While in Loiter Mode: loss of altitude when executing a Yaw

ArduCopter Copter 4.0
61

Hey Shawn, I conducted a short flt. this morn. and collected data. No hint of drift!!! I do not get it.

Anyway, TO very smooth with little TX stick corrections needed to TO. STAB, very smooth, not twitching or drift (very very slight breeze). LOITER, pretty solid, a little bit of correction taking place but for the most part an anchor in the sky! The only thing I changed prior to this flight were the params: BATT_ARM_VOLT=11; BATT_LOW_VOLT=10.8; MOT_BAT_VOLT_MAX=12.6; MOT_BAT_VOLT_MIN=9.9
I posted a word doc with screen captures of several graphs from the .bin file and this doc. has other info. of you are interested.
Could you look at the file and see if anything gives you a hint of why it is no longer drifting.
https://drive.google.com/drive/u/0/folders/1oJu4VgNgmyL_zD6jq1j20pBN_rVmGqK6

62

Z axis vibrations are high

image
image1325×353 21.3 KB

You’ll have to see if you can find the cause and fix it, or it will eventually lead to a fly-away or loss of vertical control. X and Y are good though.

Attitude control is reasonable for starting out, it can be a lot better yet :slight_smile:

image
image1294×356 72.1 KB

See if you can get up some more flight time to check reliability, then try Autotune, probably one axis at a time at first.
Lets see where the PIDs go and then you should be able to set this and do Autotunes again:

INS_GYRO_FILTER,40.00

At the end of it the quad should be highly stable and very controllable.

1 Like
63

The Quad is flying as stable as before the param corruption. In fact, in STAB it is so stable that you would think it is in Loiter!

Tried rebal. of the props. this morning, did not improve the Z axis vib issue. Not sure what else to try. Weird though, I had gotten the Z axis Vib down to apprx. 10.34 mean, but about the time this param meltdown occurred the Zaxis Vibs went to hell! I hadn’t really done anything mechanically to reduce Z Vib. but flying and performing extreme maneuvers (Roll/Pitch) over several flights helped reduce the Z Vib from a mean of 20+ to 10.34. Does that make any sense to you??

I’ll try something, not sure what as yet, to get the Z axis vibs back in the 10 .xx .

64

Hey Shawn, I was able to reduce the Z vib from a mean of 35.88 to 21.67! The Quad is slowly refining the INS_GYRO_ X,Y and Z. Does this look like the Z Vib is now low enough to begin AutoTune. Up loaded .bin
So, to recap:
1 - I will finalize your table updates prior to conducting autotune
2 - Notch filter after autotune, not before
https://drive.google.com/drive/u/0/folders/1oJu4VgNgmyL_zD6jq1j20pBN_rVmGqK6

65

It’s actually preferred to run set up the Harmonic Notch filtering first, since Autotune will have a less noisy environment to work with. Since you’ve got reasonably reliable flight going, now could be the time. In the past I’ve probably said to do autotune first because the quad seemed like it really needed that to become flyable.

You just have to set these and do another flight like you just did:
INS_LOG_BAT_MASK,1
INS_LOG_BAT_OPT,0
…then let’s see that log.

66

Hey Shawn, ok I conducted a very quick flight this morning. Changes to Param included the remaining changes per the table you provided, and like last time (just prior to the param melt down) the Quad becomes unstable. So unstable that landing is challenging!

Pre- the table changes the Quad was VERY stable, in STAB MODE it was so stable you would think it was in Loiter. Post- table changes (see below the changes) the Quad once again is exhibiting jerky motions (unstable). The instability seems like it is Roll only, but I am not positive of this since the Quad becomes a real handful when flying with these changes. Is there any compatibility issue with the legacy Pixhawk i.e. < 1M and these table changes? Will the legacy Pixhawk i.e. <1M handle the INS_GYRO_FILTER=42? What about the ATC_ACCEL_P_MAX=116700 and ATC_ACCEL_R_MAX=116700?

image
image614×739 43.8 KB

.BIN from the fight uploaded.
https://drive.google.com/drive/u/0/folders/1oJu4VgNgmyL_zD6jq1j20pBN_rVmGqK6

67

OK, that’s disappointing. The only parameter changed affecting the stability is INS_GYRO_FILTER (and the ACCEL values, but not so much).
I’d like to be able to leave it as-is for now and try another flight with these altered:
ATC_ACCEL_R_MAX,80000
ATC_ACCEL_P_MAX,80000
ATC_ANG_RLL_P,6.5
ATC_ANG_PIT_P,6.5

And from the Harmonic Notch filter data you’ve collected, set these before the next test flight:
INS_HNTCH_ENABLE,1
INS_HNTCH_MODE,1
INS_HNTCH_REF,0.22
INS_HNTCH_FREQ,92
INS_HNTCH_BW,46
INS_HNTCH_ATT,40
INS_LOG_BAT_OPT,2

I’d keep the battery failsafe settings at the recommended levels or you can easily damage the battery:
BATT_CRT_VOLT,10.5
BATT_LOW_VOLT,10.8
They dont affect attitude control.

If there’s something that wont work on a pixhawk<1M then that feature becomes unavailable, we cant enable it. All the tuning parameters are common across all hardware types and it’s just maths, in Ardupilot priority and loop rates are given to functions that need it most. Great minds have worked on that stuff :slight_smile:
A pixhawk<1M works equally well on a large multirotor with 20+ inch props, or a race style quad with very high filter values and high rates of change.

1 Like
68

Shawn, the Quad is pretty unstable, like I said the TO/flt./landing were a hand full. Do you think changing the below listed params will settle out the Quad handling?:
ATC_ACCEL_R_MAX,80000
ATC_ACCEL_P_MAX,80000
ATC_ANG_RLL_P,6.5
ATC_ANG_PIT_P,6.5

69

Hi Greg, That’s the plan.

What ESCs do you have?

70

The ESC are inexpensive Turnigy Plush.

71

Hey Shawn, ok here is the .bin. The flt. was better this morn., still seeing / hearing jerky / erratic behavior. The Quad is certainly more flyable than yesterday i.e. I can manage flight / landing. Again, I think the jerky / erratic behavior is limited to Roll, might be some pitch/yaw component but it does not look like it. Let me know what you think.

I did not set the Harmonic Notch filter params as yet because I wanted to get your input regarding the ongoing jerky / erratic behavior, then I will proceed with the Notch params.

Also, left the INS_LOG_BAT_MASK=1 turned on for the flight yesterday and today.
https://drive.google.com/drive/u/0/folders/1oJu4VgNgmyL_zD6jq1j20pBN_rVmGqK6

72

You are going backwards. By almost all metrics this craft is performing poorly. Plot these and check it out:
Pitch/Roll vs Des Pitch/Roll
RCOU-Motor outputs thrashing max to min.
Rate>Rout. Roll attitude controll has high noise.
Z-Vibes. For a simple in place Hover they are high.
PSC Velocity and Acceleration Position Control bad.

These F450’s, even the cheap plastic frames, generally will fly on default parameters. I would reset everything to default (load Rover then Copter) then set the Motor Thrust scaling Max/Min and enable the Notch Filter with the parameters as suggested. Leave the Gyro Filter at 20Hz and don’t change any of the PID’s.

Balance the props obviously. Are those collet prop adapters? Looks like it from the photo but not sure. If they are those suck on Multirotors.

73

You can definitely set these, they help to take the “noise” out of attitude data the flight controller is trying to work with:
INS_HNTCH_ENABLE,1
INS_HNTCH_MODE,1
INS_HNTCH_REF,0.22
INS_HNTCH_FREQ,92
INS_HNTCH_BW,46
INS_HNTCH_ATT,40
INS_LOG_BAT_OPT,2

I’d say, like Dave says, there’s some sort of physical issue that we’re not seeing. A quad of this size should just fly reasonably with default parameters. I’ve mentioned those prop adapters before, I’ve had them come off plane motors in flight before and never used them on a multirotor (but plenty of people have no doubt).

If you want to keep going with how it is now, set INS_GYRO_FILTER back to 20 and try these too:

ATC_RAT_RLL_D,0.007
ATC_RAT_PIT_D,0.007
ATC_RAT_PIT_FLTD,15.00
ATC_RAT_PIT_FLTT,15.00
ATC_RAT_RLL_FLTD,15.00
ATC_RAT_RLL_FLTT,15.00
ATC_RAT_YAW_FLTT,15.00

I think that will get it back to as smooth as it’s been so far.

1 Like
74

Shawn / Dave, I am not sure how to proceed now. I am wondering if the history of this Quad has become lost in the translation? Perhaps a quick review of the history would be appropriate.

The Prop adapters are tight! If I attempt to tighten them any more it will likely strip the threads. I cannot budge the props while holding the motor case.

The Quad was at a VERY VERY stable flight condition (8mar21 some of the table changes incorporated, but not all). The reason I did not include all the table param mods, is from experience, some of the changes introduce instability into the Quad. i.e. Stab and Loiter, the Quad exhibits a lot of twitching / erratic motion (I think this is primarily Roll axis). The instability is large enough to make it difficult to control the Quad on TO/Flight/Landing.

So, as I say, from the first time I attempted in implement the full table param changes the Quad went from being a very very stable Quad (with TX trim) to a VERY VERY unstable Quad. So, around the 4mar I went back to square one i.e. reloaded the params without change and resolved the issue with TX trims so that the Quad now flys VERY VERY stable without TX trims and without any table param modifications. Then I began introducing table param changes around the 8th of Mar (see below). The Quad is still VERY VERY stable after the initial table param changes. The instability begins when I introduce the remaining table param changes (see below).

Dave, are you aware that there was a point (partial inclusion of Shawn’s Tbl. params) where the Quad was Solid As A Rock, Z Vib low, All TX Trims centered! It was so stable in STAB mode that it looks like it is in Loiter mode! The param (Shawn’s tbl) all flight modes, the Quad is VERY VERY stable (see below) shown below.

8mar21
Table params. partial modified, QUAD nice to fly, Flys VERY VERY stable, Z Vib mean apprx. 21.67, TX trims now centered :
*Quad is VERY VERY stable, in all modes of flight:
ATC_ACCEL_Y_MAX = 27000
ATC_RAT_PIT_FLTE = 0
ATC_RAT_RLL_FLTE = 0
ATC_RAT_YAW_FLTD = 0
INS_ACCEL_FILTER = 20
MOT_THST_EXPO = .65
MOT_THST_HOVER = .2209667
BATT_ARM__VOLT = 11
BATT_LOW_VOLT = 10.5 (I realize this is lower than recommended)
MOT_BAT_VOLT_MAX = 12.6
MOT_BAT_VOLT_MIN = 9.9

When things begin to deteriorate i.e. instability introduced in all flight modes:
**** STAB and Loiter (twitching / erratic roll), when I revised the following parameters
ATC_ACCEL_P_MAX = 116700, was 110000 when stable 8mar21
ATC_ACCEL_R_MAX = 116700, was 110000 when stable 8mar21
ATC_RAT_PIT_FLTD = 21, was 20 when stable 8mar21
ATC_RAT_PIT_FLTT = 21, was 20 when stable 8mar21
ATC_RAT_RLL_FLTD = 21, was 20 when stable 8mar21
ATC_RAT_RLL_FLTT = 21, was 20 when stable 8mar21
ATC_RAT_YAW_FLTE = 2, was 2.5 when stable 8mar21
ATC_RAT_YAW_FLTT = 21, was 20 when stable 8mar21
INS_GYRO_FILTER =42, was 20 when stable 8mar21
The table below shows the bat / fence settings, green are changes following the Stable Flight.

image
image1137×187 7.68 KB

12mar21:
In an attempt to reduce / eliminate the severe twitching / erratic roll, per instructions I changed:
ATC_ANG_RLL_P = 6.5, was 4.5
ATC_ANG_PIT_P = 6.5, was 4.5
ATC_ACCEL_P_MAX = 80000
ATC_ACCEL_R_MAX = 80000
This reduced the twitching / erratic behavior, but did not get me back to the stable flight.

Not sure what course of action to pursue? Please advise.

75

Shawn / Dave, I am not sure how to proceed now. I am wondering if the history of this Quad has become lost in the translation? Perhaps a quick review of the history would be appropriate.

The Prop adapters are tight! If I attempt to tighten them any more it will likely strip the threads. I cannot budge the props while holding the motor case.

The Quad was at a VERY VERY stable flight condition (8mar21 some of the table changes incorporated, but not all). The reason I did not include all the table param mods, is from experience, some of the changes introduce instability into the Quad. i.e. Stab and Loiter, the Quad exhibits a lot of twitching / erratic motion (I think this is primarily Roll axis). The instability is large enough to make it difficult to control the Quad on TO/Flight/Landing.

So, as I say, from the first time I attempted in implement the full table param changes the Quad went from being a very very stable Quad (with TX trim) to a VERY VERY unstable Quad. So, around the 4mar I went back to square one i.e. reloaded the params without change and resolved the issue with TX trims so that the Quad now flys VERY VERY stable without TX trims and without any table param modifications. Then I began introducing table param changes around the 8th of Mar (see below). The Quad is still VERY VERY stable after the initial table param changes. The instability begins when I introduce the remaining table param changes (see below).

Dave, are you aware that there was a point (partial inclusion of Shawn’s Tbl. params) where the Quad was Solid As A Rock, Z Vib low, All TX Trims centered! It was so stable in STAB mode that it looks like it is in Loiter mode! The param (Shawn’s tbl) all flight modes, the Quad is VERY VERY stable (see below) shown below.

8mar21
Table params. partial modified, QUAD nice to fly, Flys VERY VERY stable, Z Vib mean apprx. 21.67, TX trims now centered :

  • Quad is VERY VERY stable, in all modes of flight:
    ATC_ACCEL_Y_MAX = 27000
    ATC_RAT_PIT_FLTE = 0
    ATC_RAT_RLL_FLTE = 0
    ATC_RAT_YAW_FLTD = 0
    INS_ACCEL_FILTER = 20
    MOT_THST_EXPO = .65
    MOT_THST_HOVER = .2209667
    BATT_ARM__VOLT = 11
    BATT_LOW_VOLT = 10.5 (I realize this is lower than recommended)
    MOT_BAT_VOLT_MAX = 12.6
    MOT_BAT_VOLT_MIN = 9.9

When things begin to deteriorate i.e. instability introduced in all flight modes:
**** STAB and Loiter (twitching / erratic roll), when I revised the following parameters
ATC_ACCEL_P_MAX = 116700, was 110000 when stable 8mar21
ATC_ACCEL_R_MAX = 116700, was 110000 when stable 8mar21
ATC_RAT_PIT_FLTD = 21, was 20 when stable 8mar21
ATC_RAT_PIT_FLTT = 21, was 20 when stable 8mar21
ATC_RAT_RLL_FLTD = 21, was 20 when stable 8mar21
ATC_RAT_RLL_FLTT = 21, was 20 when stable 8mar21
ATC_RAT_YAW_FLTE = 2, was 2.5 when stable 8mar21
ATC_RAT_YAW_FLTT = 21, was 20 when stable 8mar21
INS_GYRO_FILTER =42, was 20 when stable 8mar21
The table below shows the bat / fence settings, green are changes following the Stable Flight.

image
image1137×187 7.68 KB

12mar21:
In an attempt to reduce / eliminate the severe twitching / erratic roll, per instructions I changed:
ATC_ANG_RLL_P = 6.5, was 4.5
ATC_ANG_PIT_P = 6.5, was 4.5
ATC_ACCEL_P_MAX = 80000
ATC_ACCEL_R_MAX = 80000
This reduced the twitching / erratic behavior, but did not get me back to the stable flight.

Not sure what course of action to pursue? Please advise.

76

Shawn, since the Quad was once at a very very stable situation i.e. 8mar time frame (partial inclusion of the the table params you provided), does changing these additional params take me in the right direction?

Have you considered reverting back to the 8mar params list (this is when I had included some of the table param changes but not all and the Quad was VERY VERY stable)? I am not sure why Dave is suggesting that I am proceeding in the wrong direction. Can you please elaborate so I can decide how to proceed.

77

Yes, it should go back to as stable as it ever was, plus maybe a slight improvement.

INS_GYRO_FILTER,20 ← the main one
ATC_RAT_RLL_D,0.007
ATC_RAT_PIT_D,0.007
ATC_RAT_PIT_FLTD,15.00
ATC_RAT_PIT_FLTT,15.00
ATC_RAT_RLL_FLTD,15.00
ATC_RAT_RLL_FLTT,15.00
ATC_RAT_YAW_FLTT,15.00
INS_HNTCH_ENABLE,1
INS_HNTCH_MODE,1
INS_HNTCH_REF,0.22
INS_HNTCH_FREQ,92
INS_HNTCH_BW,46
INS_HNTCH_ATT,40
INS_LOG_BAT_OPT,2

But set all those others too, leave everything else I didnt specify here and see what we get.
If it is particularly wobbly, reduce these from 6.5 back to 4.5
ATC_ANG_RLL_P,4.5
ATC_ANG_PIT_P,4.5

When you say solid as a rock, there’s clearly still stability issues in the logs. I’m sure you’ll find if it does come good with more tuning that you’ll be able to visibly and audibly tell the difference, apart from huge improvements in the logs.

1 Like
78

My point was to re-establish a baseline to methodically work from. Clearly you are not there now and the fact that you loaded parameters file on parameter file is cause for concern. This is bad practice as Shawn pointed out. I haven’t looked at all your logs but I have yet to see one where the Tune is acceptable. It might feel “rock solid” to you until it doesn’t after it reaches a point of increasing instability.

I wasn’t suggesting the collet adapters would come loose although I have had that happen on a Plane (would never use them on a multirotor). They can cause imbalance from non-concentricity even with a balanced prop. Those adapters typically come with the cheapest low quality kits so you are starting off 2 points down. Of course the fix is to replace the motors and props which granted is an expense.

Would you happen to have a heavier/higher capacity battery? Getting the MOT_THST_HOVER out of the mud may make tuning a bit easier. I have tuned many craft with high thrust/weight but it’s a bit more challenging. The severe motor output oscillation seen on that last log (and others) has to be dealt with before any progress can be made.

I dug this Junker out of the closet and it occurs to me I can just mount a Pixhawk (have a few) and get this thing flying. Might prove useful for comparative analysis.
F450

79

Comment removed, regarding INS_HNTCH_…

80

The F450 Junker flies fairly well on Default PID’s. Loaded Rover, then Copter to be sure, set the Motor Thrust Scaling and made a flight to determine Notch Filter settings. Set those and it’s flying. No PID changes from default. Does it need Tuning? Of course it does but it’s fundamentally flyable and a good baseline to work from.

The After photo. Really crappy self-tightening props but it’s all I had to fit these motors. Actually all the parts are poor quality accept the Rx.

F450 Junker

Log if there is interest.