The advice from the expert.
The approximation of INS_HNTCH_REF if you are using a throttle-based harmonic notch. It can be fine-tuned with the filter review tool if necessary.
This may be helpful if you experience multiple noise/vibration peaks.
I guess he is using latest version thatās why i wrote it.
@Goran_F450 do not assume that, ask it directly. @prsh8ya are you using ArduCopter 4.6.0-beta1 or newer?
1 Like
I am using the latest stable version.
That one has no multi source throttle notch filter
I am struggling with setting up a big copter (2-3 kg, props 16 x 5.5)ā¦ I have had many small issues and problems, but the most dangerous one is this: when descending at about 4 m/s in loiter mode, the copter was descending fine. Once I applied power to stop the descent, the copter oscillated violently 3 times in pitch axis. The PIDs were obtained by the Autotune. What I am not getting at all, is that the desired pitch signal started to oscillate, and the real pitch followed with some overshooting. So why did Aurdupilot command these oscillations?
.
Another thing I am not getting is the following: initially, I setup the copter with 13 x 5 props and did autotune, obtained PIDs which were very similar for pitch and roll axis (something to be expected with weight distribution and geometry of the copter I have).
Now, after I changed to 16 x 5.5 props and did Autotune again, the roll PIDs remained pretty similar, maybe reduced by 10ā¦15 %. But pitch PID values were all reduced by about 50 %. I repeated the autotune for the pitch axis thinking that I may have made a mistake somehow, but obtained very similar resultsā¦ And with yaw, autotune struggled for more than 15 minutes, also reducing the P and I values,
Can you share your current LOIT_BRK_DELAY value?
Do you know both of these configurations have different initial values to begin with? Do you change them before Quiktune or Autotune?
LOIT_BREAK_DELAY is set to 1
The descent was vertical, without any stick input on pitch or roll, with a side wind of 10 km/h. The descent was quite stable until I put the throttle to center. I also noted small ābumpsā in altitude graph during the oscillation, although visually from ground, it seemed that copter was oscillating but not changing altitude. These bumps could be caused by prop wash entering the underbelly where the FC is located, not sure if it could contribute to the problem.
I forgot to change the initial parameters, I will redo the autotune now. The only parameter I changed was MOT_THST_HOVER (was 0.35 and I changed it to 0.2, and after flying it remained pretty close to this value). After tune I realized that I did not change FLTE, and INS_GYRO_Filter, and I changed them, but I did not redo autotune. So autotune was done with FLTE 18 (for prop 13), and after autotune it was changed to 14 (for prop 16).
All autotune ATC_ANG_XXX_P result in high values (12ā¦16), and I always reduce them manually to about 8. I am using the 0.1 aggressivness, but reduce manually the ATC_ANG_XXX_P, not sure if this correct and can be done or whether one should use a lower aggressivness value.
I tried to do an autotune with aggressivness of 0.05 (with 13 inch props), but the values the autotune gave me made the copter behave very mushy and to me it seemed that it was not stable enough
If you are using ArduPilot methodic configurator software how did you manage to forget to redo initial tuning and the other steps required to be done before autotune?
I need to understand how users missuse the software so that i can change the software to prevent that.
Try setting it to 0.3. and redo the flying with the set of initial parameters that produced the Autotune values that match the propeller size.
I would suggest you do a Quiktune first. Fly it in Loiter with LOIT_BREAK_DELAY, 0.3. If it is good, proceed with Autotune. If it is not ideal, find out the issue. It could be high noise or vibration in your drone.
Quiktune is easy to carry out; prerequisites are calm and low wind conditions, similar to Autotune.
That was sort of a lapse of judgement. I spent one day setting up with 13 x 5 and a small battery, than I started increasing weight and doing just Auto retune, and then I decided to swap the propsā¦ and did just auto retune again, so no fault of anyone thereā¦ I am quite new to quads (have a lot of experience with fixed wings)
I will try to change LOIT BREAK DELAY. I did a complete autotune with initial parameters modified for the 16 inch props and 0.09 aggresiveness. So far, the flying behaviour seems OK, and the roll-pitch parameters seem to be more or less similar.
I have noted very high Stabilize Roll and Pitch Error to Rate values, the ardupilot instructions suggest to reduce them from 10 to 7 for big copters, but in this case these are around 20 and 14ā¦ On previous autotunes the general behaviour was too twitchy and I had to reduce them from 15 to 7. In this autotune setup (as seen on the image), it seems that the overall behaviour is good, so I am not sure if I should reduce these values or leave them as is. Also, there is a formula in arducopter tuning instructions for FLTE which for the props I have should be 14. However, there is another mention that for very big props it is better to set FLTE to 10 before doing autotune. Not sure if this is correct
( 3. For large copters (with props at least 13inch or 33cm diameter) set the Rate Roll and Pitch filters to 10hz, these are: ATC_RAT_RLL_FLTT, ATC_RAT_RLL_FLTD, ATC_RAT_PIT_FLTT, ATC_RAT_PIT_FLTD
Did another test flight. I reduced Stabilize Roll-Pitch error to 9, not sure if it improved something.
The issues I have right now are these:
LOITER. When I move sticks abruptly, the copter oscillates but with limited amplitude which dies in about 3 oscsillations. In AltHold mode, I have zero behaviour like this, i.e. I move sticks to extreme and release in AltHold, and visually the copter does not oscillate at all. In POSHOLD mode I would say the behaviour is intgermediate.
|LOIT_ACC_MAX|500|
|LOIT_ANG_MAX|0|
|LOIT_BRK_ACCEL|150|
|LOIT_BRK_DELAY|0,3|
|LOIT_BRK_JERK|500|
|LOIT_SPEED|1500|
When descending, the copter seems to be unstable, wobbles from side to side. It seems to be much worse in LOITER than in ALTHOLD. The wobble does not seem to exceed 20Āŗ, but is often constant, i.e. it starts to oscillate, the amplitutde does not increase, but nor does it decrease, and ceases only after applying power. When applying power, for a short moment the wobble may get worse for one oscillation but then seems to die.
When flying at moderate speeds or faster (30ā¦50 km), the copter seems to wobble in the pitch axis slightly (by a few degrees), in all modes (Althold and Loiter)
Did you complete the methodic procedure until the end?
If not, no wonder it wobbles, right?
In so many drone tunings I have not done such steps. Before Quiktune was born, I depended on Autotune. May I ask have you done steps 18 and 19? Have you checked the drone vibration level?
Which Flight Controller Unit are you using?
Because Ardupilot just lets it drift, under Loiter it will try to stay at the position where you release the stick to the middle. If you are certain noise and vibrations are low, keep LOIT_BRK_DELAY 0,3, reduce Loit_speed, and try a small change to other Loit_xxx values to see if that is the way to get what you want.
What are your current PILOT_SPEED_DN and PILOT_SPEED_UP?
from your descriptions, it seems that the PIDs tuned (Quiktune or Autotune) are not close to perfect for position-locked flying.
I have checked the Vibe XYZ, XY mean is around 5, but with some occasional peaks of up to 40. Z is lower, 3 with max peak 20. The values increase when flying at high speed (70 km/h I get around 30 VibeX,Y)
Pilot speed up 4.2 m/s speed down 4.2 m/s (VALUE 420)
FC Speedybee F405 V4
under VIBE, there is another check, is the clip throughout zero?
F4 processor, a very old processor. limited supported RAM size.
Clipping ZERO.
I know F4 are more lilmited, but I do not see how they can influence basic behaviour of copterā¦ Especially for a larger size copter wheree the loop times are not so critical as for a small 5 inch racerā¦
I tried these on several test builds of 5 inch copters, and I was able to get setups which fly well.
I also did a tricopter VTOL on F405 which is also OK, and I am finishing a QUAD VTOL (which has already flown in QUAD copter mode), all this on the cheap speedybee f405 a 4x 1 60 A ESC. With QAUD VTOL I had issues with ESC overheating (it can support about 5 kg max hover in long term, I have built one of 9 kg), so I am splitting the ESCs in two and with forced cooling, to stay safe.
With the copter I am trying to set up now I did not have much luck. In fact, I have lost one similar during first serious mountain flightā¦ It flew well for the the first 30 minutes, until I started to descend, and then it lost contol, the copter was falling for more than 30 seconds and I tried to recover it in different modes (but not in ACRO!), and finally it crashed where I could bever find itā¦
Which Arducopter version are you using on the F4 chip? I notice CubeBlack becomes warmer than usual when running version 4.3.8 and above.
That also explain why you didnāt do Quiktune, F4 + limited RAM cannot run Lua script.
Since then, I have migrated to the H7 chip.
I donāt think loop times will be different for different sizes of drones.