How best to manage CCD leakage from a Nikon LS 5000 scanner

[frank_bilki]

I've recently purchased a new Nikon LS 5000 scanner and have been

disappointed to discover some fairly prominent artefacts caused by

CCD leakage. Some of my scans, particulary those from Velvia

transparencies containing highlights adjacent to dark areas, show

noticeable vertical bloom (when the image is viewed in its un-

rotated orientation). I've read several posts on this and other web

sites that suggest this is a limitation of desktop CCD-based

scanners that we as users will just have to live with. What I'd like

to know, then, is:

 

1. Just how much bloom/leakage is acceptable? Is there a threshold

value above which we should consider a particular scanner to be

defective?

 

For example, the attached file shows the first image where I noticed

the effect. Admittedly, the original trannie was underexposed and

I've contrast stretched this thumbnail to within an inch of its

life, but I've since seen the effect on normal daylight photographs

with absolutely no curves adjustments at all (just don't have one on

me at the moment).

 

2. Given that it's an inherent hardware limitation, does anyone have

any particular workflow suggestions for minimising its effects on an

image? Burning in with the brush about the same size as the

highlight seems like a solution, but my experiments suggest it would

be suitable for only the simplest 'point' highlights, and would be

difficult to use for anything more complex.

 

What are your thoughts?

 

Frank<div></div>

[carl smith]

There was a common problem like this in the older Nikon scanners, which was caused by an accumulation of dust on the mirror in the optical path. It was easily cleaned, however I don't remember it being more prominent in any particular direction, highlights bled evenly in basically all directions.

 

Regardless of that, this is a new scanner and should not be displaying this behavior. I don't know if it is common with the CoolScan 5000, which you may want to look in to, but it may warrant a replacement unit.

[johann_fuller]

It's a problem with the 9000 as well - I get it with negs where the clear film between frames bleeds an orange streak onto the image. I've sent a few e-mails to Nikon technical support - never got a reply.

[dfdncithekxlbn8kaglf33]

I get this with my Canon, too, and it is a pain. Infact, just like your example, neon seems to be the problem more than anything. You might try rotating the slide so the 'glow' heads into the hotspot as much as possible. I have found that streaking can go all the way across a frame in certain scenes.

[frank_bilki]

Johann,

 

What do you do to manage it? Are you merely marking time until you get a response from Nikon? (I've also emailed Nikon and am yet to receive a reply.) Or, have you figured out a way to get around it?

 

The local camera store where I purchased the unit has suggested that a multi-pass scan may reduce the effect. I'm yet to try this, but it kinda makes sense that if the leakage is in any way random, a multi-pass scan may be useful. Of course, if it isn't random then no number of passes will work. (The offending scans were only single pass.) Beyond that, no-one seems to have discussed this issue past the point of, "such-and-such scanner is/isn't affected by ccd leakage."

 

Frank

[johann_fuller]

My workaround is simple - with the standard plastic film holder - don't use the frame with the cutout. With the glass rotating holder I made my own mask to just the size of the film frame with no clear film area showing. The later requires scanning in 16 bit as the scanner can't calibrate itself - turn off curves and trim the histogram in Photoshop levels. You will find loads more detail in the highlights of negs this way as well but also loads more noise so use at least 4x multisample and turn on fine scan ( Single CCD) to avoid banding. Scan times are very long but the results are well worth it.

[frank_bilki]

Thanks, Johann.

 

I also sent the sample image to my local camera store, who have suggested that I suffer from camera shake! At least they've agreed to test the scanner side-by-side with the LS 5000 they use for their own production work, so hopefully the problem will reveal itself as being unique to my particular unit. I'll post a summary once I've been through the process.

[helenbach]

I've had a look through my scans of slides made with the 5000, and can find no sign of similar smearing.

 

Helen

[charles_miller]

I and several other folks went through this with earlier Nikon scanners. The bottom line is that there is very little CCD leakage in Nikon sensors, nothing even close to your example.

 

Turn the transparency 90 degrees and scan it again. If the artifact moves with the rest of the image, it is on the slide. If the artifact is still in the same direction, it is from the scanner.

 

Nikon scanners have a front-surface mirror which tends to collect dust, particularly if the scanner sits upright rather than on its side. The dust causes a DIRECTIONAL diffuse glow around highlights. I posted a mirror-cleaning procedure on photo.net twice already. Of course, you can just let Nikon clean it for you for a few hundred dollars.

 

Do your final evaluation of any potential "leakage" from a print, because the glass in CRT and LCD monitors smears highlights a little bit too.

[carl smith]

This is on a brand new scanner however, I never experienced this on any of the Nikons until they'd been in use for a while and had a chance to accumulate dust.

[frank_bilki]

Charles, Carl,

 

You're right, the scanner is brand new -- it's been out of the packing for all of three weeks.

 

I've been scanning from uncut strips (why mount trannies that will never be projected?), but I'm going to mount my two worst offenders. (An extract from the second one -- shot during the day -- is presented here with no curve manipulation at all. If you lift the curve you'll see the flare become even worse.) At least by mounting them I'll be able to rotate the image and see what effect that has on the scan, although I don't see what would cause the same flare on a tripod mounted night shot and a day shot in bright sunlight with a different lens.

 

My camera store is willing to take back the scanner and test it with one of their 5000's, which is another reason to mount the offending trannies. I had hoped for a more direct response along the lines of, "yes, we can see that it's faulty and will replace it immediately", but at least I'm getting somewhere.

 

I hadn't wanted to jam up the forum with progressive updates, which is why I've gone a little quiet, but I'll be sure to let you know how I get on.

 

Thanks for your suggestions.

 

Frank

[frank_bilki]

Some test results with my scanner:

 

Single / 4 / 16 pass -- no effect

Digital ICE on / off -- no effect

8 / 16 bit output -- no effect

Rotating the trannie -- flare orientation changed relative to the image in both of the test images.

 

Obviously a problem with the scanner. Interestingly, after viewing the rotated trannies, the flare appears to be more of an 'echo' of the highlight than just a smear of light. Hmmm.

[www.graemehird.com]

Hi Frank,

 

That's definitely wrong. The step-wise nature of the flare suggests to me the sensor is experiencing flare, in that it is "remembering" the highlight after the sensor has moved to the next step in its scan. The memory fades over time as the sensor moves along its path. It shouldn't show that effect, but it indicates the level of the light source is too bright. If you can find a way to turn it down and do multiple passes to get your exposure correct, you might be able to work around it.

 

Let's hope PD comes to the party with the replacement scanner.

 

Cheers,

Graeme

[frank_bilki]

Hi Graeme,

 

An interesting theory, but if I understand correctly, the CCD chip scans the entire short axis of the trannie in each step as it travels along the long axis. If your idea is correct, I'd expect the flare to extend *sideways* along an un-rotated landscape format scan. However, I see *vertical* flare, which to me says that the flare/leakage is extending to adjacent pixels in the CCD during any given step in its travel. (When I rotated the trannies the flare appeared horizontal but it was still vertically oriented within the scanning aperture.) Also, if the CCD were remembering charge from a previous step, I'd expect the smear of light to fade away more or less continuously, but I'm quite surprised to see it 'echo' away.

 

I dropped the scanner off at PD today but will have to wait until at least Monday for them to assess it. Here's hoping they accept that it's faulty, otherwise I could be inventing creative ways to manage that amount of flare!

 

Cheers, Frank

[www.graemehird.com]

Hi Frank,

 

Ya got me there .... I didn't realise the orientation of the flare was in that direction.

 

Given that info, it looks like internal reflection between two parallel mirrors. I'm sure you've seen the effect when you enter a room with mirrors on opposing walls. Reflections seem to step off into the distance, each slightly offset from the previous image. The scans shown above have that look about them.

 

How to deal with it? Is there a text shortcut for a shrug of the shoulders? Again, I think the best solution is to turn the light down a little, since the scanner is boosting the level to attempt to see through the dark areas of the film. That's not a real solution though - there will still be some reflection occurring.

 

Cheers,

 

Graeme

[frank_bilki]

Hi Graeme,

 

I was beginning to get an inkling that the mirrors might somehow be involved, but I'd have thought the nimble-fingered engineers at Nikon would have considered the need to scan dark trannies when they designed the scanner! It seems a bit too simple, don't you think, to say that my poor old (new!) scanner can't cope with a dark trannie?

 

It's interesting that the posts in this thread have ranged from "nup, never seen it" to "get it all the time". I wonder if it's a quality control issue?

 

I like your idea of dropping the light intensity, and will keep it in mind if I get no satisfaction from PD. (It's easy with NSCAN4 -- just drop the analogue gain. Does Vuescan have a similar option?)But, given the number of trannies in my collection that would potentially be affected by this (lots of twilight urban shots and astrophotos, both of which are dominantly dark with lots of bright highlights!)I'd rather have a scanner that can cope in the first place.

 

Cheers,

 

Frank

[frank_bilki]

Conclusion to this saga:

 

Just spoken with my local camera store. They've tested my scanner with another 5000 'out of the box' and an older 4000. Turns out they were able to get perfectly useable scans with my own scanner by manipulating the histogram endpoints in a different way than what I've been doing (using the LCH curves rather than the RGB curves). Certainly nothing as flare affected as what I've been getting. The results from my scanner exactly match those from the second (new) 5000, and in both cases they were much better than those from the 4000.

 

It seems that the problem has been due to operator error. I am, of course, yet to reproduce their 'new' results myself, but they are willing to send the unit to be tested under warranty if I remain unconvinced.

 

So much for CCD leakage and internal reflection. I don't know whether to feel relieved or not. Think I'll try vuescan though.

 

Thanks all for your feedback and suggestions.

 

Frank

[boku_iwanai]

I get the same problem. Actually I posted a message about it today before I found this. Is the Nikon worse than other scanners in this regard? The "CCD leakage" hypothesis sounds fair to me by the way, as it is definitely occuring along the CCD.

[beau 1664876222]

I have seen this with both my Nikon scanners, but since I do 100% b&w negs, the effect is of course a dark smearing near dark areas. I think this phenomenon is one of the reasons the scanner is designed to give a little less than full-frame -- the black edges of the neg will, if scanned, bleed into the image area quite noticeably.

 

I manage it as stated above, by not getting into situations where I have to blast the film with light. It's one of the many reasons I try to make my negs be as thin as possible for scanning. Dense negatives or slides are a nightmare for scanning in general.

[boku_iwanai]

Does anyone know how other makes compare in this regard? Are they better or worse. Is this a Nikon trait?

[grahame]

Hum?. I have been thinking of purchasing a 5000ED scanner or equivalent for a while now. I especially need the ability to scan multiple slides as I've many thousands of them (I've certainly not enough time left in my life to stand there shoving slides in one by one).

 

I've put off buying a decent slide scanner because it seemed to me that the state of the art wasn't quite up to it (except for drum scanners), perhaps it's still not. However, I don't know for sure such is the paucity of information on this subject.

 

It is had to get real concrete information on the limitations of scanner CCDs. All the scanner reviews that I've read contain subjective comments and few real engineering comparisons, or reasons for CCD limitations, or even the real state of the art between film and CCD-to-digital storage conversion. In other words, after we've scanned can we afford to throw our film emulsions away one and for all?

 

To develop the argument a little further I need to digress?. I seem to recall that the spatial frequency response (the practical limit of resolution) of a Kodachrome slide is about 100 lines per mm (keep in mind that some newer films are considerably better). As the actual size of a 35mm slide is 36mm x 24mm (1.412 x 0.944 inches), it more or less corresponds to 3600 x 2400 pixels [let's not argue about pixel aperture and such here]. So then, what is the necessary resolution of a scanner if it is to capture all the information within, say, a Kodachrome slide?

 

As the slide is 36mm in length then it would seem that we'd need 3600 corresponding pixels to span the full 36mm (i.e.: at 100 lines/mm). Well not so: very simply, a theoretician in information theory, Nyquist, said "a signal should be sampled at twice the frequency of its highest frequency component--i.e.: of the finest detail in an image--in order to be converted into an adequate representation of the signal in digital form." (It's the same reason why the sampling frequency in your CD player is 44.1kHz instead of about 20kHz).

 

Doing a quick calculation, this then corresponds to a Kodachrome slide requiring a scanner with a total resolution of 7200 pixels, or more commonly 5081 dpi (7200 x reciprocal of 1.417"). Any scanner with less resolution has to discard some of the slide's information. Similarly, better emulsions would require even higher resolution scanners.

 

Ok, what does this have to do with CCD smear? Not much in this context but it does indicate how little information about film scanner requirements that we're actually told. Amongst all the oodles of bumph on scanning on the Net, or for that matter, what is published scanner manufacturers' specifications really good technical information about scanners is difficult to come by.

 

As I see it, we really need some definitive and objective comparisons of film scanners. Such comparisons ought to include such aspects such as:

 

* scanner resolution [real not interpolated]

 

* depth of field (centre and outside of slide)

 

* true scanning speed

 

* gamma

 

* true (measured) dynamic range of scanner

 

* the real advantages (or otherwise) of using an extra infrared sensor to aid Digital ICE etc.

 

* CCD artefacts such as:

 

- smear/leakage

 

- problems scanning Kodachrome

 

- memory effects

 

- noise

 

- noise increasing with corresponding decrease in CCD pixel size [i.e.: resolution trade-off versus noise]

 

Etc. etc.

 

If anyone can point us to some really good references sans dross and fluff, especially with respect to how the various manufacturers measure up against each other, then it would be greatly appreciated.

 

(My apologies for this long post, but I see no other or simpler way to get my point across.)

[dan_isaacs]

Hi --

 

I know this is an old thread, but I've got a good CCD leakage (and "dark noise", in general) workaround I use with my Canoscan 2710. It should work with any scanner, really.

 

Requirements:

Vuescan 8.x

Photoshop CS

AIM RGB Pro or RGB Pro Wide Gamut profile (download-> http://aim-dtp.net)

 

1.) In Viewscan: (the first scan)

a) be sure you set Autofocus to "always"

b) multiple passes (8 or so) is good

c) make sure Long Exposure Pass is DISABLED

d) set the CCD exposure to 1.0

e) Disable TIFF and JPEG output --- scan to Raw .tif only.

 

2.) Once first scan is complete... (the second scan)

a) switch focus to "Manual"

b) set multiple passes to 4 or so

c) set the CCD exposure to 4.0

d) scan to Raw file again (be sure not to overwrite the other scan)

 

3.) Open both scans in Photoshop. Assign your scanner profile to the images, then convert the images' profiles to AIM RGB Pro (or AIM RGB Pro Wide Gamut)

 

4.) Duplicate the Background layer of the second scan on top the the first scan. Close the the second scan (you won't need this anymore). You may need to nudge the layer position a pixel or two for proper alignment.

 

5.) Apply a curve to the "blown-out" top layer to bring the shadows back in range. I usually make one node at 64 and remap that to 32 (or so). You may also want to reduce the saturation of this layer (using Hue/Saturation). Saturation -10 to -25 is good.

 

6.) Set the layer transfer mode to "Darken"

 

7.) Flatten image and convert to profile of your choice!

 

This often works quite well. Optionally, you can use Layer sets and Adjustment layers for the Curves and Saturation. This allows more control of the blending.

 

I hope this helps you (or someone else)!