Averaging multiple scans

[David Blanchard]

You will consistently find Nikon scanners reproduce KC darker than an
equivalent EC slide than other manufacturers ...[snip[ ... However none
of these are quite as tolerant of film source as, say, the Minolta or
Canon ranges which do not even require separate settings for KC film
to yield similar results to EC or other related emulsions.

I've been following this (and other NikonScan/Kodachrome) threads with
interest and am curious about your comment concerning Nikon vs. other
manufacturers regarding KC. Is it because of the LED light source in
the Nikon scanners that causes the problem? Or something else?

Considering that >98% of my nearly 2500 slides are KC, maybe I should
trade in my LS-2000 and buy something else.

David

 

[Don]

You will consistently find Nikon scanners reproduce KC darker than an
equivalent EC slide than other manufacturers ...[snip[ ... However none
of these are quite as tolerant of film source as, say, the Minolta or
Canon ranges which do not even require separate settings for KC film
to yield similar results to EC or other related emulsions.

I've been following this (and other NikonScan/Kodachrome) threads with
interest and am curious about your comment concerning Nikon vs. other
manufacturers regarding KC. Is it because of the LED light source in
the Nikon scanners that causes the problem? Or something else?

Considering that >98% of my nearly 2500 slides are KC, maybe I should
trade in my LS-2000 and buy something else.

Yes, it's the LEDs. Kennedy will do doubt jump in to tell you about
their spectral characteristics which cause this, and he's much better
at it so I'll leave it to him... ;o)

However, there are advantages to LEDs too. For example an LED "never"
burns out, but more importantly their brightness remain constant while
other light sources dim with time. That's what sold me on my LS-30
when I bought it all those years ago. As far as I know no other
scanner manufacturer uses LEDs as the light source.

So, even though LS-2000 suffers from the "Kodachrome complex" once you
have scanned in all your Kodachrome stock it's not likely you'll be
shooting any more. And for negatives, Nikon scanners work like magic.

Finally, unlike my LS-30 your LS-2000 even has single-pass
multiscanning. So with some extra work (far less than me) you can
wrestle those Kodachromes to the ground.

Don.

 

[Kennedy McEwen]

You will consistently find Nikon scanners reproduce KC darker than an
equivalent EC slide than other manufacturers ...[snip[ ... However none
of these are quite as tolerant of film source as, say, the Minolta or
Canon ranges which do not even require separate settings for KC film
to yield similar results to EC or other related emulsions.

I've been following this (and other NikonScan/Kodachrome) threads with
interest and am curious about your comment concerning Nikon vs. other
manufacturers regarding KC. Is it because of the LED light source in
the Nikon scanners that causes the problem? Or something else?

Pretty much. Ideally, if you want a sensor to provide the same output
from a range of colours that is similar to what you see by eye then the
spectral response of the sensor has to match that of the eye. The
further that you depart from this ideal, the more you become susceptible
to a type of metamerism, where the colours appear different to the
sensor from how you view them. This is similar to the more commonly
encountered metamerism where a material can appear to take on a
different colour under alternate types of illumination, just due to the
spectral content of the light. My car, for example, looks a very deep
turquoise under direct sunlight but when viewed under normal white
streetlamps (eg. mercury or high pressure sodium) appears a very deep,
dark blue. The colour you see is an interaction of the spectral
reflectivity of the dye, the spectral content of the light source and
the spectral response of the sensors in the eye. Free tip - if you are
buying a used car, its a good idea to have a look at it under street
lighting - colour mismatch easily shows replaced and workshop sprayed
panels that look perfectly matched under daylight. ;-)

It turns out that you can actually depart from the ideal sensor spectral
response quite a long way and still get very good colour representations
- fortunately, or most colour sensors just would not work at all.
However this is only possible because many of the subjects we view have
a broad spectral reflectivity. If the spectral responsivity of the
chemicals is plotted on a graph then you would see it is full of very
fine structure with peaks and troughs. However, provided that the
response of the sensor is wide enough to cover many of the peaks and
troughs, the general trend can be assessed and, consequently, the output
from each colour sensor type can be adjusted to match the average
response that the eye would see.

If however, you view the subject with a sensor which has a very narrow
spectral range (or use a narrow spectral range illumination source under
which to view it) then it really depends on whether that sensor's
response coincides with a peak or a trough in the spectral response of
the dye which determines how dense each colour it appears - and how much
correction is required to make it appear as it would to a broad spectral
response eye. If you change the dye form one type to another, even
though it appears to have exactly the same density to the eye, the
narrow band system will probably see a significantly different density.
Its just the same as the colour of the paint on my car - the spectral
reflectivity doesn't have any peaks which correspond to the green
emissions in mercury vapour or high pressure sodium street lamps - just
a little bit of blue. So it looks like a completely different hue and
density.

That is why older Nikon scanners tend to have a problem with Kodachrome
emulsion - the dyes used in KC are different from the dyes used in
Extachrome and compatible emulsions. That doesn't mean they don't work
with Kodachrome - it just means they have not been set up with the
correct balance of exposures to get a matched colour response with
Kodachrome dyes. Newer Nikon scanners (LS-40+) have a newer LED system
and a specific Kodachrome balance, and work much better than the older
units.

Considering that >98% of my nearly 2500 slides are KC, maybe I should
trade in my LS-2000 and buy something else.

Frying pan and fire really. The LED illumination has this relatively
minor weakness that changing dye types risks metamerism, and with older
scanners you are on your own when it comes to colour balance with
Kodachrome. However, for exactly the same reason as you encounter this
problem (the narrow band of wavelengths used in illumination) you also
get excellent colour purity that is far better than could possibly be
achieved with a filtered CCD. So the overall result is that, once you
have balanced the system for your media, you get much purer, cleaner,
more saturated colours than you could achieve by any alternative means.
It is no coincidence that Nikon scanners are always praised for their
colour saturation.

Then you have the stability of the illumination source - LEDs are just
like most semiconductor devices and have a "bathtub" statistical life. A
proportion of them fail early in life, before they even get incorporated
into the scanner, but those that survive keep the same performance for
most of their life before eventually just fading out due to contaminants
and carrier mobility. Compare that to a fluorescent (cold cathode)
lamp, which changes its illumination level as soon as it is turned on
for the first time, due to deposition on the phosphors, and filters,
which are merely dyes etch resist, that fade with age and light. And
the LED life, if the heat is appropriately removed from the junction,
can be tens or hundreds of years (operational life, not just
chronological life). Compare that to the two threads currently on this
newsgroup about replacement lamps for scanners, one after 2500 scans,
the other after 3500! I would not go so far as to suggest that these
are typical - these lamps are also more susceptible to mechanical damage
than solid state LEDs, so the units may have experienced some trauma in
transit - and there would be many more than two threads running if it
were, but it is a concern.

Then there is the calibration issue. A single CCD with three light
sources is just much easier to match the response than three CCDs and a
white source which changes spectral output over time and temperature.
Residual nonuniformity (which exists on all CCD devices, its just a
matter of how good or bad it is) with the Nikon will never produce
colour banding.

So, whilst there are certainly better scanners on the market now, some
of them Nikon some not, than the LS-2000 I would not recommend ditching
your LS-2000 just because your film is mainly Kodachrome. If you are
getting on fine with it now - and you can save preferred settings for
the exposure and adjustments in the Nikonscan software - then there is
no reason to assume that it will present you with problems part way
through your archive. The same cannot be guaranteed for
an alternative scanner that you might consider "trading up" to.

 

[David Blanchard]

On 1 Mar 2004 02:15:12 GMT, (e-mail address removed) (David Blanchard)
wrote:

I've been following this (and other NikonScan/Kodachrome) threads with
interest and am curious about your comment concerning Nikon vs. other
manufacturers regarding KC. Is it because of the LED light source in
the Nikon scanners that causes the problem? Or something else?

Considering that >98% of my nearly 2500 slides are KC, maybe I should
trade in my LS-2000 and buy something else.

Yes, it's the LEDs. Kennedy will do doubt jump in to tell you about
their spectral characteristics which cause this, and he's much better
at it so I'll leave it to him... ;o)
[...snip...]

So, even though LS-2000 suffers from the "Kodachrome complex" once you
have scanned in all your Kodachrome stock it's not likely you'll be
shooting any more. And for negatives, Nikon scanners work like magic.

Not true in my case. I still shoot a great deal of Kodachrome 64.
Old habits die hard...

-db-

 

[David Blanchard]

Pretty much. Ideally, if you want a sensor to provide the same output
from a range of colours that is similar to what you see by eye then the
spectral response of the sensor has to match that of the eye. The

[...snip...]

Thanks for the very informative discussion on sensitivity, dyes, and
more. Very "enlightening."

Frying pan and fire really.

Another great discussion on LED vs. other types of light sources.

So, whilst there are certainly better scanners on the market now, some
of them Nikon some not, than the LS-2000 I would not recommend ditching
your LS-2000 just because your film is mainly Kodachrome. If you are
getting on fine with it now - and you can save preferred settings for
the exposure and adjustments in the Nikonscan software - then there is
no reason to assume that it will present you with problems part way
through your archive. The same cannot be guaranteed for
an alternative scanner that you might consider "trading up" to.

Between NikonScan 3.1.x and Vuescan 7.x, I have been able to scan
almost all of my selected Kodachrome slides with respectable
results. As I use both pieces of software, I get better at the whole
scanning process and will often go back and rescan slides done earlier
in the learning curve. So far, there have been only a handful of
slides that defy quality scanning. I'm still working on those and
do not plan to give up.

Thanks for the informative discussion.

-db-

 

[Don]

[...snip...]

So, even though LS-2000 suffers from the "Kodachrome complex" once you
have scanned in all your Kodachrome stock it's not likely you'll be
shooting any more. And for negatives, Nikon scanners work like magic.

Not true in my case. I still shoot a great deal of Kodachrome 64.
Old habits die hard...

In that case, if you can afford it, you should definitely look into
newer Nikon models (if you want to stick with Nikon). From what I
understand, they don't do Kodachrome perfectly (there still seem a
slight blue cast) but a whole lot better (no dark scans anymore).

Don.

 

[Stephen Rogers]

On 2 Mar 2004 03:47:02 GMT, (e-mail address removed) (David Blanchard)
wrote:

In that case, if you can afford it, you should definitely look into
newer Nikon models (if you want to stick with Nikon). From what I
understand, they don't do Kodachrome perfectly (there still seem a
slight blue cast) but a whole lot better (no dark scans anymore).

No doubt Kennedy is right (of course) when he says the newer Nikons
have been improved in this respect. I have a Nikon Coolscan V ED and
it handles Kodachrome superbly. There is a "but" however. The blue
cast seems entirely due to the NikonScan 4 software and absolutely not
due to the scanner itself. If I use Nikon Scan to scan Kodachrome, I
get vastly better results by setting its film type to Positive rather
than Kodachrome and setting analogue gain as follows:
R 0.5
G -0.33
B -0.6

Master anywhere between -0.3 and +0.3 depending on whether highlight
or shadow is more important.

Those analogue gain settings virtually eliminate the blue cast. Using
the Positive setting instead of Kodachrome also has the advantage that
digital Ice actually works. (Nikon seem to have got round the
"problems" of Ice on Kodachrome by making the normal Ice setting do
almost nothing on Kodachrome. Whereas on Positive it seems to deal
with Kodachrome rather well.)

Having said that the analogue gain almost gets rid of the blue cast, I
can also say that VueScan produces a pretty well perfect colour
balance on Kodachrome (much better than NikonScan), using the VueScan
Kodachrome setting. By means of that, I can produce scans from
Kodachrome, straight out of VueScan, that look on my monitor exactly
like the slides do in my slide projector, with no blue colour cast
whatever. (Sometimes a slight red tinge in highlights, easily dealt
with in an image editor). (They print well too - also with pretty well
perfect colour balance relative to the projected slide). And the
VueScan "light" IR clean is even better than Ice.

As a matter of interest, if I use my analogue gain settings in
NikonScan, apply a gamma of 0.46 (=1/2.2) with its curves tool, and
save the output as a 48 bit RGB file, I can then have VueScan read
that in as a raw file. The result is somewhat better colour balance
than Nikon Scan manages by itself but still not as good as VueScan
manages by itself.

So I would say the Coolscan itself is close to perfect, especially as
it seems to generate almost no noise in dark areas of slides (unless
you go looking for it by steepening the curve on the low luminosity
part of the image and view at full resolution. Then you sometimes see
a tiny amount of noise but so what?).

Thus, provided you use VueScan and steer clear of NikonScan, colour
balance on Kodachrome should be no problem at all with the Coolscan V
ED.

 

[Don]

The blue
cast seems entirely due to the NikonScan 4 software and absolutely not
due to the scanner itself.

Could it be the Nikon Color Management (NCM)? I find it adds *tons* of
blue when scanning on my LS-30. With NCM off I still get a major blue
cast but considerably less than with NCM on.

If I use Nikon Scan to scan Kodachrome, I
get vastly better results by setting its film type to Positive rather
than Kodachrome and setting analogue gain as follows:
R 0.5
G -0.33
B -0.6

Master anywhere between -0.3 and +0.3 depending on whether highlight
or shadow is more important.

I have no choice but to use Positive as there is no Kodachrome setting
when using LS-30 and NikonScan (NS).

I'm still trying to figure out empirically (read: guess...) what
Analog Gain (AG) values to use and my figures so far are:
R +1
G +0.2
B 0
As you say Master AG must also be adjusted and in my case it's
anywhere between 1 and 2. Sometimes even that is not enough so I have
to boost individual channels by a fixed factor (e.g. R +2, G +1.2, B
+1).

What really ticks me off big time is that Nikon is unwilling to
provide this value instead of us guessing. Kennedy thinks that is
impossible and we went around the block on this, but the more I test
the more I'm convinced a generic compensation is possible. The only
variable after that would be the amount of Master AG boost.

Before I started playing with removing the cast with AG I settled on
the following curve settings in Photoshop to eliminate the blue cast
(one point per curve):

RGB: flat
R: Input=32, Output=40
G: flag
B: Input=32, Output=17

Of course, the brightness must also be boosted and I settled for:
RGB: Input=32, Output=52

The problem with this approach (other than it's all guesswork) is that
the original image data is of very limited range. Because of that I'd
like to do the cast removal at the AG stage and boost the range.

As a matter of interest, if I use my analogue gain settings in
NikonScan, apply a gamma of 0.46 (=1/2.2) with its curves tool, and
save the output as a 48 bit RGB file,

Kennedy will probably say that the two gamma conversions (from 1 to
2.2 and back) leads to some loss... ;-)

I can then have VueScan read
that in as a raw file. The result is somewhat better colour balance
than Nikon Scan manages by itself but still not as good as VueScan
manages by itself.

I'm currently giving VueScan (VS) a second look (very mixed feelings
so far, some major bugaboos) and I have tried something similar. I
used NS to scan an image with gamma at 1 and then feed that to VS to
test the color balance.

The problem there is that NS refuses to extract 10 bits but I end up
with an 8 bit image, so I'm forced to up this is Photoshop (PS) and
save as 48 bit RGB. Still, some improvement is possible when such a
file is imported into VS. The major problem is the lack of the IR
channel.

One big advantage of VS is its ability (willingness, actually) to
extract full 10 bits out of the LS-30 and, of course, the IR channel.
But this advantage is totally obliterated by VS inability to adjust AG
of individual channels. Unforgivable!

Thus, provided you use VueScan and steer clear of NikonScan, colour
balance on Kodachrome should be no problem at all with the Coolscan V
ED.

LS-30 on the other hand, is a different story... )

Don.

 

[Bart van der Wolf]

SNIP

One big advantage of VS is its ability (willingness, actually)
to extract full 10 bits out of the LS-30 and, of course, the IR
channel. But this advantage is totally obliterated by VS
inability to adjust AG of individual channels. Unforgivable!

As I said earlier, I'd prefer more user control there as well, however...
If you optimize exposure, the VS "Advanced Workflow" will offer you
unclipped data for the most transparent channel (Blue in your case) and,
given the 10-bit ADC output, 6-8 bits of excess room at the dark end in the
Raw data for that channel. The Green channel will have some headroom left
before it clips and some 6-7 bits unused room, Red will have 1-bit of head
room and 5-7 bits of excess room, at the dark end. You can check the playing
field on the top histogram or when you select the Color balance option
"None", because it doesn't apply Black/White point adjustments. You can use
"Pixel Colors" to check for clipping.

Also, adjusting an overall colorbalance is NOT performed by post-processing
the channel Brightness or Gamma! It is done by adjusting the channel (Black
or) White point. VueScan does that correctly if you Right Mouse Button-click
on a known neutral tone in the image. Residual midtone corrections can be
done by varying the channel Brightnesses, or more accurately in Photoshop
with Curves.

Only if the film suffered significant damage by adverse storage conditions,
a better result can potentially be expected by fine-tuning the channel
exposures, but that depends on the actual amounts of dye bleaching.

I still would prefer having the full amount of channel exposure control,
like you do, but not entirely for the same reasons. Part of your color issue
comes, as Kennedy McEwen explained, from the absorption characteristics of
the Kodachrome dyes at the emission wavelengths of the LEDs. The narrowband
LED emission will have difficulty matching the eye sensitivity curves for
some particular dye absorptions, there is no way around that. However, even
if that means a less than optimal color accuracy for some hues, it should
still allow a reasonable overall colorbalance (although it may require a bit
more tweaking than with other film dyes).

Bart

 

[Kennedy McEwen]

[...snip...]

So, even though LS-2000 suffers from the "Kodachrome complex" once you
have scanned in all your Kodachrome stock it's not likely you'll be
shooting any more. And for negatives, Nikon scanners work like magic.

Not true in my case. I still shoot a great deal of Kodachrome 64.
Old habits die hard...

In that case, if you can afford it, you should definitely look into
newer Nikon models (if you want to stick with Nikon). From what I
understand, they don't do Kodachrome perfectly (there still seem a
slight blue cast)

....when scanned on the same setting as conventional positive slide
films. From the images I sent you, you can see that the image with the
slight blue cast is the "Positive" scan - the "Kodachrome" scan is
almost spot on colour balance wise.

 

[Kennedy McEwen]

No doubt Kennedy is right (of course) when he says the newer Nikons
have been improved in this respect. I have a Nikon Coolscan V ED and
it handles Kodachrome superbly. There is a "but" however. The blue
cast seems entirely due to the NikonScan 4 software and absolutely not
due to the scanner itself. If I use Nikon Scan to scan Kodachrome, I
get vastly better results by setting its film type to Positive rather
than Kodachrome and setting analogue gain as follows:
R 0.5
G -0.33
B -0.6

But the reason you have to make these adjustments at all is because the
scanner hardware is responding slightly differently to the Kodachrome
dyes - you are compensating for the difference between the transmission
in the narrow LED band between E-6 and KC dyes. It is less difference
than the earlier scanners, but a relatively minor difference is still
there.

When scanned under the Kodachrome setting of either NS3.1.2 or NS4.01, I
do not find any blue cast at all - and ICE works pretty well on all of
the KC slides I have. In fact I would say that, with the last two
iterations of scanner & software design, Nikon have pretty much cracked
the Kodachrome issue completely. It is only a problem for folks like
DOn who are struggling with older hardware, even if they can access the
newer software.

 

[Stephen Rogers]

Could it be the Nikon Color Management (NCM)? I find it adds *tons* of
blue when scanning on my LS-30. With NCM off I still get a major blue
cast but considerably less than with NCM on.

I never tried turning off NCM as I stopped experimenting with Nikon
Scan, Vuescan giving me such consistently good results. Looks likely
though!

I have no choice but to use Positive as there is no Kodachrome setting
when using LS-30 and NikonScan (NS).

I'm still trying to figure out empirically (read: guess...) what
Analog Gain (AG) values to use and my figures so far are:
R +1
G +0.2
B 0

Since AG is a logarithmic factor, what is important is the
arithmetical difference between the R, G and B values. So if I add 0.6
to all of my values to bring mine to a reference level of B=0, my
values would be
R +1.1
G + 0.27
B 0

Considering that our tests are on different Nikon models and your
Kodachromes are not my Kodachromes, I think our AG values are
remarkably similar.

As you say Master AG must also be adjusted and in my case it's
anywhere between 1 and 2. Sometimes even that is not enough so I have
to boost individual channels by a fixed factor (e.g. R +2, G +1.2, B
+1).

Since it is the arithemtic differences between RGB that are important,
I chose my individual AGs so that the master AG correction would lie
symmetrically about zero and at any rate within the limits set by
Nikon Scan (of -2 to +2). You could do the same thing. If you were add
your "boost" of 1, you would get individual AG of
R +2
G 1.2
B +1
as you say, and then have to apply a Master AG between 0 and 1. You
would then have some adjustment flexibility left.

impossible and we went around the block on this, but the more I test
the more I'm convinced a generic compensation is possible. The only
variable after that would be the amount of Master AG boost.

I think your view is confirmed by the similarity of your AGs to mine,
given that we have different scanners.

Before I started playing with removing the cast with AG I settled on
the following curve settings in Photoshop to eliminate the blue cast
(one point per curve):

I had tried similar curve settings but found it an irritation to do it
that way.

I'm currently giving VueScan (VS) a second look (very mixed feelings
so far, some major bugaboos) and I have tried something similar. I
used NS to scan an image with gamma at 1 and then feed that to VS to
test the color balance.

My only reason for having applied the 1/2.2 gamma was that the raw
files that Vuescan creates for itself are "dark" because it does not
apply the 2.2 gamma to the raw data before saving the raw file. So I
thought I would have the best chance of fooling it into thinking the
Nikon file was one of its own raw files if I first applied the 1/2.2
factor to the Nikon 48 bit file.

One big advantage of VS is its ability (willingness, actually) to
extract full 10 bits out of the LS-30 and, of course, the IR channel.

And, with the Coolscan V, its other big advantage is that it manages
virtually perfect colour balance -

But this advantage is totally obliterated by VS inability to adjust AG
of individual channels. Unforgivable!>

Even if Ed Hamrick is convinced we don't *need* the individual channel
exposure, I can't see the objection to making it available. Like you,
I would prefer to have it. It would at least give the user more
control even if the user then elected to allow Vuescan to calculate
the exposures anyway. After all, if we only did what we *need*, we
probably would not even be taking photos, let alone scanning them!

 

[Don]

As I said earlier, I'd prefer more user control there as well, however...
If you optimize exposure, the VS "Advanced Workflow" will offer you
unclipped data for the most transparent channel (Blue in your case) and,
given the 10-bit ADC output, 6-8 bits of excess room at the dark end in the
Raw data for that channel. The Green channel will have some headroom left
before it clips and some 6-7 bits unused room, Red will have 1-bit of head
room and 5-7 bits of excess room, at the dark end.

The problem is that the gap between blue and red is so large that
individual control is essential. Obviously I would boost master gain
until blue starts clipping but that only goes so far.

Also, adjusting an overall colorbalance is NOT performed by post-processing
the channel Brightness or Gamma!

You misunderstood. I did not suggest it is. If a raw file is used as
input VS expects this RAW file to have a gamma of 1 (at least that's
what is says in the User Guide). That's all the gamma reference meant.

It is done by adjusting the channel (Black
or) White point. VueScan does that correctly if you Right Mouse Button-click
on a known neutral tone in the image.

At least two problems with that. First, it assumes there is a neutral
tone in the image. Second, it assumes that VS does it correctly. My
current test slide proves both of those assumptions wrong. The scan
produced by VS (Media type: Slide film, for consistent color) is a
caricature of the original image. It looks like a false color image,
not a genuine scan. Changing media type to: Image improves things as
far as exposure and tone goes but, of course, it adds the heavy blue
cast and is dark. Back to square one...

I still would prefer having the full amount of channel exposure control,
like you do, but not entirely for the same reasons. Part of your color issue
comes, as Kennedy McEwen explained, from the absorption characteristics of
the Kodachrome dyes at the emission wavelengths of the LEDs. The narrowband
LED emission will have difficulty matching the eye sensitivity curves for
some particular dye absorptions, there is no way around that.

The source of the color problem is really not relevant as both Stephen
and I have proven empirically that a generic setting is indeed
possible. The problem there is that this is only guesswork, and I'd
like recommended values, which Nikon refuses to provide. But that's
another issue and we're not going to go over that again, are we...?
;o)

Don.

 

[Don]

Since AG is a logarithmic factor, what is important is the
arithmetical difference between the R, G and B values. So if I add 0.6
to all of my values to bring mine to a reference level of B=0, my
values would be
R +1.1
G + 0.27
B 0

Considering that our tests are on different Nikon models and your
Kodachromes are not my Kodachromes, I think our AG values are
remarkably similar.

I know! It's too much of a similarity to be just a coincidence.

Since it is the arithemtic differences between RGB that are important,
I chose my individual AGs so that the master AG correction would lie
symmetrically about zero and at any rate within the limits set by
Nikon Scan (of -2 to +2). You could do the same thing. If you were add
your "boost" of 1, you would get individual AG of
R +2
G 1.2
B +1
as you say, and then have to apply a Master AG between 0 and 1. You
would then have some adjustment flexibility left.

Great minds think alike! ;o) That's exactly my train of thought and
what I have been struggling with for months: A reliable and repeatable
method to be applied consistently i.e., eliminate as many variables as
possible up front instead having them percolate and get worse, as well
as, fixing things at the earliest possible stage.

I think your view is confirmed by the similarity of your AGs to mine,
given that we have different scanners.

My thoughts exactly! But until now I have been a lone voice...

To be fair, both Kennedy and I may be right. In theory there may be
differences but they may be simply too small to affect the settings.

I had tried similar curve settings but found it an irritation to do it
that way.

Bingo! Not only that but using curves in post-processing is further
handicapped by a narrow dynamic range (limited by blue clipping way
before red even comes into play). That's what lead me to try to modify
AG directly and do the corrections at the earliest possible stage.

My only reason for having applied the 1/2.2 gamma was that the raw
files that Vuescan creates for itself are "dark" because it does not
apply the 2.2 gamma to the raw data before saving the raw file. So I
thought I would have the best chance of fooling it into thinking the
Nikon file was one of its own raw files if I first applied the 1/2.2
factor to the Nikon 48 bit file.

In that particular test I tried to skip that step by setting gamma in
NS to 1. That way NS (or the scanner's firmware?) does not do the 1 to
2.2 conversion, which then must be "un-done" with 1/2.2.

And, with the Coolscan V, its other big advantage is that it manages
virtually perfect colour balance

Lucky you! ;-)

Even if Ed Hamrick is convinced we don't *need* the individual channel
exposure, I can't see the objection to making it available. Like you,
I would prefer to have it. It would at least give the user more
control even if the user then elected to allow Vuescan to calculate
the exposures anyway. After all, if we only did what we *need*, we
probably would not even be taking photos, let alone scanning them!

Hear! Hear! To paraphrase what I write in another thread: "How can
choice (i.e. flexibility) be bad?".

Don.

 

[Don]

...when scanned on the same setting as conventional positive slide
films. From the images I sent you, you can see that the image with the
slight blue cast is the "Positive" scan - the "Kodachrome" scan is
almost spot on colour balance wise.

In that particular case I was referring to the Kodachrome scan of my
test slide sent to me by Nikon after months of wrestling (they never
sent me a Positive scan, though, even after repeated requests!).

Your Kodachrome scan, indeed, did not have that problem but - unlike
Nikon - you know what you're doing... ;o)

Don.

 

[Kennedy McEwen]

The source of the color problem is really not relevant as both Stephen
and I have proven empirically that a generic setting is indeed
possible.

How can you say that Don when, from what you have said, you have never
managed to adequately get rid of the blue cast and Stephen is using a
scanner with 16x as much dynamic range! You can't judge how much the
colour is off with the blue cast of the level your demo scans indicate
being present, only guess!

The problem there is that this is only guesswork

There you are!

, and I'd
like recommended values, which Nikon refuses to provide.

For good reason.

 

[Don]

How can you say that Don when, from what you have said, you have never
managed to adequately get rid of the blue cast

That's not entirely correct. I've managed to get rid of the blue cast,
with both the Analog Gain and before that with curves using values I
mentioned last time - as far as I can see... And that's the key. It's
a subjective judgment, while I wanted (and still do) an objective
method.

and Stephen is using a
scanner with 16x as much dynamic range! You can't judge how much the
colour is off with the blue cast of the level your demo scans indicate
being present, only guess!

I said that. That's why I want an objective, reliable and repeatable
method.

There you are!

See... ;-)

For good reason.

Marketing? ;-) Juuuust kidding! We don't want to get into that again.

But how do you account that both Stephen and I arrived at very similar
values even though we have different hardware and, presumably,
different LEDs as well as different film? I don't think all that can
be dismissed as mere coincidence.

Don.

 

[Kennedy McEwen]

But how do you account that both Stephen and I arrived at very similar
values even though we have different hardware and, presumably,
different LEDs as well as different film? I don't think all that can
be dismissed as mere coincidence.

I don't just dismiss it as coincidence Don, I look at the differences
and similarities and recognise them. The differences are far more
significant than either you or Stephen recognise. In contrast, the
similarities are, to be frank, based on a statement by Stephen which
does not appear to be verified by others using the same or similar
hardware. I refer you here not only to the scans I sent you, but to
those posted by Rafe and others in the earlier thread. As such, the
similarities *can* be little more than coincidence.

Similar values: well certainly, if you ignore half of them then they
might appear to be reasonably similar, but try them and you should
notice the difference - I just have and it is clear enough visually,
without needing to examine the histograms. That was just examining the
difference in balance between the RGB channels of 0.1:0.07:0.

However there are major differences between your settings and Stephen's
on overall level. Specifically, Stephen quotes settings with a master
AG of between -0.3 and +0.3 to suit the result, presumably based on some
standard view of the histogram limits. Lets say that is, on average, a
master AG of 0. Meanwhile you report values for the colour channels
which are, on average, 0.543 higher than Stephen's *and* a master AG
which is "between 1 and 2", so lets say 1.5 on average. That computes
to an average difference of master gain in excess of 2 - more than the
range that either scanner could adjust on its own under NS! The only
way Stephen's channel balance looks similar to yours is by adding 0.6 to
each channel - which means his master AG would range from -0.3 to -0.9
to compensate, with an average of -0.6.

Even the minimum difference between the two settings is a very
significant 1.3EV. You must have seen a few chromes that were
over/underexposed by a stop and a third, do you think they even look
acceptable, let alone almost right? With that magnitude of difference
in overall level, would you call the colour balance (even ignoring the
visible differences) anything more than coincidence?

However, I have a more fundamental issue with Stephen's settings and his
reasons for needing them at all, which I have already addressed. Stephen
considers the blue cast to appear on his LS-50 as a consequence of the
Kodachrome setting and NCM. I have already sent you images from a KC
scan with and without NCM and, as stated previously, the "blue cast"
does not exist - it is present only in the image of KC scanned with
"positive" setting (on which NCM was off, by your request). Furthermore,
Stephen's settings on my LS-4000 (almost identical internally to his
LS-50) result in an excessively warm cast to the image and, more
concerningly, substantially unpopulated areas of histogram, which are
certainly not acceptable.

Since I get similar results to those posted by others, referenced
previously, using similar equipment I don't dismiss the similarities as
coincidence at all, I dismiss one set of results as questionable.

 

[Don]

Similar values: well certainly, if you ignore half of them then they
might appear to be reasonably similar, but try them and you should
notice the difference - I just have and it is clear enough visually,
without needing to examine the histograms. That was just examining the
difference in balance between the RGB channels of 0.1:0.07:0.

To be fair my results were only preliminary ("so far" as I put it).
But even if we take them at face value, I would not consider above
differences to be large *if* we compare them to the difference between
a raw scan (with KC cast) and a corrected scan (without KC cast).

In other words, I'd be perfectly happy with a statement (if it were
*authoritative* and not based on guesswork) that the red channel needs
to be boosted by between 1 and 1.1 AG. I'd just split the difference
and consistently apply 1.05 across the board.

Why is that not such a big deal (for me)? Because even after applying
this generic KC correction, all that does is remove the (scanner
introduced) KC cast. In other words, all it does is it gets back to
"status quo ante". And since no image is perfect at that stage some
further correction will inevitably be necessary.

And the difference between the final image and raw image (without KC
cast) is far greater than then difference between my and Stephen's
settings.

To avoid confusion, or maybe increase it ... ;-)
Image 1 = scan with flat AG (with KC cast)
Image 2 = scan with Stephen's AG correction (without KC cast)
Image 3 = scan with my AG correction (without KC cast)
Image 4 = Image 2 or 3 with regular corrections, levels, balance, etc.

What I'm saying above is that the difference between Image 4 and
either Image 2 or 3 is far greater than the difference between my
(Image 2) and Stephen's (Image 3) settings. In other words, the
difference between my and Stephen's settings gets lost in the final
image correction.

And yet, the difference between Image 4 and either Image 2 or 3 is
still far less than the difference between either Image 2 or 3, and
Image 1! In other words, such is the magnitude of the KC cast.

Even the minimum difference between the two settings is a very
significant 1.3EV. You must have seen a few chromes that were
over/underexposed by a stop and a third, do you think they even look
acceptable, let alone almost right? With that magnitude of difference
in overall level, would you call the colour balance (even ignoring the
visible differences) anything more than coincidence?

You're assuming that my slides are considerably more underexposed than
Stephen's (or ~ vice versa) and you're basing this on my scanner's
response, which I don't think can be taken as an objective baseline.

Even without seeing his slides, I'd say that's very unlikely, simply
for practical reasons. So let's assume that (on average) most of the
time both Stephen and I know how to shoot slides and take that out of
the equation.

So how do I explain the difference? Well, my scanner, of course!

However, and this is the key, even with this significant 1.3EV
difference in absolute terms, we still came up with similar KC cast AG
correction values in relative terms! That (to me) is very important.

Stephen's settings on my LS-4000 (almost identical internally to his
LS-50) result in an excessively warm cast to the image and, more
concerningly, substantially unpopulated areas of histogram, which are
certainly not acceptable.

I suspect you'd probably consider my corrections also too warm (the
result of massive red channel boost). I'll try to post an example of a
correction if I have time this weekend.

I know, fewer words, more images... ;-) But posting images is a very
convoluted process for me...

Don.

 

[Kennedy McEwen]

To be fair my results were only preliminary ("so far" as I put it).
But even if we take them at face value, I would not consider above
differences to be large *if* we compare them to the difference between
a raw scan (with KC cast) and a corrected scan (without KC cast).

My point is that you *cannot* take them at face value because you are
specifically picking on one very small aspect of the comparison which is
similar, but swamped by at least an order of magnitude (on a log scale!)
by the areas of dissimilarity - even ignoring the fact that Stephen's
experience does not correlate with *anyone* else's using similar
equipment!

You are looking for an LS-30 equivalent of the "Kodachrome" setting on
and LS-40/50 or LS-4000/5000. Stephen is telling you that not only
doesn't that work (it leaves a blue cast) but that his AG settings on
the "Positive" setting give better results - something which is
completely the opposite of what you have been looking for and which does
not correlate with any of the comparative "Kodachrome" scans which you
have been shown.

In other words, I'd be perfectly happy with a statement (if it were
*authoritative* and not based on guesswork) that the red channel needs
to be boosted by between 1 and 1.1 AG. I'd just split the difference
and consistently apply 1.05 across the board.

We've been through this till it is so full of holes I am surprised you
still find something to cling to. There is *NO* authoritative (ie. one
which will be endorsed by Nikon) statement because such a statement is
not possible to cover all of the possible combinations currently
existing in the field. You already have the most authoritative
statement that is necessary - your own results. If that isn't good
enough for you then I am afraid that you have a much bigger problem than
just scanning film. Neither Nikon nor anyone else are going to provide
you with the crutch you crave to underpin your own work - if you don't
have confidence in it, why should they?

You're assuming that my slides are considerably more underexposed than
Stephen's (or ~ vice versa) and you're basing this on my scanner's
response, which I don't think can be taken as an objective baseline.

I am not assuming anything of the sort! I am assuming that both of you
are attempting to obtain a scanned image which has an equivalent density
to the original slide. Stephen quotes a variance of +/-0.3EV to achieve
this, whilst you quote a variance of +/-0.5EV. However you both quote
an average difference which is of the order 2.1EV - a difference which
simply cannot be bridged by the variances you both quote.

So how do I explain the difference? Well, my scanner, of course!

Too easy - except it ignores the fact that your scanner works fine with
E-6 type slides, negatives and monochrome film. Blaming your scanner
implies that it somehow magically detects your Kodachrome slides from
other media and then adjusts its own exposure *equally* on all 3
channels to achieve a 1.3EV error. If the firmware is that smart you
really ought to spend all the time you have disassembling it, because
the algorithm could win you a fortune at the races!

However, and this is the key, even with this significant 1.3EV
difference in absolute terms, we still came up with similar KC cast AG
correction values in relative terms! That (to me) is very important.

That is simply irrelevant given the situation. If you introduce that
level of general adjustment on the master AG then it is almost
inevitable that most scans will have saturated histograms in one, if not
all three, colours. Under those circumstances all bets as to the
resultant optimal balance are void because you simply have no clue as to
where the actual exposure will end up. The similarity is therefore
insignificant.

I suspect you'd probably consider my corrections also too warm (the
result of massive red channel boost).

Possibly, although I doubt it, having already seen some of your scans
and heard your comments on my own. However there is a vast difference
between your scanner and mine, which would explain your adjustments,
whilst there is very little difference between Stephen's scanner and
mine.