise a question about whether there could be a slight registration problem between the two scan passes that multi-exposure creates, or an inordinate exaggeration of the dark pixels relative to the lighter edges - I'm purely drawing inferences or hypotheses here, and even if either were true, not clear whether it is software or hardware-related, but perhaps interesting avenues of further investigation - provided the image names correctly identify the differences of scan treatment.
Mark, whoa, hold on.... The scans were made with different versions of SilverFast. You can see the '6' and '8' in the file names. (Sorry for the cryptic names; I probably should have used longer file names.)
- left image: SilverFast 6.6 with 8x multisampling (there is no multi-exposure option in SF 6, as far as I know)
- right image: Silverfast 8 with multi-exposure on (since there is no multi-sampling option in SF 8 )
As I said earlier, SF8 with multi-exposure (M-E) on has negligible effect on the jaggedy lines, but I am using M-E to provide the best possible scan. So this is a SF 6.6 vs SF 8 conundrum. SF 6.6 appears to work; SF 8 does not. The Nikon part of this discussion might be irrelevant, but I included it for comparison (now that it appears to work on my Snow Leopard machine).
(And I knew you knew I wasn't hallucinating, but sometimes I'm not sure myself. Besides, I wanted you to see the phenomenon.)
-
Mark, whoa, hold on.... The scans were made with different versions of SilverFast. You can see the '6' and '8' in the file names. (Sorry for the cryptic names; I probably should have used longer file names.)
- left image: SilverFast 6.6 with 8x multisampling (there is no multi-exposure option in SF 6, as far as I know)
- right image: Silverfast 8 with multi-exposure on (since there is no multi-sampling option in SF 8 )
As I said earlier, SF8 with multi-exposure (M-E) on has negligible effect on the jaggedy lines, but I am using M-E to provide the best possible scan. So this is a SF 6.6 vs SF 8 conundrum. SF 6.6 appears to work; SF 8 does not. The Nikon part of this discussion might be irrelevant, but I included it for comparison (now that it appears to work on my Snow Leopard machine).
(And I knew you knew I wasn't hallucinating, but sometimes I'm not sure myself. Besides, I wanted you to see the phenomenon.)
Oops - missed that! OK - problem seems not to be mutii-exposure related. Back to the drawing board.
-
I hope you labelled the images correctly, because the Nikon Scan image looks the worst of the lot - the sky looks like coarse sandpaper, not the case for the two to the right.
On second thought (to hijack my own thread briefly), check out these quickly corrected images: Nikon Scan vs SilverFast 6.6. Ignoring the color mismatch, to my eye the "sandpaper" coarseness is about the same for both scans.
OK, that's all for now. I'll wait for our scheduled discussion of SF 6.6 vs SF 8 at the end of the month.
-
I agree.
On another matter, somewhere above you raised the issue about the monitoring and response on the SilverFast Forum. I checked on this, and the first item in the Forum rules says:
<This forum is for generic talking and not for supporting your special problems!. Please use our support assistant to get special help for your problem.> They want people to go to tech support for tech support issues and use the forum for everything else, and it would appear from what you are saying that they do monitor for this quite closely!
Cheers,
Mark
-
On another matter, somewhere above you raised the issue about the monitoring and response on the SilverFast Forum. I checked on this, and the first item in the Forum rules says:
<This forum is for generic talking and not for supporting your special problems!. Please use our support assistant to get special help for your problem.> They want people to go to tech support for tech support issues and use the forum for everything else, and it would appear from what you are saying that they do monitor for this quite closely!
Oh, yes. The notorious "generic talking" rule. I know it well. But I could never figure out what they meant by "generic" because the forum is full of very specific technical questions. (The forum even has topics such as "Installation Problems," "Host Related Problems," "OSX Problems," "All Other Problems," etc.) They also have these specific rules:
Before you post.
* Use our search function in this forum to find similar topics.
* Use our FAQ to find solutions for your problems.
* Use our support assistant to get help if nothing of the above points help you.
That would seem to imply that someone else with "your (special) problem" was successful in getting their problem posted—an event that should be prohibited by the "generic talking . . . not . . . your special problems!" mission of the forum.
It would be comical if it weren't so frustrating. We're talking Alice in Wonderland territory here.
In my case, the support assistant couldn't answer the problem, and further correspondence with their team came up empty. One might think they would let forum members post difficult cases to the wider community (i.e., leveraging the value of a true forum, like Microsoft and Phase One did/do with Expression Media/Media Pro). Instead I got shut down and my question never was answered. Very peculiar company, LaserSoft.
-
OK, I understand how it can appear rather confusing!
Back to the main business.
Cheers,
Mark
-
I have EXACTLY the same set-up that you have, I have done tons of work with it, and I have not noticed this problem. I am not in a position to do so now, but come the end of the month (I know, not instant gratification) I shall be, and I would be pleased to run some tests to see whether I can reproduce the observations you've made here. My preliminary assessment based on what I've and seen before is that this should not happen.
........
Hi Mark,
I am now back at home, and as promised I ran several tests to try to reproduce your observations and I could not. I used a Kodachrome slide with both light and dark areas (contiguous) and did one scan unadjusted with no multi-exposure, another scan unadjusted with multi-exposure, and another scan adjusted. All 48-bit tiff 4000 PPI, 100% magnification ratio, using LSI's canned profile for Kodachrome. Opened in Photoshop and magnified to 100%, none of them displayed the condition you explained and demonstrated, either in Normal mode, or in Screen mode (which opens darks and exposes defects); hence as I cannot reproduce it I cannot confirm a systemic issue either with the application or with the scanner and unfortunately cannot offer a viable explanation of what may be causing the problem you are having. Sorry I could not be more helpful with a solution.
Cheers,
Mark
-
Mark,
Thank you for your time and effort. The mystery continues.
I can only assume there is a problem on my end. The first obvious variable is the actual slides. Perhaps my slides are unusually difficult to scan (too old? too dense? too dark?). The next most obvious variable is my scanning method, but I think we've established we are using the same settings and profile, and the folks in Germany looked at my log file and didn't find any issues. So if it's not the slides or my scanning method, then there is a hardware or software issue. The first test I can try is to install SF8 on my laptop and run a test with the same slide. I'll report back when I have results. If you have any other suggestions, let me know.
(BTW... I'm not familiar with "Normal Mode" vs "Screen Mode." My Photoshop CS2 has a View > Screen Mode > Standard Screen Mode or Full Screen Mode with Menu Bar or Full Screen Mode, but I don't know what these views do or if changing them would make a difference in troubleshooting my issue.)
Is there any reason to suspect a 32-bit vs 64-bit mode issue? I assume my SF8 is running in 64-bit mode since the "Open in 32-bit mode" checkbox is unchecked. I know that I had to change Nikon Capture NX2 setting from 64- to 32-bit mode to get it to work properly. A shot in the dark.
Thanks,
(the other) Mark
-
Mark,
Thank you for your time and effort. The mystery continues.
I can only assume there is a problem on my end. The first obvious variable is the actual slides. Perhaps my slides are unusually difficult to scan (too old? too dense? too dark?). The next most obvious variable is my scanning method, but I think we've established we are using the same settings and profile, and the folks in Germany looked at my log file and didn't find any issues. So if it's not the slides or my scanning method, then there is a hardware or software issue. The first test I can try is to install SF8 on my laptop and run a test with the same slide. I'll report back when I have results. If you have any other suggestions, let me know.
(BTW... I'm not familiar with "Normal Mode" vs "Screen Mode." My Photoshop CS2 has a View > Screen Mode > Standard Screen Mode or Full Screen Mode with Menu Bar or Full Screen Mode, but I don't know what these views do or if changing them would make a difference in troubleshooting my issue.)
Thanks,
(the other) Mark
Mark - I was referring to the Blend Modes in Photoshop (and CS2 has them), when you add a Curves Adjustment Layer, you can select from a large number of Blend Modes. After adding the Curves Adjustment Later, at the top of the Layers Panel, you will see a pane with the word "Normal". That's the default Blend Mode. Click on it and a bunch of options open up. Select Screen and most of the image turns much lighter, allowing you to see through otherwise dense material. It helps to reveal hidden defects.
From your first post, I suspect it is something to do with the slides. If you got somewhat decent results using high multisampling in SilverFast 6.6, this indicates to me that perhaps a lot of nasty stuff is being "averaged out" by the multisampling. I suspect Multi-Exposure in SF8 is not as strong as the strongest multi-sampling was in SF 6.6. To test whether it may be the slides, try slides from a different camera or a different era - whatever you can get your hands on and see whether the problem reproduces itself with media from a different vintage/camera/era.
-
This may be a long shot, but multi-sampling would give time for any mechanical vibrations to settle out and average the result. Multi-exposure might resemble a multi-sample of 2 from a time perspective- you would have to talk with the engineers at Lasersoft. i forget if they make two complete passes or hold position and expose twice. From what i remember, multi-sampling holds a set position and scans x times. So ccd registration offset could still be the issue, and i think more likely. There were some guides on the net on how to open up the scanner and look for dust/dirt issues, but it would be better if Nikon ( still ? ) serviced the unit. At one point, SF replaced the Nikon Maid drivers (it was in the 6.x series if i remember correctly) on the Mac with their own. Otherwise, the SF software would be using the same drivers as Nikon. Ed reverse engineered the drivers for the Nikon and did not use the Maid modules for Vuescan.
Frank
-
To test whether it may be the slides, try slides from a different camera or a different era - whatever you can get your hands on and see whether the problem reproduces itself with media from a different vintage/camera/era.
Attached are samples from a 1982 Kodachrome slide, the most recent year in my collection, shot with a different camera. (My previous sample was 1975 Kodachrome, and the problem occurs in my 1930s and 40s slides, too.)
I've included full, 100x, and 300x screenshots. SF6.6 on the left, SF 8 on the right. As you can see, the scan line problem is present in this more recent sample.
-
Until one gets to a 300% screen magnification any difference is not detectable on my high res NEC PA271 display. At 300%, the edge of one of the grass blades looks slightly more pixellated in the SF8 version versus the SF6 version. Have you printed both to see whether there is a practical difference on paper t the resolution and linear dimensions you would normally require?
-
Until one gets to a 300% screen magnification any difference is not detectable on my high res NEC PA271 display. At 300%, the edge of one of the grass blades looks slightly more pixellated in the SF8 version versus the SF6 version.
Odd. I can see them in the 100% version if I look closely (my monitor is NEC MultiSync LCD 1960NXi, if that's of importance). I've attached a marked-up version of the 100% shot to draw your attention to the many jagged areas that I can see.
Have you printed both to see whether there is a practical difference on paper t the resolution and linear dimensions you would normally require?
Not yet. I just assumed the differences I am seeing would translate to paper.
-
Mark, this is really very interesting. I downloaded your image and opened it in Photoshop, attached a Curves layer in Screen mode so I could really see what's going on, and at 100% differences are hardly detectable - my eyes and my display - both very good dare I see, but perhaps not as good as yours' - that we'll never likely know! :-). Then I magnified the image to 200%, and indeed, at that magnification, the degree of edge pixellation is greater in the SF8 image than in the SF6 image. Of course anything magnified beyond 200% (which is one screen pixel per one image pixel) is bound to show some pixellation, so all that matters here is the relative appearance between the two application versions. As to what's causing it, I can't say. BUT, what you see at these kind of magnifications is NOT necessarily what you will see in a print - in fact hardly ever. I recommend that you print both versions at your standard print settings and just examine the photos at normal viewing distance, and let us know what you find.
-
This may be a long shot, but multi-sampling would give time for any mechanical vibrations to settle out and average the result. Multi-exposure might resemble a multi-sample of 2 from a time perspective- you would have to talk with the engineers at Lasersoft. i forget if they make two complete passes or hold position and expose twice. From what i remember, multi-sampling holds a set position and scans x times. So ccd registration offset could still be the issue, and i think more likely. There were some guides on the net on how to open up the scanner and look for dust/dirt issues, but it would be better if Nikon ( still ? ) serviced the unit. At one point, SF replaced the Nikon Maid drivers (it was in the 6.x series if i remember correctly) on the Mac with their own. Otherwise, the SF software would be using the same drivers as Nikon. Ed reverse engineered the drivers for the Nikon and did not use the Maid modules for Vuescan.
Frank
Thanks for the input, Frank. I'm inclined to think that the removal of multi-scanning from SF 8 is contributing to this issue. I don't know enough about drivers to comment, but perhaps the driver replacement is also an issue.
I should also revisit the VueScan question, since I have noticed a similar problem of jagged scan lines in my demo version of VueScan. Attached is a comparison of SF 6.6 vs SF 8 vs VueScan (in fine mode). I'm including both 100% and 300% views. Curiously, VueScan appears to produce fewer jagged edges/scan lines than SF 8, but it produces some new jagged scan lines on different parts of the image (most notably the blade of grass angled upward left to right).
-
Mark, this is really very interesting. I downloaded your image and opened it in Photoshop, attached a Curves layer in Screen mode so I could really see what's going on, and at 100% differences are hardly detectable - my eyes and my display - both very good dare I see, but perhaps not as good as yours' - that we'll never likely know! :-).
OK, I tried to do the same thing (open the 100% image, add a Curves layer in Screen mode). I see the same jagged/toothy edges, only now they're more pronounced because the Curves layer brightens the image. I can't comment on your eyes, but my eyes see the jaggies. And they're obvious.
Then I magnified the image to 200%, and indeed, at that magnification, the degree of edge pixellation is greater in the SF8 image than in the SF6 image. Of course anything magnified beyond 200% (which is one screen pixel per one image pixel) is bound to show some pixellation, so all that matters here is the relative appearance between the two application versions.
Agreed. I included the 300% shots to show the relative difference (and to prove that there is a difference, even though the effect is exaggerated).
As to what's causing it, I can't say. BUT, what you see at these kind of magnifications is NOT necessarily what you will see in a print - in fact hardly ever. I recommend that you print both versions at your standard print settings and just examine the photos at normal viewing distance, and let us know what you find.
Done. And I've attached two scans of the resulting print (300 ppi and 800 ppi).
My conclusions: The effect is less noticeable in a print than onscreen. In fact, the average viewer who knew nothing of this experiment probably would not notice any differences. But, to my eyes, the scan lines in the original SF 8 scan create a toothier/jagged roof line in the print. In the scans of the print (especially the 800 ppi scan), you can see that, where light and dark areas meet, small light and dark spikes intrude into each others' space, just as in the direct Kodachrome scan I posted earlier.
Ultimately, I still think SF 8 is not producing quality scans. (Or to be more accurate, it's producing flawed scans compared to SF 6.6; other than my jagged line problem, the quality is great.) I recently asked Lasersoft to help troubleshoot the problem again. They still couldn't reproduce the problem. But interestingly, in his most recent email, the support manager told me, "One of our developers is currently working on improvements for the Nikon scanners. He will look out for anything possibly causing this effect." I won't be holding my breath, but perhaps they will figure this out. But, to be honest, until they restore multi-scanning to SF 8, I seriously doubt this will be fixed.
-
One should not need multi-sampling to get smooth edges from such subject matter. The problem is elsewhere, so let us hope they can determine what's going on. Good to see that in print any differences are indeed not noticeable unless one is preconditioned to look real hard for them.
-
One should not need multi-sampling to get smooth edges from such subject matter. The problem is elsewhere, so let us hope they can determine what's going on. Good to see that in print any differences are indeed not noticeable unless one is preconditioned to look real hard for them.
I admit I'm not an expert on scanning technologies, so maybe I'm misunderstanding something, but according to the Nikon Scan reference manual (see attached screenshot), one of the benefits of multi-sample scanning is "smoother changes in tone" (in addition to noise reduction). And my tests show that multi-sampling clearly makes a difference. When I increase the multi-sampling levels from 1x to 4x to 8x in Nikon Scan and SF 6.6, I do see corresponding improvement in edge smoothness. The fact that multi-sampling is the one feature that Lasersoft specifically dropped from SF 8, which creates jagged edges in my tests, can only lead me to conclude that its absence is a factor. And VueScan, which has only one setting for Fine Mode (on/off), exhibits similar issues.
As for the print looking acceptable. . . I suppose that's good news, but I'm more likely to use my images for the Web than for print, and I often need to highly crop an image, which would make the jagged edges more apparent. So neither SF 8 nor VueScan is usable for my purposes.
-
No, I wasn't suggesting that multi-sampling is not useful. It can be. What I was trying to say is that one shouldn't need to depend on multi-sampling in order to avoid the problem you came across. My experience with scanning oodles of slides and negatives over the past 12 years indicates that I never needed a multi-sample setting greater than 2x to achieve all the tonal smoothness and edge clarity the equipment and processes could deliver. So what I'm suggesting is that if you need such high levels of multi-sampling to avoid these problems, the issue is elsewhere and not because SF8 and Vuescan don't provide such high levels of multi-sampling.
-
No, I wasn't suggesting that multi-sampling is not useful. It can be. What I was trying to say is that one shouldn't need to depend on multi-sampling in order to avoid the problem you came across. My experience with scanning oodles of slides and negatives over the past 12 years indicates that I never needed a multi-sample setting greater than 2x to achieve all the tonal smoothness and edge clarity the equipment and processes could deliver. So what I'm suggesting is that if you need such high levels of multi-sampling to avoid these problems, the issue is elsewhere and not because SF8 and Vuescan don't provide such high levels of multi-sampling.
Ah, OK. I think we are in agreement. But then should I be looking at driver issues or (gasp) hardware issues such as CCD registration, as Frank suggests above? I'm not sure Nikon will be able to advise me, considering I haven't noticed any differences using their software. It's only with SF 8 that I noticed the problem.
-
You've noticed it with both Vuescan and SF8. Recall, sadly Nikon stopped producing these scanners some time ago. You could try contacting them - it would be interesting to see what kind of support they are still offering. I wouldn't rule out the possibility of some kind of minor hardware issue that your former use of SF 6.6 with a heavy overlay of multisampling compensated.
-
You've noticed it with both Vuescan and SF8. Recall, sadly Nikon stopped producing these scanners some time ago. You could try contacting them - it would be interesting to see what kind of support they are still offering. I wouldn't rule out the possibility of some kind of minor hardware issue that your former use of SF 6.6 with a heavy overlay of multisampling compensated.
Nikon still maintains basic troubleshooting documents for scanner hardware and software on its Support website. But as I suspected, they only go so far, and they can't troubleshoot issues that appear isolated to SilverFast or VueScan. My scanner passes all the nominal hardware tests (basically limited to making sure the green light blinks as expected when powering up with no other devices attached and the USB cord connected to the rear USB jack). My scan results from Nikon Scan are the same now as they were six years ago.
I opened a ticket with Nikon tech support and, after I told them my scanner was working properly with Nikon Scan, this was their final response:
"As long as your scanner is working properly with Nikon Scan, that would indicate that your scanner is working properly. While we recommend that you try third party software, unfortunately, we cannot guarantee that they will work properly or provide support for those programs."
As I think about this more, I guess I'm not convinced that my use of multi-sample scanning would be considered "overcompensation" for a hardware problem as you suggest. It would seem that multi-sample scanning is simply "compensating" for difficult circumstances. In other words, it's working as intended. I understand that in your experience you never needed more than 2x multi-sample scanning to produce acceptable results. But as this thread has suggested, you and I have different reactions to the same image. Recall that you didn't see jagged edges on my 100% images where the jagged edges were obvious to me. We have different monitors, different eyes, and perhaps different thresholds of acceptability.
-
Nikon still maintains basic troubleshooting documents for scanner hardware and software on its Support website. But as I suspected, they only go so far, and they can't troubleshoot issues that appear isolated to SilverFast or VueScan. My scanner passes all the nominal hardware tests (basically limited to making sure the green light blinks as expected when powering up with no other devices attached and the USB cord connected to the rear USB jack). My scan results from Nikon Scan are the same now as they were six years ago.
I opened a ticket with Nikon tech support and, after I told them my scanner was working properly with Nikon Scan, this was their final response:
As I think about this more, I guess I'm not convinced that my use of multi-sample scanning would be considered "overcompensation" for a hardware problem as you suggest. It would seem that multi-sample scanning is simply "compensating" for difficult circumstances. In other words, it's working as intended. I understand that in your experience you never needed more than 2x multi-sample scanning to produce acceptable results. But as this thread has suggested, you and I have different reactions to the same image. Recall that you didn't see jagged edges on my 100% images where the jagged edges were obvious to me. We have different monitors, different eyes, and perhaps different thresholds of acceptability.
Just to be accurate about this, at 300% we are seeing the same issues. At 100% we are not. I'm using an NEC PA271 at resolution of 2560*1440. with L calibrated at 100. Recall I examined your images in Photoshop in Screen Blend Mode to make sure I could see detail that may be otherwise smothered in darkness. When I wear my computer glasses my vision is very well corrected and I've been examining image detail in photographs over a time span of 50+ years, so I am quite confident that I know whato to look for and whether I am seeing properly. That doesn't mean I can't miss a trick or two, or that your problem isn't real - it is.
It is good that Nikon is willing to examine your scanner, but I understand your reluctance to send it in, because it really isn't clear yet whether the root cause of this problem is hardware or software. Even with no multisampling and no multiexposure you should not be getting those artifacts, so let us sweep that issue aside. I am inclined to think there could be a hardware issue simply because two different scanning applications are producing similar effects with the same unit. But if you are not experiencing this problem with Nikon Scan, that could mean there is something in Nikon's driver that controls the scanner in a manner that neither SilverFast nor Vuescan got right. So that would point back to software - see what I mean? Back to the drawing board.
I'm working on a good candidate image just now - has the kind of contrast that would show up such problems. The left side of the bottom image is the whole frame showing the small white rectangle that becomes the content at 300% display magnification, shown on the right side and larher in the upper image. The bright vertical material, being part of a window ledge and a drain pipe on the building wall do show vertical stripes. Not clear, however, whether these are display pixelation artifacts, or image artifacts. To assess that, one needs to print the image, so the display magnification is no longer in contention. I printed the 4000 PPI scan from LR4 on a US Letter size sheet in my Epson 4900 on gloss paper, and examined those sections under a 7x aspherical loupe. No artifacts showed.
Pushing this further, I downsized (not downsampled) the image to 180 PPI so I could make a letter size print of the area in the yellow box, then examined those same vertical brights using the 7x aspherical loupe, and guess what - those lines are there loud and clear as I see them on display; hence the issue not due to display pixelation. So correct me if I'm wrong, but I think I've now replicated your problem; to see it however, requires extreme kind of magnifications and close-up inspection that would seldom figure in most photographic work most people do, indicating that your requirements are quite exceptional, but of course no less real or legitimate. I think I've taken this as far as I can - i.e. I don't know the causes or how abnormal.
-
So correct me if I'm wrong, but I think I've now replicated your problem; to see it however, requires extreme kind of magnifications and close-up inspection that would seldom figure in most photographic work most people do, indicating that your requirements are quite exceptional, but of course no less real or legitimate.
I interpret your samples as depicting normal pixelation at 300% enlargement, not the artifacts that I witness in my scans, so I would suggest this test has not replicated my problem.
I think I've taken this as far as I can - i.e. I don't know the causes or how abnormal.
Well, I appreciate your sticking with me this far, and all of your time and effort to try to help me figure it out. I'm done too.
I will post just one more image to illustrate my SF 8 artifacts in normal viewing (100%), and to put a cap on the thread (obviously, leaving the mystery still unsolved).
-
Mark, true, what we've each observed is somewhat different. That said, the reason I did the print tests was to isolate whether or not what I found was due to display pixelation and I demonstrated that it is not. Those lines exist in a print of a non-interpolated image simply resized within normal bounds, and they become visible when seen with a loupe on the print. Nothing to do with display pixelation from 300% magnification. The lines are very feint but they are in the image and the Epson 4900 can reproduce them (I tell you, this printer is UNCANNY). I attach minimal practical significance to my findings for the reasons I mentioned. Soooo, onward and upward..........
:-)
-
One way to try to replicate what SF8 is doing is to make two separate scans of the same slide and create a differences layer in PS to see if there is a pixel registration error.
You could also make two different exposure scans and combine them in PS to get the multi-exposure effect
If you want to track it down to the driver software, see if you can use a windows machine and install the demo of SF6.6 on it. The windows version uses the Nikon Maid driver (which is the same as NS) whereas the MAC version uses the re-written driver.
If this is color moire being created ( the grass examples reminded me of it, BTW reply #34 triple it is visible to me on all three - the color balance is hiding some on the left 2 i think), try scanning at different resolutions -and/ or at non-divisible by two of 4000 ppi to force interpolation which will act as a filter.
Bart Van der Wolf on these forums (BartvanderWolf) is expert in this area.
Frank
-
In my case, the support assistant couldn't answer the problem, and further correspondence with their team came up empty. One might think they would let forum members post difficult cases to the wider community (i.e., leveraging the value of a true forum, like Microsoft and Phase One did/do with Expression Media/Media Pro). Instead I got shut down and my question never was answered. Very peculiar company, LaserSoft.
There is a thread about this:
http://forum.silverfast.com/disgusted-with-this-forum-t9023.html (http://forum.silverfast.com/disgusted-with-this-forum-t9023.html)
Very peculiar indeed.
-
in my observation, there were a couple things i saw about the SF forum before the switch.
- they were having a very difficult time getting bugs accounted for since many were using the forum instead of the reporting utility
- a few started rants and were being unprofessional if not downright rude and harassing and LS shut it down by switching to a moderated forum
- it was consuming a lot a time and effort basically trying to answer support questions that should have been in the bug reporting system or already had answers in the FAQs.
So though i disagreed with what they ended up doing, i can sympathize with their plight.
Frank
-
I am new to the forum (registered today) but have a few comments to add to the mix, since I had a similar experience with SF 8.
I purchased SF 8 recently and had no experience with SF 6.6. I immediately noticed the banding problem after some test scans. It is not obvious but readily visible at 100%. It is most noticeable in areas of strong contrast-a white flower edge on a dark background, for example. I have been scanning Velvia so the problem is not isolated to Kodachrome. I didn't mess with multi-exposure either. I at first tried to get help on the Silverfast forum but was promptly ejected and my case was dealt with behind the scenes. SF first suggested that the problem was with the hardware (sensor). I downloaded the trial version of VueScan and scanned the same slide-no banding, then went back to an old scan made with NikonScan and sent them all back to SF. There was noticeably more detail in the Nikon and VueScan scans and no banding. The tech attributed the problem to over sharpening and claimed the SF version was delivering the true information from the scanner. I respectfully disagreed and to test the idea I suggested that I "upgrade" to version 6.6. The tech agreed and allowed me to download. The scans I made with 6.6 were fine-sharper than version 8 and no banding. I think SF 8 is a work in progress and is still a bit raw. At the time I purchased it they didn't even have documentation.
I am currently scanning happily with SF 6.6 and see no reason to go back to SF 8 at this point.
-
It sounds to me like it might be worth starting a proper forum about it.
-
I don't understand - LULA isn't a "proper forum"? What more or different is needed to discuss this?