PROCESSOR-SDK-TDAX: Extracting sensor embedded data

Yes it is possible to receive image/non-image data with different data type into a single buffer. This is the default mode of operation. Are you facing any issues with this?

Sorry for the confusion. Will make it clear.

Thanks for explaining. I will check if the CSI-2 driver has the support for DataType field.

Can you please check if the sensor(IMX390) is correctly configured to send embedded data and video data as different data types?

Yes. The sensor is configured properly and this will send data as 

1. top front embedded data with data type 0x12,

2. bottom rear embedded data with data type 0x35, and 

3. image data with data type 0x2C.

Hi Athira,

The driver supports most of the standard data-type. The capture link might require an update to support some formats

Regards, Sujith

Yes I agree. But we need to know if it is possible or not to get the image/non-image data with different data types into different buffers using same virtual channel.

Currently "gSensorInterfaceInfo" structure defined in iss_sensor_tda2px.c as data type is 0x2C(image data only) and virtual channel num as 0.

Hi Athira,

You don't require to update the sensor layer for this. Before the capture link is created, add config required to receive different dt with same VC.

Regards, Sujith

Hi Sujith,

Can you share few more information regarding how to configure to receive data(with different DT) into separate buffers.

Hi Athira,

Referring vision_sdk\apps\src\rtos\iss_usecases\iss_capture_isp_simcop_display\chains_issIspSimcop_Display.c, consider updating the function chains_issIspSimcop_Display_SetAppPrms () with

1. Data type pUcObj->IssCapturePrm.inCsi2DataFormat to data format specified by the sensor for meta data

2. Virtual channel pUcObj->IssCapturePrm.inCsi2VirtualChanNum to virtual channel number of the meta data

Also, check if you require to increase the allocated memory to store the received buffers.

Regards, Sujith

Hi Mayank,

Have you got a chance to check this ?

Hi Sujit,

Thanks for your reply. But we are not looking for this exactly. Will explain it.

We need to receive the image and non-image data with different data types separately into three different buffers parallel (in same virtual channel). And then we need to pass this image data and embedded data to our algorithms. Sensor is supporting single virtual channel only.

Currently there is no support in SDK for handling the multiple data types at the same time. So we need to know if this is feasible or not to implement same using current framework or else by modifying it.

1. top front embedded data with data type 0x12,

2. bottom rear embedded data with data type 0x35, and 

3. image data with data type 0x2C.

Please check this point. I guess you got my query. 

Hi Athira,

Yes, this can be supported. You will have to choose an multiple channel usecase

1. modify to not configure multiple sensors

2. Update to use the same VC for all channels

3. Update the data-type for all channels.

Regards, Sujith

Hello Athira,

Sujith has checked the support and explained it in the posts above. Please let me know if there is anything else you need.

Regards,

Mayank