B&W neg-as-pos scan on Minolta, use color or B&W? (ping Bart)

[false_dmitrii]

Well, I'm no closer to fathoming the definitive negative-to-positive
color workflow, but in the meantime here's a question regarding B&W
negatives. The SE5400 seems to clip these just as heavily as color
negs. When scanning B&W negatives as positives, does it make sense to
stay in B&W to keep file size down and simplify post-processing? Or is
it better to scan in RGB like a color negative, with per-channel
exposure adjustments, and adjust to neutral afterward? My one thought
was that the Minolta software might make invisible per-channel exposure
adjustments in B&W mode that would be fine for a B&W slide but throw a
negative further out of balance. As usual, I don't know how all of
this would have played out in the darkroom.

And as usual, thanks for any advice.

Off-topic, has anyone here tried ICE when scanning B&W in color? Has
it ever had a useful effect? The film I was looking at was labeled
Ilford HP5 Plus, if that matters.

false_dmitrii

 

[Bart van der Wolf]

Well, I'm no closer to fathoming the definitive negative-to-positive
color workflow, but in the meantime here's a question regarding
B&W negatives. The SE5400 seems to clip these just as heavily
as color negs.

The Minolta software does so on all negatives, color or B&W. So the
only real option with the Minolta software is to scan as linear
positive. The scanner itself doesn't clip unless the exposure is
wrong.

When scanning B&W negatives as positives, does it make sense
to stay in B&W to keep file size down and simplify post-processing?
Or is it better to scan in RGB like a color negative, with
per-channel
exposure adjustments, and adjust to neutral afterward?

Again if your exposure is optimal, i.e. maximum exposure per channel
without clipping, there should be little difference between the R, G
or B channels. You can verify/adjust channel exposures until you get
neutral filmbase color. In fact, having 3 'identical' channels will
allow to average them and get a little better Signal/Noise ratio.
Theoretically, slight misalignment between channels might compromise
sharpness a little, but in my experience with the DSE5400 that is not
a real issue.

My one thought was that the Minolta software might make invisible
per-channel exposure adjustments in B&W mode that would be
fine for a B&W slide but throw a negative further out of balance.

Little is known what that software does under the hood. Scanning
negatives as a linear RGB positive seems to offer a small (average
noise) benefit, and the file size can be reduced to 33% in that
process. Inverting and applying a gamma adjustment and tonemapping is
what remains, some of which could be done automatically by attaching a
profile, or by simply loading a predefined curve which inverts,
linearizes the film's characteristic curve (if you like that
decompression of shadows and highlights), and gamma adjusts for
display.

The only potential caveat I see is that the software used to average
the three monochrome channels should not add artifacts due to
impresice calculations. You can verify that by comparing single and
average channel standard deviations.

SNIP

Off-topic, has anyone here tried ICE when scanning B&W in color?

The silver in the B&W film will not allow ICE to work reliably, in
fact it introduces artifacts. However, the DSE5400 (1st generation)
does have a built-in diffuser, and that not only reduces graininess,
but it also reduces visibility of scratches and dust. It is one of the
huge benefits of the DSE5400 (Gen1) for B&W scanning.

Has it ever had a useful effect? The film I was looking at was
labeled Ilford HP5 Plus, if that matters.

ICE and silver won't work together. HP5 is high speed silver based B&W
film, so use the grain dissolver in the DSE5400 instead !!!

Bart

 

[Raphael Bustin]

Again if your exposure is optimal, i.e. maximum exposure per channel
without clipping, there should be little difference between the R, G
or B channels. You can verify/adjust channel exposures until you get
neutral filmbase color. In fact, having 3 'identical' channels will
allow to average them and get a little better Signal/Noise ratio.
Theoretically, slight misalignment between channels might compromise
sharpness a little, but in my experience with the DSE5400 that is not
a real issue.

I wonder about the best strategy for scanning BW
with a standard 3-color CCD array.

IIRC, the green channel has by far the best
sensitivity; red is middling and blue is the worst.
Look at typical output sensitivities from CCD
data sheets.

With a white illuminant and a neutral base color
(ie for BW negatives) you might be better off using
just the green channel. Mixing in Red or Blue
might even drag down the S/N.



rafe b
www.terrapinphoto.com

 

[Bart van der Wolf]

SNIP

I wonder about the best strategy for scanning BW
with a standard 3-color CCD array.

Yes, that would be different for a Bayer CFA type of sensor, because
with a single exposure the raw data doesn't necessarily get equal RGB
signal levels. The theoretically best solution is to adjust the
illumination color, and/or filter the lens in such a way that
spectrally uniform reflection produces the same (Raw) signal levels in
all three channels. For a three-shot (filter wheel) type of
'monochrome' sensor, the individual channel exposures can perhaps be
changed. This will ensure that the Photon shot noise levels will be
identical for each channel.

Now comes the unknown bit, has the Raw data had different ADC gain
applied to each channel when quantizing, and what does the
Rawconverter do with the channel data?
To answer that, just look at the noise levels of a evenly lit flat
image area per channel. You may find that one channel has a much
poorer Signal to Noise ratio. In that case I would use the channel(s)
with the lowest noise.

IIRC, the green channel has by far the best
sensitivity; red is middling and blue is the worst.
Look at typical output sensitivities from CCD
data sheets.

With my Canon cameras the red channel is usually the noisiest, so it
might differ between brands/models.

With a white illuminant and a neutral base color
(ie for BW negatives) you might be better off using
just the green channel. Mixing in Red or Blue
might even drag down the S/N.

Yes, and that's in addition to resolution differences that may be
present between R/B versus G (depends on Raw converter and should be
minimal for equal channel exposure). Also lens aberrations,
particularly for Red, may cause unsharper image data.

Bart

 

[false_dmitrii]

Hi, Bart, thanks for the answers. Couple of questions buried below:

The Minolta software does so on all negatives, color or B&W. So the
only real option with the Minolta software is to scan as linear
positive. The scanner itself doesn't clip unless the exposure is
wrong.

That's what I meant to say. For whatever reason, their DSLR does the
same thing to JPEGs.

Again if your exposure is optimal, i.e. maximum exposure per channel
without clipping, there should be little difference between the R, G
or B channels. You can verify/adjust channel exposures until you get
neutral filmbase color. In fact, having 3 'identical' channels will
allow to average them and get a little better Signal/Noise ratio.
Theoretically, slight misalignment between channels might compromise
sharpness a little, but in my experience with the DSE5400 that is not
a real issue.

In my experience, one channel usually ends up a couple of points above
or below the others unless all three are pushed all the way to 255 or
beyond. I may also be confusing the B&W scan with the issue of very
different per-channel curves of a color negative. Since the B&W
negative is (I assume) a single monochrome source, do the separate RGB
scan channels have identical tone response once the exposure gain is
equalized? Whatever the cause, I was seeing a slight cast on my
original B&W-as-color scans and wasn't sure how to interpret it. Might
also have come from trying out ICE on some of them.

Little is known what that software does under the hood. Scanning
negatives as a linear RGB positive seems to offer a small (average
noise) benefit, and the file size can be reduced to 33% in that

If the per-channel hardware gain in B&W-posi mode was adjusting *away*
from a good balance for a B&W negative, would the image even remotely
resemble a proper scan? The output looked okay to me after adjustment,
so either it's not an issue or I'm not judging the right things.

profile, or by simply loading a predefined curve which inverts,
linearizes the film's characteristic curve (if you like that

Could you elaborate a bit on this sentence? I'm not sure what such a
linearizing curve would look like in the workflow, let alone how to
match it to the characteristic curve. The film gamma and CC stuff is
still going over my head, sorry to say. Right now I get decent B&W
output in PSCS2 by assigning a gamma 1.0 profile, then stacking a
Levels black-as-whitepoint adjustment, a single inverted gamma curve
(somewhat arbitrary value), and a subjective s-curve for contrast.

Do you have an example of a proper characteristic curve correction
lying around?

ICE and silver won't work together. HP5 is high speed silver based B&W
film, so use the grain dissolver in the DSE5400 instead !!!

Thanks, I wasn't quite sure about ICE. Grain dissolver is on all the
time anyway.

false_dmitrii

 

[Bart van der Wolf]

Hi, Bart, thanks for the answers. Couple of questions buried below:

Bart van der Wolf wrote: SNIP
Since the B&W negative is (I assume) a single monochrome
source, do the separate RGB scan channels have identical
tone response once the exposure gain is equalized?

The B&W film (assuming proper processing) is quite neutral in its
absorption across the visible spectrum. However, the scanner lamp has
a certain color that needs to be eliminated as good as possible to
obtain similar channel response.

Whatever the cause, I was seeing a slight cast on my original
B&W-as-color scans and wasn't sure how to interpret it.
Might also have come from trying out ICE on some of them.

If the film processing was sub-standard, there may be residual
anti-halo dye left, or there may be a (photo)chemical reaction when
fixation and washing was too short. In addition there is the lamp
color to deal with.

If the per-channel hardware gain in B&W-posi mode was
adjusting *away* from a good balance for a B&W negative,
would the image even remotely resemble a proper scan?

Yes, but the channel differences would produce a color cast and thus
would be less accurate and more noisy.

SNIP

Could you elaborate a bit on this sentence? I'm not sure what
such a linearizing curve would look like in the workflow, let alone
how to match it to the characteristic curve.

Film has a rather non-linear response to light. It compresses the
shadow and highlight information with lower contrast than midtones,
which allows to chapture a larger range of luminances with more
apparent sharpness in the mid-tones.

If that non-linear response can be quantified, e.g. by measuring the
response to a stepwedge target with known step increments, then that
non-linearity can be taken out of the film response (by increasing the
shadow and highlight contrast and/or lowering mid-tone contrast). The
result would be that the equal stepwedge steps would result in similar
equal increases in RGB numbers.

In fact the step increments are multipliers (each step is the same
factor lighter or darker than its neighbor), but when represented on a
Log/Log scale (logarithmic exposure time vs Optical Density) there
will be a straight line (which makes it easier to check for
irregularities).

<http://www.xs4all.nl/~bvdwolf/main/downloads/SE5400+GD.jpg> shows the
response of my DSE5400 to a transparent stepwedge, which produces a
reasonably straight response (except for the highest densities).

The response of film itself would exhibit a somewhat more S-shaped
curve, slowly increasing its response, followed by a more rapid
increase in the midtones, and then a gradually decreasing response,
AKA the Characteristic Curve.

In Photoshop or similar photo-editors you can use a Curves control to
boost the contrast in shadows and highlights until the response
resembles the response shown above, and apply an inversion from
negative to positive. A given profile on the other hand would require
the exact same exposure level on each film frame (which is impossible
with outdoor scenes) to be valid.

The film gamma and CC stuff is still going over my head, sorry to
say. Right now I get decent B&W output in PSCS2 by assigning
a gamma 1.0 profile, then stacking a Levels black-as-whitepoint
adjustment, a single inverted gamma curve
(somewhat arbitrary value), and a subjective s-curve for contrast.

Which is fine if it produces pleasing results, afterall the final
result is more art than science in most cases.

Do you have an example of a proper characteristic curve correction
lying around?

No I don't, but it could (depending on lots of of factors) look
something like this:
<http://www.xs4all.nl/~bvdwolf/main/downloads/FilmCalibrInversionCurve.gif>
Don't take the exact curve shape too seriously, I just drew up
something approximate that inverts from negative to positive, applies
a given (probably wrong) gamma correction, and adjusts for shadow and
highlight contrast.

Bart