Missing "System_Time" message to reference time_boot_ms to time_unix_usec

[TangoLima58]

Most telemetry messages contain a timestamp that provides the time since boot. The unix_epoch_time message contains absolute time but not the time since boot.

So there seems to be no way to directly correlate the time since boot to unix_epoch time within a single message and get their offset as it is determined by the flightcontroller itself. Yet this would be very useful to correlate all the telemetry data to unix time if the offset would be known.

Mavlink provides a system_time message that contains both timestamps and therefore the clock offset as well. As far as I can tell this is not yet implemented in mavsdk-python.

I hope I have not overlooked any way to get the timeoffset.

Is is feasible to implement the system_time message in mavsdk-python?

[julianoes]

Right, so what you would want is a way to do time sync? And between what and what? Can you give overall context what you're trying to do? Then, we'll have to do some thinking how to achieve it. MAVSDK is pretty terrible when it comes to timesync and timestamps at the moment.

[TangoLima58]

Thanks for your reply.

I want to perform measurements with a drone where part of the measurement system is on the drone and there are several other devices on the ground. All ground based systems are connected by Ethernet to a PC running a custom python code to gather data and control the drone.
I need to know the GPS position and especially the orientation (yaw pitch roll) during the measurement wich happens almost instantly.

If I could reference the timestamps of the telemetry data to GPS time I could timestamp the ground based measurements to GPS time as well and therefore determine the drones position and orientation well enou

[julianoes]

Ok, so some sort of time sync to get roll, pitch, yaw in sync with GPS UTC time. I wonder if MAVSDK should do the time sync internally and then expose not just PX4 boot time but also UTC time with attitude (and other) updates.

[TangoLima58]

Exactly! As per the documentation the timestamp of the telemetry messages can be either boot time or utc time and one should infer which one it is by the magnitude of the number. I expected to automagically get utc timestamps after gps fix but that does not happen. Additionally, there should be the "system time" message in the mavlink protocol that gives boot and utc time but as far as I understand it is not implemented in mavsdk python yet.

Any ideas how to solve the time-sync/stamp promblem?

[julianoes]

I expected to automagically get utc timestamps after gps fix but that does not happen.

Yeah, so I think that's up to the flight stack. If we look at HIGHRES_IMU for example, we can see that PX4 always just uses the boot timestamp:

https://github.com/PX4/PX4-Autopilot/blob/f7269c9c220cbd009c2031354a6da2a3ea3bb80f/src/modules/mavlink/streams/HIGHRES_IMU.hpp#L183

Any ideas how to solve the time-sync/stamp promblem?

I think I want to:

1. Add boot timestamp and Unix UTC timestamp to all telemetry messages.
2. Populate timestamp given what autopilot sends (usually boot timestamp)
3. Infer the other, e.g. Unix timestamp using timesync implemented internally by MAVSDK.