[sane-devel] Acer Scanwit 2720S backend -- alpha version

[Max Ushakov]

Hi,

I have written a preliminary backend for Acer ScanWit 2720S film
scanner. This is very, very ALPHA.

You can download it from

ftp://mccme.ru/users/ushakov/scanwit

Even if you do not have an Acer ScanWit, you can have some ideas
of how to resolve problems listed in the attached README file. I
would be glad to hear from you :-)

Also, this of course needs many, many testers. Please test it
and send me comments.

This is usable right now for me, but is still far from a
complete backend.

--Max Ushakov.

-------------------README file from the tarball-----------------
This is an awlfully ALPHA release. You have been warned.

This is a SANE backend for Acer ScanWit 2720S film scanner. I
have written it having little or no documentation, based on the
following sources:

  -- SCSI trace log from evaluation version of VueScan
  -- SANE backend for Acer/AGFA SnapScan scanners
  -- SCSI-2 specification
  -- protocol specifications from AGFA for the SnapScan

For now the backend lacks the detection routines (too lazy to
write them), documentation (even the .desc file) and many, many
necessary features (see below).

It can be fetched from

ftp://mccme.ru/users/ushakov/scanwit/

INSTALL
=======

Copy the *.c files to the sane backend directory, apply a patch
to Makefile.in, reconfigure, recompile.

FEATURES
========

(The main one) It scans.

Scanning at 2700 and 1350 dpi in both preview and scan modes
seems to work. (At least full frames (24x36mm); little testing
has been done on partial scans.)

Autofocus is implemented in software (seems like the scanner
cannot do autofocus in hardware, but can only set the desired
focus value). The autofocus is as fast as in VueScan and
win-driver.

Custom gamma table uploading. Useless, I think, since the
scanner gives the full 12 bit range to the output. You can just
download the full frame and do any color correction later.

Scan speed measure is implemented. This gives a bit of speed to
1350dpi-scans and preview mode.

Shading Calibration is implemented. This seems to sometimes give
random color stripes along the picture. Maybe I should have done
this selectable...

With a simple script you can batch-scan a whole holder of film
(6 frames / 4 mouted slides) -- still needing to make color
adjustment later (especially for negative film).

BUGS (and lacking features)
===========================

The main bug is a lack of low resolutions. If one chooses a low
resolution, the scanner seems to double each scan row at least
for the first half of the picture (I did not try to read more
from the scanner, sorry). I tested resolutions 2700 and 1350 --
they are ok, but 675 has this problem.

Sometimes the "low resolution bug" shows up even in 1350 dpi.
Sometimes it can be dealt with by running a VueScan with the
same scan parameters (resolution and mode) -- the bug often goes
away.

The custom gamma table uploading is not tested well.

It does not do bpp other that 16 (12bpp really, shifted << 4).
This breaks preview in xscanimage.

The scan speed could be improved? At least win-driver (and
VueScan too, I think) scans a bit faster...

Quite many options are missing. (I.e. place where autofocus is
performed cannot be changed).

The backend does not have a detection routine. You have to
specify the device directly (e.g. xsane scanwit:/dev/sg0).


AUTHOR
======

Max Ushakov <ushakov at mccme.ru>