LS4000 depth of focus problem--Nikon LS5000 also, or other alternatives?

[Gary L Hunt]

Nope, the lens is the same speed as the LS-4000. The LS-5000 is twice
as fast as its predecessor (20sec v's 38sec) because it uses a double
CCD - the same technique as the LS-8000 used for its high speed mode,
but with two lines instead of three. Hopefully it does not suffer from
the same mechanical resonance which made that (default!) option unusable
in the LS-8000.

Didn't mean to imply that I thought the LS-5000 had a faster lens than the
LS-4000, only that my LS-4000 makes single scans much faster than my old
Polaroid SS4000 and apparently faster than most other comparable scanners
(e.g. the Minolta 5400). But what I've read (and your description) strongly
suggests that the LS-5000 design has not changed anything that will make my
depth of focus annoyance go away. So I'm thinking of trying the Minolta
anyway. Besides, the LS-4000 produces incredibly grainy looking scans
(for me anyway) from B&W negatives, and I can't help thinking that the light
source has something to do with that.

Gary Hunt <[email protected]>

 

[Bart van der Wolf]

SNIP

Didn't mean to imply that I thought the LS-5000 had a faster lens
than the LS-4000, only that my LS-4000 makes single scans much
faster than my old Polaroid SS4000 and apparently faster than most
other comparable scanners (e.g. the Minolta 5400).

Don't underestimate that the DSE-5400 scans in excess of 82% more pixels.
The interface can also become more important to accommodate that volume of
data. It may be more meaningful to compare the number of pixels/sec or
bytes/sec for a scan speed benchmark.

But what I've read (and your description) strongly suggests that the
LS-5000 design has not changed anything that will make my depth
of focus annoyance go away. So I'm thinking of trying the Minolta
anyway. Besides, the LS-4000 produces incredibly grainy looking
scans (for me anyway) from B&W negatives, and I can't help
thinking that the light source has something to do with that.

The amount of light source diffusion versus collimation plays a role in the
visibility of very fine detail. You will also find the Minolta to show
graininess, partly because it is magnified more and the lens is *very*
sharp, but you can use the "Grain Dissolver" (a diffuser) to reduce the
visibility of grain. The GD does more than double the exposure time though.
An additional benefit of the GD for B&W negatives is the reduced visibility
of scratches and dust on a film that doesn't allow IR based ICE.

IMO a noise and graininess removal tool like "Neat Image" is indispensable
for film scans.

Bart

 

[nikita]

Gary,

I'm using the 4000 with Silverfast in RAW highbit mode.( based on the
RAW output for later Negafix workflow in Silverfast HDR). However, I'm
skipping the HDR completely. Instead I'm working on the RAWs directly
in Photoshop. This works extremely well with any negative. I'm using
Reala 100 for studio and portraits, Fuji 400 for the documentary and
800/1600 for pushdeveloping. The B&W stuff is everything from old
Neopan 400, TMZ 3200, T-max C41, 400 &100. The Nikons have always been
tough to do traditional BW work. The Polaroids have been very easy to
team with BW. But using the RAW (Negafix) output is much better with
Nikon. The bad experinces you do have with the Nikon 4000 and black
and white films will go away.

The RAW negafix output is not a pure RAW in the traditonal meaning.
But it uses some parts of the NikonScan driver to "prepaire" the basic
RAW (that you get from using positive mode) to be a better
startingpoint for the negafixworkflow in HDR. What you get from this
is a wonderful "RAW" output of a negative without any clipping of the
ends. ( It's like getting it all as with Vuescan compared to a normal
NikonScan ). Note that there is a big difference between Postive mode
RAW and Negafix RAW. You can use the positive mode too, but the colors
will be a lot easier to get into the ballpark with the negafix RAW
output. I know how to use Vuescan in and out since ages, but the
controlled workflow I get by using Silverfast this way beats it. You
do get it all into Photoshop and from that you just choose how to work
on it. The starting point is better than a Vuescan "autoprocessed" as
soon as you set the endpoints in R-G-B channels to clean up the mask.
Nothing has been done to the RAW before you're on it – except for the
slight "preparation". When Vuescan nails the colors, fine. When
Vuescan doesn't nail the colors, then you've got work to do within
that interface.

The "preparation" for the Negafix will be almost the same for each
RAW. The tool you use is Levels or Curves to get it into ballance –
but that's in Photoshop. It is a very very fast workflow. You don't
even have to crop in Silverfast or use ANY of the fantastic tools it
has. Load the neg, max rez, highbit. Preview and scan. You can do it
with hangover, no brain is needed for the scanning part. The exposure
of the scan is calculated from the allover frame. So, a cropping will
not change anything with exposure and that's why it can be left out at
scanningstage. The exposure reads the filmbase that shows up on the
side of the filmholder and calulates on that. No shadow
clipping......A crop will, of course, clip out a few MB from the
filesize, but that just matters if you scan directly to disk (which is
possible to do with Silverfast).

You should NOT adjust the Nikon "analog gain" to bump up all the three
componentchannels for "maximizing the dynamic range". No, the allover
ballance is much better when left on auto analog gain. You can use the
black and white or RGB. But I really doubt that you will gain anything
by using the RGB for BW. You'll have a very good "headroom" in both
ends to clip as much as you want by using Levels with the Video LUT
(alt+ dragging the sliders). Just INVERT in Photoshop first.
Then go into the separate comonentchannels for the endpointsettings.

When scanning colornegs with this flow the Bruce Fraser method with
Autocolor works wonderful. You know, stepping the values from zero
with the arrowkeys on the keyboard, using the snap midgrey in combo
with the colorpicker. This creates a controlled semiauto workflow that
really speed things up. The Autocolor with your own tweakings, is like
it was created for the negafix RAW ouputs.

See if the Silverfast Ai6 DEMO can create the negafix RAW highbit
outputs. The BW with nikon 4000 will not be a problem anylonger. But
the problem with blurred grain is still there <grin>.

nikita

ps. the "negafix system" itself isn't used in anyway – it's just the
preparation for it in the RAW that you will use. It makes the RAW
output a bit softer and not so "dense". The colors in colornegs will
be better ballanced regarding hue and saturation than using positive
RAW for the negs. Another good thing is that you get a RAW highbit to
archive. This Raw is much much easier to work directly with, than for
example the Vuescan RAW.

 

[Kennedy McEwen]

Besides, the LS-4000 produces incredibly grainy looking scans
(for me anyway) from B&W negatives, and I can't help thinking that the light
source has something to do with that.

It is a collimated light source and collimated light results in a higher
lens MTF at Nyquist and consequently more grain aliasing.

Personally, I think that Nikon could have produced, and still may, an
8000ppi scanner from exactly the same design as the LS-4000 instead of
the rather meagre upgrade to the LS-5000 for no cost increase
whatsoever. The lens and illumination source clearly has excessive
resolution for the sampling density of 4000ppi. Just use the staggered
CCD approach instead of the aligned CCD they have used. The overall
result would probably have been closer to 3000-3500cy/in resolution, or
6-7000ppi equivalence. This would be limited by the lens, not the
sampling density, with the total elimination of grain aliasing as a
bonus. Ideal for B&W negatives, where all LEDS are on at once in any
case. All that would be sacrificed is the 20s of scan time for colour
film - negligible given the time required to implement ICE, GEM and ROC
afterwards, even on a fast PC.

Which raises an interesting question: did Nikon marketing get it right,
pushing for faster scan speeds instead of resolution? I know I would
have preferred a virtually grain free 8000ppi desktop scanner rather
than what the LS-5000 offers, and I suspect that I am not in the
minority, let alone alone. ;-)

 

[Gary L Hunt]

Personally, I think that Nikon could have produced, and still may, an
8000ppi scanner from exactly the same design as the LS-4000 instead of
the rather meagre upgrade to the LS-5000 for no cost increase
whatsoever. The lens and illumination source clearly has excessive
resolution for the sampling density of 4000ppi. Just use the staggered
CCD approach instead of the aligned CCD they have used. The overall
result would probably have been closer to 3000-3500cy/in resolution, or
6-7000ppi equivalence. This would be limited by the lens, not the
sampling density, with the total elimination of grain aliasing as a
bonus. Ideal for B&W negatives, where all LEDS are on at once in any
case. All that would be sacrificed is the 20s of scan time for colour
film - negligible given the time required to implement ICE, GEM and ROC
afterwards, even on a fast PC.

Which raises an interesting question: did Nikon marketing get it right,
pushing for faster scan speeds instead of resolution? I know I would
have preferred a virtually grain free 8000ppi desktop scanner rather
than what the LS-5000 offers, and I suspect that I am not in the
minority, let alone alone. ;-)

It certainly would provide a much more convincing reason to upgrade my
LS-4000 than increasing the scan speed, especially since I can't keep up
with mine (using the SF-200 feeder) in Photoshop anyway, even with it
producing 120 MB 14-bit files. Of course the combination of *not' getting
the faster scan speed along with double the resolution would *really*
slow down the throughput on most computers. Does the present CCD arrangement
require half the step size of the LS-4000, or do they scan both arrays at
each step? If the latter, this would work against the idea of doubling
the resolution, at least in the direction of motion.

Gary Hunt <[email protected]>

 

[Kennedy McEwen]

Does the present CCD arrangement
require half the step size of the LS-4000, or do they scan both arrays at
each step? If the latter, this would work against the idea of doubling
the resolution, at least in the direction of motion.

In the LS-5000 they expose both CCD lines at once and then step 1/4000"
to get the next exposure. Each point in the image is interrogated by
both lines of the CCD, so they only need to expose half as long as in
the LS-4000.

The technique is called "Time Delay and Integrate", or TDI and is
particularly well suited to CCDs to achieve a longer effective exposure
time than the dwell at each position permits.

 

[Robbafett]

Hi!

I had terribly HUGE problems with the depth of focus with the Nikon LS5000ED, could never get any sharpness in the corners/edges.

Then I cleaned the objective and voila, really great sharpness all over the picture.
Just a tip.

Kind Regards
R