Fresh Tune - Occasionally Descends in Loiter. Why?

I have a fairly new build that will be put into service in a couple of weeks.

The drone will occasionally keep descending after I force a semi-fast descent in Loiter. I can hold the throttle stick up to around 60-70% and it will continue to maintain altitude.

Can someone take a look at my logs and assist with getting this resolved?

Logs and Params: Mission Planner – Google Drive

Looks like the throttle input mismatched at where the altitude drop

And the achieved altitude is slightly different from desired altitude.

You may try to recalibrate the hovering throttle.
As you also mentioned

I can hold the throttle stick up to around 60-70% and it will continue to maintain altitude.

I guess the hovering throttle setting has been 60-70 in stead of 50, such that it requires extra thrust to hold

I did notice that Hover_Learn was still set to “Learn and Save” from initial tuning flights

Would this have caused the descending issue? To be on while flying around rather than hovering?

Is it the Cortex folder?

I think there’s some sort of motor mount or frame twist going on here, predominantly the CCW motors are working harder to counteract some physical yaw issue.

Since it’s relatively stable and not doing any fly-away, you could set these and it might help with the altitude issue too:

The Harmonic Notch filter looked a bit wrong, like for smaller props or you’ve chosen the wrong peak to base it on. Try these and retest:
Also I’d probably test with INS_ACCEL_FILTER,20 and maybe just lower it a little if you have to.

I would standardise to these:

And for anything bigger than a small race quad, I would set these:
FENCE_ALT_MAX,120 <- set to suit safety and local laws
FENCE_RADIUS,600 <- set to suit safety and local laws
it will make you wait until a home position can be set before arming, but that’s a small price to pay.
And see if GPS_GNSS_MODE,67 helps with time to 3D fix and update rate.

Attitude control could be a little better.
ATC_ACCEL_P_MAX and ATC_ACCEL_R_MAX could be reduced to 78000, but that’s not a big issue.

I suspect ATC_RAT_PIT_D and ATC_RAT_RLL_D could be a bit higher
ATC_RAT P and I terms lower in conjunction with ATC_ANG_PIT_P and ATC_ANG_RLL_P being higher.

But Autotune will sort all that out and should get really good results.
Autotune with the lightest take off weight, then when you add a payload change:
ATC_ACCEL_P_MAX x (min_TOW / max_TOW)
ATC_ACCEL_R_MAX x (min_TOW / max_TOW)
ATC_ACCEL_Y_MAX x (min_TOW / max_TOW)

Keep MOT_HOVER_LEARN,2 unless you’ve got a special reason to change it.

Download a copy of this spreadsheet and use it to set the initial params. I updated it recently.
MissionPlanner has the Alt A key combination to do the same thing (mostly).

1 Like

Hi Shawn.
Thank you for your work in creating this file.
Is your spreadsheet equally valid for 12S 30" diameter prop heavy lift quads and hexes?

1 Like

Yes, it suits nearly all multirotor sizes within reason.

1 Like

Shawn, thanks for all of your help! I’ll try these values this weekend!


Here is my initial before/after notch work:



Do you still see a need for me to change the notch frequency?


Have you got .bin logs for that before and after notch filter testing?