[sane-devel] SANE -backends-1.0.19-10.fc9 and HP SJ 5590 ADF - with gscan2pdf - A4-duplex scan - paper dimensions wrong

Ilia Sotnikov hostcc at gmail.com
Fri Jul 4 18:31:55 UTC 2008

On 6/28/08, Ari Latvala <a.e.latvala at gmail.com> wrote:
> 1.  Note:  gscan2pdf using SANE backend for HP LJ 5590 with duplex ADF scan
> needs definition, that the reverse side must be rotated 180 degrees,
> otherwise the scan result gives that upside down. Page sizes are defined as
> A4 in the GUI of gscan2pdf.

The ADF duplex scanning was discovered occasionally, and that was a
first and straight thought on how to implement it. That's far from
ideal, I agree.

> 2.  Page 2 (reverse side of  A4 size aper original (209 mm wide, 297 mm
> height) scanned) starts in output image  PDF result as 6 mm late (clipped
> from top of page 2),  12 mm shifted left on page 2 and ends about 24 mm
> shrinked at bottom.

According to the Windows driver logic (seen from USB sniffing), the
scanner lacks (or at least, doesn't use) hardware paper sensor
correctly implemented. Strictly saying, the scanner outputs one
continous data stream for multiple pages. Thus, the only way to
distinguish between the pages is the software calculations.
Interestingly enough that changelog for the Windows driver (on hp.com)
said that they fixed image size calculation in ADF mode. That confirms
software calculations idea.

> 3. The phenomenon accumulates in multi-page ADF feeded double-size
> documents, stepping up with about 12 mm on each

To avoid (or at least minimize) one should set exact paper size
(perhaps, doing several scans). Not a perfect way, I agree as well.
> 4.  The end result is, that multi-page double-sized ADF scans get cutted
> from wrong places and  the result is a PDF document, where the  page images
> are combination of a cropped end-of-page  parts of previous page and the
> current page.  Due to the ADF duplex processing or reverse side, then some
> parts of the text are upside down.

I thought about backend modification in that way, when each page would
result in the separate image, with correct orientation. Shouldn't be
hard to implement, but not at this moment because of time pressure.

Anyway, thanks for your feedback.

 Ilia Sotnikov

More information about the sane-devel mailing list