[sane-devel] Scannning trouble with HP 2520A

[Brett (Mare) Henley]

 With abel deuring and Peter Kirchgessner helpful pointers and ideas,
the scanning problems I was having are solved. Thank you very much for helping
me with this problem it was really driving me nuts.
 It took a little doing but the end solution was to rethread the drive belt on
my scanner. The scanning unit was creeping from inertia causing the lines. By
taking apart the scanner I found I could easily move the scanning bar about
10 mm back and forth.
 End result, after several hours of head banging, the purchase of a torks
driver and help from this list. I have a solid 600 dpi scanner working. I
suspect what may have happened is the shipping lock had falled down and when
the scanner reset it hosed the way the scan head was attached to the drive
head. It's a friction attatchment kind of thing and just wasn't doing it's
thing.

Thanks everyone.

Mare



abel deuring said:
> Hi Mare,
>
>
> Brett (Mare) Henley wrote:
>> Hi Abel,
>>   I tried increasing the scan buffer using the configure file but that
>> didn't
>> help.
>>   The noise looks like a buffer problem of somekind I think. the data
>> looks to
>> me like 2 pixel lines were added at the end of the buffer block. but what
>> or why I don't know. the fact that the lines almost disappear with a solid
>> colour background might indicate that the pixels are out of order. But I
>> don't have experience with graphics on this scale.
>>   I'm sending an example scan so you can decide. I did my best to make the
>> file as small as possible. this was scanned at 300 DPI 8b/colour depth.
>
> Never mind; i have DSL line at home, and receiveing somewhat larger  emails
> at my workplace is also no problem.
>
> Now, for the image: Firstly, I guess that you scanned a printed image;
> perhaps from a magazine, but not a photo, and that the resolution used  for
> the print was 100 dpi. Since the skin color has not 100%
> cyan/magenta/yellow/black, we have "real" dots, not larger areas
> completely covered with one or more of these "base colors". This means  that
> neighouring 300 dpi pixels will "see" different colors, which is  very good
> in our case. If you look onto the image at 200% or 300% zoom,  you will
> notice diagonal lines (well, not at an angle of 45 degree, but  the lines
> come roughly from "nort-west" and go to "south-east". These  lines are
> caused by one of print colors (the black raster is printed at  45 degree;
> the other colour are printed at 15, 75 and ??? degree --  sorry, I forgot
> the last angle...) Anyway, I guess that the "diagonal  lines" we see are
> caused by the raster points of either cyan, magenta or  yellow.
>
> Now, if you look at one of the "distortions", and if you try to align a
> ruler or the edge of a sheet of paper parallel these lines, you'll see  that
> the "distortion" looks roughly like this (sorry, my English is not  good
> enough to describe the effect with words; and Gimp does not provide  an
> option to draw straight lines -- or I am too dumb to find the correct  Gimp
> toll...):
>
> \
>   \
>    \
>     \
>      |
>      |
>      \
>        \
>          \
>           \
>            \
>
> The point is that the distortion is a "dent" in a straight line.
> Moreover, the vertically adjacent pixels in the "vertical" part of the  line
> show above are not identical. This indicates that each pixel line  is really
> scanned, not duplicated by some "buffer mismanagement". (Ok,  an alternative
> explanation are JPEG compression artifacts.)
>
> As a really "long shot" conclusion, I'd blame the scanner's mechanics  for
> the distortions (based on the additional assumption that the
> scanhead stops more or less frequently during a scan): As you wrote in  an
> earlier mail, you get more distortions with higher resolutions; this
> indicates a relation with the SCSI buffersize or some other limit for  the
> amount of data transferred with one SCSI command.
>
> Many scanners stop the scanhead between the data transfers of two SCSI
> commands, caused by a tiny buffer memory of the scanner's
> microcontroller, and perhaps a too long delay between the termination of  a
> "read data" SCSI command and the start of the next "read data"
> command. Especially multitasking OSes can easily cause relatively large
> delays.
>
> I assume that the stepper motor of your scanner stops properly, but I  guess
> that the scanhead moves 3 or 4 pixels, i.e. ~0.01 inch, forward,  when the
> motor stops. This could be caused by a somewhat "slack" belt or  wire
> connecting the motor and the scanhead.
>
> When the motor starts again, the scanhead remains for these 3 or 4  pixels
> in the position until the belt/wire is again tightened, and then
> accelerates during 2 or 3 more scan lines to "normal speed", which would
> causes the lower part of the distortion.
>
> Abel
>
> PS: I needed a bit help from a dictionary to write this mail, and
> "all-round" dictionaries are not very precise when it comes to technical
> terms. So I hope the stuff I wrote is halfway understandable and a
> quarterway correct ;) If it isn't, let me know.