Subject:
Re: Plustek 9636T scans color/grey incorrectly FURTHER
From:
kevin ####@####.####
Date:
23 Jan 2001 16:38:30 -0000
Message-Id: <3A6DB496.EC22EC8F@bellsouth.net>
Mark Bratcher wrote:
> Kevin,
>
> I tried doing this more (repeatedly) and I eventually got one good scan.
>
> Seems tedious. Does this normally work better than this? I was just
> wondering if this problem is fairly unique to my system, or is it really
> the current state of the driver.
>
> Mark
>
Hi Mark, Gerhard, and list,
I did a number of tests yesterday with my 9636T, and at first
thought I was seeing a pattern. I did some more tests today, and
now I'm not sure what I'm seeing. I'm going to give a summary of
what I have done. I suggest that other people try some of these
tests, and post their results. Maybe then we can find a pattern.
The necessary info would be: your scanner type, Linux distribution
and kernel version, sane version, plustek driver version, and
comments on how it worked. You might also include how you did the
scans (from command prompt, or in X (KDE or GNOME or ??) using
Xsane or Xscanimage), and whether you compiled the driver with #if
0" or "#if 1" in the ../h/plustek-scan.h file. For example.
I have been testing on two computers (Paws and Cricket, for future
reference). The hardware is nearly identical (same MoBo, BIOS
version, CPU, memory amount and type, scanner (switched through a
switchbox), video card, sound card, NIC, ISA parallel port card).
I did a few scans in Windows 98 (both computers have Win98 and
several versions of Linux), so I know there is no hardware problem
with ports, cables, or scanner. The PP cards are set to EPP on
Paws, and ECP+EPP on Cricket (uses EPP mode for scanning). For the
most part, I used the tarball version of sane-1.0.3 and scanned in
KDE (1.2 and 2.01) using xscanimage. The one exception to this is
on Cricket, (see below). I did use the plustek-sane-0_37_20.tar.gz
version of the plustek driver (Gerhard, my BUILD says 21?) for ALL
scans on both computers. This is the common element (other than
the scanner). I tried using both "#if 0" and "#if 1" in the
../h/plustek-scan.h file; I seemed to get slightly better results
using #if 0 (which uses the parport_pc functions instead of the
delayed inb/outb functions), but at this point I can't see that
using one over the other makes a big difference.
OK, now the results:
Paws, Mandrake 7.1 (original kernel 2.2.15, upgraded by me to
2.2.16), KDE 1.2, other factors as above. When trying to do a
preview scan, usually the first attempt would give just a mess of
either vertical or horizontal line. Second and subsequent attempts
would usually work correctly. I noticed that when a scan fails,
either there is no motion of the scanner head at all, or it does
the scan but does not return to the top of the scanner (normal
position). This is true for all failed scans on both computers.
Also, sometimes if I exit xscanimage and then restart it, I get
messages about i/o errors and the scan fails. I then usually have
to unload the driver, turn the scanner off and then on, reload the
driver. Sometimes this fixes things, sometimes I have to do it
several times. This behavior can occur on either
computer/distribution.
Paws, Mandrake 7.2 (kernel 2.2.17), KDE 2.0.1. I had a lot of
problems yesterday getting a scan at all. Today, it seems to be
working fairly well. Again, sometimes I can do several prescans in
a row with no problems, but usuallly the first attemp fails, and
subsequent attempts succeed.
Cricket, RedHat 6.1 (kernel 2.2.12), KDE 1.2. On this
configuration, using #if 0 worked quite well; it almost never
failed. When I recompiled the driver with #if 1, I could not get a
good scan - i/o errors, or just long periods with no scanner
activity and finally lines or a blank page in the preview scan
box. When I reset to #if 0 and recompiled, the scanner worked
again quite well as before. So far, this is the only case in which
changing the #if 0/1 made a noticeable difference. As this is
based on a rather old kernel, the problem may have something to do
with the kernel parport code (remember parport and parport_pc
modules must be loaded for the driver to work).
Cricket, Mandrake 7.2 (kernel 2.2.17), KDE 2.01. This
configuration is virtually identical to Paws on Mandrake 7.2,
EXCEPT that I used the precompiled (for Mandrake) rpms for
sane-1.0.4. I also used Xsane-0.69, because in sane 1.0.4,
xscanimage is a separate package and wants an older gimp library
(I am using gimp-1.2 on this system). BTW, I did not try scanning
from gimp on any configuration. I can Xscanimage/Xsane directly.
Usually the first prescan failed, and subsequent ones worked. I at
first thought that running a regular scan, then a prescan would
work, but that might just be because I had to do a second scan of
any type to get a good result. Today it seems to be working rather
well, but the first attemp still usually fails.
As the only common sofware seems to be the scanner driver, when I
get a chance I will go back to an earlier version (the only other
0.37 driver I have is 0_37_12) and see if this makes any
difference. I hope some of you can do similar tests and post the
results.
Kevin
