[sane-devel] Re: Shading problem on HP2100C and Negative scanning on EPSON

[Till Kamppeter]

I have found a possible reason for the stripes on the Epson Perfection 
1260: It is a self-calibration problem. Before the scanner scans the 
negative, it calibrates itself by scanning under the calibration hole. I 
don't know exactly whether it is only supposed to scan inside the hole 
to get a white reference or also outside the hole to get a black 
reference. After scanning the calibration hole it drives back the 
printhead to and starts again moving it to the negative in the scanning 
hole. I assume that the stripes come because of the scanner scanning the 
border of the calibration hole when it is supposed to be completely 
inside or outside the hole. The following experiment lead me to this 
assumption:

To exclude any artifacts I made a fresh, clean compilation of 
sane-backends 1.0.9pre1 with plustek-backend 0.44-8. Then I did two 
experiments, before very experiment I did:

rmmod scanner
modprobe scanner
rm -rf ~/.sane/xsane
xsane &

In both cases I set "negative" as scanning source and "full range" as 
source type. I used the colour inversion button (2nd button from the 
left in main window) to invert the negative's colors.

I did both a preview (50 dpi) and a real scan (1200 dpi). For both scans 
I proceeded the same way with the TPA.

First experiment:

During the auto-calibration of the scanner I removed the negative frame. 
I put the TPA lamp directly onto the glass, at the position where it is 
normally when it is on the frame. So the scanner sees always white 
during the calibration. As soon as the scanhead returns, I put the frame 
in place and the lamp onto it, so that the scan itself is done with the 
normal setup.

Second experiment:

I close the calibration hole with a material where no light can pass 
through and which lower side is nearly black. During the 
auto-calibration I put the TPA lamp away from the scanner, with the 
light down, so that the light getting onto the scan head is as low as 
possible. So the scanner should see always black during the calibration. 
After the calibration I put the lamp back onto the frame so that the 
negative is scanned with the light.

The second experiment gives the stripes about which I reported. I got 
the also on a test with 0.44-6. It also gave a worse color quality (very 
buish-greenish) than the first experiment:

http://www.linuxprinting.org/till/tmp/EpsPerf1260TPA_calnolight.png

The first experiment shows NO STRIPES and a better color quality:

http://www.linuxprinting.org/till/tmp/EpsPerf1260TPA_calwithlight.png

I scanned also the corresponding photo in 300 dpi to compare:

http://www.linuxprinting.org/till/tmp/EpsPerf1260TPA_photo.png

    Till


Jaeger, Gerhard wrote:
> Hi Till,
> 
> the stuff becomes more and more strange!
> Let's make a summary and please correct me if I'm wrong.
> 
> 0.44-6: The version works, but you'd like to have raw pixel data
>             (non-inversion)
> 0.44-8: Works, but we have stripes and wrong colors
> 
> So my problem is currently the stripes. I know, that there's a problem
> with negative scanning. I can reproduce this on my OpticPro UT24,
> but I never got those stripes...
> 
> One last test please. Can you reinstall the 0.44-6 and check the
> stripes stuff.
> 
> Thanks gain for your patience, but currently you are my only
> Epson TPA tester...
>   Gerhard
> 
> On Wednesday, 16. October 2002 15:20, you wrote:
> 
>>I have taken the plustek-usbmap.c from 0-44.6 and did not change the
>>rest (being 0.44-8). The vertical stripes on the negatives are still
>>there and the colors for negative scanning are much worse than with pure
>>0.44-8.
>>
>>    Till
> 
> 
>