Linux/AM5728: MT9V032 camera module questions

The MT9T111 sensor on TI GP EVM is configured to capture in 8 bit YUV mode, so the data output pin used is from D2 to D9 pins (D0 and D1 is ignored). Are you planning to capture using mt9v032 in 10 bits parallel RAW RGB mode? If so -

1. You will need to connect all the 16 data pins to capture 10 bit output.
2. You will need corresponding modifications in your sensor driver to use 10 bits mode when using MT9T111x driver as reference as it is programmed for 8 bits capture
3. *.dtsi file will need additional entry on <bus-width> = 16. This is missing in example dtsi file as the default bus-width is  8 bits and that's what is sensor is programmed for.
4. The VIP driver will need to be configured to capture in 16 bit RAW mode. Follow this wiki page link 

feng said:

1. If my idea mentioned above is right . A dedicated clock oscillator is provided for my camera module,  it works at 27MHz  and  provids 27 MHz clocks to the sensor mt9v032.  Then the camera's pixel clk (pclk)  is     27MHz too, which  is connected to the BVIN3A_CLKO of P9  (pin num : 28).   Do i have to modify the "pixel-clock-freq=<96000000>" to "pixel-clock-freq=<27000000>"  in am57xx-evm-common-cam-mt9t111.dtsi?   what does the "input-clock-freq=    <32000000>" means ?  when the linux kernel gets the information of "pixel-clock-freq = <96000000>" and "input-clock-freq=<32000000>" , what kind of work will it do?  whether TI has a detailed document on the description of the device tree?

Please check following documentation under board-support/linux-xxx/Documentation/devicetree/bindings/media folder

video-interfaces.txt, ti-vip.txt, i2c/mt9t11x.txt, i2c/ov2659.txt document

- input-clock-freq(Source clock freqquency) and pixel-clock-freq(Target pixel clock frequency) are sensor board specific. Please check if your sensor board needs configuration of these properties. 

feng said:

2.After reading the mt9t11x.c file, my understanding is that the mt9t11x just acts as a subdev of the V4L2  framework.Does this means when we want to add a new camera driver ,we just have  to add a new subdev to the V4L2 framework?  If the ansower is yes, then:     As far as I know , the complete V4L2 framework consists of  video_device,  v4l2_device and v4l2_subdev . Is it  because TI has implemented the work of video_device and v4l2_device according to am5728 platform , so we just have to complete the v4l2_subdev subsection?

Yes, this is correct understanding.

feng said:

3. Is the video capture framework for TI only based on V4L2 or anything else? For example, the soc_camera framework?  I want to read this part of the code in detail, but I don't know what documents to look at, and whether TI has a detailed document on the description of this piece?

Yes, it is only based on V4L2 based framework. soc_camera framework is obsolete. You will need to read the MT9T111x driver code and refer to V4L2 APIs document available on internet.