Pixhawk 2.1 - big hexacopter - wobbles

Tarot X6 (960 class) Multistar Elite 5010/274KV motors with 18" props.Hobbywing X-Rotor Pro 40A escs.40,000mAh of batteries,Taranis and X8R.HERE GPS.Mauch power stuff.Pix 2.1 using the stock pads supplied.Firmware 3.5.RC8

This used to fly nice and stable using a Pixhack 2.83 and up to 3.4.6 firmware but since swapping to the Pix 2.1 (first issue,Edison capable, with stipulated mods done) it wobbles all over and generally handles like a sack of ferrets.Altitude hold doesn’t and Position hold is downright scary.Stabilize is managable but twitchy due to the tight tolerances at centre position (which I haven’t relaxed yet).So autotune is out of the question until I get it less jumpy.I can see some clipping in the 3.3 vibes section but have no idea what those numbers mean.I’m guessing some form of vibe correction will be needed ?

Any ideas, directions or help gratefully received.The new 3.5.RC8 logs are a bit beyond me at the minute.Log of the first test flight on 3.5.RC8 included.It scared me.

A bit of additional ammo.

Just attempted an autotune which was pretty exciting.I didn’t manage to save the values as it was too uncontrollable to land and disarm in autotune.It seemed t want to take off to the next village quite badly when I went back to autotune to land it.I include the aborted log as it has all the steps of the tune saved,just not the values.

I had this wobble problem with a apm that loose the antivibration system

It was too scary

After i put new antivibration at the fc everything work ok

Thanks for that but the Pix 2.1 is a very different animal to an APM.Never had a problem getting an APM up and happily runiing nor,for that matter,a Pixhawk of various flavours.But the 2.1 has a lot more sensors and therefore a lot more requirements and I have no idea how to give it what it wants or how to figure out what it wants.Hence the question on here.It’s mounted with a couple of the pads that were supplied with the kit and has internal damping for the sensors.I’ve dealt with that before with a couple of Pixhacks but the 2.1 is another step up in complexity.It may be a step too far for me but I’ll give it a go to get it running well before I bin it and go back to the Pixhacks.

Sounds like something is amuck? I have a 2.1 side mounted to the frame of a large 800mm helicopter with thin foam tape and it flies mint with very low vibes. No isolation mounts with rubber balls etc.
Are you using some form of vibration dampening system? I know you said “stock pads”, but just wondering if there is anything else in the mix. I can take a look at your log when I get to a Windows machine, although I am certainly no expert but maybe adept enough to spot some high vibration issues.

Thanks Tim.
The stock pads are not very thick so not a lot of damping from them.Nothing else in the mix.That’s meant to be done inside the cube.And a Pixhack mounted very similarly had no vibe issues or clipping.I’m wondering if I got a duff early unit.It already needed a new cube cover and the channel 1 modification to stop it shorting out and who knows if there’s more gremlins in there.I know I’m not impressed with it yet and support appears to be DIY in the extreme.I’m leaning toward replacing it with a Pixhack as I know they work well but I’ll try for a little while longer to get it tweaked before binning it.

As a comparison,this is the last flight with a Pixhack 2.83.A tiny bit twitchy but it’s stock loadout PIDs and totally flyable.The exact same copter with the Pixhawk 2.1 is not nice at all.I didn’t video it with the 2.1 but it’s comfortably worse than this config even when I’d forgotten to calibrate the radio.It’s a bit long but shows a perfectly acceptable set up and is totally different to the Pixhawk 2.1.I must be missing a setting or two somewhere but figuring out which ones is tricky without expert help.I really am tempted to put a Pixhack back in it but keep getting told how good the Pixhawk 2.1 is.I haven’t yet seen any evidence that backs that theory up.

While I am thinking of it, you were using Copter 3.4.6 with your other controller and now are using 3.5rc8 correct?
There are some differences between them that could affect things, namely the Gyro and Accel filtering being a couple.

INS_GYRO_FILTER likely changed to a lower value in 3.5rc8
INS_ACCEL_FILTER likely changed to a lower value in 3.5rc8

Those were lowered in the defaults I believe in some of the release notes, but either way if you are using defaults or setting you used with an earlier release it could definitely lead to a twitchy larger frame like a big X6, X8 or in my case helicopter which is far less forgiving to such things… :confused:

In any case, they used to be set to 20hz on both parameters. You could either leave them at 5 or 10hz where they likely are or bring back up up to 20hz and tune there. I believe the theory in running them lower is to reduce the noise the sensors see and thereby getting a cleaner flying setup after tuning.

As you may know already, those parameters are the single biggest contributor to the snappiness and ossicillatory issues in flight. Lower values will reduce the tendency to be too snappy, but also in the mix are ATC_RAT_PIT_I and ATC_RAT_PIT_P as well and of course across all axis although YAW axis is far less sensative in my experience to gains that are too high
I still havent had a chance to look at the log, I will a bit later today but if there is alot of noise in the signal I could see some issues running the GYRO and ACCEL filters lower than 20hz.
I looked at your video, that is with your Pixhack and 3.4.6 correct?

I can say with certainty that the Pixhawk 2.1 flies well, that being said there are likely a couple things working against you with vibration and changes to 3.5 that you should be aware of noted above.
Ill look later as I said, and maybe someone else might chime in as well?

Thanks a lot Tim.

I think I was on 3.4.3 with that flight on the Pixhack.As you can see it flew ok.Even with the radio not calibrated it was more manageable than on 3.5.RC8 with the 2.1.

Just checked and those two values are at 20Hz which must be the stock loadout now.I vaguely remember seeing some reference to them in a past post but it didn’t lodge in the grey matter as to what the details were.

PARM, 69968038, INS_GYRO_FILTER, 20
PARM, 69968508, INS_ACCEL_FILTER, 20

Both ATC_RAT_PIT_P and I are stock loadout or 0.135

I did a full ground up install with the Pix 2.1 so everything bar the two autotune filters that were softened (as per the big copter post a while back) should be stock loadout too.

PARM, 70062325, AUTOTUNE_AGGR, 0.05
PARM, 70062341, AUTOTUNE_MIN_D, 0.001

I’d test out the throw mode but it weighs over 8Kg with the big batteries on board (simulating a camera weight underneath). :slight_smile:

Looks like Wednesday before I get another chance to autotune.This time I’ll video it but it will be a video nasty and parental guidance will be recommended. :movie_camera:

Okay, hmm. I am at the Dojo with my son at the moment, but will get to my workstation in a little bit and look at the vibe profile. What is your ATC_RAT_PIT_P and ATC_RAT_RLL_P term set at, as well as the same for I and D terms? RC_FEEL as well? ATC_RAT_ANG terms also?

All those are stock loadout and same for P and I.

PARM, 89349869, ATC_RAT_RLL_P, 0.135
PARM, 89349912, ATC_RAT_RLL_I, 0.09
PARM, 89349955, ATC_RAT_RLL_D, 0.0036
PARM, 89350001, ATC_RAT_RLL_IMAX, 0.5

Angle are stock too

PARM, 89349304, ATC_ANG_RLL_P, 4.5
PARM, 89349340, ATC_ANG_PIT_P, 4.5
PARM, 89349611, ATC_ANG_YAW_P, 4.5

RC feel is already down at 50% but it misbehaves in Position Hold and altitude Hold too so I’m thinking maybe a dodgy barometer fluffing things up there.I can’t really describe it as razor sharp.Using a HERE GPS.

I’ve got a new Pixhack V3 on the way just in case.I can always use it in the X8 if the 2.1 passes muster in the next month.I like the CUAV gear.Very nicely put together.

Honestly, the 2.1 is a stellar controller. The heated IMU to eliminate, well cut down on changes, in gyro bias issues in of itself is worth using it for. The internal dampening seems to help matters out as well and having redundancy with the sensors is also a stro g selling point. As I noted, its flying a helicopter quite nicely and that frame is usually quite finicky.

The log you posted shows INS_ACCEL_FILTER as 10 which is what Rc8 will change it to from previous default 20.There is no change in INS_GYRO_FILTER with Rc8. I had major throttle stability issues with this at 10 on my PixRacer quad and changing it back to 20 corrected this. The sensor suite in the 2.1 is closer to the PIxRacer than Pixhawk so I would give this a try. I thought this was a case isolated to my quad but a couple others have noticed this also and corrected it by changing this back to 20. In addition it was a brief topic of discussion on the weekly Dev call. We will see what they do with Rc9. My guess is you will still have to do some manual tuning anyway.

Yeah, i can attest to issues with an otherwise nicely tuned frame just arbitrarily dropping the ACCEL filter to 10hz. I had a horrible ossicilation in loiter on Z axis as well as issues on YAW with bounce back and hunting. Also de-stabilized RLL and PIT, made them squirrely, sharp and just plain edgy feeling.

Position hold, loiter etc issues at least in my experience are attributed to the tune not being tight. I did a fair amount of testing using CH6 tuning knob and whether it was kP, kI or kD, if I adjusted it from a good tune to some other higher or lower value it held position terribly.
Barometer, issues there’s likely too much prop wash causing fluctuating pressure changes if its a genuine issue. That being said, I have one mounted right under an 800mm helicopter rotor with no shielding and it still functions admirably. The barometer may have nothing to do with it though, and if the ACCEL is indeed at 10hz there could be some Z axis issues causing problems maintining altitude?
Manual tuning may be the best route in the end anyways?

This was exactly my experience. Initially I thought it must be something specific to my quad as it didn’t garner much attention in a post I made ~10 days ago. But then a few more trickled in. It still looks to be a fairly specific set of circumstances that cause this.

Looking back the reasoning behind the issues makes sense. When you reduce the Filter to 10hz, instabilities that may be apparent can be pushed down the spectrum and pile up at the knee of the LPF. I am doing a fair bit of testing using filtering with some others getting quite deep into things and am picking up on some of the eccentricities that pop up when adjusting the filters. Some positive, some negative. Running a lower filter may be benificial but also might require alot more tuning after the fact. Likely good to start tuning from scratch with the filter lower and manually bring up the PID’s using CH6 tuning.

Yep.Thanks guys.I just spotted that. I posted values from when it was on 3.4.6 and the autotune log (that I posted),which I did after upping it to 3.5.RC8,has it set at 10.I’ll swap that out and try tuning it again during the week.I hope it is something as simple as that.And that’s why I need a second and third pair of eyes.We’ll see if it affects big 'uns as well as lillte 'uns.I shall report back on Wednesday (hopefully).I wish they’d stop slipping a load of new values in there at each upgrade.I can’t learn as quick as I used to.

You mentioning Yaw Tim also rings a bell.It was very twisty in the air and not nice to control.Very over responsive is the best description.I think I’ll relax the dead zones a touch as well.

I can’t tell you how scary it is when a 1400mm (prop tip to prop tip),8 Kg copter gets wiggly in the air.Apart from it begins with F.

Oh I hear ya. Imagine a 20lb helicopter with a 6’ rotor disk hammering up and down 3’ like a jackhammer a few feet away when you flip into loiter after lowering the ACCEL filter. :confused: i can see where a big hex or octa would be a bit intimidating when it breaks loose as well.
Scary scenarios aside, I know just the ACCEL and GYRO filters can have a large impact across the board with flight behavior as it changes alot, so I also hope it is that simole for you!

Thanks a lot Tim,and Dave.The value is already set back to 20Hz. and I’m watching treetops. :slight_smile:

1 Like

Just as a note, I looked at the log and Vibes don’t look that bad, Z axis has the worst of it but not saturated by any means. You said it was an autotune flight so the oscillations make sense but at the end the desired and actual do not line up well at all. I would certainly try again with the higher filter and I suspect that you will get better results, but honestly you might get further manually tuning PID’s and reviewing logs bringing Pdes, Rdes and Ydes closer to the actual P, R & Y under your ATT tab in your GSC?