That check box is not “locking” the PID’s from being changed it simply indicates if pitch and roll are equal. PID’s do not change during normal flight in any mode. Unless you call Auto Tune a mode.
Then you will have to manually tune it. I would use in flight tuning (Transmitter based tuning) as described here; page down to Manual Tuning of Pitch and Roll:
https://ardupilot.org/copter/docs/tuning-process-instructions.html
I did read it. I am working on the drone stability more than 3 month. I did read and test almost everything. But the reason is the same: after many elaborations I reset to the factory settings, Than change parameters following the propeller length and slowly change PIDs until copter is stable. Now I add payload and test the fly. After several take off drone starts be unstable… Nobody knows how it is possible and after check settings PIDs are different (withaut autotune). I am despearing.
Did you change these based on the change in weight?
For aircraft designed to carry large directly mounted payloads, the maximum values of ATC_ACCEL_P_MAX, ATC_ACCEL_R_MAX and ATC_ACCEL_Y_MAX should be reduced based on the minimum and maximum takeoff weight (TOW):
There is no functionality to do this. Think about how dangerous that would be if it was possible… Only Autotune, In-flight tuning (which has to be specifically configured) and some LUA applications running on the Transmitter will do this.
Make the 1st hover flight I suggested and post that log. Not much can be learned from the one you posted other than to extract the parameter file.
OK, I will do it on Monday. I will send the first flight log.
Hi, I changed PIDs and other settings according our conversation. TOW/MTOW is about 0.6. First start was on LOITER. Drone flight like drunkard. 
2020-12-07 13-31-36.bin (666.8 KB)
Could you give me some recommendations?
Thank you
Post a photo of this craft. You ay it’s an Octa-H frame and you have it configured as an OctaQuad.
download this video. It is a last “fly”
Don’t need a video, just a photo. Is this a OctoQuad (X8) frame or not?
Yes, it is V octo coax. 
Sorry this is quad. But octo is the same, only more motors
That looks like an Octoquad X.
FRAME_TYPE 4 (which you have right)
FRAME_CLASS 1 (which you have as 2/V)
And set the ATC_ACCEL values back to where they were. The modified values based on TOW factor is after doing an Autotune.
Yes you are true. But do you think that instability is caused by ATC_ACCEL?
No, it’s probably caused by the wrong frame class.
BTW-I click that link of your video and 2 different browsers give warnings so I’m not trying that service again.
Just click on it and download. Stahnout video=download video. Its public service for big files transfer.
After changing the frame class, recheck the motor order with MissionPlanner motor tests just to be very sure.
Then post you latest .bin log file and we’ll take a look before you do too much more.
I use ublock origin and privacy badger - that stops these sites linking to other bad sites.
I downloaded and watched the video OK
The frame is almost X, but the rear and front angle is 40 deg. I suppose that the V is adequate frame type. I can try pure X tomorrow a I will let you know. But as I wrote before, this drones was flying well several weeks ago and suddenly are unstable without any substantial changes. May be vibrations?
I did a fly with Quad on AltHolder. Still it is unstable and waving. Construction is the same as Vcoax octo. Could you give me some advice how to stabilize PIDs?
Thank you.
2020-12-08 10-58-01.log.param (18.6 KB)