[sane-devel] Canon LiDE 90

[Pierre Willenbrock]

Hi,

Guillaume Gastebois schrieb:
> Hello,
> 
> 
>> What about only setting register 0x7f? that one should do something
>> without needing to setup reg 0x1a.
> 
> Not better I think. Result : 
> http://ggastebois.free.fr/lide90_snoop/10_test0.tar

I forgot that there is a switch-on-bit for that, too: bit0 of reg[0x7d].
Please try with that one set.
(And that tar seems to be truncated)

>> I didn't expect reg[0x1a]=0x24 to work without setting the corresponding
>> clock bit masks. What happens if you leave line 1159 commented and set
>> regs 0x74-0x7d(my guess: works without changes in behaviour)? Does
>> setting regs 0x71-0x73 change anything (line 1159 still commented)?
>>
> Result of setting only 0x75 0x76 0x79 0x7c 0x7d (not 0x7f) : 
> http://ggastebois.free.fr/lide90_snoop/10_test1.tar
> 
> Result of setting 0x71 0x72 0x73 0x75 0x76 0x79 0x7c 0x7d (not 0x7f) :
> http://ggastebois.free.fr/lide90_snoop/10_test2.tar

The same as without setting them at all. So the bits in 0x1a are
probably switching from some internal bit masks to those in 0x71 to
0x7c. And changing the switching point of RS didn't change anything, either.

>> (I have only little understanding of the actual relative timing and use
>> of all clock signals going out to the ccd/afe, so i am guessing and
>> doing experiments.)
>>
>> But this seems to be basically working. Please send your changes leading
>> to a usable scan, so i can integrate them.
> For now, my code is ugly. I only modified lide60 to lide90. But you can 
> find genesys_gl841.c and genesys_devices.c in 
> http://ggastebois.free.fr/lide90_snoop/sources

Will merge that.

>>>>>>> Another thing : when I make several scan with sane backend and sane 
>>>>>>> command line, I have alternatively brite and dark images !!! Why ???
>>>>>> The calibration is probably giving widely differing results with
>>>>>> different starting conditions. It swings between two states. But you
>>>>>> shouldn't see this after the gl842 did its shading correction. Then, the
>>>>>> problem is probably overexposure of the ccd cells. Try reducing the
>>>>>> upper threshold in genesys_gl841.c:4383
>>>>>>
>>>>>> 	  if (avge > 2000) {
>>>>>> 	      expr = (expr * 2000) / avge;
>>>>>> 	      expg = (expg * 2000) / avge;
>>>>>> 	      expb = (expb * 2000) / avge;
>>>>>>
>>>>>> Reducing the lower threshold may be needed, too. The current values for
>>>>>> your scanner are:
>>>>>> expr: 1235
>>>>>> expg: 1235
>>>>>> expb: 675
>>>>>>
>>>>>> No guarantee that this helps at all.
> I tryed that (upper threshold to 1500 and lower to 250 and it doesn't 
> work. You can find the same test as test2 without the same result on : 
> http://ggastebois.free.fr/lide90_snoop/10_test2b.tar (I just make two 
> consecutive scanimage without recompiling sane).

That is not unexpected.

> One other thing : we can see vertical lines with different contrast on 
> result images. What is it ?

The shading correction not doing its work correctly. I see similar
behavior when using the method of scanning a single line multiple times
with/without lights for acquisition of dark/bright levels. I don't see
this when using a scan over my black+white calibration area. Currently,
i don't know what causes this difference.

> Regards
> Guillaume

Regards,
  Pierre