Problem scanning very dark Kodachrome

[Dan]

I have tried unsuccessfully for days to get a decent scan from some
very dark Kodachrome slides from 1978.

I get a good preview using Nikon scan v. 4.01, but when I do the final
scan, the result is unusable. Vuescan doesn't even get me a good
prescan. I've tried tweaking and changing all the settings, turning
off ICE, turning off ROC, etc., but nothing helps. I suspect the main
problem is with ROC, which makes the preview look great, but won't
carry through to the final scan. Anyone have any suggestions? I
suspect I will go the internegative route for this batch, but would
much rather do direct scans. I'm using a Coolscan V.

I've posted examples:

http://www.danmurano.com/kchrome/kchrome.htm

Thanks!

 

[Jim]

The only suggestions I have are:
1. Restore the default settings.
2. Increase analog gain
3. Use the curve tool in NikonScan to see if it helps
4. Don't use ICE. It usually only makes Kodachrome scans worse.
5. You should not need ROC on such a recent image
Jim

 

[Dan]

Thanks Jim. The only one I haven't tried is resetting to default,
which I will try next time I can do this. I do have my doubts though,
since I've been all around the default settings with the various
tweaks.

w/o ROC, the image is just heavily blue and very dark -- ROC, in the
preview, makes the colors appear as in the preview sample. Without
ROC, the final scan is far darker than the actual scan shown on the
right.

Hopefully, I will achieve a decent scan at some point and can then
save the settings. Otherwise, I'll go with the interneg route.

I have made scans from other slides in this batch that came out well.
Those were not so heavily underexposed. I have also tried raising the
gain all the way to 2.

Dan

 

[Gary L Hunt]

I can't explain the source of your problem, but I've been scanning 1970s-80s
Kodachromes for a week now (a couple of thousand so far) with very consistent
behavior--my scans look pretty much identical to the preview images. I'm
using NikonScan 4.0.1 with a 4000ED and the SF-200 auto-feeder, with ICE,
ROC etc. all turned off, "Kodachrome" as the film type, and Color Management
turned on. (I accidentally left ICE turned on for a while, but it read all
my deep shadow areas as defects and tried to "correct" them.)

The scans are so-so at best, even though I've tweaked the analog gain settings
and the curves settings in NikonScan to some extent--there just isn't enough
dynamic range for K64 in the sunshine, and the shadows are generally too dark
to bring back effectively in Photoshop without posterization. (People tend to
look as though they have a bad rash when I lighten them up.) But the good news
is, the scans look just like the previews

You might double-check all your Preferences settings. I'm using auto-exposure
for both the preview and the scans, and they can be set independently. Also
I'm using a custom monitor profile, although that shouldn't make any difference
as long as you use the same profile for both NikonScan and whatever you're
viewing your scans in.

Gary Hunt

 

[Ed Hamrick]

I get a good preview using Nikon scan v. 4.01, but when I do the final
scan, the result is unusable.

The problem is that ROC is resolution dependant, and the
results can vary with resoltion.

Try using VueScan, making sure "Input|Media type" is
set to "Image". Press the Preview button, then click
on a neutral color with the right button (or while
holding down the control key on Mac OS X). Then press
the Scan button - this should produce a good scan.

Regards,
Ed Hamrick

 

[Don]

I can't explain the source of your problem, but I've been scanning 1970s-80s
Kodachromes for a week now (a couple of thousand so far) with very consistent
behavior--my scans look pretty much identical to the preview images. I'm
using NikonScan 4.0.1 with a 4000ED and the SF-200 auto-feeder, with ICE,
ROC etc. all turned off, "Kodachrome" as the film type, and Color Management
turned on. (I accidentally left ICE turned on for a while, but it read all
my deep shadow areas as defects and tried to "correct" them.)

Turn Nikon Color Management OFF! This is very important because Nikon
Color Management *boosts* blue and *cuts* red (you want exactly the
opposite for Kodachrome).

In my experience VueScan doesn't work because you can't adjust
individual Analog Gain. This may not be a big problem for bright
slides but VueScan is useless with darker ones. I don't like VueScan
(for many other reasons too) so take this with a grain of salt...

I've been wrestling with Kodachrome for a year now. If you check the
archives there are quite a few threads on the subject.

Don.

 

[Dan]

Thanks Gary. The auto exposure might be the thing that's throwing it
off so badly. I'll give it a try.

 

[Ed Hamrick]

In my experience VueScan doesn't work because you can't adjust
individual Analog Gain.

That's just silly (and just plain dumb).

Regards,
Ed Hamrick

 

[Dan]

Thanks Ed. That explains a lot. At least knowing why something happens
makes it more tolerable .

I'll try Vuescan with the settings you recommend. BTW, I've been
scanning images since the very first Nikon and Leaf scanners came out
in the early 90's. Vuescan is a terrific product.

Dan

 

[Don]

That's just silly (and just plain dumb).

Yes, the fact that VueScan refuses to adjust individual Analog Gain is
just silly (I'm too polite to call it plain dumb - but I take your
word for it...). ;o)

Don.

 

[Jim]

Thanks Jim. The only one I haven't tried is resetting to default,
which I will try next time I can do this. I do have my doubts though,
since I've been all around the default settings with the various
tweaks.

I suggested restoring all of the default settings because it seemed that you
have lost track of the effect of all of those settings.

The samples that you showed have pixels throughout the range. It just
happens that far more of them are concentrated toward the dark end of the
spectrum.

I would intepret this as meaning that additional analog gain would not help.

Instead, you need to investigate what judicious use of the curve tools would
do.

Jim

 

[Erik Krause]

Hello, Don
you wrote...

In my experience VueScan doesn't work because you can't adjust
individual Analog Gain. This may not be a big problem for bright
slides but VueScan is useless with darker ones.

In my experience VueScan is much better for dark slides than nikon
scan, since it supports long exposure pass and multi pass scanning.
Both features are not available with nikon scan for the LS 50.

 

[Mike Engles]

I have tried unsuccessfully for days to get a decent scan from some
very dark Kodachrome slides from 1978.

I get a good preview using Nikon scan v. 4.01, but when I do the final
scan, the result is unusable. Vuescan doesn't even get me a good
prescan. I've tried tweaking and changing all the settings, turning
off ICE, turning off ROC, etc., but nothing helps. I suspect the main
problem is with ROC, which makes the preview look great, but won't
carry through to the final scan. Anyone have any suggestions? I
suspect I will go the internegative route for this batch, but would
much rather do direct scans. I'm using a Coolscan V.

I've posted examples:

http://www.danmurano.com/kchrome/kchrome.htm

Thanks!

Hello

It is interesting that you say this. I have had a long running argument
with a guy at Nikon support in the London.
He denies that that something like this is happening.
I have sent him innumerable screen grabs showing the discrepency between
Preview and scan. This problem only exists in scans with bit dept
greater than 8 bit.

My solution is this.

I turn up the Analogue gain to 1.5 and then apply curves in LCH, to give
a good preview. I then scan at 14 bit with my LS 50. In Photoshop I do
my black and white points. This avoids the severe shadow clipping that
seems to happen in Nikon scan when doing white and black points.

ROC gives a good preview, but the shadow clipping happens in the scan.

Mike Engles

 

[Dan]

Thanks Mike. I was getting frustrated from wasting so much time and
having no sense of what might be happening. I look forward to trying
out these solutions, hopefully over the weekend. Ed's explaination
about ROC being resolution-dependant makes a lot of sense too.

 

[Roger Halstead]

I have tried unsuccessfully for days to get a decent scan from some
very dark Kodachrome slides from 1978.

I get a good preview using Nikon scan v. 4.01, but when I do the final
scan, the result is unusable. Vuescan doesn't even get me a good
prescan. I've tried tweaking and changing all the settings, turning
off ICE, turning off ROC, etc., but nothing helps. I suspect the main

If ROC makes the preview look OK then it should make the final look
OK.
I take it you are referring to the image as seen in Nikon Scan and not
Nikon View?

What format are you using? JPG, NEF?...

problem is with ROC, which makes the preview look great, but won't
carry through to the final scan. Anyone have any suggestions? I
suspect I will go the internegative route for this batch, but would
much rather do direct scans. I'm using a Coolscan V.

I've found that much of the Nikon soft ware seems to have some
idiosyncrasies, but this sounds strange.

I've posted examples:

http://www.danmurano.com/kchrome/kchrome.htm

Looking at the examples and taking a SWAG... As a trial with a very
dense slide I'd set ROC to ZERO and play with DEE AND the Analog gain.

The confusing thing is that what you see in the "processed" preview
panel should be what you see in the final scan.

How do the <processed> and <natural> preview compare?
Does the processed look line what's on the above link and the natural
look like the one on the right, or the other way around. There
certainly should be a substantial difference between the two.

I did find that there are combinations of ROC with DEE that come out
terrible.

But I was able to get around those without ROC and using some
combination of DEE and analog gain.

Good Luck,

Roger Halstead (K8RI & ARRL life member)
(N833R, S# CD-2 Worlds oldest Debonair)
www.rogerhalstead.com

 

[Don]

In my experience VueScan is much better for dark slides than nikon
scan, since it supports long exposure pass and multi pass scanning.
Both features are not available with nikon scan for the LS 50.

I always acknowledge the ability of VueScan to go beyond 4 AG as a
plus. However, this is nullified by the inability of VueScan to adjust
individual RGB AG values.

Not to mention, the totally incomprehensible decision to use a
proprietary (and undefined!) "measurement unit" instead of the
accepted EV. But that decision is in keeping with VueScan's overall
confused "design".

Multipass multiscanning is really a gimmick (at least on my scanner)
as images don't align. Only single pass multiscanning is consistently
reliable.

Don.

 

[Bart van der Wolf]

SNIP

Not to mention, the totally incomprehensible decision to use a
proprietary (and undefined!) "measurement unit" instead of the
accepted EV.

What is so incomprehencible about a simple multiplier of calibration
exposure? Scanners usually calibrate themselves during startup, and VueScan
improves on that by automatically increasing exposure while trying to avoid
saturation of the sensors, based on the film used.
From there you have the possibility to change the calibration exposure with
a simple multiplier, and you get VueScan's improved exposure suggestion
presented on a silver platter to start with.

The Exposure Value only 'complicates' things because you have to raise 2 to
a power EV, and only then you can multiply the calibration exposure.
Although I am a photographer by original profession and have no problem with
converting integer and half EVs by head, a simple multiplier is much less
trouble, but YMMV.

Multipass multiscanning is really a gimmick (at least on my scanner)
as images don't align. Only single pass multiscanning is consistently
reliable.

Yes, it does depend a lot on the quality (repositioning accuracy) of the
hardware, which often is less than perfect. A solution like
http://aberrator.astronomy.net/registax/ for registering multiple scans
probably suits the hard cases better.

Bart

 

[Don]

What is so incomprehencible about a simple multiplier of calibration
exposure?

You misunderstood. What "incomprehensible" refers to above is the
decision (to use a non-standard unit) not the unit itself. Of course,
the fact that this confusing "unit" isn't sufficiently described in
the documentation just makes matters worse.

If it were just this one thing, fair enough. It can be overlooked. But
it's from a long list of similar "unusual" VueScan design decisions.

The Exposure Value only 'complicates' things because you have to raise 2 to
a power EV, and only then you can multiply the calibration exposure.
Although I am a photographer by original profession and have no problem with
converting integer and half EVs by head, a simple multiplier is much less
trouble, but YMMV.

The point is not how easy or difficult it is, the point is this
"complication" is totally unnecessary. Why not do it in software and
spare the user? What's the purpose of using a non-standard unit
instead of converting to EV (other than lazy programming)? What
overwhelming advantage for the user does it have over the familiar EV
to warrant such a radical departure?

It's yet another symptomatic example - like the dogmatic declaration
that no user shall ever need to adjust individual Analog Gain - of how
idiosyncratic, haphazard and inconsistent VueScan "design" is.

But I hasten to add (as I always do) many VueScan fans are prepared to
live with all these numerous "quirks", and that's fine. But just
because they're willing to put up with all this doesn't mean the
problems aren't there.

Don.

 

[Dan]

If ROC makes the preview look OK then it should make the final look
OK.
I take it you are referring to the image as seen in Nikon Scan and not
Nikon View?

What format are you using? JPG, NEF?...


I've found that much of the Nikon soft ware seems to have some
idiosyncrasies, but this sounds strange.


Looking at the examples and taking a SWAG... As a trial with a very
dense slide I'd set ROC to ZERO and play with DEE AND the Analog gain.

The confusing thing is that what you see in the "processed" preview
panel should be what you see in the final scan.

How do the <processed> and <natural> preview compare?
Does the processed look line what's on the above link and the natural
look like the one on the right, or the other way around. There
certainly should be a substantial difference between the two.

I did find that there are combinations of ROC with DEE that come out
terrible.

But I was able to get around those without ROC and using some
combination of DEE and analog gain.

Good Luck,

Roger Halstead (K8RI & ARRL life member)
(N833R, S# CD-2 Worlds oldest Debonair)
www.rogerhalstead.com

I tried combinations of DEE and Analog Gain and Scan enhancer (can't
remember its precise label -- I'm not at my computer now). I've been
scanning using the Twain plug-in directly into Photoshop. In the
preview window, the Natural tab shows a very dark image, the Processed
image is the one I copied to show how the preview looks, which,
considering the density of this slide, is pretty good. I'll keep
trying and let you know the solution, if I can find one.

Dan

 

[Erik Krause]

Hello, Don
you wrote...

I always acknowledge the ability of VueScan to go beyond 4 AG as a
plus. However, this is nullified by the inability of VueScan to adjust
individual RGB AG values.

I don't understand at all why this should be necessary to scan dark
slides. Could you please explain in detail?