[Elphel-support] problem with stereo setup

[Andrey Filippov]

Jorick

I'm working now to make software to distinguish between eyesis camera and
other cameras that have 10359, tomorrow it will likely be done - then it
will be easier to use latest software.

*Jan  1 00:01:00 Elphel353 kernel: JFFS2 notice: (599) check_node_data:
> wrong data CRC in data node at 0x03b9d800: read 0xfceecc3d, calculated
> 0x8a87afb7.
> Jan  1 00:01:00 Elphel353 kernel: JFFS2 notice: (599) check_node_data:
> wrong data CRC in data node at 0x00b2e000: read 0x2fcf066a, calculated
> 0xc1b83d1c.
> Jan  1 00:01:00 Elphel353 kernel: JFFS2 notice: (599) check_node_data:
> wrong data CRC in data node at 0x00b23000: read 0xe045f277, calculated
> 0x76ac2023.*
>

Not really serious problems. It is what happens when you write to flash and
then power down the camera without "sync" or "reboot" (in most cases
internal camera software performs "sync" automatically). It is similar to
"Safely remove" on a computer. JFFS2 handles such problems, but those
warnings will stay there until those blocks will be overwritten or camera
reflashed.


> ***multisensor_pgm_sensorphase **** ERROR adjusting SDRAM clock phase in
> arch/cris/arch-v32/drivers/elphel/multisensor.c:1227:multisensor_pgm_sensorphase,
> result=0xffffffff
> *
>

That seems to be also not so critical problem. With XIlinx FPGA it is
possible to adjust phase by 90-degree steps and fine steps (~22ps) in the
range of more than 90 but less than 180 degrees (at the frequencies we
use).  And the 90-degrees steps are not so precise, so it is difficult to
compare full phases (defined by a pair of 90-degree steps and 22 ps ones)
When the memory clock phase is adjusted, camera tried to find to margins -
high and low and then use the middle. On the 10359 board frequency now is
low - just 96MHz (vs 160 on 10353 board) so "good" phase range is rather
wide and in some cases software fails to find one of two margins (but it is
still in "good" phase range).


> * Jan  1 00:01:14 Elphel353 kernel: Resetting MT9X001 sensor^M*
>

Yes, that is OK - it just falls to the single-sensor initialization of a
single sensor (at this stage all the sensors are initialized in the same
way).

Andrey
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20100810/c98d68b3/attachment-0002.html>