[sane-devel] Q: Performance Tip wanted, USB vs. SCSI HP 63xx Scanner

[abel deuring]

Juergen Sauer wrote:
> 
> Am Freitag, 23. November 2001 00:19 schrieben Sie:
> > option enable-image-buffering
> 
> Somewhat thing was what I'm looking for.
> 
> It is not nessessary to have the same Device-Bandwith to the ram as with
> SCSI. Scanning time is not important. My topic is more to save the
> scanner's mechanical lifietime.
> 
> But this did not helped here.
> During preview scan xsane .81+ sane 1.06 I recognize a ugly misbehavior:
> the scanner moves forward-stop-little backward-stop- etc. instead of
> moving counitnuesly until buffer full - stop transfer - go on.
> 
> This behavior sounds like a cruel mechanical torture - HP's hardware
> quality is not the same as for a few years, this way the scanner will
> probably break soon.
> 
> Any idea what's going on there - any idea ?

Since none of the USB experts seems to be available at present, here my
$0.02. (disclaimer: I haven't ever even connected any USB device to a
Linux box...)

Below is a copy from a recent mail to the Sane list. In another mail you
wrote that you use an SMP kernel. Assuming that the Epson scanners use
the same USB <-> SCSI "translation" as HP scanners, it might help to try
a non-SMP kernel.

Abel

> Subject: Re: [sane-devel] S..L..O..W USB Scanning with Epson 1650
> Date:    Sat, 17 Nov 2001 11:11:37 -0500
> From:    "Justin S. Peavey" <jpeavey+sanedevel@peaveynet.com>
> To:      sane-devel@mostang.com
> 
> On Fri, Nov 16, 2001 at 09:50:02AM -0500, Justin S. Peavey wrote to To
> sane-devel@mostang.com:
> > On Thu, Nov 15, 2001 at 07:11:26PM -0500, Karl Heinz Kremer wrote to Cc
> sane-devel@mostang.com:
> > > I read your post on the usb group, and your suspicion that this problem
> > > is caused by the USB subsystem is very likely correct. The scan times
> > > you are listing are definitely not right. You should get about the 
> > > same times as you list for the Windows driver. I suspect that it has
> > > to do with your dual CPU setup.
> > > 
> > > Sorry, but I can not help to solve this problem.
> > > 
> > > Karl Heinz
> > > 
> > 
> > I'm going to try a non-SMP kernel tonight to see if that does
> > anything.  The other thing that is tickling my mind is the shared
> > interrupts, USB, SCSI, and Network card all sharing the same one.  I
> > understand this is much more common now, but the problem looks similar
> > to a slowly or mis-handled interrupt.  Anyone know how to force USB to
> > it's own interrupt?
> > 
> > -JSP
> > 
> 
> On the nose Karl, the scanning works fine when I boot with a non-SMP
> kernel.  I'll continue this on the linux-usb group.  Thanks for your
> help!
> 
> -JSP