Hexa unstable at 55 seconds

Hi guys I’m testing a hex, with a wheelbase of of 1610mm, 30 inch props, pixhawk1. I am using Lightware SF11, and IRLock for loiter, u-blox GPS.

I have tested this drone with loiter mode, and the first time it was flyable, without any issues. That was using [3.5.3], default PIDs. Heres the video link:

and the params

I want to show the other result here and please see the flight logs. In the (unstable) log [3.5.4] there is a sudden diverge between the desired roll and actual roll. Auto analysis shows GOOD for most items.

Did anyone experience this? And also what is the best way to tune a large wheelbase hexa? I have read other posts saying ATC_RAT_PIT_FILT and ATC_RAT_RLL_FILT should be set to 10. I have not tried that yet. But in the good flight they were at 20 Hz and that has me puzzled. Please advise, thanks guys!

Definitely drop the ATC_RAT_PIT_FILT and ATC_RAT_RLL_FILT to 10.Don’t forget AUTOTUNE_AGGR = 0.05 as well.I was particularly impressed with Position Hold autotune on my Tarot X6.I did seperate axes for the first autotune then all together.

I’m scanning the log and I already see that roll and pitch went the other way from the input.It looked almost reversed controls.

Also,channels 1 and 2 on the RCIN show skewed values.Mid point is logged as 1512 and 1514 respectively and that’s not what they were showing during that hover.The Pitch got closer than the Roll to the actual mid points.There’s some noise showing too.Reminds me of when a connector came loose in my Taranis.Then it crashed.And also of the time I didn’t calibrate the radio.It flew,but in an unpleasantly exciting manner.

That’s all I can see so far.

Thanks Jagger, I am going to set the ATC_RAT_PIT_FILT and ATC_RAT_RLL_FILT to 10 and test again. I will also calibrate the radios again to make them center. Yes, the roll and pitch definitely became opposite from the input in the end before the motor cutoff. That’s something I still have no clue yet.

Can you share your PID gains for your tarot x6? I am testing a tarot x6 as well, but with an extended arm to a 1600 wheelbase.

I’ve been thinking of extending mine to take 20" props.They’d suit the motors better than 18".

I can share the PIDs but they won’t mean much.Totally different set ups will produce individual PIDs.These started high (around 0.66 P and have come down to 0.4ish after three autotunes.No manual tuning at all so far.

PARM, 115304115, ATC_RAT_PIT_P, 0.401874
PARM, 115304164, ATC_RAT_PIT_I, 0.401874
PARM, 115304218, ATC_RAT_PIT_D, 0.02477124

PARM, 115303496, ATC_RAT_RLL_P, 0.4038906
PARM, 115303542, ATC_RAT_RLL_I, 0.4038906
PARM, 115303585, ATC_RAT_RLL_D, 0.02235605

PARM, 115304572, ATC_RAT_YAW_P, 1.660267
PARM, 115304627, ATC_RAT_YAW_I, 0.1660267

Also,have a look at your servo numbers.The motors are servos 1-6 (33-38).I’ve seen this before and have no idea if these numbers are in effect when used as motor outputs but there’s a mix of different values in there.Maybe someone can chime in ?

PARM, 667621819, SERVO3_FUNCTION, 35
PARM, 667621912, SERVO4_MIN, 1104
PARM, 667622188, SERVO4_MAX, 1924
PARM, 667622282, SERVO4_TRIM, 1518
PARM, 667622450, SERVO4_REVERSED, 0
PARM, 667622876, SERVO4_FUNCTION, 36
PARM, 667622974, SERVO5_MIN, 964
PARM, 667623092, SERVO5_MAX, 2064
PARM, 667623187, SERVO5_TRIM, 964
PARM, 667623320, SERVO5_REVERSED, 0

I think it`s not very good idea. On the second video i clearly see low stiffness of arms and frame (arms goes up at takeoff). Later it can cause
big problems when you start flying more active, not just hovering in steady air.

With the amount of arm movement during flight it is a surprise that it even flies for a while. Make that arm stiff, no bending no moving. Otherwise it will never fly stable