[sane-devel] [PATCH] Progress on HP2400 (genesys backend)

[stef]

Le jeudi 9 juin 2011 10:49:37 Alexey Osipov, vous avez écrit :
> В Срд, 08/06/2011 в 18:36 +0700, Alexey Osipov пишет:
> > В Втр, 07/06/2011 в 22:20 +0200, stef пишет:
> > > Le mardi 7 juin 2011 21:07:08 Alexey Osipov, vous avez écrit :
> > > > Hi, dear SANE developers!
> > > > 
> > > > Proposed patch enable support of 50/150/600dpi Color/Gray/LineArt
> > > > modes for Hewlett-Packard ScanJet 2400 scanner.
> > > > 
> > > > This was made by parsing UsbSnoop logs with stef's scripts (thank
> > > > you!).
> > > > 
> > > > 100/300dpi modes were already good, so I didn't touch them.
> > > > 
> > > > I have also refined sensor and motor values for 1200dpi modes, but
> > > > there are some problems with them.
> > > > 
> > > > Here is a list of remaining problems I can't solve:
> > > > 
> > > > 1. 1200dpi Gray mode works, but the resulting image is interlaced
> > > > vertically with black one-pixel width lines. Why this can happen?
> > > > Example: http://lion-simba.github.com/1200dpi_gray_lines.tiff
> > > > 
> > > > 2. 1200dpi Color mode works, but very slow. Scanner warm-up and
> > > > calibration takes approx 5 mins before head even moves. Then, when
> > > > scanning begins, head moves forward and backward each few lines. I
> > > > have made timestamped logs of this situation (truncated log
> > > > attached). You can see - problem is in "gl646_bulk_read_data" which
> > > > regularly hangs for approx 10 seconds like this:
> > > > 
> > > > during warm-up and calibration:
> > > > ---------------------
> > > > 1307470777: [genesys_gl646] gl646_bulk_read_data: trying to read
> > > > 65472 bytes of data 1307470778: [genesys_gl646] gl646_bulk_read_data
> > > > read 65472 bytes, 679896 remaining 1307470778: [genesys_gl646]
> > > > gl646_bulk_read_data: trying to read 65472 bytes of data 1307470778:
> > > > [genesys_gl646] gl646_bulk_read_data read 65472 bytes, 614424
> > > > remaining 1307470778: [genesys_gl646] gl646_bulk_read_data: trying
> > > > to read 65472 bytes of data [9 seconds delay]
> > > > 1307470787: [genesys_gl646] gl646_bulk_read_data read 65472 bytes,
> > > > 548952 remaining 1307470787: [genesys_gl646] gl646_bulk_read_data:
> > > > trying to read 65472 bytes of data 1307470787: [genesys_gl646]
> > > > gl646_bulk_read_data read 65472 bytes, 483480 remaining
> > > > ---------------------
> > > > 
> > > > during scan:
> > > > ---------------------
> > > > 1307471061: [genesys_gl646] gl646_bulk_read_data: requesting 30464
> > > > bytes 1307471061: [genesys_gl646] gl646_bulk_read_data: trying to
> > > > read 30464 bytes of data [9 seconds delay]
> > > > 1307471070: [genesys_gl646] gl646_bulk_read_data read 30464 bytes, 0
> > > > remaining 1307471070: [genesys_gl646] gl646_bulk_read_data: end
> > > > ---------------------
> > > > 
> > > > Any ideas of what is wrong? Proprietary driver running on the same
> > > > system scan well in 1200dpi Color, so I think this is not libusb bug.
> > > > Maybe some GL646 miscofiguration?
> > > > 
> > > > 3. In all modes, there are extra pixels to the left of the image,
> > > > which are beyond of glass window of scanner. Meanwhile, right side
> > > > of image is cropped by (i guess) same number of pixels. I think it
> > > > can be fixed by tuning CCD_start_xoffset/dummy_pixel/sensor_pixels
> > > > in Genesys_Sensor, but I'm unsure what is for exactly? What is
> > > > dummy_pixel?
> > > > 
> > > > Regards,
> > > > Alexey.
> > > 	
> > > 	Hello,
> > > 	
> > > 	good job, if I understand well, 50, 100, 150, 300 and 600 dpi modes
> > > 	are
> > > 
> > > correct. That's great.
> > 
> > Yes, this is correct.
> > 
> > > 	The 1 & 2 issues are surely related. Some registers maybe incorrectly
> > > 
> > > set. What you can do is to compare all the register set at scan time
> > > with the one from USB log at 1200 dpi. If you look at the code there
> > > some places in genesys_gl646.c where some actions are taken depending
> > > of the CCD. Maybe a case for the HP2400 CCD should be added.
> > 
> > Ok, I have made some more research and found out several different
> > registers.
> > 
> > After setting register 0x1e from 0x80 to 0x81 (set 1 CCD dummy line
> > according to GL646 datasheet) the issue 2 has gone (patch attached).
> > 
> > I have also refine geometry values of Genesys_Model hp2400c_model in
> > genesys_devices.c as you suggested, so the issue 3 also gone (patch
> > attached).
> > 
> > However, issue 1 is still there (for Color, Gray and LineArt modes).
> > There are several registers with different values in UsbSnoop logs and
> > genesys logs:
> > 
> > I need an advice what to try next.
> 
> Ok, I need not any advices. :)
> 
> Happy to say that I got working 1200dpi mode now. :)
> 
> The problem was in wrong value in frontend register number 3. The value
> 0x02 is good for 50-600 dpi, but gives interlaced result for 1200 dpi.
> In UsbSnoop logs there is a value 0x32 for this register, but I found
> out that scanner works good with value 0x12 (just like HP3670 in 1200
> dpi mode). So I left the value 0x12 for both scanners.
> 
> GPO12 output port also need to be manipulated: in 1200dpi it must be set
> to zero, while in other modes it must be one.
> 
> So, I had to add some workaround code in genesys_gl646.c. Results in
> genesys-hp2400-1200dpi-done.patch.
> 
> After solving this I notice another bug: if start scanning position (y)
> is not zero, the head moves double length way and eventually goes beyond
> bottom border. I fix this by changing optical_yres of MOTOR_HP2400 from
> 2400 to 1200. For fine tuning I create a new case in
> gl646_setup_registers() for MOTOR_HP2400. Values seems good, except,
> head positioning after calibration (shading) slightly different from the
> case when no calibration was done. Also I had to change y_offset in
> "Genesys_Model hp2400c_model" from 7.5 to 2.5.
> 
> Results in genesys-hp2400-fix-head-positioning.patch.
> 
> 
> That way I have fully functional HP2400 scanner with genesys backend.
> All modes works fine: 50, 100, 150, 300, 600, 1200 dpi; Color, Gray,
> LineArt; 8, 16 bits per channel.
> Head positioning works well. XSane works well.
> 
> 
> Cumulative changes I made against current git master of sane-backends is
> in genesys-hp2400-complete.patch.
> 
> 
> Regards,
> Happy Alexey. :)


	Hello,

	good job. I'm going to apply it soon.

Regards,
	Stef