4990 C41 Scanning Issues

[rafe b]

Hello gang. Received my new Epson 4990 yesterday and
am having serious issues with regard to scan exposure,
gamma, and color consistency.

I've been scanning film for years with lots of other scanners,
but this one's giving me fits so far.

I'm scanning C41, let's say Portra 160VC or NC. I'm scanning
4x5 film at, say, 1200 or 2400 dpi.

I started with Epson's driver, both in TWAIN and standalone.
Bottom line, the final scan looked nothing like the preview,
generally with one or more channels having bottomed-out
histograms in Photoshop. Not consistently wrong, either.
Just no palpable correlation between the controls in the
scan driver and the final image in Photoshop.

Eventually gave up and tried Silverfast SE -- the freebie
that comes with the scanner. This yields smooth,
consistent histograms -- consistently clipped at both
ends. No image content within 10 codes of either end,
and a hard cliff at both ends. It's a pleasant image
to look at on screen but IMHO a very flawed scan.

Vuescan (8.2.06) gives the best scans so far, but it's too
early to tell if the fix will hold. I've never made my peace
with VueScan, try as I might...

Looking for any/all suggestions for scanning C41
transparencies on the Epson 4990. Thanks in advance.


rafe b
www.terrapinphoto.com

 

[Dan M]

I'm having the same problem. Have had some success so far playing with
the settings under "Configuration." So far, Vuescan has given me the
best results, though I can't find the option there to use Digital Ice
with prints. Just got the 4990 yesterday, so I'm still setting it up.

Dan

 

[Don]

I started with Epson's driver, both in TWAIN and standalone.
Bottom line, the final scan looked nothing like the preview,
generally with one or more channels having bottomed-out
histograms in Photoshop. Not consistently wrong, either.
Just no palpable correlation between the controls in the
scan driver and the final image in Photoshop.

Silly question, but here goes anyway... Are you sure you're looking at
the appropriate preview window?

I'm not familiar with Epson drivers but sometimes there are two
preview windows "raw" and "processed". Looking at the "wrong" one
(i.e. raw) would certainly not be consistent with the final scan.

If that's not the case, then please ignore.

Eventually gave up and tried Silverfast SE -- the freebie
that comes with the scanner. This yields smooth,
consistent histograms -- consistently clipped at both
ends. No image content within 10 codes of either end,
and a hard cliff at both ends. It's a pleasant image
to look at on screen but IMHO a very flawed scan.

In order to judge the image by the histogram you need the histogram at
the same bit depth as the scan (not to mention you need a raw scan!).

If the scan is at 16-bit but the histogram is at 8-bit (the usual
case) *all* scans produce "smooth" histograms because such an 8-bit
histogram is a rounded down approximation.

If you are not using raw scans then the histogram will further be
"corrupted" (read, smoothed out) by all the image editing the scanner
software applies. In such a case the histogram is only useful as an
editing tool, but it says absolutely nothing about the actual scan.

Indeed, even though a histogram in an invaluable tool for both
scanning and editing, it has to be interpreted correctly, in context.
Just having a "smooth" histogram by itself is pretty meaningless.

Vuescan (8.2.06) gives the best scans so far, but it's too
early to tell if the fix will hold. I've never made my peace
with VueScan, try as I might...

You're in good company! ;o) Nobody who cares for even a semblance of
quality can make peace with Vuescan...

Don.

 

[rafe b]

In order to judge the image by the histogram you need the histogram at
the same bit depth as the scan (not to mention you need a raw scan!).

If the scan is at 16-bit but the histogram is at 8-bit (the usual
case) *all* scans produce "smooth" histograms because such an 8-bit
histogram is a rounded down approximation.

If you are not using raw scans then the histogram will further be
"corrupted" (read, smoothed out) by all the image editing the scanner
software applies. In such a case the histogram is only useful as an
editing tool, but it says absolutely nothing about the actual scan.

Indeed, even though a histogram in an invaluable tool for both
scanning and editing, it has to be interpreted correctly, in context.
Just having a "smooth" histogram by itself is pretty meaningless.

My standard and reference for the Histogram tool is
Photoshop's, though I implicitly trust the tool in
NikonScan as well. I'm not worried about gaps
or missing codes at this point. I'll be using the
4990 at 2400 dpi and 24-bit color, scanning only 4x5.
For anything smaller I use the Nikon.

You're in good company! ;o) Nobody who cares for even a semblance of
quality can make peace with Vuescan...

I'm not going there... not with you, Don. It's never
been my favorite tool but lots of folks apparently
find it useful.

It was nice to have it on hand as an alternative to
the stuff that came bundled with the Epson. In
this case, I found VueScan's image processing
much less capricious than Silverfast's.


rafe b
www.terrapinphoto.com

 

[Bart van der Wolf]

SNIP

I understand Rave's reduced (from native) 2400 ppi resolution (huge
300 MB full-frame vs 1.2 GB files @ 4800 ppi, times two for
16-bit/channel), although Vuescan allows to average a 4800 ppi scan to
a 2x2 pixel average (which reduces scanner noice by a factor of 2) for
the same output file size. It won't solve Rave's current clipping
issue, but it will increase accuracy.

But then "good" in this context, in Don's vocabulary, tends to be
"questionable" in more balanced circles, although I apparently don't
need to tell you (I'm just mentioning it for newbies, to create some
balance in the feedback).

VueScan, although not perfect for all scanners in all situations (but
then which software is?), allows a lot of user intervention, which may
be a bit mind boggling to some.

Bart

 

[Bart van der Wolf]

Hello gang. Received my new Epson 4990 yesterday [...]

Congratulations, it's supposed to be quite a capable scanner (for its
price/technology), I might get me one myself in due time (after
Photokina 2006).

SNIP

Vuescan (8.2.06) gives the best scans so far, but it's too
early to tell if the fix will hold. I've never made my peace
with VueScan, try as I might...

Before (hopefully) receiving some additional (hands-on) scanner driver
guidance, it might also help to first update to the latest version
(VueScan 8.3.20 as of this writing, or other (Epson/Silverfast)
software versions). It may get to become increasingly
confusing/difficult if mixed VueScan versions are discussed, and it'll
also reduce your need to adapt to a changed feature set after getting
somewhat re-acquainted.

I've written a kind of Usenet 'Negatives to Raw' scanning suggestion
for the Minolta DSE 5400 (model 1)which also clips negative scans, and
that might well be somewhat applicable to the Epson, but since I have
no personal experience with the 4990 I can't tell for sure, yet.

One of the basic scanner "unknowns" for me at this time is whether
the "Input tab, lock exposure time" allows to input anything for the
4990. If that is possible (can be verified with clocking/timing the
actual scan time"), then VueScan will probably allow to optimize the
per channel exposure times (=getting rid of the mask density, while
optimizing exposure/integration time).

Bart

 

[-]

Eventually gave up and tried Silverfast SE -- the freebie

that comes with the scanner. This yields smooth,
consistent histograms -- consistently clipped at both
ends.

Did you try the HDR's raw-like output? That should eliminate the clipping
although the file will not be "printer ready" and require a separate final
preparation/ finishing workflow.

Doug

 

[rafe b]

Did you try the HDR's raw-like output? That should eliminate the clipping
although the file will not be "printer ready" and require a separate final
preparation/ finishing workflow.

I've never worked with HDR files... Can Photoshop 7 deal
with them?

All my image files are edited to some extent so "printer
ready" isn't an issue.

I'm beginning to think the scanner has some basic
hardware issues. I took the same 4x5 chrome, same
crop, using "No Color Correction" in the Epson scan
driver -- and scanned it repeatedly, every 5 or 10
or 15 minutes over the course of a couple hours.
Between scans, I changed nothing but the resolution
(1200/2400) and bit depth (24/48bit.)

Here's the data. "M" = mean, "SD" = standard
deviation. The total time span from first file to
the last is maybe an hour or two.

Between ep110 and ep111, the mean value of
the green channel shifted 20 points.


file RGB M RGB SD RED M RED SD GRN M GRN SD BLU M BLU SD

ep102 70.41 29.87 59.85 25.55 71.63 32.71 92.62 30.71
ep103 70.53 29.87 59.88 25.56 71.58 32.7 93.92 30.79
ep104 70.38 29.87 59.94 25.59 71.57 32.7 92.45 30.64

ep105 77.1 30.49 56.78 25.51 84.75 33.9 91.44 30.6
ep106 77.04 30.54 56.69 25.55 84.69 33.95 91.45 30.68

ep107 77.65 27.11 56.82 20.52 85.58 31.11 91.85 27.49
ep108 78 27.12 57.97 20.6 85.6 31.09 91.81 27.44
ep109 78.95 27.99 57.19 20.31 87.46 32.88 92.63 27.35

ep110 80.08 28.07 58.24 20.18 88.48 33 94.61 27.89
ep111 69.13 26.75 59.3 20.4 68.64 30.56 98.54 28.36
ep112 66.62 26.47 59.57 20.39 64.07 30.07 99.51 28.5
ep113 70.97 26.98 59.07 20.4 72.25 30.99 96.59 28.13

rafe b
www.terrapinphoto.com

 

[degrub]

just a couple thoughts.
lamp variation ? which may have two sources - internal power supply or
external power fluctuation

looking at the SD's , it appears the noise decreased slightly over time
- scanner temperature change ? lamp stabilized ?

is the scanner self calibrating each time ?

would changing the resolution change the data the scanner driver uses to
generate the output file ? ditto for bit depth ?

regards,

 

[Don]

My standard and reference for the Histogram tool is
Photoshop's, though I implicitly trust the tool in
NikonScan as well.

A potential problem is that both are 8-bit histograms. If the scan is
also 8-bit then everything's fine, but when the scan is 16-bit then
the 8-bit histogram is only an approximation.

The trouble is both Photoshop's and Nikon's way of reducing 16-bit
histograms to 8-bit may actually distort them due to the way those
16-bit bins are combined.

I'm not worried about gaps
or missing codes at this point. I'll be using the
4990 at 2400 dpi and 24-bit color, scanning only 4x5.
For anything smaller I use the Nikon.

Missing codes and gaps are not really a problem if the histogram is of
a raw scan. Indeed, it's only to be expected especially after gamma
was applied.

....

I'm not going there... not with you, Don.

Come on, I'm just kidding. Don't lose your sense of humor.

It was nice to have it on hand as an alternative to
the stuff that came bundled with the Epson. In
this case, I found VueScan's image processing
much less capricious than Silverfast's.

I tried Silverfast on two occasions but on a very cursory basis. It's
exactly the opposite of what I want from scanning software i.e. it
puts a lot of emphasis on heavy image editing and automation. I don't
think it's even possible to get a raw scan with Silverfast. So, I
didn't pursue it any further.

Don.

 

[Don]

But then "good" in this context, in Don's vocabulary, tends to be
"questionable" in more balanced circles

Still obsessing with Don's messages, eh, Bart?

Pretending not to read them but then taking every opportunity to
piggyback on a comment exposes the hollowness of your protestations.

And, of course, you missed the irony completely...

Not at all surprising since you apparently read the message through a
red haze...

Don.

 

[Surfer!]

On Mon, 2 Jan 2006 01:44:15 +0100, "Bart van der Wolf"

<Snip>

At times this becomes like two 5yo boys yelling at each other 'I was
right' and so on...

 

[Don]

At times this becomes like two 5yo boys yelling at each other 'I was
right' and so on...

Hello?

*Rafe* complains about Vuescan.
Bart snarls, and lashes out at *Don*!?!?

I just pointed out how irrational this unprovoked Bart's outburst was.

Don.

 

[rafe b]

*Rafe* complains about Vuescan.

Not incessantly, as you are wont to do.

I don't complain that much, and in fact it's
saved my bacon of late with my Epson 4990.
Damn the user interface, all I need is the raw
data. The drivers that came with this scanner
are worthless.


rafe b
www.terrapinphoto.com

 

[Don]

Not incessantly, as you are wont to do.

I am not want to, I just state the facts. Like this:

I've never made my peace with VueScan, try as I might...

==> So, the question still stands: Given the above, why does Bart's
head explode when I comment with:

You're in good company! ;o)

The logical thing (inasmuch as Bart's outburst can be called logical)
would be to attack you for making the original statement, not me for
posting a *tongue-in-cheek* comment!

The only rational explanation appears to be that Bart has a huge chip
on his shoulder about Vuescan and a pathological obsession with Don's
messages.

Don.

 

[Roger S.]

Bart wrote: "One of the basic scanner "unknowns" for me at this time
is whether
the "Input tab, lock exposure time" allows to input anything for the
4990. If that is possible (can be verified with clocking/timing the
actual scan time"), then VueScan will probably allow to optimize the
per channel exposure times (=getting rid of the mask density, while
optimizing exposure/integration time). "

Is there evidence that this can be done on any scanner? Scanhancer
Erik seemed to think this was a myth and done by software in this post:
http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00EiGI

 

[rafe b]

Bart wrote: "One of the basic scanner "unknowns" for me at this time
is whether
the "Input tab, lock exposure time" allows to input anything for the
4990. If that is possible (can be verified with clocking/timing the
actual scan time"), then VueScan will probably allow to optimize the
per channel exposure times (=getting rid of the mask density, while
optimizing exposure/integration time). "

Is there evidence that this can be done on any scanner? Scanhancer
Erik seemed to think this was a myth and done by software in this post:
http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00EiGI

The Nikon LS-8000 and LS-8000 have individual
analog gain controls that directly control exposure.

I use these controls routinely.


rafe b
www.terrapinphoto.com

 

[Bart van der Wolf]

SNIP

Is there evidence that this can be done on any scanner?

Yes, several of my scanners will allow to adjust the exposure time (as
witnessed by longer scan times and reduced S/N ratio).

Bart