if anybody has (or is writing) a driver for the Fujitsu M3091,
please send me an eMail and disregard the remainder of this message.
Otherwise, read on:
Reasonable C programming experience but no SCSI drivers or SANE
Fujitsu M3091DCd, a scanner that looks more like a fax machine and
is ADF-only, supports 50-300 by 50-600 dpi in line art, grayscale
and 24bit color, 128 kB cache and duplex (!). I am in possession of
the full SCSI command documentation for that specific scanner.
To get as much as possible of that working with SANE. Below is a
report of my current progress, and I would be very happy about any
and all sort of hints, tips, tricks, pitfalls, spiritual guidance,
links to documentation etc :-)
With minor changes, I got Randolph Bentson's M3096 driver to work with
this scanner supporting the full resolution but neither color nor
duplex. I was thinking about using that driver as a basis and make
extensions as necessary, but I have some questions for the SANE
experts here before I go on.
Question 1 - How to do duplex scanning.
To use the duplex mode, I have to send an extended SET WINDOW command
to the scanner which should be no problem. The question is, how do I
then feed the two images into SANE? As far as I can see from the
existing driver, a subprocess gets started to read more-or-less raw
binary data from the scanner, which is then passed via a pipe to the
calling program; it seems to "leave" the backend via the sane_read
function. Since the scanner has dual scan heads and only 128 kB of
cache, in duplex mode it will send a data stream that contains two
interspersed images. I would be able to separate them on reading, but
what next? How do I tell the SANE middleware (or front-end for that
matter) that I actually have TWO images? How would a front-end (like
scanimage, the only one I've used so far) cope?
If SANE doesn't support that, I could introduce a new parameter that
controls whether the front or rear side is to be scanned, and each
page would have to be run through the scanner twice. A waste, but
I've grepped for "duplex" in the existing backends (1.0.4) but found
Question 2 - How to do the color thing.
This is just a quick question in case somebody has noticed something
similar with their scanners. When I read color data from the scanner,
I get one row of "R" values, then one row of "G" values, then one row
of "B" values, and so on. These are slightly offset, e.g. the first
row of "R" values have to be paired with the fifth row of "G" and the
ninth row of "B" values or so (the documentation is detailed but at
the same time not very clear on that - http://www.fujitsu-europe.com/
home/support/scanner/manuals/M3091dc.scsi.en.pdf, pp. 75-77). So far,
so good, but in reality it seems that after a certain number of these
R row/G row/B row groups I suddenly get TWO R rows following each
other, and then on like before, and the same phenomenon again every x
blocks. I haven't fully investigated that and don't really know what
to make of it yet.
I do understand that I'll have to get all the bytes into the right
order before passing them on to the SANE middleware. Once I found out
what the correct order is, that should be no problem. However I don't
know how the scanner deals with the specified window size in relation
to the sensor offset. If, as stated above, the R/G/B sensors are
offset by 4 lines each, that means that there are 8 lines of
incomplete data at the beginning and the end of the data stream. If I
request an image of, say, 100 pixels in height, will the scanner then
give me 100 of which only 84 are fully usable, or will it give me 116?
Randolph has suggested looking at the SP15C driver which contains
color support, and I'll do that.
Question 3 - How to support the "hot button(s)"
This is probably beyond the scope of SANE but maybe somebody knows
something. The scanner has the following controls: a "scan" button, a
"duplex" button, a "send to" button, and a number dial that can select
numbers 0-9. The scanner does nothing with these buttons; the obvious
purpose (as supported by some Windows software that comes with it) is
that you can insert a document and then, by pressing one of them, have
your computer launch a program that does something. (My idea of use
would be assigning categories to the numbers, i.e. 1=bills, 2=bank
letters, 3=personal letters, etc., and then use the number dial and
"send" button on the scanner to store documents away in the proper
folders on my Linux box.)
The SCSI docs tell me that there are some extended fields in the
scanner's response to the INQUIRY command that tell you if one of
these buttons has been pressed. The respective fields are "true" for
three seconds after button press. So I imagine I'd need some sort of
"inquiry daemon" that sends the command to the scanner every 2 seconds
or so and then launches whatever program has been defined when certain
fields in the response data have certain values.
Does such a program already exist? If not, can somebody point me to a
very very simple program that (outside of the SANE context) sends a
simple SCSI command to a device and evaluates the response, so that I
could work from there?
-- Frederik Ramm ## eMail firstname.lastname@example.org ## N57°48.10' W005°40.32'
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail email@example.com
This archive was generated by hypermail 2b29 : Tue Feb 27 2001 - 05:27:08 PST