I should have mentioned it in my first email--in the HP Windows software,
the PhotoSmart (PS) does as expected: it generates a preview as the media goes
in, and does the final scan as the media comes back out. I'm not sure
why xscanimage is not doing this. I'll investigate further tonight.
The PS also behaves itself when driven (in Windows) by PScan32 and by
VueSmart (VueScan now?), two third-party programs for driving the PS.
VueSmart is a shareware program that acquires the 10-bit data from the PS
and preforms orange mask removal (for negatives) and gamma correction
on that, before downsampling to 8-bit output. It does a terrific job,
and even has canned calibration curves for well over 100 different
types of film (Kodak Gold 100, Fuji Reala, etc.).
(I've been using the scanner in Windows--with HP software--for several
months now with success. I just want to run it in Linux using SANE now,
and have the opportunity due to the new SCSI card.)
> > - The vertical (y) postion of the slide seems to slip by about 1mm each
> > time the slide is fed in and out. In other words, if I draw a crop box in
> > the preview window, I get a same-size box in the output, but it is of
> > a rectangle shifted downward in y 1mm or so. Subsequent feeds shift it
> > more each time, as if the backend or the scanner are loosing track of
> > where the slide is physically.
> >
>
> As from above I suppose that the slide is removed from the scanner
> between preview and scan. So there is no possibility to get a correct
> placed box.
The slide isn't completely ejected from the scanner between scans--if I
hit the "unload media" button in xscanimage, the slide comes all the way
out and is "released" by the PS's rollers.
> Firmware version R032 seems to have no support for tone maps (at least
> greyscale tonemap which are used with firmware version R029 for gamma
> correction).
> I can show you some places in the hp-backend where you can try to add
> those code for testing. I can not test 30bit support on my own
> Photosmart,
> because R029 does only support 24 bit.
I would be very interested in trying this. Over the weekend, I played
with some code for doing PNM with a MAXVAL == 1023. There is an
assert depth == 8 somewhere in scanimage, so it's not quite ready for
10-bit PNM output. Is that correct?
> > Are there
> > any docs from HP specific to the Photosmart?
> >
>
> No. And if they have documentation, it must be specific to the firmware.
True. I wonder how the authors of PScan32 and of VueSmart came up with the
proper SCL codes for driving the PS. Maybe I could learn something by
running those programs in debug mode and checking the log of SCL commands.
John
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com