LS-50: Initial impressions

[Bart van der Wolf]

SNIP

You can find KC films where some frames scan almost
perfectly without loss of sharpness and others look positively
soft until ICE is turned off. Just a word of warning before
your assumptions become enforced.

Another benefit of VueScan ;-)
It'll show you the amount of IR channel data, either as a separate channel
or as a signal color.

Bart

 

[Kennedy McEwen]

Another benefit of VueScan ;-)
It'll show you the amount of IR channel data, either as a separate channel
or as a signal color.

Sure, it will show you the IR channel, but it won't show you how ICE
interprets the IR channel. It is a commonly held misconception that the
IR channel only contains defects but, as anyone who looks at the IR
channel in Vuescan can see, it contains a significant amount of image
detail as well - even with chromagenic films. Interpreting which parts
of that IR channel are defects and which are not isn't merely a
thresholding exercise. If it was, then small defects (which I find the
most objectionable in any case) would actually be missed by ICE entirely
because they also appear at a low contrast.

 

[Don]

It can do, if it interprets the grain as defects and attempts to conceal
it as such. Remember that the resolved contrast of the image recorded
on the film has deteriorated virtually to zero by the time it reaches
granular dimensions (ie. MTF is zero well before the spatial frequencies
of a granular scale are reached) so the effect on the final image may
not appear noticeably softer.

That's very interesting! If I understood correctly, what you're saying
is that it's really physics (at this scale) which results in sharpness
being maintained rather than some fanciful ICE4 programming. On the
LS30 - i.e. ICE (1) - even non-Kodakchromes and negative scans were
softened (distorted, is probably a more apt term) far beyond my
tolerance level.

However, be careful with Kodachrome since
this effect of ICE can vary to a much more severe level, literally from
one image to another, simply due to the amount of retained silver in the
final emulsion. You can find KC films where some frames scan almost
perfectly without loss of sharpness and others look positively soft
until ICE is turned off. Just a word of warning before your assumptions
become enforced.

Oh, no danger of that... Remember, this is the guy who didn't
believe neither his own nor other people's assumptions about the
LS-30. That's why it took over a year of empirical tug-of-war until I
finally wrestled it to the ground. And then I got a new scanner!

In any case, that's exactly why I liberally sprinkled numerous caveats
all over my original message. It's early days, and all initial
impressions are to be taken with a boulder of sodium chloride... ;o)

Don.

 

[Don]

Nikonscan is now updated to 4.02. They have cleared up the problem with
the colour and contrast difference between preview and 14 bit output.
Dark kodachromes in conjunction with ROC/GEM are very good.
I really do think that Nikonscan is now very good. They have also had a
look at the problems with multiple images on a strip, but I have not
looked at that yet.

I know, I just saw Kennedy's note and went to get the new version
right away. The Nikon site must be quite busy because my download
timed out twice half-way - but, third time lucky...

I haven't spent a lot of time with ROC/GEM because I prefer to do that
sort of thing manually in Photoshop.

First impressions of dark Kodachromes is that they still need to be
boosted beyond what Auto Exposure determines, but at least there is no
need for individual RGB Analog Gain adjustments. Well, at least so
far...

Even if there is a cast, it appears relatively minor. It certainly
does not affect the histograms like it did on the LS-30, i.e. there is
enough headroom for adjustments in Photoshop without banding or other
artifacts. And 14-bits of dynamic range doesn't hurt either!

So, it's still a honeymoon, but rigorous tests continue...

Don.

 

[Roy G. Biv]

First impressions of dark Kodachromes is that they still need to be
boosted beyond what Auto Exposure determines, but at least there is no
need for individual RGB Analog Gain adjustments. Well, at least so
far...

For dense Kodachromes, I suggest trying the shadow
enhancement in Nikon Scan before boosting the analog gain.
Sorry but due to spam I shall not provide a valid e-mail address.
Please reply to group with questions or comments.

 

[ThomasH]

I would venture a guess that we have a widely different definition of
"blessed"... ;o)

Warts and all, but I'd pick NikonScan any day.

The day will come you will pick Vuescan, discover how to
use it and begin to love its ability to reproduce color
and to control it!!

Thomas

 

[Kennedy McEwen]

I would venture a guess that we have a widely different definition of
"blessed"... ;o)

Warts and all, but I'd pick NikonScan any day.

The day will come you will pick Vuescan, discover how to
use it and begin to love its ability to reproduce color
and to control it!!
[/QUOTE]
Hasn't arrived for me yet, and I suspect I have been using Nikonscan in
its various implementations for somewhat longer than you, Thomas. ;-)

I would suggest that the statement should be reversed.

 

[Kennedy McEwen]

That's very interesting! If I understood correctly, what you're saying
is that it's really physics (at this scale) which results in sharpness
being maintained rather than some fanciful ICE4 programming. On the
LS30 - i.e. ICE (1) - even non-Kodakchromes and negative scans were
softened (distorted, is probably a more apt term) far beyond my
tolerance level.

Not quite. It is certainly a programming difference that means that ICE
on your LS-50 is better than ICE on the LS-30 (or the LS-2000 for that
matter). However it can still be fooled by certain film aspects, to
which Kodachrome is particularly prone. Even some chromagenic emulsions
can cause similar problems, particularly Kodak and Fuji from a couple of
years ago. Look up references to pepper noise if you are interested.
This was discovered soon after the LS-4000 was introduced and eventually
traced to microscopic bubbles in the film base, which were detected and
corrected by ICE. However, because of their density on some films, it
resulted in ICE softening. I understand that both Kodak and Fuji have
now modified their manufacturing process to eliminate these defects.

 

[Bart van der Wolf]

Sure, it will show you the IR channel, but it won't show you how ICE
interprets the IR channel.

You can toggle the option on/off and see the final result in the preview,
and the signal color will show the affected pixels (after thresholding,
etc.), which may be helpfull because the differences may be too small to
notice at first glance.

If the preview is done in the final scan resolution, there can be no
surprises before saving the full res preview (which is identical to the
final scan if the exposure is locked first).

One can also save the Raw data file and postprocess with the IR channel mask
later, either in VueScan or in e.g. Photoshop. Lots of built-in flexibility.

Bart

 

[Don]

Not quite. It is certainly a programming difference that means that ICE
on your LS-50 is better than ICE on the LS-30 (or the LS-2000 for that
matter). However it can still be fooled by certain film aspects, to
which Kodachrome is particularly prone. Even some chromagenic emulsions
can cause similar problems, particularly Kodak and Fuji from a couple of
years ago. Look up references to pepper noise if you are interested.
This was discovered soon after the LS-4000 was introduced and eventually
traced to microscopic bubbles in the film base, which were detected and
corrected by ICE. However, because of their density on some films, it
resulted in ICE softening. I understand that both Kodak and Fuji have
now modified their manufacturing process to eliminate these defects.

I was going to leave ICE testing for later, but I just ran 3 simple
scans: no ICE, ICE normal and ICE fine. I overlaid the three images
and then toggled between them to spot the differences.

One irregularity - which caused me to drop ICE (1) completely on the
LS-30 - is still there albeit very much reduced. I'm referring to the
transition in high contrast areas. Viewed at 1:1 magnification (or
higher) the edges are very regular and straight in the "no ICE" scan
but there are definitive and highly irregular "protuberances" of
bright areas into dark on both ICE4 scans. Although considerably
reduced from ICE (1) it's still noticeable.

I wonder how much this improvement is due to ICE4 and how much simply
to the higher resolution of the LS-50?

As far as this effect is concerned, there seem very little difference
between "normal" and "fine" ICE setting. However, this is all when
viewed at 1:1 or higher so, to be fair, for all intents and purposes
it shouldn't be an issue. Nevertheless, the effect is there.

Regarding the above mentioned softening, somehow I don't find it very
objectionable on the LS-50 while I really resented it on the LS-30.
This is all very subjective, but I wouldn't even call it softening. It
reminds me the most of "grain reduction". However, unlike all grain
reduction methods I've seen which just seem to (as far as I'm
concerned) totally destroy the image by making it appear as if viewed
through semi-transparent plastic - ICE4 seems to smooth out the grain
without noticeable loss of sharpness! And then to boot, it gives this
appearance of multiscanning in dark areas by reducing noise! Well, at
least so far...

Don.

 

[Don]

For dense Kodachromes, I suggest trying the shadow
enhancement in Nikon Scan before boosting the analog gain.

On closer inspection (those were just first impressions) I actually
don't think I need to modify the exposure at all!

The dynamic range of the LS-50 is just "luverly"! There's more dynamic
range than I can shake a Photoshop at! ;o) Even at simple "Auto
Exposure" all scans (so far...) have at least about 10 pixels worth of
headroom at both ends.

After all the agonizing on the LS-30 with contrast merging, my reflex
was to first try and improve the exposure on the LS-50 in the same way
(by boosting exposure until the highest histogram bumps up against the
right edge).

But there seem to be no point of doing that because there's plenty of
dynamic range headroom at both ends even at Auto Exposure! Well, so
far...

Don.

 

[Don]

Hasn't arrived for me yet, and I suspect I have been using Nikonscan in
its various implementations for somewhat longer than you, Thomas. ;-)

I would suggest that the statement should be reversed.

Indeed, although I doubt that (in general) people who are happy with
VueScan have the capability to truly appreciate NikonScan.

They remind me somewhat of that guy in Spinal Tap who insists on
buying an amplifier because "it goes to 11"... ;o)

Don.

 

[ThomasH]

Indeed, although I doubt that (in general) people who are happy with
VueScan have the capability to truly appreciate NikonScan.

As much one can appreciate what Kennedy said--its a different
opinion--you shoot over the target. NikonScan is a simplistic
product. Search google archives to find numerous article written
on the topic here for many years. Me, I used Nikon scan in
classes to teach people how *not to make a gui*. Its truly an
exceptional collection of all mistakes and "crimes" against
ergonomics one can make.

Aside of this gui aspect, there is the image processing side.
Vuescan provides excellent autocrop, film and scanner profiling
mechanism, white balance algorithm, night exposure heuristics,
film database, raw file scan ability and many many more. This is
a small, uniquely (controversial) designed tool, which is
accepted for its abilities to scan from virtually every thinkable
source there is. (So is Silverfast, by the way, but its very
expensive and you must pay extra for things like batch scan.)

None of this is available in NikonScan, except for some
simplistic bit manipulations in awkward menus and a rather
arbitrary curve tweaking. I am not counting the ICE**4 modules
provided by the ASF, now belonging to Kodak. Both ICE and GEM
have proven to be invaluable tools. I make usually 48bit
*.tiff raw files using NikonScan, but processed by ICE and
GEM, and later I use Vuescan to generate whatever I like from
these raw files.

Thomas

 

[David Blanchard]

Hasn't arrived for me yet, and I suspect I have been using Nikonscan in
its various implementations for somewhat longer than you, Thomas. ;-)

I would suggest that the statement should be reversed.

I have used both for a few years now. I greatly prefer the user
interface of Nikonscan compared to Vuescan. I have found that making
adjustments with Vuescan takes considerable effort compared to
Nikonscan. And, after sufficient experimentation, I have learned how
to use Nikonscan to get quality results.

On the other hand, I am investigating Silverfast since it is currently
the only scanning software that is still providing updates for my
current scanner and my current OS.

David

 

[Don]

Me, I used Nikon scan in
classes to teach people how *not to make a gui*. Its truly an
exceptional collection of all mistakes and "crimes" against
ergonomics one can make.

With all due respect, but this should be interesting. Would you mind
outlining a few? (And which classes would that be, exactly?)

Especially when compared to VueScan's... ahem... "interesting"
so-called user interface. From confusing interaction of settings on
different tabs (did you say "ergonomics"?) and *annoyingly* flickering
screen (you did say "ergonomics", right?), to "random" redraws
*interrupting* user activity and constantly changing "menu du jour" as
options are being reshuffled (I could've sworn you mentioned
"ergonomics")...

Surely, you're not implying that VueScan has a superior (or indeed,
*any*) GUI design?

If you really know about GUIs, as you claim, and hold VueScan as an
example, then just one word: Muscle memory! OK, two words... ;o)

No wonder VueScan's User Guide reads like a Trivial Pursuit
question...

Vuescan provides excellent autocrop, film and scanner profiling
mechanism, white balance algorithm, night exposure heuristics,
film database, raw file scan ability and many many more.

Most of those are either irrelevant (as they are a function of
multiscanner "support") or are subjective preferences and arguing
about taste is futile...

Instead, I have posted various VueScan "mutilations" in the past based
on objective histograms rather than subjective personal preferences.
Histograms don't lie.

I also posted scans where VueScan does work (for balance and
objectivity) and if VueScan serves your purposes (as it obviously
does) then that's great. But to claim any objective superiority or
even quality - especially regarding GUI design - is just not
realistic.

As stated earlier, VueScan is like a point-and-shoot throwaway
Instamatic camera. If it does what the user wants, great (for that
user) but to claim it's a Hasselblad goes a bit far...

This is
a small, uniquely (controversial) designed tool, which is
accepted for its abilities to scan from virtually every thinkable
source there is. ....
None of this is available in NikonScan, except for some
simplistic bit manipulations in awkward menus and a rather
arbitrary curve tweaking.

And there's a good reason why not. NikonScan is for... erm... Nikon
scanners?

As for image editing, that's what a dedicated program like Photoshop
is for. Adding such "features" unrelated to the program's purported
usage (scanning!) just leads to a bloated design. After all, NikonScan
is a scanner *driver* and can be used as a TWAIN module to be invoked
by another program which is expected to have such features.

But, market place being what it is, adding all this bloat to NikonScan
had to be done because not everyone can afford, nor wants, a dedicated
image editing program.

Finally, in some aspects (dynamic histogram display, to name one)
NikonScan curves are actually *superior* even to Photoshop (at least
version 6 I have here)!

And, if curves are what you want, NikonScan has them, which can't be
said of VueScan...

If you reply that VueScan doesn't need curves because of "superior
color handling design". Well, that's like saying an Instamatic doesn't
need all the adjustments of a Hasselblad because of Instamatic's
"superior design"...

Both ICE and GEM
have proven to be invaluable tools. I make usually 48bit
*.tiff raw files using NikonScan, but processed by ICE and
GEM, and later I use Vuescan to generate whatever I like from
these raw files.

Let's just backtrack a little... Didn't you just say above that
NikonScan can *not* do raw scans?

Vuescan provides ...
...raw file scan ability and many many more.

None of this is available in NikonScan

It's also quite revealing that you apparently use NikonScan for quite
a few of its functions (ICE, GEM) considering your dislike for it.

Anyway, I really don't have much time for this. I know, hard to
believe, eh? But I have a brand new scanner sitting invitingly
right here next to me with that "come hither" look - so let's just
agree to disagree agreeably...

Kennedy and I, as well as many others, will continue to use NikonScan,
you and other VueScan fans will not - except to use ICE or GEM or...
;o)

That's all there is to it, so let's get back to subjects relevant to
both.

Don.

P.S. Please forgive and ignore my, in places, seemingly mocking tone.
I meant no disrespect but have no time to go back and edit it out...

I gotta go... LS-50 is calling me! ;o)

 

[Kennedy McEwen]

NikonScan is a simplistic
product. Search google archives to find numerous article written
on the topic here for many years. Me, I used Nikon scan in
classes to teach people how *not to make a gui*. Its truly an
exceptional collection of all mistakes and "crimes" against
ergonomics one can make.

What an incredibly arrogant piece of complete bunkum!

The NS interface is completely fine ergonomically and implements most of
the image processing functions, such as curves or levels, in exactly the
same manner as Photoshop, making it a virtually transparent interface
for anyone using that final image editing software package. That,
sadly, cannot be said for Vuescan.

 

[ThomasH]

With all due respect, but this should be interesting. Would you mind
outlining a few? (And which classes would that be, exactly?)

Sure: Lets show one classic.
Load film into the SA-30 and generate thumbnails. Even now with
the just released 4.02 update NikonScan places every of them a
bit to the left or maybe a bit to the right. Thus, you must click
on each of them to select one by one, move mouse over to the Tool
Palette "Scanner Extras", click on arrows, or select the text
input and enter offset, click "reload thumbnail." On my screen
its on average 7" mouse movement in one direction, 14" in both
directions. Times 36 for the roll of film, totals to 504", 42 feet.
You hand has to work for 42 feet to adjust thumbs in every each
roll of film. Scan 10 rolls, you move mouse for 420 feet! If you
click on a thumb to select it and happen to "double click," an
unwanted preview begins and it is difficult to interrupt.
Scanner will move film and perform physical scan. (Vuescan
can perform many operations based on one and the same
physical scans.)

Better possible way:
Now assume that left and right of each thumb is an area on
which you could click to shift it left or right.


Classes were directed to people developing aviation, space
and ground operator equipment and controls. We used research
in human cognition to provide screen elements and controls
known to be best recognizable and we analyze time needed to
complete tasks. We were investigating frequency of occurring
mistakes, of which some had fatal impact on safety, and
showing how make user interfaces which work. You might
find lots of useful information on this topic online.
Search for human factors in computing.

Especially when compared to VueScan's... ahem... "interesting"
so-called user interface. From confusing interaction of settings on
different tabs (did you say "ergonomics"?) and *annoyingly* flickering
screen (you did say "ergonomics", right?), to "random" redraws
*interrupting* user activity and constantly changing "menu du jour" as
options are being reshuffled (I could've sworn you mentioned
"ergonomics")...

Do not confuse bugs and deficiencies with gui design problems.
My solution to the "auto refresh" was always to disable it,
ever since Ed introduced it. I will not indulge in a debate
about this. Frankly, its Ed's task to take position to his
tool.

NikonScan is backed by an entire department and has still a
ton of bugs. For example, make a preview and shift the "Scale"
slider on the Crop tool and observe how your crop change size,
than jumps sideways to the upper left corner. You must go
back to the preview window and correct crop position.

Or try to use SA-30 (SA-21) with slide and than with negative.
Figure out how to change the settings and apply them correctly
on all thumbs.



Looking at the text below I can only repeat: search the news
archives, you will countless reports of problems with NikonScan
over the years. We left behind NikonScan 2 and 3... They were
worse than NS4.

Try to apply a film profile in NikonScan. Try to lock film color
base or image color to make a few images for stitching.

Last but least, stay with the tool you prefer. The bottom line
is: If you know it all better anyway, ...do not ask questions.

Thomas

 

[Kennedy McEwen]

With all due respect, but this should be interesting. Would you mind
outlining a few? (And which classes would that be, exactly?)

Sure: Lets show one classic.
Load film into the SA-30 and generate thumbnails. Even now with
the just released 4.02 update NikonScan places every of them a
bit to the left or maybe a bit to the right. Thus, you must click
on each of them to select one by one, move mouse over to the Tool
Palette "Scanner Extras", click on arrows, or select the text
input and enter offset, click "reload thumbnail." On my screen
its on average 7" mouse movement in one direction, 14" in both
directions. Times 36 for the roll of film, totals to 504", 42 feet.
You hand has to work for 42 feet to adjust thumbs in every each
roll of film. Scan 10 rolls, you move mouse for 420 feet! If you
click on a thumb to select it and happen to "double click," an
unwanted preview begins and it is difficult to interrupt.
Scanner will move film and perform physical scan. (Vuescan
can perform many operations based on one and the same
physical scans.)[/QUOTE]

Err, it sounds like its *you* that is having the problem Thomas. The
command you are using is for the adjustment of *individual* thumbnails
in the gate, not the global crop position. Of course, Vuescan doesn't
display thumbnails in any case, so the alternative that you are
comparing it with is simply not to bother with thumbnails at all.

If the shift on the film is as consistent as you state ("you must click
on each of them to select one by one..") then you should adjust the
default crop position once and for all. The gate is deliberately
oversized in the film strip direction to permit this and the process to
implement it is intuitively obvious for any Windows user.

However, in my experience (and also as indicated by the examples on your
web site) the thumbnail positioning error of the previous version was
only restricted to the first and last couple of frames of any roll at
most, so you *would* need to adjust these individually, irrespective of
the user interface. This appears to have been fixed now though, as far
as my tests show.

 

[Don]

Sure: Lets show one classic.
Load film into the SA-30 and generate thumbnails. Even now with
the just released 4.02 update NikonScan places every of them a
bit to the left or maybe a bit to the right.

What does that *unintentional* effect have to do with the
*intentional* ergonomics of GUI design*!?

Classes were directed to people developing aviation, space
and ground operator equipment and controls. We used research
in human cognition to provide screen elements and controls
known to be best recognizable and we analyze time needed to
complete tasks. We were investigating frequency of occurring
mistakes, of which some had fatal impact on safety, and
showing how make user interfaces which work.

Which - besides, all of a sudden, starting to make me queasy about
flying - also brings us back to your original statement. Again, what
parts of NikonScan *GUI design ergonomics* can you point to as
"criminal"?

Do not confuse bugs and deficiencies with gui design problems.

Funny you should say that when your example of a GUI design problem
above is a programming bug!? (A hint: a bug is unintentional,
ergonomics are intentional.)

All the VueScan problems outlined above *are* GUI design problems and
fly in the face of *all* ergonomics guidelines:

The screen flashing is not a programming bug but an intentional
function. The author *wanted* the screen to flash.
Incomprehensible option interactions are not programming bugs but
*documented* "features". The author *wanted* this incomprehensible
interaction and documented it (well, he attempted to...).
Ever-changing menus which interrupt user input are not a programming
bug but (misguided) "user feedback". The author *wanted* constantly
changing menus.

With all that in mind, I have a sneaky feeling that your professed
expertise in GUI design and ergonomics - how shall I put it politely -
may not be as extensive as you would lead us to believe. Especially if
you can't even tell (unintentional) programming bugs from
(intentional) ergonomics design decisions.

Since you seem to be so fond of Google as a replacement for professed
knowledge may I be so bold as to, respectfully, suggest you look up
the definition of "ergonomics"?

Don.

 

[ThomasH]

Me, I used Nikon scan in
classes to teach people how *not to make a gui*. Its truly an
exceptional collection of all mistakes and "crimes" against
ergonomics one can make.

With all due respect, but this should be interesting. Would you mind
outlining a few? (And which classes would that be, exactly?)

Sure: Lets show one classic.
Load film into the SA-30 and generate thumbnails. Even now with
the just released 4.02 update NikonScan places every of them a
bit to the left or maybe a bit to the right. Thus, you must click
on each of them to select one by one, move mouse over to the Tool
Palette "Scanner Extras", click on arrows, or select the text
input and enter offset, click "reload thumbnail." On my screen
its on average 7" mouse movement in one direction, 14" in both
directions. Times 36 for the roll of film, totals to 504", 42 feet.
You hand has to work for 42 feet to adjust thumbs in every each
roll of film. Scan 10 rolls, you move mouse for 420 feet! If you
click on a thumb to select it and happen to "double click," an
unwanted preview begins and it is difficult to interrupt.
Scanner will move film and perform physical scan. (Vuescan
can perform many operations based on one and the same
physical scans.)

Err, it sounds like its *you* that is having the problem Thomas. The
command you are using is for the adjustment of *individual* thumbnails
in the gate, not the global crop position. Of course, Vuescan doesn't
display thumbnails in any case, so the alternative that you are
comparing it with is simply not to bother with thumbnails at all.

If the shift on the film is as consistent as you state ("you must click
on each of them to select one by one..") then you should adjust the[/QUOTE]

It is *inconsistent*, every each is different. If you do not do
that, your raw files will cut off parts of the image and store
a black strip on one of the sides.

default crop position once and for all. The gate is deliberately
oversized in the film strip direction to permit this and the process to
implement it is intuitively obvious for any Windows user.

I have completely no idea what are you talking about. Could
you refer to "The Gate" in the NikonScan Manual?

However, in my experience (and also as indicated by the examples on your
web site) the thumbnail positioning error of the previous version was
only restricted to the first and last couple of frames of any roll at
most, so you *would* need to adjust these individually, irrespective of
the user interface. This appears to have been fixed now though, as far
as my tests show.

Oh no, I will post another gallery with misfits.

Thomas