I agree about the pedestal mounts. If I had a set of the collet prop adapters I would use those instead. I too balance all my props.
The next quad (or hex) I build will use these motors. I haven't run them yet but they appear to be built to higher tolerances, no slop or play whatsoever.
I just thought I'd add my own ELEV-8 flight. I'm sure we will see a lot of this, since the ELEV-8 platform is perfect for cameras. This was filmed with a GoPro HD.
That's some pretty fine flying! How do you keep your orientation when flying that high? It's a little easier with a regular heli with it's unique profile.
I think with the self- leveling enabled it's probably less of an issue than without it. The biggest reason I want it on mine is so I can go ludicrously high and not worry about it tipping over before I can figure out the orientation. Or maybe Kevin just has awesome eyesight and Mad Skillz - I can't deny that possibility.
For some reason I thought the Sport had the auto-leveling as well, but I just checked the specs and it does not. I spoke to Kevin this morning and he was indeed using the auto-level feature. He says at that height it's pretty much a requirement because it's really hard to tell which direction to correct in if you're not level.
In related news, I have my DCM code incorporated into a new version of my flight code, and have auto-leveling "working". It's still rough, and in dire need of tuning, but I expect I'll have something I can share within a week or two, depending on how much time I can put into it.
There shouldn't be much of an issue with it (mine aren't centered). It means that during a stationary yaw your accelerometers will read a small centrifugal force, but that's about it. If you place them really far from center it would make the effect much more pronounced, and might cause you grief.
Or maybe Kevin just has awesome eyesight and Mad Skillz - I can't deny that possibility.
I don't know. He's a very BOLD pilot and we're watching him closely right now. Let's be honest - with 30+ connections on an ELEV-8 there's room for failure. And if that happens then his craft will drop to the ground very quickly.
Though that flight was not at Parallax and conducted on his own time, I think I better set up some new rules for flights at Parallax and the office park (fly only over "our" airspace or the field next door, for starters!).
I'm all excited - I got my DCM-based self-leveling code tuned to the point where it's quite flyable. It still needs some tweaking, but a few changes I made today made a huge difference in the stability. No compass yet, so it won't hold a true heading, but it's pretty steady. I still have some undershoot (it doesn't quite hit the level it's supposed to) but I think a little more tuning will clear that up.
Change #1: After reading a bunch of papers on PID design and tuning last night, I found that it's very common to make the system ignore the control input (set-point) when calculating the derivative values. This avoids large spikes in the derivative caused by step changes in the control input. Changing mine in this way makes the response to control inputs much smoother.
Change #2: Start tuning using derivative, not proportional. I found this paper: http://igor.chudov.com/manuals/Servo-Tuning/PID-without-a-PhD.pdf and it details a quick and dirty way to tune your PIDs, and gave me a much better understanding of the way they behave in general, which helped me tune the system and close in on good numbers much faster than I had been doing before.
I can't wait until I can put the controller on the ground and walk around the thing while filming it. I'm hoping that's going to be soon.
I started a thread in the prop forum to ask a few questions about standardising the parameters passed between objects in the QuadCopter code. It turned into a bit of a discussion about my code which I gratefully acknowledge is base on Jasons code (and others for other modules). Then followed by an attempt to run it up.
Thanks heaps to Jason for the reality checks and his algorithms.
Jason: Looking forward to seeing those vids too! This is really fun once you get over the worry about destroying something.
Ken: I must post a photo of our backyard... Lots of ways to get into trouble there... a pool, followed by a large canal and our boat. The front is better but it is a reasonably busy road for our little island.
I'd like to see the first minute of that flight from the ground - it looked like you were really high, and that banked turn was pretty much vertical. Looked like fun.
Jason:
I was having a think overnight about the drift of the gyro... When the quad starts it calculates the gyro home position. Now, supposing the quad is not purely horizontal in both axes. The gyro would be set with a slight offset and this would cause the quad to drift in one direction while it hovers, because the quad would not be perfectly horizontal. This would occur, no matter how efficient the gyro and code. Without another sensor, accel etc, this should be impossible to correct. Does this make sense???
It seems like with the HoverFly Sport that any stick movement moves the home position of the gyro. Even if it starts out all askew as soon as you lift off and correct it, it will hold the last orientation.
Mine behaves the same way - All gyro readings are added together to produce an estimate of the current orientation. All stick inputs are added together to produce a desired orientation. The two results are fed to the PIDs to control the flight. The craft has no idea what "level" is since there's no reference (accelerometer).
The zero being read at startup is to compensate for the fact that a gyro may have a slight offset - IE the readings at rest may not be exactly zero, but those readings are angular rate, not orientation. Calibrating that on startup means that I get the most drift-free zero value I can, so if the craft isnt rotating, the gyro output with that offset subtracted should be an actual zero, or very close to it.
Comments
The next quad (or hex) I build will use these motors. I haven't run them yet but they appear to be built to higher tolerances, no slop or play whatsoever.
That's some pretty fine flying! How do you keep your orientation when flying that high? It's a little easier with a regular heli with it's unique profile.
In related news, I have my DCM code incorporated into a new version of my flight code, and have auto-leveling "working". It's still rough, and in dire need of tuning, but I expect I'll have something I can share within a week or two, depending on how much time I can put into it.
Kevin - Remind me not to park my car where you're flying. :-) I didn't think you were ever going to stop climbing. What an awesome flight!
You guys are making me jealous. I've got to get mine together. Thanks for the great posts.
...Tiger
...Tiger
What, if any, consequences would there be to mounting the control board (such as the HoverFly Pro - gyros and accelerometers) off center on a quad?
I don't know. He's a very BOLD pilot and we're watching him closely right now. Let's be honest - with 30+ connections on an ELEV-8 there's room for failure. And if that happens then his craft will drop to the ground very quickly.
Though that flight was not at Parallax and conducted on his own time, I think I better set up some new rules for flights at Parallax and the office park (fly only over "our" airspace or the field next door, for starters!).
Ken Gracey
Thank you for making that contribution!
Ken Gracey
Change #1: After reading a bunch of papers on PID design and tuning last night, I found that it's very common to make the system ignore the control input (set-point) when calculating the derivative values. This avoids large spikes in the derivative caused by step changes in the control input. Changing mine in this way makes the response to control inputs much smoother.
Change #2: Start tuning using derivative, not proportional. I found this paper: http://igor.chudov.com/manuals/Servo-Tuning/PID-without-a-PhD.pdf and it details a quick and dirty way to tune your PIDs, and gave me a much better understanding of the way they behave in general, which helped me tune the system and close in on good numbers much faster than I had been doing before.
I can't wait until I can put the controller on the ground and walk around the thing while filming it. I'm hoping that's going to be soon.
For those interested, here is the thread
http://forums.parallax.com/showthread.php?136897-QuadCopter-(Tri-Quad-Hex-Octa)-propeller-chip-code-discussion&p=1061975#post1061975
I am getting close to takeoff now
Thanks heaps to Jason for the reality checks and his algorithms.
Jason: Looking forward to seeing those vids too! This is really fun once you get over the worry about destroying something.
Ken: I must post a photo of our backyard... Lots of ways to get into trouble there... a pool, followed by a large canal and our boat. The front is better but it is a reasonably busy road for our little island.
J
Jason:
I was having a think overnight about the drift of the gyro... When the quad starts it calculates the gyro home position. Now, supposing the quad is not purely horizontal in both axes. The gyro would be set with a slight offset and this would cause the quad to drift in one direction while it hovers, because the quad would not be perfectly horizontal. This would occur, no matter how efficient the gyro and code. Without another sensor, accel etc, this should be impossible to correct. Does this make sense???
The zero being read at startup is to compensate for the fact that a gyro may have a slight offset - IE the readings at rest may not be exactly zero, but those readings are angular rate, not orientation. Calibrating that on startup means that I get the most drift-free zero value I can, so if the craft isnt rotating, the gyro output with that offset subtracted should be an actual zero, or very close to it.