Scanning 4x5 on epson 4870 at 16-bits/channel

[jjs]

There are two different modes for reading pixels from the scanner,
one reads all three colors per line, and another reads one color
at a time. VueScan uses the former on the 3170 and the latter
on the 4870.

The limit in VueScan on the 4870 is pixel_width*bytes_per_sample <= 32752
The limit on the 3170 is bytes_per_line <= 65520 bytes.

On the 4870, for 16 bits per sample, the maximum pixel width in
VueScan is 16376, and at 4800 dpi this is 3.411 inches.
At 8 bits per sample, this is 6.822 inches.

On the 3170, for 16 bits per sample, the maximum pixel width
in VueScan is 10920, and at 3200 dpi this is 3.413 inches.
At 8 bits per sample, this is 6.826 inches.

If the requested scan area is too wide, VueScan
first tries to reduce the bits per sample to 8 bits. If this
is still too wide, VueScan reduces the scan resolution one
step.

Thanks so very much for clarifying exactly what your great program does, Ed.

So there you go, folks - the authoritative source! More at
http://www.hamrick.com/

 

[Bart van der Wolf]

SNIP

The alleged need for 48 bit scans has been debated
often on this and many other forums, and by luminaries
such as Jeff Schewe and Dan Margulis. I side with Dan.

Good for Dan ;-), but although he's a knowledgeable person, many of his
views are too (CMYK) printing press oriented for my taste. If anything
restricts continuous tone imaging, printing most definitely will.

If you do the coarsest of your tonal manipulations in
the scanner driver - and do them well - the resulting
24-bit file will be robust enough to withstand (or thrive
on) fine tweaks later on in Photoshop.

Possible, especially since Photoshop only uses 15-bit channels. But the
issue at hand is getting the 48-bit in the first place, even before the
coarsest of tonal manipulations. After that, the OP seems experienced enough
to choose a workflow suiting his needs.

Bart

 

[Tim]

The maximum scan width of the Perfection 3200 at 3200 dpi and 48 bit
colour would be 4 inches assuming a 74Kb limit per scan line with the
EPSON Scan driver.
As the holders are the other way around in this scanner it is possible
to scan 4 x 5 inch material at 3200 dpi and 48 bit colour.

 

[Leonard Evens]

There are two different modes for reading pixels from the scanner,
one reads all three colors per line, and another reads one color
at a time. VueScan uses the former on the 3170 and the latter
on the 4870.

The limit in VueScan on the 4870 is pixel_width*bytes_per_sample <= 32752
The limit on the 3170 is bytes_per_line <= 65520 bytes.

On the 4870, for 16 bits per sample, the maximum pixel width in
VueScan is 16376, and at 4800 dpi this is 3.411 inches.
At 8 bits per sample, this is 6.822 inches.

Ed,

Suppose I choose a custom resolution on the 4870 of 3200 ppi (rather
than 4800 or 2400)? How do you go about reading the pixels and storing
the information? Can one scan wider than 3.411 inches at 16 bits per
channel and 3200 ppi?

Also, how hard would it be for you to modify the program so that 4 x 5
can be scanned completely at 4800 ppi and 16 bits per channel? At
present, this would seem to be unnecessary, but hardware does tend to
improve at a rapid pace, so it won't be too long before that is
feasible. As usual, Roger Clark seems to be pushing the boundaries.

Finally, does it make any difference whether one is operating under
Windows or Linux in these matters?

 

[No One]

4x5 @ 3200ppi 48 bit is 1.14GB, but who's counting?

Curious - what would you do with a file that large? I regularly print 40x50
from 4-500 meg files and do not believe I am losing anything.

Cheers -

 

[Mac McDougald]

Curious - what would you do with a file that large? I regularly print 40x50
from 4-500 meg files and do not believe I am losing anything.

Cheers -

Well, if this were knocked down to 24 bit (which is all printers can use
anyway) it would be about 550MB at same pixel dimensions.

Mac

 

[Roger N. Clark (change username to rnclark)]

HiYa

so roger ... does this mean we can expect to see an update of

http://www.clarkvision.com/imagedetail/hp7400-drum_compare.html

with this scanner?

I've been thinking of getting a scanner for my own 4x5's and have been
considering the same scanner.....

Yes, and the results I've gotten on the same test spot are very close
to the drum scan. However, the drum was only 8-bits/channel, and I'm
able to bring out more detail in the shadows with 16-bit, so in that
sense, it is better than the drum scan.

An update. I've scanned several 4x5s in two pieces each, at 3200
ppi and 16-bits/channel. When I laid the second half on the first
it was essentially perfect in tonality and range (I used the same
settings on each scan). Spatially there are 1-pixel offsets in
blocks, meaning one section will be right on, and then the next section
of a couple hundred lines will be offset one pixel. By carefully
erasing the edge of the overlay so it is feathered and not a straight
line, one can't see the 1-pixel offsets.

I have vuescan, but an older copy. I'll update it and try
it out. I have 9 large prints to deliver soon, so I need
to get a good procedure down.

The scans on the 4870 are slow. With ICE turned on, 3200 ppi,
48-bit, it takes about an hour per section, so two hours per
4x5. With ice off, it is only a few minutes.

Roger

 

[Roger N. Clark (change username to rnclark)]

The alleged need for 48 bit scans has been debated often
on this and many other forums, and by luminaries such as
Jeff Schewe and Dan Margulis. I side with Dan.

If you do the coarsest of your tonal manipulations in the
scanner driver - and do them well - the resulting 24-bit file
will be robust enough to withstand (or thrive on) fine tweaks
later on in Photoshop.

I do scientific imaging for a living, even working with data at a signal
to noise of 30,000 at times. In the astronomy world, multiple (e.g.
hundreds) of 16-bit images are added together to increase signal
to noise to the 10,000+ range. There is no doubt that with
higher signal to noise than 8-bit you can bring out subtle
detail in images not possible with 8 or even 16-bit capture.
(Most capture is not 16-bit signal to noise, but when added together
can exceed 16-bit; I soften work in 32-bit.)

In practice, for large format work, the information one can get
from any scanner software I've used leads a lot to be desired
compared to working with the full resolution data in something like
photoshop. Even chrome images like velvia contain more information
than covered by an 8-bit range, even with appropriate gamma. For example,
I reduced my need for split density filters, because I know I can
keep the range I want with a good scan, and detail in the high end
that one could not print with normal enlarger methods.
Here is an example:
http://www.clarkvision.com/galleries/gallery.landscape-1/web/c072099_L4_01a2-600b.html
The clouds would have come out white in a normal enlargement, and even
with traditional dodging would have no detail. But a good scan
got the information. There are no saturated pixels in the full image.

While some, even many, images may be able to work well in 8-bit,
I have seen and made many that I can do much better with 16-bit
because one can dodge and burn select areas of the image, just like
used to be done in the darkroom. E.g. read Ansel Adams "The Print."
I bet Ansel would cringe at being limited to only 256 levels.

I have had images that needed two 8-bit scans: one for the high to middle
and one for the low to middle. Merging such scans is very difficult.
With 16-bit, it becomes easier to do a better job, and I've never
succeeded with large format because the task is too great.

48 bit capture simply allows you to carry noise and trash
data further up the signal stream. Sooner or later you
have to make tonal manipulations that pare away the
useless codes and noise.

IMO, the need for 48 bit capture is hugely overrated.
Using 24-bit capture simply means doing the paring
earlier rather than later.

This only works if the image is flat enough that you don't need
to dodge and burn much.

Just my opinions...
Roger Clark

 

[Roger N. Clark (change username to rnclark)]

This might be why some people using a 4870 + Vuescan (non TWAIN software)
have reported they can scan larger files?

Maybe my feeble mind is just not grasping the concepts presented in this
thread correctly so late in the afternoon, but in theory, shouldn't you be
able to scan a 4x5 at a slightly higher dpi if it is in vertical orientation
instead of horizontal orientation since not as much data would be
transferred through with each "step" (4 in inches wide instead of 5)?

Yes, just that the epsong 4870 4x5 holder puts the long dimension
across the scanner. But even so, it would not be enough, just
frustratingly close at 3200 ppi (3.4 inches) with the epson software!
Roger

 

[Raphael Bustin]

While some, even many, images may be able to work well in 8-bit,
I have seen and made many that I can do much better with 16-bit
because one can dodge and burn select areas of the image, just like
used to be done in the darkroom. E.g. read Ansel Adams "The Print."
I bet Ansel would cringe at being limited to only 256 levels.

I have had images that needed two 8-bit scans: one for the high to middle
and one for the low to middle. Merging such scans is very difficult.
With 16-bit, it becomes easier to do a better job, and I've never
succeeded with large format because the task is too great.

I understand. In part, it's a matter of being "frugal" with data,
in light of requirements, of course. I mostly do without selective
tonal manipulations, or if I do, they're moderate.

The only real "penalty" for working with 48 bit files is... well
the sort of thing that's being discussed here in this thread,
namely the memory, bandwidth, storage, and throughput
issues that you're currently up against.



rafe b.
http://www.terrapinphoto.com

 

[Ed Hamrick]

Suppose I choose a custom resolution on the 4870 of 3200 ppi (rather
than 4800 or 2400)?

The current implementation in VueScan is to scan at 4800 dpi and
sample down once it's acquired from the scanner. This wouldn't
help.

Also, how hard would it be for you to modify the program so that 4 x 5
can be scanned completely at 4800 ppi and 16 bits per channel?

It's a firmware limitation in the scanner.

Finally, does it make any difference whether one is operating under
Windows or Linux in these matters?

No, this doesn't make any difference.

Regards,
Ed Hamrick

 

[Leonard Evens]

The current implementation in VueScan is to scan at 4800 dpi and
sample down once it's acquired from the scanner. This wouldn't
help.




It's a firmware limitation in the scanner.




No, this doesn't make any difference.

Regards,
Ed Hamrick

All this seems to mean that you can't really use the Epson 4870 to scan
4 x 5 at 16 bits per channel and higher than 2400 ppi in one fell swoop.
It would be interesting to know if it still produces better results
than the Epson 3200 at 3200 ppi, which may be the case.

 

[-]

All this seems to mean that you can't really use the Epson 4870 to scan

4 x 5 at 16 bits per channel and higher than 2400 ppi in one fell swoop.
It would be interesting to know if it still produces better results
than the Epson 3200 at 3200 ppi, which may be the case.

If the 4870 can't do better than 2400, I wonder if the 3200 can either?

Doug

 

[MikeWhy]

If the 4870 can't do better than 2400, I wonder if the 3200 can either?

The Epson 3200 can scan full frame 4x5 at 3200 dpi. I don't do it often any
more, since I can't print a 10x enlargement to use it. The novelty quickly
wore off; Photoshop is not at all responsive when manipulating half-gig plus
files.

SilverFast required an update to work with a scan that size. It was a quick
download from their website.

 

[Leonard Evens]

If the 4870 can't do better than 2400, I wonder if the 3200 can either?

My Epson 3200 has no trouble scanning 4 x 5 at 3200 ppi and 48 bit color
depth (in portrait orientation). I guess Epson designed the firmware so
that the buffer can't hold a full line about 95 mm wide. The problem
with the 4870, as explained by Ed Hamrick, is that the firmware doesn't
allow for a line large enough to accomodate the full width of the frame.
Also, apparently you can't tell the scanner to scan at anything but
4800, 2400, ... ppi. To scan at 3200 ppi, you have to tell it to scan
at 4800 ppi and then scale down to 3200 ppi.

Of course, the Epson 3200 doesn't deliver 63 lp/mm, which is the
theoretical maximum obtainable by scanning at 3200 ppi. It delivers
about half that. So one could argue that it makes sense to scale down
after scanning because that reduces file size and you don't lose much
detail since it wasn't there anyway. So it is still possible that if
you set the 4870 to scan at 2400 ppi, it will produce better results
than the 3200 scanning at 3200 ppi and rescaled to 2400 ppi.

Of course, for most purposes, 2400 ppi should be more than enough for
use with 4 x 5. And also for many purposes, scanning at 24 bit color
depth is adequate. Roger Clark, of course, is always pushing the
envelope of what can be done, so it is not surprising he discovered
these limaitations.

Still it seems bizarre for Epson to design a scanner capable of scanning
at 4800 ppi at 48 bit color depth and also scanning 4 x 5 but not of
doing both. It couldn't cost all that much to double the size of the
buffer.

 

[Tim]

This is a different issue and is a bug in the driver. There is a newer
version of the driver that resolves this issue, Revision 2.0c Version
2.02.
But this is not the issue that Roger has encountered.