CanoScan 9000F support in SANE

[shugyo]

Dear all,

I found someone who has a CanoScan 9000F (the new Canon flatbed scanner replacing the 8800F which is already supported in SANE 1.0.21), and managed to find out that it also uses the PIXMA protocol (the protocol used in the scanners for Canon's multi-function printer/copier/scanner devices of the PIXUS/PIXMA MP series (and other names depending on sales region).

Support for the scanner has therefore been added in the pixma backend. However, the initial tester is having trouble compiling and configuring SANE from source, so I am hoping there will be some people here who have access to this scanner and can try to help out as well.

After compiling SANE, if "scanimage -L" does not detect the scanner, probably "usb 0x04a9 0x1908" needs to be added to /etc/sane.d/pixma.conf and /usr/local/etc/sane.d/pixma.conf.

After that slow steps need to be taken to get support for the various resolutions, as this scanner as double the number of CCD rows as the 8800F, and can do double the resolution (up to 9600dpi) optically.

Please contact me on this thread or via PM, and I'll PM you my email address.

Best regards,
Gernot Hassenpflug
Tokyo, Japan

[shugyo]

I'm delighted to report that initial support for the CanoScan 9000F has been successfully added to the development version of SANE. As the maintainer of the pixma backend is on vacation until September the developing code will not be reviewed or committed until then at the earliest. In the meanwhile I will try to add support for all resolutions for paper scanning, plus TPU (film) support.

[shugyo]

Update:

• Scans at 75, 150, 300, 600, 1200 and 2400dpi work.
• Awaiting test results of 4800dpi scan in order to correct possible problems.

TODO:

• Implement 9600dpi mode.
• Add TPU support.

[shugyo]

Update:

• paper scans all work up to 4800dpi
• 9600dpi is only supported for film scan (TPU mode)
• TPU works (awaiting individual resolution results)

TODO:

• check all TPU resolutions (especially uncertain about 9600dpi)
• discover y-offset in TPU mode

[shugyo]

Update:

• (all paper scans up to 4800 work)
• TPU works for 300dpi, including vertical offset found correctly

TODO:

• TPU for 600dpi and above requires different color plane shifts, reconstructing from Windows UsbSnoop logs
• corresponding offsets also need to be found
• currently working on 600 and 1200dpi TPU images (2 bytes per channel per pixel)

[shugyo]

Update:

• (all paper scans up to 4800 work)
• TPU works for positives ("slides") at 300, 600, 1200 and 2400dpi, including vertical offset found correctly

TODO:

• TPU for 4800 and 9600dpi is in testing now, may require help via reconstructing from Windows UsbSnoop logs
• need to test & develop support for negative scans
• need to add mode to distinguish between slides and negatives in the driver code

[shugyo]

Update:

• (all paper scans up to 4800 work)
• TPU works perfectly up to 4800dpi

TODO:

• TPU for 9600dpi is partially done, last byte reordering of 8 lines needs to be perfected.
• need to add mode to distinguish between slides and negatives in the driver code

[barmaley]

Hello, everybody
I have a CanoScan 9000F - looks like UFO, but does not work yet in Linux.
I use Ubuntu 10.04 with this kernel 2.6.32-24-generic-pae.
I have installed the latest Sane sane-backends-1.0.21, installed the libusb-dev package, but still no go.
here is an output of sane-find-scanner: found USB scanner (vendor=0x04a9, product=0x1908) at libusb:002:003
here is a scanimage -L output: No scanners were identified.
I need this scanner to be working at its best, so I'm open to all suggestions.
Thanks a lot for your help.

[shugyo]

@barmakley: as stated, the code is not in SANE yet, not even in CVS, only on my computer. So unless you are willing to install SANE CVS and compile it with my modified code, then you're going to have to wait until it is in CVS in a few weeks' time at most and then learn how to compile SANE CVS. Alternatively, you can wait until the next stable release of SANE, which if I am correct in divining the intentions of the SANE developers, will be sometime in the late autumn. That still leaves the question of when your distro will have it, but Ubuntu should not be too long behind.

[shugyo]

Hi all,

I have good news on the Canoscan 9000F front...the enemy is in retreat, pants down and pleading for mercy. Let me tell it like it is: don't all rush and go buy this device. Canon is putting el cheapo devices on the market instead of giving a real hardware upgrade. In these times, who can blame them (MP970 <<< MP960, for example; and cameras too: this info from a good friend who knows his stuff).

So, here is the deal: the 9000F has the same capabilities as the 8000F, and then adds a quasi-9600dpi mode for TPU only. This mode does the following:

• It gives motor-controlled 9600dpi in the y direction, and 4800dpi in the x-direction: the pixels in the x-direction are also the size of 1/9600, but each pair of pixels has the same value. Software then does interpolation to give square pixels in the Windows driver.
• This is the same stunt pulled on the MP960, where the real resolution is only 1200dpi, and pairs of pixels plus motor resolution were used to give 2400dpi (in one direction only) and 4800dpi (in one direction only, the other was 2400dpi by taking two 1200dpi images at slightly-offset positions).

SANE can now support the CS9000F as soon as my code goes into CVS, and will be in the next stable release. In the meantime, I'll put up my modified file along with basic instructions on how to use it in your own compilation of SANE CVS, on my own website (home server on ADSL connection, so slow u/l to the internet) and post the link here hopefully by the weekend. My file also has support or completed support for several other devices that are not yet in SANE CVS.

[barmaley]

Thanks for your work. shugyo.
Could you, please explain which is y- and which is x-direction?
I use it at work, I'm a scientist, but in my neighbourhood for such dirty tricks as Canon did we used to beat those guys.

[shugyo]

Thanks for your work. shugyo.
Could you, please explain which is y- and which is x-direction?
I use it at work, I'm a scientist, but in my neighbourhood for such dirty tricks as Canon did we used to beat those guys.

I agree on the neighbourhood part! x-direction is the width of the scan, parallel to the CCD scan, in other words related to the spacing of the elements next to one another. y-direction is the direction of movement of the scan head, along the length of the page.
Of course the Canon Windows driver will rotate photo negatives and positives 90 degrees so may be confusing to relate x and y to length and width of a photo.
I should be back from the USA on the 18th and get back to the development during this week, only got to get the 9600dpi scan sorted out still (correct interleaving of 8 subimages).
Regards,
Gernot

[benefici]

Shugyo, thanks for your work. Is this scanner now supposed to work? I tried compiling today's git snapshot of sane-backends but scanimage -L still does not recognize my scanner. Maybe your code has not been uploaded yet to the git repository?

[shugyo]

Hi there. No, sorry, there is no code yet in the repository, not sure when that will be. The only thing I can do is send people a patched file for now.

[edg]

Hi Gernot,

it's great that you are working on the 9000F driver!
I'd be happy to test your driver. Being a (somewhat retired) SANE developer myself, I have no problem patching & compiling the sane backends.
Can I find your code somewhere?

Best regards,
Eddy

[stevey]

Hi!

Many thanks for working on this driver.
Shamefully tossing the last shred of restraint and self respect out the window I say: me to!
If a patched file might be available for testing on my trusty 9000F I'd be most grateful

-Steve

[shugyo]

Hi, just saw your post, PM me and I'll send you my working file.
Cheers,
Gernot

[stevey]

Hi Gernot,

Today I finally got to testing the hi-res TPU scans. Seems that the 9600 code might need a tweak, and (hopefully I'm wrong), but the 4800 too.

First, the 4800dpi scan:
I have compared our result to a 4800 scan done on W7
The W7 scan appears to be a true 4800 dpi with no y-axis pixel doubling. Yet the Sane result shows the duplicated pixel horizontally.
If I'm right, then we're not getting true 4800dpi yet (I'm new at this, so kindly bear with me if I'm wrong

Re: 9600dpi
File size is 4x the 4800dpi scan...
Vertical "lines" appear in the image indicating the "last byte reordering of 8 lines" is still buggered.
I see the adjacent y-axis doubling - though this time it's expected (I haven't got a 9600 W7 scan to compare though).

Is there more helpful information I can provide you?

Best regards,
Steve

[shugyo]

Hi Steve,
Thanks (and I got the PM too---replied to that) for the report.
Yes, in TPU mode 9600dpi needs work. 4800dpi seems to be that way, as far as I can tell, so maybe the Windows driver does some interpolation. I will double-check though, to be certain.

[ekb802]

For your information (I hope it can be to any help)
I'm not a programer guy but like to tamper with hardware, so I just deassemled my Canon 9000F to reveal its internals parts :0)
Photo of the circuitboard and the chip set can be found here http://www.stockholmviews.com/canon_900 ... aning.html
Look at the botom of the page
I'm looking forward to the official backen release :0)
Regards/ Stefan


http://www.stockholmviews.com