We are trying to use the DM3730 Torpedo+Wireless SOM and dev board with a custom image sensor board, and are having trouble.  Hopefully one of you can help us out!

The image sensor we're using is a monochrome, 10-bits-per-pixel Aptina MT9V033.  Our primary problem is that when we attempt to capture an image from the sensor, we only seem to get the least significant 8 bits per pixel.  We need the full 10-bit resolution.  Here is a summary of what we've done:

1) Electrical connections to the sensor.  We routed the 10 sensor data signals to the CAM_D0-CAM_D9 pins on the camera connector (J6) on the dev board, and we set jumper JP5 for 12bit operation.  We also tied CAM_D10 and CAM_D11 to ground since we're not using those bits.

2) Pin muxing.  We modified the dm3730logic_camera_init() function in board-omap3logic.c to enable the 12-bit camera interface by changing the "#if 1" line to "#if 0".  Kernel startup messages confirm this change was executed.  We also added appropriate setup functions and structures for our image sensor in this file, similar to those functions and structures defined for other image sensors in the file.

3) Driver.  We have written a driver for the sensor, basing it on the MT9V032 driver that is included in the BSP, and modified the kernel config files to include our driver in the kernel and filesystem. In the driver, we set the media bus format to V4L2_MBUS_FMT_Y10_1X10 since our sensor is a monochrome, 10-bit sensor.  The driver seems to work fine - it can communicate with the sensor successfully, and it seems to correctly support the software interface needed.

4) Command line operation.  The system boots successfully, the driver executes, everything seems to be OK.  We run the following commands from the command prompt in order to configure the image pipeline and grab an image:

a) (Set up the pipeline: sensor->CCDC->CCDC output) media-ctl -v -r -l '"mt9v033":0->"OMAP3 ISP CCDC":0[1], "OMAP3 ISP CCDC":1->"OMAP3 ISP CCDC output":0[1]'

b) (Set up the pipeline formats: grayscale 752x480) media-ctl -v -f '"mt9v033":0 [Y10 752x480 (1,4)/752x480], "OMAP3 ISP CCDC":2 [Y10 752x480]'

c) (Grab 4 images and save the data to a file) yavta -f Y10 -s 752x480 --capture=4 --file=image /dev/video2

All three of these commands seem to execute correctly.  "media-ctl -p" confirms that the pipeline is configured the way we expect.

 5) Analysis.  We expect the image file to be a series of 16-bit pixel values, with the top 6 bits set to zero.  Instead, we see the top 8 bits always set to zero.  The bottom 8 bits do match the 8 least significant bits of our sensor data stream, but the top 2 bits from the 10-bit sensor data are missing.

 We cannot figure out why the top bits are being clobbered.  Can any of you help us out?  We can provide more data, including source code if necessary.  Thanks so much for any help!

Ryan