VueScan & Nikon LS scanner exposure control

[Jonathan]

Hi,

Nikon Scan allows users to control the analog exposure of the light
(master, red, green, or blue).

Does VueScan support the exposure control in the Nikon LS scanners? I
do find the brightness control, but I understand that is the gamma
mulitplier. The exposure control is an analog control adjusting the
light intensity used in the scan.

Thanks.

 

[Ed Hamrick]

Nikon Scan allows users to control the analog exposure of the light
(master, red, green, or blue).

Does VueScan support the exposure control in the Nikon LS scanners? I
do find the brightness control, but I understand that is the gamma
mulitplier. The exposure control is an analog control adjusting the
light intensity used in the scan.

Use the "Input|Lock exposure" option to control this.

What Nikon calls analog gain is actually CCD exposure time. You can
measure this by seeing that the scan time goes up linearly with
analog gain.

You can't separately control the red, green and blue CCD exposure
time in VueScan, because the ratio of these exposure times is needed
to properly convert from one color space to another.

Regards,
Ed Hamrick

 

[Jonathan]

Ed:

While it is true that the analog gain in Nikon scanner is the CCD
exposure time, the individual R, G and B analog controls do change the
light itensity of the respective colors. I verify this by peeking in
the slide mount adapter of my LS-2000. When I set a 0 gain for the
master, +2 for Red, -2 for Green and -2 for Blue, I saw a red light
traversing the slide surface.

Logically, I presume that the R, G and B light cathods are attached in
1 piece of hardware and thus does not allow individual adjustment of
exposure time, but why can't VueScan support the adjustment of light
intensity of R, G and B given the Nikon hardware is able to do it?

Regards,
Jonathan

 

[Kennedy McEwen]

Ed:

While it is true that the analog gain in Nikon scanner is the CCD
exposure time, the individual R, G and B analog controls do change the
light itensity of the respective colors. I verify this by peeking in
the slide mount adapter of my LS-2000. When I set a 0 gain for the
master, +2 for Red, -2 for Green and -2 for Blue, I saw a red light
traversing the slide surface.

Logically, I presume that the R, G and B light cathods are attached in
1 piece of hardware and thus does not allow individual adjustment of
exposure time, but why can't VueScan support the adjustment of light
intensity of R, G and B given the Nikon hardware is able to do it?

You presume wrong then.

It is much more easy to linearly control the time an LED is switched on
for than it is to linearly control it's intensity. The "D" in LED is
for *diode*, and if you don't know the non-linear and temperature
dependent relationship between bias voltage and current for a diode (and
thus power and emitted light intensity in a LED) then you can consider
that as set homework! Almost all applications which require light
intensity control from LEDs do so using time modulation.

What you see when you look inside your Nikon scanner is the average of
the red green and blue LEDs pulsing on and off for each exposure period.
You cannot discern the individual pulses because they are too fast, so
what you see is an optical illusion - the average intensity.

For example, the LS-2000 scans a complete, non-ICE, positive frame in
20seconds. That corresponds to 3850 cycles of red, green & blue light,
or 190cycles per second. If you think you can see that well enough to
be able to distinguish pulse period from pulse intensity then I do hope
your entire home is powered by DC electricity rather than that nasty
50/60Hx mains that the rest of us get! TV is probably quite annoying to
watch and going to the cinema must be a real nightmare.

In your example, having reduced the exposure time of the green and blue
LEDs by a factor of 4, they obviously contribute a quarter of their
normal amount to that average light intensity that you see, whilst the
red has increased its on time by a similar factor - resulting in light
which is visibly red: not because the red LED is shining brighter and
the green and blue LEDs are dimmer, but because it is now on 16 times
longer than either the green or blue LEDs, and 8 times longer than both
of them together. Why would you expect it to be any other colour?

 

[Neil Baylis]

I've heard this mantra so many times.. that the NS analog gain does
nothing more than change exposure time. I can't really understand it
though.

I'm using the LS 50 with NS 4. When I increase the master analog gain,
the resulting contrast increases. I see the histogram spread out, with
the highlights moving to the right more than the shadows. The various
peaks on the histogram become further apart as you increase the analog
gain. If the scanner was simply increasing exposure, wouldn't the
histogram just be shifted to the right, rather than spread out? How
does increasing exposure increase the contrast?

When I increase the exposure on my digital camera, the entire
histogram simply moves to the right. It doesn't spread out like the
scanner does. Is this a valid comparison, or am I missing something
obvious?

 

[Kennedy McEwen]

I've heard this mantra so many times..

[OED]
Mantra: a word or sound repeated to aid concentration and understanding.

Clearly this is not a mantra as it has failed in both objectives.

that the NS analog gain does
nothing more than change exposure time. I can't really understand it
though.

I'm using the LS 50 with NS 4. When I increase the master analog gain,
the resulting contrast increases. I see the histogram spread out, with
the highlights moving to the right more than the shadows. The various
peaks on the histogram become further apart as you increase the analog
gain. If the scanner was simply increasing exposure, wouldn't the
histogram just be shifted to the right, rather than spread out? How
does increasing exposure increase the contrast?

Well, (at least) two things come into play here, and another later.

Firstly, if something is black then it is black - that means that no
matter how much you increase the exposure it will still be black. On
the other hand, if something is white then how white it appears depends
on how much light is incident on it and, in the case of a sensor, how
long it is exposed for. So, increasing the exposure will, by simple
laws of physics, stretch out the histogram - blacks remain at black and
whites increase in level as a function of the exposure. What you see in
NS4 is *exactly* what you should expect.

Note that this does not mean an increase in the contrast: contrast is
the ratio between maximum and minimum signal, not the difference! It is
only a difference in density because density is a logarithmic
measurement.

The second thing that comes into play is the system gamma and whether
the histogram plots the data in absolute terms or as corrected by the
system gamma. Usually it is the former, but not always. This can have
the effect that very small increases in the absolute luminance data of
dark objects can appear as large steps in the output luminance. What is
happening is that the linear encoding scheme of the CCD results in
insufficient levels in the dark regions when the gamma has been applied.

When I increase the exposure on my digital camera, the entire
histogram simply moves to the right. It doesn't spread out like the
scanner does.

The mantra isn't "NS AG changes exposure time", it is "True black stays
black, irrespective of exposure". If you were shooting at the bottom of
the deepest coalmine in the world without any light source, why would
you expect a perfect sensor's histogram to shift right (blacks and all)
just because you increase the exposure?

What this histogram is indicating *is* a reduction in contrast as
exposure increases but, as you correctly noted earlier, increasing
exposure does not change contrast.

This is where the other thing comes into play. It would appear that the
histogram display on your camera is indicating limited bit depth or base
noise (dark current) in the data it is based on.

Why your digicam indicates that contrast is reduced by increased
exposure is a matter between you and its manufacturer, but it is clearly
wrong, and not something you should use as a metric on which to judge
other instruments.

My guess is that it's black level limit is actually the CCD dark current
level, which is limiting the contrast achievable in your images as you
increase exposure and also the significance of the dark current present.

The CCD in the LS-50 (and LS-5000) has a very low dark current level
indeed - lower than almost any other scanner, which themselves are
usually lower than most digital cameras because they are simpler
devices.

In addition, the LS-50, like most other film scanners, is dark current
compensated at every time film is inserted into it (that is why it spits
the film back out after a period of inactivity, to *force* a new dark
calibration). By comparison, if you are lucky your digicam might
compensate for dark current *after* the exposure (so not available on a
histogram of a preview image) or, if it isn't a top of the range model,
have been dark current compensated once at the factory in the hope it
doesn't change too much through the life of the unit. It does. :-(
Consequently, your scanner can provide a much more accurate estimate of
"black" than your digicam - and hence a more accurate histogram and
demonstration of the correct effect on the image histogram of exposure.

Is this a valid comparison, or am I missing something
obvious?

Both apparently.

 

[Jonathan]

Ken:

Thanks. Your explanation definitely makes sense. It definitely
cleared up some of my misunderstanding earlier and wrong
interpretation of an observation (

The remaining question is why VueScan does not allow users to change
the individual R G B exposure time. According to Ed in the earlier
threat, the ratio of these exposure times is needed to properly
convert from one color space to another. Why is that?

Here is my thought:

1) For argument's sake, let assume the proper ratio is R = +2, G= -2
and B= -2 (these are the values as input in NikonScan), i.e., VueScan
has to see a strong red tint in order to properly convert an image
from 1 color space to another.

2) Now if I have a predominantly green or blue image (taken with a
dark blue or green filter), because of the absense of red and because
VueScan needs to see reds, VueScan will not function properly.

Is my conclusion correct? (I don't believe this is the case, but then
how to explain VueScan need to maintain certain ratio of RGB exposure
time?)

My thinking is: every image is different and they should be. After a
slide/film is digitialized, we can't tell if a certain color in the
image is directly generated from the pigment on the slide/film or it
is a result of a different hue of the light passing through the
slide/film during scanning. How could we than rely on the resulted
color information to do color space conversion.

Regards,
Jonathan

 

[Kennedy McEwen]

Ken:

Thanks. Your explanation definitely makes sense. It definitely
cleared up some of my misunderstanding earlier and wrong
interpretation of an observation (

The remaining question is why VueScan does not allow users to change
the individual R G B exposure time. According to Ed in the earlier
threat, the ratio of these exposure times is needed to properly
convert from one color space to another. Why is that?

Ed has produced a piece of software which manages to achieve, in a very
compact and efficient piece of code, what most bloatware drivers and
everything with a Micro$oft monogram on it fails to deliver. In doing
that he clearly makes certain assumptions and tradeoffs which, in the
main, are justified, valid and accurate.

However, one of the major successes of Vuescan is the ability to achieve
very good colour balance from a wide range of film stock on almost any
colourspace from almost any scanner. That is, Vuescan manages to
reproduce the scene colour balance irrespective of the film stock or the
system - and I doubt that this is something he would want to sacrifice,
but which would be inevitable if the assumption is made that the user
wishes to somehow adjust the scene colour themselves. There is a valid
argument that this should be done by the user after a correct scan has
been achieved. Ed has stated that certain the individual analogue gains
are required to enable Vuescan to interface between the various colour
spaces. I don't know why that is the case and I suspect that Ed is the
only person who can answer the question accurately - although he may
well be unwilling to explain such details as they could divulge his
proprietary methods.

However, it seems to me that the "nominal" settings could be used as a
basis on which individual channel exposure adjustments could be based.
The nominal settings would meet the requirements for color-space
conversions, the modifications would not - and a dialogue could pop-up,
warning of such colour space errors, when the adjustment was initiated.
The problem then arising is that Ed would be swamped by support requests
from people (including certain readers of this newsgroup and others -
they know who they are ;-) ) who think that this would represent a bug
in Vuescan, or that Ed was somehow holding back something that prevented
other settings from working as well as his own defaults and demand it
was fixed.

Sometimes its just better not to provide an option than to get bogged
down in the support of it to the extent that it prevents product
development and growth.

 

[Don]

The remaining question is why VueScan does not allow users to change
the individual R G B exposure time. According to Ed in the earlier
threat, the ratio of these exposure times is needed to properly
convert from one color space to another. Why is that?

It's a software design choice. To allow users to change individual RGB
exposures could unravel VueScan's automatic color balance which is its
main reason for existence.

Personally, I find that far too inflexible and overbearing (not to
mention VueScan's "unique" user interface) which is why I stick with
NikonScan.

Just my opinion, but VueScan is like a throwaway Instamatic camera.
Point and shoot. When it works, it works well, but when you get a
difficult situation (e.g. a very dark Kodachrome) you end up with
something looking like a false color image. Totally unacceptable (for
me, anyway, although many differ).

One VueScan plus is the ability to boost exposure more than NikonScan
(which is effectively limited to 4 AG). However, instead of using
Analog Gain (AG) i.e., EV values, VueScan insists on its own
"interesting" scale... :-/ And, of course, you still can't adjust
individual channel AG which renders this advantage meaningless.

If you, nevertheless, insist on wrestling with VueScan one option may
be to scan difficult images with NikonScan where you can set
individual AG values and then feed this file to VueScan for color
correction. You may want to leave all other NikonScan settings at
neutral (or off) with the exception of ICE (if applicable). For best
results you should also scan with gamma at 1 and in 16 bit.

But I'm sure VueScan fans (and as you gather I'm not one) will soon
jump in and tell you all the gory details... ;o)

Don.

 

[Don]

Sometimes its just better not to provide an option than to get bogged
down in the support of it to the extent that it prevents product
development and growth.

You're being too kind... ;o)

Limiting legitimate user needs is never a good idea. If a reasonable
option causes "problems" for the programmer then it's an indicator of
a fundamental design flaw and not the fault of the option or the user
requesting it.

Don.

 

[Kennedy McEwen]

You're being too kind... ;o)

If the cap fits...

...well, I see from your other post in this thread that you have
already taken that figure of speech onboard. ;-)

 

[Neil Baylis]

Kennedy,

thanks for the explanation. I never knew just how much I didn't know!

Firstly, I should make a correction. After reading your post, I went
back to my digicam to experiment further. I took my gray card, and put
a 3 x 5 piece of black paper in the middle of it, and then a 1 x 1
square of white paper on top of that. Then I framed it so the gray
card filled the frame on my digicam. The idea was to create a
histogram with recognizable peaks.. a wide one for the gray, a medium
one for the black, and a tiny one for the white. I shot an exposure
series from -2/3 to +1.0 on the exposure compensation, and then
examined the resulting images in the camera's LCD with histogram
enabled. I wanted to verify that the histogram did not spread out as I
increased the exposure. Well, the short answer is that it does spread
out, contrary to what I had believed based on previous, not so well
controlled experiments.

What you see in NS4 is *exactly* what you should expect.

Apparently so.


So apparently it works like this (simplified): If I increase the
overall exposure, it means increasing the number of photons that hit
the entire surface of the transparency in the scanner. Given that the
transparency has darker areas and lighter areas, more of these extra
photons will get through the lighter areas than the darker areas.
Consequently, the location of these lighter pixels on the histogram
will move right more than the location of the darker ones. Is that
basically correct?

So let me ask this: If the AG control actually did increase the gain
applied to the analog signal coming out of the CCD, how would this
look different from increasing the exposure? I mean, how would the
resulting histogram differ?

Thanks again for your input,

Neil

 

[Don]

Firstly, if something is black then it is black - that means that no
matter how much you increase the exposure it will still be black.

I also find this not to be the case if I boost AG exposure
significantly. I guess it's because blacks (we're talking slides here)
are not really black. Unexposed slides when viewed against a bright
light source are still transparent to a degree.

Like Neil, I also observed that blacks do shift into dark grays but
only very reluctantly. The whites on the other hand are very eager to
rush toward clipping (all this when examining the histograms, of
course).

Not knowing the theory, I attribute this uneven shift to gamma
conversion (no time to test it with linear gamma). Anyway, is that
right?

Don.

 

[Ed Hamrick]

Ed has stated that certain the individual analogue gains
are required to enable Vuescan to interface between the various colour
spaces. I don't know why that is the case and I suspect that Ed is the
only person who can answer the question accurately

It's fairly straightforward. Conversion from the internal color
space of a scanner to a known color space is done with a 3x3
matrix multiplication. Changing the ratio of the CCD exposure
times will change the result of this matrix multiplication if
the changed CCD exposure time isn't compensated for.

This compensation isn't complicated, but it would mean I'd have
to store the ratio of the CCD exposure times in the raw scan
files. This would require using some custom TIFF tags, and I
didn't want to do this.

The main reason is that it's completely unnecessary to change the
ratio of the CCD exposure times, since for 99% of the images
people scan, the very slight gain in dynamic range that would
result wouldn't be visible.

Regards,
Ed Hamrick

 

[Bart van der Wolf]

SNIP

Like Neil, I also observed that blacks do shift into dark grays but
only very reluctantly. The whites on the other hand are very eager to
rush toward clipping (all this when examining the histograms, of
course).

Not knowing the theory, I attribute this uneven shift to gamma
conversion (no time to test it with linear gamma). Anyway, is that
right?

Not exclusively. Basically you see a multiplication of exposure amount
(time). This means that low luminance values increase less rapidly (on an
absolute scale) than high values. On top of that there is a (reciprocal)
Gamma correction to offset the Gamma of the CRT, which will boost shadows
and midtones more than highlights.

Bart

 

[Jonathan]

Thanks Ed.

The reason I want to change the analog gain is that I have a bunch of
under-exposed Kodachrome. Currently, when I scan them in NikonScan, I
need to use a high positive gain in the master and red and a high
negative gain in the blue. And yes, it was my fault that I did not
expose those Kodachrome properly, but I really need to salvage those
images.

It sounds like the goal of VueScan is to achieve a good realistic
color balance. If VueScan had allowed users to change the CCD
exposure ratio, VueScan would have countered the change by adjusting
the 3x3 matrix. So, instead of undoing the change made by users,
VueScan simply does not allow the individual CCD exposure to be
altered. If users want to introduce a color cast to the image, he is
then encouraged to do it by adjusting the brightness of individual
color (I hope I make the right conclusion).

I think this is good information to know. For now, I will stick with
NikonScan and do my own analog adjustment because VueScan can't quite
reproduce the original color of the scene in my Kodachrome even though
I select Kodachrome as my slide type.

Thanks and regards,

Jonathan

 

[Kennedy McEwen]

Kennedy,

thanks for the explanation. I never knew just how much I didn't know!
You're welcome.

So apparently it works like this (simplified): If I increase the
overall exposure, it means increasing the number of photons that hit
the entire surface of the transparency in the scanner. Given that the
transparency has darker areas and lighter areas, more of these extra
photons will get through the lighter areas than the darker areas.
Consequently, the location of these lighter pixels on the histogram
will move right more than the location of the darker ones. Is that
basically correct?

That's the gist of it.

So let me ask this: If the AG control actually did increase the gain
applied to the analog signal coming out of the CCD, how would this
look different from increasing the exposure? I mean, how would the
resulting histogram differ?

At the histogram level, in terms of the dispersion of black and white,
it would look exactly the same, which is why Nikon get away with calling
it Analogue Gain in the first place. However it would differ in terms
of the way that noise is represented. The readout noise floor on the
CCD would be amplified by a real analogue gain stage, together with the
signal, resulting in a fixed signal to noise ratio for any given level
in the image. However changing the exposure does not change the readout
noise even though the signal increases - thus the signal to noise ratio
is significantly improved for image levels which are readout noise
limited. This means finer gradations in the blacks of a slide image.

 

[Kennedy McEwen]

I also find this not to be the case if I boost AG exposure
significantly. I guess it's because blacks (we're talking slides here)
are not really black. Unexposed slides when viewed against a bright
light source are still transparent to a degree.

Like Neil, I also observed that blacks do shift into dark grays but
only very reluctantly. The whites on the other hand are very eager to
rush toward clipping (all this when examining the histograms, of
course).

Not knowing the theory, I attribute this uneven shift to gamma
conversion (no time to test it with linear gamma). Anyway, is that
right?

Probably.

 

[Kennedy McEwen]

It's fairly straightforward. Conversion from the internal color
space of a scanner to a known color space is done with a 3x3
matrix multiplication. Changing the ratio of the CCD exposure
times will change the result of this matrix multiplication if
the changed CCD exposure time isn't compensated for.

This compensation isn't complicated, but it would mean I'd have
to store the ratio of the CCD exposure times in the raw scan
files. This would require using some custom TIFF tags, and I
didn't want to do this.

Cheers Ed. I can understand why you would like to do this - ensuring
that the colour balance that Vuescan is well respected for is retained
even after a user has dabbled with the individual analogue gain. However
I would also suggest that any user who is prepared to dabble with
individual channel exposure *probably* isn't happy with the resultant
balance in any case, so I don't think there would be any demand for you
to do this - just use the same nominal values for exposure as you
currently do in the matrix but allow the user to adjust the exposure on
each channel from this. As I suggested, a pop-up dialog warning of the
change of colour balance would be sensible and, IMO, exactly what most
users who wanted to do this would expect.

The main reason is that it's completely unnecessary to change the
ratio of the CCD exposure times, since for 99% of the images
people scan, the very slight gain in dynamic range that would
result wouldn't be visible.

True, but there is always that 1%. ;-)

 

[Don]

SNIP

Not exclusively. Basically you see a multiplication of exposure amount
(time). This means that low luminance values increase less rapidly (on an
absolute scale) than high values.

Do you mean that when the exposure of a pixel with the value of, say,
5 is doubled it's "only" 10, but when a pixel with the value of 50 is
doubled it's a 100? (These are values out of thin air for
demonstration purposes only.) Therefore, the "dark" pixel would only
move 5 points while the "light" one would move 50.

I realized that but I just thought the answer was more complicated.

For example, as Kennedy wrote, black is black, i.e. absence of
luminance, so no matter how much exposure is boosted it should stay
black.

On the other hand, as I noted earlier, the black in slides is not
really true black (i.e. impervious to light). This can be easily
demonstrated by looking at an unexposed slide against a bright light.

On top of that there is a (reciprocal)
Gamma correction to offset the Gamma of the CRT, which will boost shadows
and midtones more than highlights.

That should in theory accelerate the movement of dark pixels. It's
because of things like this I thought the answer was more complex.

Anyway, this is not really a big deal because the workflow I settled
on takes care of this automatically, but it's just interesting...

Thanks Bart!

Don.