Flip over on Takeoff - Large Copter 30 Inch

[Jeffuasa (Jeff)]

Hi all,

We are confused at a flip over on takeoff. We didnt change many settings between the last very successful flight.

PIDs werent changed. When we lifted off the copter flipped over.

We are using a herelink to control it, however i couldnt see anything in the log except for desired yaw and actual yaw divergng on lift off. Yet no control input.

I understand that loiter isnt the prefered takeoff however we had great GPS positioning an are developing this for ease of use of inexperienced operators during training.

Can anyone plese take a look?

J2021-02-07 11-56-59_yaw_crash.bin (828.6 KB)

[dkemxr (Dave)]

Was the Boost motor scale parameter one you were changing? It looks like it’s commanded to max and so was Motor 4 so Motor 3 was dropped to attempt stability. Got a log of this craft just hovering?

[Jeffuasa (Jeff)]

Thanks for getting back to me so quickly!! Had to link to a cloud file as the successful flight was obviously longer and larger. https://drive.google.com/file/d/1AmUyzFW4UlecPpU6M_QKo6Qc84XxstrE/view?usp=sharing

Yes we are acutally using Mot_Boost scaling to scale our generators servo for its throttle. Using this we are able to generate power with overall throttle percentage, it isnt being used for a lift motor. Here is the log from the previous successful flight with hovering in the initial and final stages.

The motors are usually imballanced in logs due to the COG with fuel load that burns off over its flight envelope. For reference the fuel load is baffled so it cant slosh.

It has lifted off muliple times in loiter successfully, its very over powered but controls the power nicely even in wind. Any comments about the good flight or flip overs logs are appreciated!!

[Jeffuasa (Jeff)]

We rebuilt the quad last night and decided on an electric only flight without the generator. Calibrated the ESC’s well. We performed a 90 second uneventful flight with stabilise takeoff into alt hold mode then landing in loiter.

We then logged the filters to try filter out the harmonics which we did using the following.
INS_HNTCH_ENABLE - 1
INS_HNTCH_REF - .17212
INS_HNTCH_FREQ - 81
INS_HNTCH_BW - 40.5

Our next flight was exactly the same but as soon as advancing throttle in stabilise we had another flip over!

Like our first flipover at the start of this topic you can see ATT Yaw goes in the direction then DES Yaw follows it without instead of correcting it?

To note is the direction of the flip was right and backwards yet our motor PWM suggest C4 went higher than everything else?

See logs below for frst successful flight after flip over then second flight with flip over. One is a link the other is a file.

I dont think the harmonic filter is related to the flip over as we didnt have it activated for our first unknown flip over.

Could this be a bug?

https://drive.google.com/file/d/1h85mbugA16reqS5XrTSUoXRSQqyJDEtH/view?usp=sharing

2nd flip over after filt change.zip (265.1 KB)

[dkemxr (Dave)]

@Jeffuasa

To note is the direction of the flip was right and backwards yet our motor PWM suggest C4 went higher than everything else?

Remember this is commanded output. If it’s supplying insufficient thrust it will be driven high.
I think your Rate Filters are too low. Try changing them from 8.5Hz to 11.

[Jeffuasa (Jeff)]

Thanks for that, i used the excell sreadhseet recommended figures found at this wiki page.
Initial parameters.xls (dropbox.com)

We have had many flights using the setup and it has flown nicely, obviously needed some tuning but got us in the air ok every time. These flips have started appearing sporadically.

After the crash we recognized our GPS on UAVCAN wasn’t working, and upon investigation we found that the parameter file after the tip over had almost reset all the way bac to defaults!

We have retrieved our parameters after the good flight this morning and have compared each parameter change. The first pic is from first good flight parameters to in-between flights when we were checking logs. Note the compass DEV change? Uncommanded by us.

The second is the good flights parameters to the flip over flight. Note this was post flip over so perhaps not changed before but during? I’m don’t know how this can occur on its own??

All help is appreciated! Can you see the logs where this flip over may have been commanded from? It makes sense that motor 4 was commanding so high when flipping if it wasn’t generating the thrust. As it flipped right and onto its back which would suggest low thrust out of 4. All motors ESCs calibrated beautifully for the first flight, the only change was the filter settings INS_HNTCH i mentioned above.

[amilcarlucas]

Temperature can cause bias values to tip over. If you are familiar with control tehory, your phase margin was too small.

[Jeffuasa (Jeff)]

Hi!

Would you mind expanding a bit on your explanation? Can you point me to the log showing that and perhaps some suggestions to adjust?

[amilcarlucas]

You should test ArduCopter 4.0.7-rc1 to see if it fixes the parameter reset. And you should read and follow https://ardupilot.org/copter/docs/tuning-process-instructions.html

[Jeffuasa (Jeff)]

Hi thanks,

I have followed the tuning instructions and have tried auto tune unsuccessfully with a recovery from a desync.

We were able to get the values close and continue to improve them in each flight looking at the logs to inspect the changes for our ATT Des and ATT Actuals.

My second log posted of the longer flights show the tuning to be close. It has responded well.

The problem we have is the two flips on takeoff that were uncommanded. I have posted those logs also.

These were random in between successful takeoffs. That’s the problem we are chasing down. Did you see the log with the flip over?

Jeff

[Javier_Beringola]

Hi,

This can happen to most beginners for three reasons. Sorry and leave out this reply if is not the case.

1. Wrong order of propellers
2. Wrong motor connections to controller.
3. Recalibration of ESC can provoque the wrong turn of some motors.

[xfacta (Shawn)]

Hi Jeff,
Can you detail what motors, props, ESCs you have and take-off weight please.
I assume 30 inch props from the title.

EDIT: so far I dont see anything in your logs that says “flip over randomly on takeoff”. Can you post a link to a .bin log file with good take off and flight please.

[Jeffuasa (Jeff)]

Hi Shawn,

First of all I would like to say a big thank you for your spreadsheet. It was perfect for our initial flights. To have something like that to guide builders has no doubt saved thousands of $$.

Our combo is this,
Quadrotor
UA90 KV120 (Eaglepower) motors
40A ESC’s (Eaglepower)
30 x 9.5 carbon props
TOW 12- 18KG
Current flights TOW - 12-13kg
On board generator gives 2000W continuous

Background info is we bought this setup for development to see what did and didnt work. We have been flying it at the lighter end of the spectrum (1hr fuel max) to save the generator that has a tendancy to overheat with increased electrical load.

When you see MOT BOOOST, its because we are using that as a nice little workaround for generator control as apposed to the voltage sensing arduino which caused large variences in throttle.

You can see that our model is horrendously overpowered, using a hover throttle of .172 most of the time. This is using the Mot Thrst Expo of 0.8 as suggested in the spreadsheet. I stumbled upon Leonard hall’s explanation of that and it could be related?

In terms of the flipover which is the last file in post 4, DKEMXR commented that seeing motor 4 command a high PWM didnt nescesarily mean that the motor was powering up, more so that it wanted to power up and was likely not able to produce the thrust.

On this angle we replaced motor and esc on number 4 and had much more success. We also ground tested the generator and motors with props off with large throttle inputs. This resulted in our logs seeing desired and att yaw inputs whilst staying on the ground. Sometimes in the area of 15 degrees varience. As a result we raised our HERE 3 GPS which is used as our external compass and this reduced the spikes to about 7 degrees. So still work to be done there.

Note we cant do the motor compass calculation as we cant measure our current in such a high powered and dual dirrection generator type system. We assumed with this dual direction aperage from and to the batteries it would give unwanted results.

Now we have flown this adjusted motor and compass setup we have gotten fairly good auto flights out of it in light wind. However any sort of wind and gust is causing us to have variences and sloppy control. We raised our PI’s and was met with resonance or wobble so backed it off again.

Currently we are sitting on
ATC ACCEL P MAX - 18900 (Per Spreadsheet)
ATC ACCEL R MAX - 18900 (per spreadsheet)
ATC ACCEL Y MAX - 9000 (Per spreadsheet)

Stab Roll and Pitch
P 0.13
I 0.13
D 0.038

Stab Yaw
P 0.18
I 0.15
D 0.0

These are two logs, the first log we fly an auto mission and we are attempting to make tight 20m line spacing. But on every corner halfway through the turn it starts banking more and ends up sliding backwards before the nose comes around. It is hardly coordinated.

The last flight and log we did you can see that in higher winds it cnt maintain nose direction.

On all of these flights we have our MOT SPIN MIN set to 0.15. And on the last flight you can see it slows down all the way to that base PWM trying to maintain heading. We could see the propellor just windmilling slowly at that min speed. And it obviously scared us as its not able to maintain that control.

My hunch is we are flying it too light and need to increase the weight to give more buffer between hover and low throttle to allow for torque to spin the craft. However our generator wont keep up. It would be great to have you or anyone first of all look at our tune and secondly chime in on options to help fix the torque steer.

For note, i did try autotune once and it flipped on its back at 50m. Irecovered which was lucky! The autotune never succeeded but was trying to put our P levels down at 0.078. I tried one flight with our levels down at 0.1 and whilst stabilse could control ok, in loiter any wind resulted in oscilations and overcorrecting lazily til we had to disconnect.

Jeff

First Flight lighter winds
https://airgeox-my.sharepoint.com/:u:/p/david_alam/EXJcjHekXdhHqqBRA95z4CQBfKVU6IfdyXX5JYCHwjtCjA?e=UdDjxl

Second flight Longer route
https://airgeox-my.sharepoint.com/:u:/p/david_alam/EW6nl-DcomRMuG_Hbo2kS-oBecQDkozlmDMYtHdC9t2N0Q?e=GvD6vG

Last flight couldnt hold yaw, updated roll and pitch
https://airgeox-my.sharepoint.com/:u:/p/david_alam/EYaLIxa-UCBDiG-IXRHzvIYBS1UtzgnED8wpju3PVla63Q?e=iJ0vHj

UA90 KV120 Specs with Differnt props, ours is 3095

[ThePara (Cornel Fudulu)]

Why don’t you under-prop it !? Go for 27" props.

You’ll gain alot. You’ll be hovering at a higher throttle so the descending zone will have a higher resolution. Motors will be upper in the revs, so more torque is available for adjustments and less torque is needed to accelerate or decelerate the prop. You’ll be less at risk of desynch. And your entire quad can be a bit smaller and less affected by wind.

[Jeffuasa (Jeff)]

Hi cornel,

That is something we have and are considering. 27 inch props will up our RPM by 500rpm which will definitely help.

I believe it costs us efficiency and power which will again make our generator work harder which we are at risk of an overheat already. When we can fix the generators issues we should be able to increase weight through fuel load to achieve something similar.

I was hoping perhaps some tuning could help it. Perhaps our throt expo at 0.8 isn’t helping our down low resolution. However if we can see a prop almost stopping at mot spin min 0.15, then I guess there isn’t any more ability to turn then that?

But yes, 27 inch props have been ordered to trial. Thanks

[ThePara (Cornel Fudulu)]

Jeff,

Higher rpm with a smaller prop doesn’t mean more power is being consumed. Unless you aim for a weird motor/prop pair that’s out of a decent envelope. And it’s not the case.
If you’re hovering at .172, power is not something that you lack, quite the contrary. Efficieny ? You can’t define efficiency when your craft is barely hanging in the sky. Once it’s tuned, it’s sharp-moving, agile and wind-proof and you have a baseline, then you can see if an increase in prop diameter will add efficiency without compromising flight attitude.
And, from my personal experience, those 40A ESCs are a trouble spot. I sport 50A units on a bird that’s sipping 4A per motor at hover, but it takes me 15-20 seconds to get them thermal throttling after some high-speed flying.

[xfacta (Shawn)]

Yeah smaller props are going to help at least until all aspects are sorted out.

I’d be tempted to re-assess the harmonic notch filter settings. This seems awfully high for such big props:
INS_HNTCH_FREQ,81
INS_HNTCH_BW,40.5
INS_HNTCH_ATT,15
Judging by the logs I’d try:
INS_HNTCH_FREQ,32
INS_HNTCH_BW,16
INS_HNTCH_ATT,40
and see what you get in FFT’s. There could be a few different frequencies to try, but we’d need to see the pre-filter logging. You’d need to rerun the pre- and post- filter logging when changing prop sizes.

I’d also consider increasing the PIDs a bit:
ATC_ANG_RLL_P,6.0
ATC_ANG_PIT_P,6.0
ATC_RAT_RLL_P,0.12
ATC_RAT_RLL_I,0.12
ATC_RAT_RLL_D,0.0065
ATC_RAT_PIT_P,0.12
ATC_RAT_PIT_I,0.12
ATC_RAT_PIT_D,0.0065
But be careful at takeoff and look for oscillations. Reduce all by 10% if there is instability - more than normal

Vibrations are low in X axis, but a bit higher in Y axis, indicating vibration damping is not equal in all directions, or something is vibrating against the flight controller or frame in the Y axis - ICE and generator springs to mind.
Z axis is a little high, it’d be nice to have that a bit lower.

If you can get those vibrations down around the under-10 like X axis that would be ideal and then I’d consider increasing these filter values too:
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_GYRO_FILTER,30.00
This will allow the FC to catch and correct all the small movements in the frame that it’s currently unable to do anything about except log them:

Is MOT_SPIN_MIN,0.159 definitely the lowest you can go without motors losing sync ? I guess so from previous posts. I wonder if a good BLHeli ESC might do better…
Motors are definitely going to minimum during flight, but I guess you knew that:

Motor 4 (CW) is definitely working harder than the rest, sometimes motor 3 (also CW) too. This indicates some weight imbalance but also motor mount twist or frame twist. That’s affecting your yaw authority and it wont be helping stability either.

[Jeffuasa (Jeff)]

Hi Shawn and Cornel,

Thanks for taking the time to look at those. We are down for weather here at the moment but as soon as we have a better day i will implement those changes.

I will also investigate the flight controller and any interfering wires as this could be a possibility. Here is the Electric Only Vs Generator on vibes for a short alt hold flight where we were testing filters, definately more vibes with the generator.

We have never raised our D values that high before. We raised our P’s, and that resulted in a few oscilations around 0.145. However we didnt do the D’s.

The motor imbalance unfortunately is something we struggle with due to the nature of carrying fuel and burn off over the flight envelope. We have countered the slosh by adding HDPE piping cut into cylinders which makes the whole fuel tank a honeycombe which has signifigantly reduced the slosh.

As for COG, over a flight we intend to start Nose Heavy as thats where the tank is, and burn off to tail heavy as thats where the generator is. Most of these flights are ‘close-ish’ to COG but with fuel, nothing is perfect!

You can see the bottoming out of those PID’s is in our turns. The nature of what we are seeing is the nose turn nicely initially but then it hangs in the same position as the flight controller commands roll to make the tight 20m line spacing turns. This results in our nose still pointing sideways when we have crossed our waypoint then it is able to swing around again. During these turns as you can see we are at MOT SPIN MIN.

In regards to whether MOT SPIN MIN could be reduced, at that speed they arent making any thrust and are effectively windmilling. I could reduce it and see what happens, but the low rate they are spinning a 120g prop, im not sure a smaller reduction would help? What do you think?

In terms of the ESC’s, i have yet to check if they offer active braking. Im going to try and test that this week, as its likely that that could be signifigantly affecting its ability to correct.

These images are as follows.
Image 1 is PRE FILTER and Electric motors only, no Generator. Peaks at 81HZ.

Image 2 is POST DYNAMIC FILTER which i listed above @81HZ. Again no Generator.

Image 3 is PRE FILTER Again but running the generator. It looks to be in the same frequency at 78hz.

Image 4 is POST FILTER, with the generator. We left the dynamic filter set at 81hz as it was close.

We still have static notch to play with if you think it could be worthwhile.

[andyp1per]

My guess is that with this setup you will be better off with the FFT driven notch that is in 4.1 since that way you could target the actual vibrations