@griffin , this seems like an network issue. So please try the following:

open a terminal and just run the following command to "inspect" the hires stream so that it is always going and you can check if it's running any time : voxl-inspect-cam hires_down_small_encoded start a continuous ping test from your host that is trying to receive the video feed keep testing the video feed until you see the issue occur and check if ping is dropping out and double check voxl-inspect-cam output to make sure the data is flowing.

What i expect is the output of voxl-inspect-cam should be continuously updating and you may see ping having trouble. Please let me know what you observe.

Regarding the voxl2 heating up, i will check with the team. open-vins-server is running (VIO) and it has been recently updated to use GPU for feature detection, so a slight increase in voxl2 temperature may be possible.

Alex

the fix has been committed to voxl-camera-server dev branch : https://gitlab.com/voxl-public/voxl-sdk/services/voxl-camera-server/-/commit/520bad13a573dde500d36cf1008e67102e0b6078 . I tested it for the rotated and non-rotated use case using voxl-streamer.

Alex

@brahim after looking at this some more I think it is valid to have multiple streams per camera. I guess the use case would be to have a high res stream and a low res stream that's separate or something like that.

However I think people would still like to setup multiple cameras. I haven't tried this yet, but I think it could also be valid to spawn a manager per camera, and then stagger the component ids for each app. I think this might also fix the threading issue since each application would initialize separate pipe reads and wouldn't be blocked by each other?

@Eric-Katzfey I'm using a different mavlink GCS that's not QGC, so I need to make voxl-mavcam-manager conform to the standard.

@griffin Great to hear!!

It is possible, depending on what size tip and what kind of soldering iron you were using, that 650F may actually not have been hot enough. After talking about it with my team I realized that we have set our irons up to the 750-800F range.

Nonetheless, it is likely the safest option to just splice in the new wires just as you have done

edit: use mavutil.mavlink.MAVLINK_MSG_ID_LOCAL_POSITION_NED* instead of magic number 32 for message ID

@James-Strawson

Hey James,

Thanks for the tip, installing via apt install voxl-mapper and then running with systemctl start voxl-mapper worked out of the box.

Kind regards,
Griffin

@griffin , this is expected behavior (from our point of view 🙂 ). Here is what's happening:

the image from hires large stream (which is what I assume you are using) is either 3840x2160 or larger, around 4xxx by 3xxx depending on which camera you are using voxl portal is not very efficient at processing large images and when it receives the huge 4K image (YUV or RGB), it will use turbojpeg (software JPG encoder) to convert it to jpeg. This process takes a lot of cpu for large images and especially if fps is 30, it is too much for voxl-portal to handle if you want to view hires images through portal, you can enable (if not enabled) en_small_video stream in camera config file /etc/modalai/voxl-camera-server.conf and set resolution to something smaller, usually 1280x720 or similar is good (depending on the aspect ratio that you need) once you enable the small video stream, you can view the hires_small_color stream from the voxl-portal the cpu % (assuming you are looking at output of top) will show the cpu usage percentage at the current cpu frequency . So this number is often very misleading because if the cpu is not running at max frequency, it will have higher % usage compared to perf mode (which can be set using voxl-set-cpu-mode perf. You can use tool voxl-inspect-cpu to show the total usage of each core (which takes into account the frequency of each core) and you can also see the current core frequencies (but you cannot see individual processes).

Please let us know if you have any other questions.

Alex

@griffin Correct