I am chasing bad althold performance for a long time now and jumped on every theory I could find. Still no luck. Althold seems to be a lot worse than even Phantoms. Twitching motors and poor althold performance are the main issues. It also dosn’t help a lot that there is no real explanation of how to tune the relevant parameters. A lot of effort went into roll/pitch tuning. Yaw and Althold are just neglected. Standard answer is vibrations. No matter what dampening, prop or motor balancing I do the result is always poor. I also heard about aerodynamic effects and for the sake of starting a discussion about how to improve althold I want to share a log that was done in stabilize with the craft sitting on the floor, armed and about 40% throttle just not taking off. V3.1.5 on APM2.6 with external compass in a TBS Discovery, 980kv motors and 9x5 Graupner props. Baro covered with black foam, APM in 3DR case. Pretty standard.
What is going on with the baro? It clearly changes with throttle. Hardware issue, software bug or aerodynamic effect?
Althold is highly dependent on GPS signal, so if you are performing these test inside it is likely that you get a bad result.
The APM determines altitude by a combination of barometer values and GPS altitude.
Try repeating the test outdoors with good gps HDOP values.
If you still have an unstable althold, PID settings might need to be tuned.
I have also noticed that a balanced machine will perform alot better in this mode, determine the mid point betwwen your props and move your gear so that the copter will balance in that midline.
As far as I’m aware Alt Hold doesn’t use GPS at all.
“Note: The flight controller uses a barometer which measures air pressure as the primary means for determining altitude (“Pressure Altitude”) and if the air pressure is changing in your flight area due to extreme weather, the copter will follow the air pressure change rather than actual altitude (unless you are within 20 feet of the ground and have SONAR installed and enabled). Below 26 feet, SONAR (if enabled) will automatically provide even more accurate altitude maintenance.”
If I’m reading your picture correctly your barometer is dropping 4mtr when you throttle up?
If that’s the case then to me it would imply the prop wash is effecting the barometer. Do you have a picture of your setup?
This sort of thing is usually caused by vibration issues. The easy test is to plot BarAlt against RelAlt during hover or gentle maneuvering. If they don’t track almost exactly, it’s an indicator of vibration. BarAlt uses only the baro while RelAlt mixes in the accelerometers. The true test would be to turn on logging of the IMU and see what the Z values are.