Dear 6y9,
thank you again for your extensive answer.
Sorry that I was not clear enough in the beginning. Our goal is indeed a pure sequential operation.
We will also have a separate hardware trigger for both cameras and a good knowledge of what is happening on the mipi lanes after a trigger. Hence, overlapping image sequences will not be a problem.
libcamera will also be no issue, as we are not using it at the moment and have no intention to use it in the future, like described in my last post (please correct, if my assumptions were wrong regarding this point). We are working on the video-device level and are using IOCTL and V4L2 to access the device directly.
I think the only big remaining issue will be:
So it all boils down to the question, if we can change the image geometry and maybe also the pixel-format (data type) in unicam dynamically? If it is possible maybe you can give us a rough idea how we can achieve this most efficiently? That would be great.
Thank you again!
thank you again for your extensive answer.
Sorry that I was not clear enough in the beginning. Our goal is indeed a pure sequential operation.
We will also have a separate hardware trigger for both cameras and a good knowledge of what is happening on the mipi lanes after a trigger. Hence, overlapping image sequences will not be a problem.
libcamera will also be no issue, as we are not using it at the moment and have no intention to use it in the future, like described in my last post (please correct, if my assumptions were wrong regarding this point). We are working on the video-device level and are using IOCTL and V4L2 to access the device directly.
I think the only big remaining issue will be:
I am not sure if we can provide the same bit-depth and pixel format (hence the same data type) from both cameras, but it could be possible. However, we can definitely not provide the same image geometry from both cameras.Practically you'd need to have the two sensors being identical so that you don't change the image geometry between frames. Unicam will have a line stride programmed, and will be looking for the configured number of lines per frame, although it will also react to the CSI-2 frame end short packet. It will also have a filter for which CSI-2 data type is image data, so both would need to use the same bit-depth if using Bayer or mono sensors.
So it all boils down to the question, if we can change the image geometry and maybe also the pixel-format (data type) in unicam dynamically? If it is possible maybe you can give us a rough idea how we can achieve this most efficiently? That would be great.
Thank you again!
Statistics: Posted by marcp — Tue Feb 06, 2024 2:57 pm
