I would be grateful for some advice on autotune and tail vibration

I am very grateful to Bill Geyer for his wonderful efforts, which enabled me to complete the complex tuning process of a helicopter with great success. I would like to take this opportunity to thank Bill Geyer.

The reason for this post is as follows.

  1. autotune angle p gain
    The autotune resulted in an angle P gain of 10 for all 3 axes. The message from the GCS at the time of the flight indicated autotune success, so it does not appear that autotune failed, but it does seem odd that all 3 axes got a high gain of 10. I would appreciate any advice on this.

  2. tail vibration
    Autotune on the yaw axis has given the helicopter a better response and less vibration than before, but contrary to my expectations, the tail is still vibrating (but weaker than before). The rate D-gain is 0.003, so the vibration does not seem to be caused by D-gain. I would appreciate any feedback on whether this is due to the PID gain or some other mechanical issue.

The flight log is attached below.

Flight Log

Finally, I would like to share a video of the aircraft flying after tuning. The tail oscillates a bit, but it has a nice hover performance. Very cool!

@aaaa I’m glad you found the autotune feature helpful. I am still looking through your logs but have other things going on. I will try to post my thoughts and recommendations later this week.

2 Likes

I am always grateful for your kindness!

I also have a question that is separate to this topic.

My copter is using align 380mm blades, I assume they are similar to naca 0015(as I don’t have data from naca 0014) as they are 5mm thick on a 35mm chord.

So I was using a pitch of about 5 degrees as a hovering target, which is the L/D max point of naca 0015 at 100,000(mean) - 200,000(tip) reynolds number. (=head speed : 2100rpm)

However, I soon realised that I had not taken into account the velocity of the upper inflow to the disc.

So I calculated the hover rpm to keep the blade’s AoA around 5 degrees and found that the headspeed was 1700 rpm.

My copter is 1.9kg, so 1700rpm is expected to cause difficulties in controlling the copter. Do you think the ArduPilot will be able to maintain control under these extreme conditions?

always thank you for your time

@aaaa After reviewing your logs, here are some suggestions.

In the pitch axis, it looks like autotune did not find the correct value of the FF gain. I determined this by looking at the second frequency sweep while it was conducting the Rate P and D tuning. At the low frequency the gain is around 0.7. So I would suggest raising the ATC_RAT_PIT_FF to 0.135. The rate P and rate D gains that the autotune determined should be fine.

The roll axis looked pretty good. The roll FF gain might be slightly low but don’t change it for now.

For both the pitch and roll angle P gains, reduce them to 6. Typically if during the test, the response gain doesn’t change much but it raises the angle P gain to the max value, then I’m not inclined to believe the results. An Angle P gain of 6 is typically a good value for disturbance rejection.

As for the yaw axis, I would recommend lowering the Rate P gain to 0.4. The value found during autotune might be a little high. I would recommend lowering the Angle P gain to 6 for the same reason as the pitch and roll axes.

Please conduct a tune check (AUTOTUNE_SEQ = 16) for all three axes and post the log files.

I don’t know that I have enough experience to say whether it would or not. You would need to lower it incrementally and then retune the control system. I think the biggest concern would be blade stall. You could also check the hovering blade pitch angle and see that it corresponds to what you predicted. If it is close, then you have some confidence in your prediction. Once you determine the rotor speed that you want to keep it at, then you’ll want to retune to ensure the gains are set properly.

2 Likes

I appreciate your advice. I’m always grateful for the time you spend helping others.

However, I’m afraid to say this because it seems to make your efforts in vain… A recent flight test led me to change the headspeed.

As I mentioned before, I tested the theoretical optimum of 1740 rpm and was able to increase the hover efficiency to 9.5 g/W (at 2100 rpm, the hover efficiency was 7.4 g/W).

At that headspeed, the control was very loose, but surprisingly, the autotune worked, and I was able to reach some flyable levels as a result of the tuning.

However, as you mentioned, I did experience a vibration at a high angle of attack that I presume was a blade stall.

The hover pitch at this time was around 8 degrees, presumably as a result of my failure to account for the 3D effects of the blades.

After roughly applying the 3D effect to 80%, I lowered the target angle of attack a bit and finally settled on a headspeed around 1860 rpm.

At that headspeed, after autotune, I’ll set the angle P-gain to 6 and if there is any vibration, I’ll lower the yaw rate P-gain a bit and do a tuning check flight for all three axes and post the logs.

I’ll probably be able to fly it sometime over the weekend.

Thanks for your response :slight_smile:

Not completely in vain. It is always worthwhile for me to review the autotune logs because i can see where it is not tuning well. I have an update in the works. It doesn’t automate it any more but it adds some level of safety and cleans up the code quite a bit.

2 Likes

I am saddened to report that… There was a crash during autotune.

The airframe is the same as before, with changes to the FC and mounts.

I switched to a Pixhawk 6x pro instead of the Pix32v6 I was using, and changed the FC mount accordingly. The photos are shown below.

After carefully transferring the parameters, I went straight into autotune without any testing, as the existing settings were sufficient.

After the VFF autotune, when adjusting the rate P&D gains for the roll axis, the aircraft suddenly exhibited uncontrollable behavior in the roll, pitch, and yaw axes.

I switched from autotune to althold and it was still uncontrollable, and on stabilize it was still uncontrollable.

Fortunately, the collective was controllable, so I grounded the uncontrollable vehicle the moment it leveled out and landed it. Thankfully, I didn’t have to carry any of the burning debris home :slight_smile:

During the subsequent log analysis, I noticed that the vibration was quite severe (despite a better FC and better mounts) and initially thought it was a ‘lean’ due to vibration.

However, when I went through the logs a little more slowly, I realized that unlike ‘lean’, the attitude estimation was fine, but the attitude was not following the desired attitude.

The same thing happened yesterday when I first replaced the FC and tried to autotune, but at the time I thought it was due to excessive VFF gain.

I would appreciate it if you could review the logs and give me some advice on the issue.

Vibration is never a good thing, so, first, I’m going to try to find the source of the vibration and think more about isolating the FC.

I’ve attached the logs below.
Flight log

I look forward to your response, thank you.

1 Like