[sane-devel] Experiences with the Minolta Scan Dual II

[Alfred]

Hi everyone,

I'd like to document my experiences in getting my Minolta Dimâge Scan Dual 
II (also AF-2820U or FS-V1) to run with sane-avision. I started at zero but 
got a working setup by following these steps.

1. After plugin, the 'hpusbscsi' module was claiming ownership of my 
scanner, so I disabled it by adding it to /etc/hotplug/blacklist. (I still 
run kernel 2.6.24; newer kernels don't even have this module.)

2. Avision could now grab ownership of the scanner, but the calibration 
step failed. I followed in ZP and Dietmar's footsteps and added the 
AV_ONE_CALIB_CMD flag to the Scan Dual II device description in avision.c, 
which did the trick and got me past calibration.

3. When trying to scan some frames, the sane-avision driver seemed to 
timeout before the scanner had moved the film holder to the right frame. I 
had to change STD_STATUS_TIMEOUT from 2500 to 20000 to give the film holder 
enough time to travel to its position. (The 20000 is trial-and-error; I 
started at 5000 and went up in steps of 5000 while trying to scan frame 6, 
which requires the furthest travel.) Error log of the original situation 
available on request.

I have a working scanner now and am fairly happy! Some random comments:

- I do calibration with 'scanimage -d avision > /dev/null', which is not 
very elegant but fairly painless. I do scanning with 'scanimage -d avision 
--frame 4 --resolution 2820 | pnmtopng > outfile.png'.

- Scanning with '--mode Gray' is broken on this scanner. Three different 
parts of the original image are interlaced vertically, like:

   A
   B
   C
   A
   B
   C
   ...

Sample images on request. This means I have to do full RGB scans on my b/w 
negatives...

- How do I enable 16-bit output? The Scan Dual II has 12-bit A/D 
conversion, but sane-avision seems to have no command switch (as per 
'scanimage -d avision -h') to choose between 8-bit and 16-bit output. 
Having the full dynamic range is important with film scanning because of 
the high dynamic range of the negatives themselves and the needs of 
postprocessing.

- How do I bypass the scanner's internal gamma correction and get my hands 
on the raw 'untouched' image data? I find that the internal gamma 
correction loses a lot of information that I could previously obtain with 
Minolta's Windows software if I scanned the negatives in as slides. (This 
probably applied a high gamma to the original 12-bit data, which preserves 
the extremes.) For archiving purposes, I want as much dynamic detail in my 
negatives as possible. I rather get the uncorrected image and do the gamma 
correction myself, than getting an acceptable-but-not-super pre-corrected 
image back from the scanner. Yes, I'm a terrible perfectionist ;-)  Do I 
have to set 'gamma' to '1.0' in the source code? Or maybe the 
'--gamma-correction' switch may help here, but I'm not sure how it works. 
Do I have to supply 255 values on the command line?

- The Scan Dual II has support for autofocus, manual focus and 
autoexposure. The autoexposure is probably done in software, but what about 
the autofocus? Is it possible to control or enable it?

- A command switch '--negative' or '--invert' would be welcome for scanning 
negatives, together with maybe an option to write landscape-oriented images 
instead of portrait ones. But maybe these are typical frontend 
responsibilities.

All in all, the Linux support for this device is quite good and I would 
really like to thank everyone who made it possible! Not being able to use 
my expensive hardware under Linux was one of the last things that was 
keeping me from going 100% Linux :-)

Cheers,
--A. Klomp

The Netherlands