I'm using 16.09+r782~g8c101ad
That is one of the "Next" versions it is 16.09 ... + 782 sets of changes.
I just configured fast yaw response, I like it. Pitch and roll are much slower.
I didn't checked VelocityRoam, could You say more about VelocityRoam mode? Description in the wiki is not clear for me...
VelocityRoam: Just imagine Attitude mode that becomes Position Hold when you center the sticks. It is better than that though. The roll/pitch stick controls horizontal velocity. Throttle stick controls vertical velocity with a large center deadband so you don't have to be real careful to maintain the same altitude. If you have max horizontal speed set to say 5m/s then that is what you get at full stick. At half stick you get 2.5m/s. This is true whether you are flying into the wind or with the wind. Full throttle straight up is only whatever you have max vertical speed set to, and low throttle is that same speed, but coming down (NOT falling as you would with normal throttle). Be careful because zero throttle stops the motors (a safety feature that can be disabled, but I leave the motor kill there.
Try VelocityRoam. If it isn't the easiest flying imaginable, then there is a problem. I am guessing that the problem is that your Attitude mode response needs to be "instant" but it isn't.
There is a thing with autonomous flight modes. The quad needs to be set up so that Attitude mode works almost instantly. The code assumes this. If it doesn't have it, it gives a bank command and if it doesn't see a response soon, it asks for more and more bank and then it is too much. It gets confused.
When flying Rate mode you need a slower aircraft response for a given stick position, because your eye and brain is in the control loop. You hold this much stick for this long to get it to move 45 degrees, even if that is measured in 10ths of a second, you develop a feel for it that you can't develop if it moves too quickly with a small movement. If it is moving instantly, you can't reliably/repeatedly stop at 45 degrees.
When flying Attitude mode, although you could configure slower response, you don't have to because X amount of stick always produces Y amount of bank, and as I said before, the code assumes the Attitude mode response is quick unto instant. When you are flying Attitude mode, if you want it smooth, move the sticks slowly. Get to where you think "stick angle equals bank angle with no delay"
Think of a steering wheel on a car. You don't want it to delay so that you only get a turn a second after you ask for it, and you are used to it having instant response so you move the steering wheel smoothly.
In GPS flight modes like VelocityRoam, the code is flying in Attitude mode and you are asking the code to change speed from 0 to 5 m/s. The code initially banks a lot to accelerate quickly, but soon reduces the bank to almost zero when it gets up to speed. You don't even need to know that. This is like pressing the resume button on a car cruise control. It uses more throttle till it gets up to speed, but then it reduces it to just enough to maintain speed.
VelocityRoam: For instance, if you are flying forward slowly downwind with a small amount of forward pitch stick with a fair amount of wind at your back, it is probably actually banking backwards to hold it from drifting downwind quickly, but you don't have to know this. You just know that X amount of stick always produces Y amount of speed, regardless of which way the wind is blowing.
How did you limit the roll/pitch response? Using the Stabilization page response sliders? That is the correct way, but again, I think at least part of the problem is slow response in Attitude mode. I shudder to recall that some folks have detuned the PIDs in an effort to reduce the response.
Hmm I remember one case when RTB was slow than expected. RTB has "speed" settings, I figured out that above some values (6 m/s) increasing it did not reflect in speed change...why?
It may be that the max speed set in VtolPathFollowerSettings won't let you go any faster? It may be an issue in the flight code. Again, I recall that a guy got two different RTB speeds depending on which flight mode he was in before RTB, and that was probably a bug. I kind of assume that the issue had been fixed in next. I am guessing a lot here.
I've already done it, we discussed about it in the thread about my problems with banking take off. After AutoTune quad is (in my opinion) now quite stable durning flying.
I didn't know if this was a different aircraft. When someone uses AutoTune I expect that it will make the Attitude mode response nice and crisp, but that isn't what I think I saw in the video.
AutoTune was done with smaller (lighter) battery (1500mAh) and pathplanner on the video was done with 2200mAh, a bit bigger. Does it matter?
You should tune (auto or manual) with the heaviest, slowest configuration, but if you don't, it isn't usually a big problem. Maybe some roll/pitch small oscillations. Using the heavy tune in a light configuration just makes it fly like it is heavy, a little less responsive. Using a light tune in a heavy configuration causes small roll/pitch oscillations / ringing. This is one reason we have multiple PID banks. You can have one bank tuned for heavy and another for light if you want it perfect. I don't bother with it.
There is no problem with climbing (even with 4200mAh 3S battery which is pretty big for such 250 quad) or fast banking (120km/h with banking fly). Hard to say about coming back to level, I guess is OK but I have no experience to judge it, this is my first build. I'll try upload a video. The only difference I see watching other people's videos with "free style" flying is that my quad (rather my flying) has a little "deadband" style (don't know how describe it correctly and I don't mean GPS assist fly mode - it's perfect stable in the air), for me is hard to achieve precise flight especially between obstacles like trees etc - I need a space for safe flight (or just my flying skills are not yet high).
A crisp, tight, response in Attitude mode is a joy to fly and it makes you feel more in control, connected to the aircraft, and easier to fly.
For me and I suspect most people, hovering very slowly (especially with FPV) through a gate is harder than going a bit faster. You line it up from a ways out and by the time you get there, it's perfect and you just go straight through. It is like balancing when walking is easier than balancing when standing.
There is a possible invisible yaw oscillation (very small and very fast) that causes the throttle to respond as if the stick were closer to center than it is. High throttle isn't as high as it should be and low isn't as low. Motors get hotter than you would expect. If manually tuned, reduce the D terms (yaw is the important one). If AutoTuned (no need for retune flight) set the System -> Settings -> SystemIdentSettings -> DerivativeFactor to 0.5 or 0.25 (to save: click once on DerivativeFactor to highlight it. Then click the red up arrow (Save) at top of screen). Then re-export your PIDs with SmoothQuick (make sure you rebooted the FC after putting AutoTune on your Flight Mode Switch, you are disarmed, quickly toggle into AutoTune mode and out of it, 3 times, quickly, 6 clicks, I recall you have 2 seconds). You can be watching PID bank 3 (default destination of AutoTune PIDs) in Configuration -> Stabilization -> Bank3 -> Advanced, and you will see the new PIDs appear immediately after the 6x FMS toggle, especially the reduced yaw D term. If you set the DerviativeFactor to zero, then the D terms will be zero. That is OK and a good way to verify that the PIDs did get changed.
Uh...
Yea, don't worry about this. It is way down the list of possible causes.
I know I keep mentioning that Attitude mode response needs to be instant. If you intend to use Rate mode at all, be aware that there are settings that allow Rate mode to have a slower response and still allow Attitude mode to have instant (well much quicker anyway) response.