[sane-devel] Canon LIDE 90

[Pierre Willenbrock]

Ralf Haueisen schrieb:
>> -----Ursprüngliche Nachricht-----
>> Von: "Pierre Willenbrock" <pierre at pirsoft.dnsalias.org>
>> Gesendet: 27.11.07 22:23:19
>> An: Ralf Haueisen <ralf.haueisen at web.de>
>> CC: sane-devel at lists.alioth.debian.org
>> Betreff: Re: [sane-devel] Canon LIDE 90
> 
> 
>> Ralf Haueisen schrieb:
>>>> Can you send a small image?
>>>>
>>> The image shows the cover of a book...
>>>
>> This looks like the shading correction(corrects for the properties of
>> each ccd cell) is making the image worse. please disable it by changing
>> gl841_init_optical_regs_scan to only disable it:
>>
>> /*    if (flags & OPTICAL_FLAG_DISABLE_SHADING)*/
>> 	r->value &= ~REG01_DVDSET;
>> /*    else
>> 	r->value |= REG01_DVDSET;*/
>>
>> The resulting image may give better information. Another possibility is
>> to look at the generated black_white_shading.pnm when
>> SANE_DEBUG_GENESYS=255 is set in the environment.
>>
> 
> It is not the shading.

Well, it does look better. My current guess: either some gpio is
incorrect, or the sensor needs some other clock settings. The noise may
stem from too few light or the data bytes from the frontend are swapped.

The sensor does not seem to switch the read out pixel, but it does
capture line data, so you end up with the first pixel over and over again.

First, i'd make sure that i actually get the most significant byte from
the frontend(by trying the possible byte positions in register 0x52) if
you get a noise free image from the scanner, you got the msb right. The
lsb is not that important at the moment(It would be the byte position
where you get a different noisier image). The "correct" msb may be very
dark(not 0 though), you are scanning pixels that are under the scanner
cover).

Then i'd try to find out which of the clock/gpio registers needs to be
set. An usb log may reveal this.

Hope this helps,
  Pierre