Hard for this old guy to tell if this is the really bad horizon tilt or just bad editing.
What are your thoughts?
What are your thoughts?
Opinions on this video
- Thread starter Jagerbomb52
- Start date
I'd say there's an issue, but it's not huge. This should be a fairly simple firmware update.
It looks like gimbal control loop is using too long a time constant, because it does eventually basically recover. Having a long time constant will make the video feel smoother, at the expense of taking longer to recover after a maneuver.
I expect it will take two more firmware fixes to fully resolve. The next firmware fix will be a little too jittery, and then they'll get it right.
Honestly, this is the sort of issues I expect with early release tech products. There's nothing functionally wrong, just an issue with tuning preferences. That sort of thing is really hard to sort out without using things like focus groups and dramatically slowing development.
It looks like gimbal control loop is using too long a time constant, because it does eventually basically recover. Having a long time constant will make the video feel smoother, at the expense of taking longer to recover after a maneuver.
I expect it will take two more firmware fixes to fully resolve. The next firmware fix will be a little too jittery, and then they'll get it right.
Honestly, this is the sort of issues I expect with early release tech products. There's nothing functionally wrong, just an issue with tuning preferences. That sort of thing is really hard to sort out without using things like focus groups and dramatically slowing development.
HiloHawaiian
Well-Known Member
To be honest, I was expecting worse. If you asked a group of non-flyers to watch a pretty video, I bet very few would notice. That said, it is an issue. Is it a gimbal issue or editing? My vote is gimbal, and like GSA said, it’s a tweaking kinda fix, which drives coders nuts. One “fix” can affect other systems in unexpected ways, it’s a fine-tuning effort which takes a lot of resources and time. Autel has limited resources, so it could be awhile (or they could stumble on a fix tonight!)
So while I could not tell them how to fix it, I think I understand the issue. There are effectively two yaw axis. There is aircraft yaw and there is camera yaw. I think the evo is either tracking one out of sync with the other or not tracking one yaw at all. I'd wager on the second. When it is what appears as aircraft yaw everything is fine. But if the camera is deflected downward and then the camera is turned in relation to the body you are no longer moving on a single plane. Instead of a clean left to right motion you are on a pan and a sweep at the same time. That likely has to be accounted for in software more than they are or at a faster sampling interval. What I would hope they would have is a testing rig where they can rapidly apply motions and freeze the gimbal to built a relation table of movement intervals.
Geometric coding is right up there with date math for me as a developer under the "thing's I prefer to avoid" list.
Geometric coding is right up there with date math for me as a developer under the "thing's I prefer to avoid" list.
Similar threads
