Accessing the frame number from MIPI-CSI2 short packets on the iMX93

[enrique001]

Hello Community,

We are attempting to develop a nano drone which uses Visual-Inertial odometry to autonomously navigate in indoor environments. We have decided to use NXP's iMX93 and are attempting to integrate an OV9282 MIPI CSI-2 camera.

We need to timestamp images from the camera precisely for good VIO performance. We have decided to timestamp strobes from the camera using the iMX93's TPM module and correlate these timestamps to images received in userspace using the frame sequence numbers.  Our timestamping strategy requires us to get frame numbers from the camera for each image that is received by the iMX93. Looking at AN5305 (Section 2.2.2.2), we think these frame numbers may be included in the MIPI CSI-2 short packets, in the "word count" field.

The MIPI CSI-2 and DPHY drivers do not seem to have any reference to this field and I was unable to find information in the iMX93 Reference Manual which would allow me to modify the DPHY driver to get the frame count directly from the raw CSI-2 packets.

Is it possible to read the "word count" field directly from the raw MIPI CSI-2 short packets on the iMX93, or does NXP provide any additional information (E.g. register maps) about the DPHY peripheral which would allow me to make the relevant driver changes to read it ?

Thank you for any assistance,

Enrique

[joanxie]

you should mean word count as below, right?

based on my understanding, the The CSI-2 low-level protocol turns the pixel data, VSYNC, and HSYNC pulses into a series of packets. Pixel/Byte Packing/Unpacking Layer unpacks bytes from the Low Level Protocol layer into pixels before sending the data to the Application layer. so data packing and unpacking is done in the layer level

[enrique001]

Hi joanxie,

Yes, the word count field in your screenshot is the data that I want to access. I need to access the data within this field in Linux kernel-space.

It is my understanding that on the iMX93, the byte-to-pixel unpacking is being done in hardware that is controlled by drivers (DPHY or CSI-2) running in kernel-space. Is this correct? If so, is the word-count data stored in any register that I can read by modifying the DPHY or CSI-2 drivers?

If this data is not accessible in kernel-space that would be helpful to know as well.

Thank you,
Enrique

[joanxie · Accepted solution]

I already replied to you in your the similar internal case, I don't find there registers are exposed to the customer in the imx93 reference manual, if you have further issue or finding, pls mail to me in the new case

[enrique001]

Thank you for your help.