Re: [linux-usb-devel] "Rainbow" striping on USB HP ScanJet 5300C

From: Scott Gifford (
Date: Wed Dec 27 2000 - 09:44:57 PST

  • Next message: Per Aalrust: "Epson Perfection 1200 Photo USB crashes"

    jhall@UU.NET (Jeremy Hall) writes:

    > WAY KULE!
    > I haven't tried my driver with super-recent kernel, I'm still running
    > test8. What version of sane are you running?


    > Since I am a visually impaired person, perhaps the stripes don't interfeer
    > with OcR.

    I would think they would; when I print out a scanned image with the
    stripes, about every 10th letter is completely illegible.

    > I could scan a page with my scanner and provide the image for your
    > examination.

    Sure. Feel free to email an image to me, or put it somewhere
    publicly accessible, and I'll see if it looks at all similar to what
    I'm seeing.

    > Is this in colour or gray?

    I see the strips both when scanning in color, greyscale, or line-art
    mode. Of course, they're grey or black-and-white in the other modes.

    > I have noticed that the scanner behaves quite differently when
    > trying to scan at say 600dpi instead of 300dpi, sane's defaults try
    > to produce a horridly LARGE image, trying to produce a 24-bit colour
    > image.

    I'll play with that some more; I tried different resolutions, but
    mostly tested with the defaults.

    > I've been mostly scanning gray 300dpi images for OCR purposes.
    > I have considered the possibility that scsi errors may not be getting back
    > to the scsi handler, but I find this difficult to understand why since the
    > code seems to read correctly. Have you tried running with an USB analyzer
    > both in Linux and Windoze to determine if they chat differently?

    I have not tried that; this is the first time I've ever tried to use a
    USB device. I'll try to find a good analyzer and see if I can make
    any headway with it.

    > Basically what happens is the following sequence:
    > We send a command to the OUT endpoint. We may then retrieve data or the
    > status handler may be called. Depending on the results of the status
    > handler, we may or may not send back an error to the scsi layer.
    > There may be some race conditions not yet realized, this is why testing is
    > suggested.

    Thanks for your help!


    Source code, list archive, and docs:
    To unsubscribe: echo unsubscribe sane-devel | mail

    This archive was generated by hypermail 2b29 : Wed Dec 27 2000 - 09:31:03 PST