Minolta 5400 streaks

[Robert Feinman]

It seems that some people are bothered by streaks with
the Minolta 5400 and Vuescan and some (like me) are not.
It is possible that this is a function of manufacturing
differences between various copies of the scanner.
It would seem that there is some problem with Vuescan
compensating for variations in sensor sensitivity. I
would think that this shows up only as a streak when
scanning at the highest resolution, since that is the
only case where a single scan pixel can be resolved.

This would imply that the user is attempting to make about
an 18x enlargement or is trying to extract a fairly small
area from the image.
I suspect that most users seldom have a need to actually
scan at 5400 dpi and thus I'm wondering if they are being
unnecessarily worried. Scanning at a lower dpi or
resampling in an image editor later would tend to obscure
any streaks.

One idea that occurs to me is for users troubled with
streaks to make a scan of a blank frame and then use
this as an adjustment layer in Photoshop. With the right
combination of blend mode and opacity for this layer it
should be possible to "calibrate" your scanner post-scan.
This assumes that the streaks are always in the same place,
which would be the case if they are caused by sensor
sensitivity issues.

Perhaps someone who is having streaking problems could try
my layer idea and report back as to the degree of success.

 

[Fernando]

It seems that some people are bothered by streaks with
the Minolta 5400 and Vuescan and some (like me) are not.
It is possible that this is a function of manufacturing
differences between various copies of the scanner.
It would seem that there is some problem with Vuescan
compensating for variations in sensor sensitivity.
I
would think that this shows up only as a streak when
scanning at the highest resolution, since that is the
only case where a single scan pixel can be resolved.

Yes, this is very possible, at least I think so.

This would imply that the user is attempting to make about
an 18x enlargement or is trying to extract a fairly small
area from the image.
I suspect that most users seldom have a need to actually
scan at 5400 dpi

Honestly, I don't see the point in shelling out €700 (in my case) to
buy a 5400dpi scanner, without actually needing such resolution.
I obviously purchased the DSE5400 precisely because I needed all the
resolution I could extract from my film... I'd have happily purchased
an Imacon 929 if I'd had $20000 of spare cash.
I've seen through microscope analysys that even the DSE5400 manually
focused within Silverfast could not extract all the little details I
have in some of my shots (tripod, primes, high resolution slide film).
Anyway:

One idea that occurs to me is for users troubled with
streaks to make a scan of a blank frame and then use
this as an adjustment layer in Photoshop. With the right
combination of blend mode and opacity for this layer it
should be possible to "calibrate" your scanner post-scan.
This assumes that the streaks are always in the same place,
which would be the case if they are caused by sensor
sensitivity issues.

This is extremely interesting, and actually I had such an idea in the
past, but I wasn't able to put it in practice.
First of all, the streaks are not only luminance-related: there are
color variations too (so a trivial luminance adjustment layer would
not suffice). So, I don't know how to procede! Any hints? A guide
would be helpful.

I also have to say, that the streaks are only one of the three
problems I'm experimenting with Vuescan, and the other two are equally
show-stoppers for me: 1) low DMax (inability to read through shadows
as much as another software -Silverfast- does), with too little
dynamic range between absolute black -green, actually: see later- and
absolute white, and 2) heavy green casts in the shadows, that I was
not able to fix even with custom profiles. I see those casts even when
scanning completely opaque originals, such as the Slanted Edge Test
target slide. I don't see this cast with Silverfast, nor with Minolta
Scan Utility.

I think Bart van der Wolf could put this into tech words much better
than me (he experienced the very same problems with his DSE5400
and Vuescan, and reported to Ed many times).

Thanks for your help!

Fernando

 

[Alan Browne]

It seems that some people are bothered by streaks with
the Minolta 5400 and Vuescan and some (like me) are not.

The only way I see any streaks with the 5400 (using Vuescan or Dimage) is if I
scan at 5400dpi, zoom in on a dark area (400%) and turn up the monitor
brightness. Then there are faint streaks.

I believe the 5400, as wonderful as it is, suffers great variance in some areas
when it is last touched by Minolta and shipped. This has led to some people
having to return units (in some case several times) until they get a "good one".
I'm happy that I appear to have gotten a good one on the first go.

It is possible that this is a function of manufacturing
differences between various copies of the scanner.
Yes.

It would seem that there is some problem with Vuescan
compensating for variations in sensor sensitivity. I
would think that this shows up only as a streak when
scanning at the highest resolution, since that is the
only case where a single scan pixel can be resolved.

This would imply that the user is attempting to make about
an 18x enlargement or is trying to extract a fairly small
area from the image.
I suspect that most users seldom have a need to actually
scan at 5400 dpi and thus I'm wondering if they are being
unnecessarily worried. Scanning at a lower dpi or
resampling in an image editor later would tend to obscure
any streaks.

It doesn't matter that someone has a need or not. The scanner is spec'd to do
somehting, and in some cases is failing to do it very well. It is true, however
that a lot of the noise people are seeing on their monitors will likely not be
visible on a print.

One idea that occurs to me is for users troubled with
streaks to make a scan of a blank frame and then use
this as an adjustment layer in Photoshop. With the right
combination of blend mode and opacity for this layer it
should be possible to "calibrate" your scanner post-scan.
This assumes that the streaks are always in the same place,
which would be the case if they are caused by sensor
sensitivity issues.

Perhaps someone who is having streaking problems could try
my layer idea and report back as to the degree of success.

Not a bad idea. But I'll decline the test opp, thanks!

Cheers,
Alan

 

[David O'Rourke]

I've been coping with this issue for sometime now. I work with older
Minoltas, Scan Elite, and Elite II. Both do the streak thing 90% of the
problems occur in the shadows. The streaks are one pixel wide, different
colours, and the line goes from the start of the scan to end. Some are on
occastion bad enough to be visible in the lower mid-tones.

Unfortunately resolution isn't the problem. I was pretty sure that I had
encountered the streaks at other resolutions. To be sure I just scanned a
very dark slide at 2820 and again at 1410. I pumped up the brightness in
Photoshop. I was able to find the streak at the same location in the image,
and still one pixel wide.

It may well be some kind of internal "pacing" issue. The Scan Elite is
SCSI. The Elite II is USB and firewire. On one occastion in the past I had
a USB problem and the result was all kinds of coloured streaks just like in
the shadows of a normal scan, but maybe 100x as many.

When can't res-down the image in Photoshop to mask the problem, I do copy
and pastes of single rows of pixels to cover the problem.

David

 

[Wilfred van der Vegte]

David O'Rourke wrote:

It may well be some kind of internal "pacing" issue. The Scan Elite is
SCSI. The Elite II is USB and firewire. On one occastion in the past I had
a USB problem and the result was all kinds of coloured streaks just like in
the shadows of a normal scan, but maybe 100x as many.

Maybe USB is the problem: I'm using the 5400 with FireWire and haven't
seen any significant streaks. My Dimage Scan Speed (SCSI, too) did show
these streaks every now and then.
Perhaps it would be good to make an inventory whether the people with
problems are using USB?

 

[Fernando]

Maybe USB is the problem: I'm using the 5400 with FireWire and haven't

I have the DSE5400, see the streaks (only with Vuescan) and use
Firewire.
For once, I tried USB 2, and found the very same streaks at the very
same positions.

Fernando

 

[Matthew Dranchak]

My SE5400 is Firewire connected.
I did not observe the streaking problem when I tried to use Vuescan (trial
version).
I do not use Vuescan because of other problems.
I have a slightly under exposed slide (Kodachrome) of my children in a dark
shadow area wearing bright red caps that really show up. There is a
pronounced smooth shading of the red caps when scanned using the Minolta
software or when placed in a slide projector for viewing. When scanned using
Vuescan, the caps show up all mottled with areas looking like an elevation
scan of a mountain with various shades of reds. I spent umpteen days trying
to get Vuescan to give me a good scan.
Because the Vuescan scan did not meet my expectations I just did not
purchase Vuescan and have no intention to do so.

 

[Fernando]

pronounced smooth shading of the red caps when scanned using the Minolta
software or when placed in a slide projector for viewing. When scanned using
Vuescan, the caps show up all mottled with areas looking like an elevation
scan of a mountain with various shades of reds.

Go under the "Filter" tab and deactivate "IR Cleaning" (default is
"Light": set it to "None"). This should solve your problem (KC
emulsion is partially opaque to IR light, and densest zones are seen
by the software as dust accumulation to be "cleaned").

Bye!

Fernando

 

[Robert Feinman]

I've been coping with this issue for sometime now. I work with older
Minoltas, Scan Elite, and Elite II. Both do the streak thing 90% of the
problems occur in the shadows. The streaks are one pixel wide, different
colours, and the line goes from the start of the scan to end. Some are on
occastion bad enough to be visible in the lower mid-tones.

Unfortunately resolution isn't the problem. I was pretty sure that I had
encountered the streaks at other resolutions. To be sure I just scanned a
very dark slide at 2820 and again at 1410. I pumped up the brightness in
Photoshop. I was able to find the streak at the same location in the image,
and still one pixel wide.

If you are getting 1 pixel wide streaks at less than full scan resolution
this may be from one of two problems. The first is that the down sampling
may be achieved by just using less of the sensor spots during a scan.
For example, half max would just use every other sensor. If you were
unlucky enough the miscalibrated sensors would be the one used.

The other possibility is that the downsampling is done in software
using a simple algorithm such as averaging adjacent pixels. Then a
bright one next to a normal one would create a brighter average as well.
It should be possible to test this if you change your downsample ratio.
For example on the 5400 instead of 2700 use 2699 or 2701 or the like.
In one test I did this way the scan times increased dramatically showing
that the software needed to do much more processing.

As to the poster who claimed he needed the full resolution, I can't speak
for your requirements, but in general 35mm photography is not usually
seen as being able to withstand such a large degree of magnification.
Having said that, I have an example in the tips section of my web site
showing quite acceptable detail with 5400 dpi scan. If you are not
maintaining full resolution to your final output then I'm not sure you
will really see the difference in the final print.
If you can consistently archive such high quality results, I'm sure all
of us would appreciate you describing your workflow.

Another poster said that he didn't think my layer idea would work since
the streaks were not of a uniform color. This would seem to point to some
other problem, rather than miscalibration of the scanner. Changing colors
implies that there is some noise being added to the scan as well.
The way to test for this is to scan an opaque image and thus rule out
film effects. If the streaks are still colored than it would seem that
there is a noise problem in the A/D converter. Other possibilities
include powerline noise or electromagnetic interference from adjacent
equipment.

Perhaps some of us just have an electrically quieter locale than others.
I would suggest those having problems try moving the scanner and/or
getting the power feed from a different outlet. In addition some power
strips do much better at filtering out noise. The noise may also be
getting in through the USB wire or port. A better quality cable and the
use of a port with no other devices running might be worth a try.

If the streaks continue to be colored than I don't see how the scanner
software could be enhanced to compensate for this. Perhaps the packages
that seem not to streak are doing some sort of low level clipping to
remove the noise.

 

[Nils Heidorn]

Hello,
as one who knows quite some scanners & Scanner Softwares ( I know
SilverFast best but also others ), maybe i can be of help.

To get hold of this "streak" Problem, you can do some steps to clarify
the situation:

-1- Please tell me, are the "streaks" directed in the movement
direction "vertical" or in the CCD Sensor direction "horizontal" ?
( I would assume they are vertical, but havent read it here so i
better ask ).

-2- Are the "streaks" always at the same position ?

-3- You keep saying that the stripes occur in Vuescan, but not in
SilverFast or the native Software, are the stripes really *only*
visible in Vuescan or are they just less obvious in the latter
Softwares ( find out by sharpening the image and/or stretching
contrast in the dark areas to find out ).

If you could try out & answer the above Questions i may have a
specific idea whats the problem...

Thanks,

Nils

 

[Fernando]

As to the poster who claimed he needed the full resolution, I can't speak
for your requirements, but in general 35mm photography is not usually
seen as being able to withstand such a large degree of magnification.

"In general" is quite a generic statement, isn't it? ;-)
I need as much resolution I can get, and the DSE5400 was the only
scanner with "enough" resolution that I could afford.
That's why I'm trying to have it work properly to its advertised
specifications. It does with Silverfast AI, and it does with Minolta
Scan Utility when I'm lucky enough to get a properly focused scan.
I hope that I will get optimal results also from Vuescan, one day.
Who knows.

Another poster said that he didn't think my layer idea would work since
the streaks were not of a uniform color. This would seem to point to some
other problem, rather than miscalibration of the scanner. Changing colors
implies that there is some noise being added to the scan as well.
The way to test for this is to scan an opaque image and thus rule out
film effects. If the streaks are still colored than it would seem that
there is a noise problem in the A/D converter. Other possibilities
include powerline noise or electromagnetic interference from adjacent
equipment.

I don't know about that, but I'm seeing differently colored streaks
when scanning my Slanted Edge Test target, which is not a film: just a
framed razor blade. I have colored streaks that are 1 pixel wide and
run for the whole length of the frame.
Scanning the same target seems to get the same streaks of the same
color and in the same position.
Seems a calibration problem to me, because Minolta Scan Utility and
Silverfast don't show any streaks, under the same conditions.

Perhaps some of us just have an electrically quieter locale than others.

If EM interference were the cause, I should see the streaks with
Minolta Scan Utility and Silverfast as well.
I don't.

If the streaks continue to be colored than I don't see how the scanner
software could be enhanced to compensate for this. Perhaps the packages
that seem not to streak are doing some sort of low level clipping to
remove the noise.

They don't. Streaks (with Vuescan) also appear in midtone areas under
certain conditions (proper amount of sharpening). No amount of
clipping would eliminate them. They don't appear with Minolta and
Silverfast software, no matter how much sharpen I try.

Minolta Scan Utility and Silverfast just detect those streaks via some
kind of blank scanning, than store the defect map (it's just an
integer array after all) in some area of the scanner memory.
When they scan the actual film, they somehow mask out the defects
because they know the deviation of each single photodiode of the
linear CCD sensor.
Vuescan tries to do the same thing, but for some reason, it fails to
obtain results that are as good. Sure its calibration is better than
nothing, but not as good as Silverfast's or MSU's.

Now, there is a way to build and use such "colored" defect map with
Photoshop?
I think so.
I'm not very proficient with Photoshop layers, masks, blending options
and so, but you are, for sure (I've visited your site many times).
Could you help?

Thank you!

Fernando

 

[Matthew Dranchak]

Go under the "Filter" tab and deactivate "IR Cleaning" (default is
"Light": set it to "None"). This should solve your problem (KC
emulsion is partially opaque to IR light, and densest zones are seen
by the software as dust accumulation to be "cleaned").

Tried that. It doesn't help.
I tried all the avenues that Vuescan allowed and no combination of features
provided a good scan. Your suggestion even made things worse.
Vuescan is not for me, period.
Matt D

 

[Fernando]

Hello,
as one who knows quite some scanners & Scanner Softwares ( I know
SilverFast best but also others ), maybe i can be of help.

To get hold of this "streak" Problem, you can do some steps to clarify
the situation:

Actually, we already repeated all this over and over just during the
last few days anyway, my answers:

-1- Please tell me, are the "streaks" directed in the movement
direction "vertical" or in the CCD Sensor direction "horizontal" ?
( I would assume they are vertical, but havent read it here so i
better ask ).

They are along the movement direction, that is, along the LONG side of
the frame.

-2- Are the "streaks" always at the same position ?

Yes, they always seem to be in the same position.

-3- You keep saying that the stripes occur in Vuescan, but not in
SilverFast or the native Software, are the stripes really *only*
visible in Vuescan or are they just less obvious in the latter
Softwares ( find out by sharpening the image and/or stretching
contrast in the dark areas to find out ).

They only appear in Vuescan at typical sharpening levels (and I
sharpen my scans quite a bit). They are not present at all with the
other two software tools, at the same sharpening settings.
I never tried "extreme" (unreasonable) sharpening settings... maybe
when exaggerating, they could also appear with other software. If you
think this is an important point, I may do some tests. Is there an
optimal test for you that I can try?

Thanks for your contribution.

Fernando

 

[Pasi Savolainen]

I never tried "extreme" (unreasonable) sharpening settings... maybe
when exaggerating, they could also appear with other software. If you
think this is an important point, I may do some tests. Is there an
optimal test for you that I can try?

Say, do you scan at 16bit/channel or 8bit/ch?
For whatever reason, vuescan seems to produce much better results when
scanning in 16bit, even if actually saving to 8bit. (This is on another
scanner)

 

[Wilfred]

It seems that some people are bothered by streaks with
the Minolta 5400 and Vuescan and some (like me) are not.
It is possible that this is a function of manufacturing
differences between various copies of the scanner.

It may be interesting to know that a on a German filmscanner forum, a
Nikon LS-5000 user has reported similar streaks when using VueScan. He
claims that the streaks started to appear after version 8.0.4.
For those who understand German:
http://f27.parsimony.net/forum67102/messages/5833.htm
For those who don't:
http://tinyurl.com/5uzcp

 

[Fernando]

Say, do you scan at 16bit/channel or 8bit/ch?

Never scanned at 8 bit/channel in my life. ;-)

Fernando

 

[Alex Stols]

It may be interesting to know that a on a German filmscanner forum,
a
Nikon LS-5000 user has reported similar streaks when using VueScan.
He
claims that the streaks started to appear after version 8.0.4.
For those who understand German:
http://f27.parsimony.net/forum67102/messages/5833.htm
For those who don't:
http://tinyurl.com/5uzcp

The English translation especially is worth while to read

Regards, Alex
alexUNDERSCOREstolsATxs4allDOTnl

 

[Wilfred]

The English translation especially is worth while to read

Yes - I discovered that, in the translation, you can also click on the
replies to this message, which are translated as well. These
translations are even funnier;-)

Anyway it's probably better than nothing. The translator at
http://dictionary.reference.com/translate/text.html is slightly better
but it requires cut-and-paste into a form - it doesn't work with URLs
directly.

 

[Alex Stols]

Yes - I discovered that, in the translation, you can also click on
the
replies to this message, which are translated as well. These
translations are even funnier;-)

Anyway it's probably better than nothing. The translator at
http://dictionary.reference.com/translate/text.html is slightly
better
but it requires cut-and-paste into a form - it doesn't work with
URLs
directly.

Sites like that are always useful for us, foreigners ))

Regards, Alex
alexUNDERSCOREstolsATxs4allDOTnl

 

[Don]

It may be interesting to know that a on a German filmscanner forum, a
Nikon LS-5000 user has reported similar streaks when using VueScan. He
claims that the streaks started to appear after version 8.0.4.
For those who understand German:
http://f27.parsimony.net/forum67102/messages/5833.htm
For those who don't:
http://tinyurl.com/5uzcp

Even without this conclusive evidence, it has been clear for about a
year that this is a VueScan bug and not a (Minolta) hardware problem.

I strongly suspect this to be a communications problem. Here's why:

Communicating with peripherals directly is not something casual
programmers can - or should try to - do. Writing low level drivers is
not for the faint of heart - or for the petulant... ;o)

There is a (good!) reason why scans with Minolta software take
significantly longer. Commercial software is usually very conservative
by making sure the communication is reliable and that takes time (be
it due to handshaking or timing). By cutting corners one can speed
this up, but at the risk of dropping data. Furthermore, when this
direct communication routine itself is sloppy the calling program may
not even be aware of the missing data.

Hence, if you examine the raw VueScan streaked scan by looking at
individual channels (at 300-400% magnification) I would strongly
suspect you'd see black stripes in some channels and not in others. It
may be necessary to perform this test with early versions of the
program because in latter versions (I suspect) the author may have
tried to "mask" this by averaging nearby pixels/lines thus making
these streaks "faint". As history teaches VueScan bugs are rarely (if
ever?) really fixed, instead they are simply masked, so ultimate
caution is advised!

Another (circumstantial) evidence that the VueScan's author is
incapable of writing low level drivers is when he floated a trial
balloon of incorporating TWAIN drivers. I strongly suspect this is
because tests using Minolta TWAIN drivers produced streak-free scans
exposing his programming shortcomings.

Streaking solved? You're welcome! ;o)

Don.