Copter-3.4-rc3 is available for beta testing. It can be downloaded through Mission Planner and APM Planner 2.0 beta firmware link.
Release notes are here and copied below:
We are getting close to the final release but we still have three issues to resolve which will be included in –rc4.
Landing detector improvements:
a) checks that descent rate is less than 1m/s (avoids disarms when vehicle is obviously still descending)
b) tradheli fix for landing detector
WP_NAVALT_MIN parameter limits maximum lean angle when vehicle is below the specified alt while taking off or landing. Only applies in LAND mode, final stage of RTL and during take-off in AUTO mode
GND_EFFECT_COMP parameter allows turning on EKF compensation for ground effect during take-off and landing
AutoTune minor improvement to reduce step size on vehicles with very low hover throttle (i.e. high power-to-weight ratio)
Bug fixes, minor improvement:
a) OneShot ESC calibration fix
b) LAND_SPEED_HIGH allows specifying vertical descent speed separate from WPNAV_SPEED_DN during
the initial descent in LAND mode (and final stage of RTL)
c) improved USB driver reliability for windows users (NuttX change)
d) report dataflash logging health to ground station
e) Guided mode accepts yaw rate from SET_ATTITUDE_TARGET message
Thanks so much to our beta testing team.
Warning #1: please check the MOT_THST_HOVER parameter and ensure it’s between 0.2 and 0.8. If it’s
higher than that, please reduce it back to a reasonable value (check CTUN.ThO) and set MOT_HOVER_LEARN to 0 to disable the learning of this parameter. The issue is that it’s possible (although unlikely) that during user ground testing, the hover-throttle can be learned to be too high and that when entering AltHold (or any mode in which the autopilot controls the throttle) it may climb and climb until the pilot re-takes control in Stabilize or Acro. This will be fixed in -rc4. The problem can also be resolved by increasing the ACCEL_Z_IMAX to 1000.
Thank’s Randy. After hover learn finish and the param was auto set, It’s convenient to disable auto learning?
No issues up to now with this param in my little 270 racer and rc2; now I update to rc3.
Oneshot esc calibration can we do it with old wiki method? (rise thr --bep bep, down trh etc)?
Hi Cala.
Yes, it’s fine to disable the hover learning although theoretically it should be ok to leave learning on. I’m going to update the wiki page on throttle-mid to tell people the parameter to use (although for people here it’s written above in the warning anyway).
For OneShot calibration, yes the old method works. I’ve also updated the ESC calibration wiki page so it tells people to use the MOT_PWM_TYPE to “1” or “2” if they’ve got OneShot or OneShot125 ESCs. I also added a description on the “Semi Automatic ESC calibration”… which is nothing too special, it just raises and lowers the throttle for you… which is not really a very difficult thing to do anyway but… this method was even present in AC3.3 I think but it was never documented :-).
Thanks for testing!
Thank’s Randy, that semiautomatic method I try it and the old one but didn’t work with oneshot 125, I’m going to try again now and share results, thank’s for solve the issue.
@rmackay9 I noticed this change in the landing algorithm:
Landing detector improvements:
a) checks that descent rate is less than 1m/s (avoids disarms when vehicle is obviously still descending)
What if LAND_SPEED is less than 1 m/s? We specify .5 m/s right now for LAND_SPEED. What if the copter stops descending temporarily due to ground effect?
Seems like 1 m/s may be too high a value; that’s still over 2 MPH.
(I’ll take a look at the code as well to see if there are other mitigating checks to prevent premature disarming.)
Ideally, all copters will eventually have a low altitude ground detection sensor to prevent having to guess at when you’re on the ground.
Looks like there is an acceleration check that might handle the ground effect case. The throttle check may prevent the descent rate check from causing a landing complete, but if not, it may be better to lower the descent rate to like 5-10 cm/s just to be sure. Why was 1 m/s picked?
Yes, there’s an acceleration check that’s already there. No checks have been removed, we’ve added one more check so that if the vehicle’s actual descent rate is more than 1m/s it shouldn’t ever think it’s landed. When it’s less than 1m/s it may or may not think it’s landed based on the existing checks.
During landing, the ground effect may cause some noise to appear on the descent rate temporarily so it’s possible this additional check could slow the detection of the landing for some people, I think we will just have to see. I think 1m/s was just a guess as to what a good value should be. If it’s too high then it becomes a meaningless test, too low and it will lead to people’s landing detector not detecting the landing for a few seconds. We can adjust it based on testing.
Thanks for testing.
In case anybody noticed, I’ve removed a warning that I had added yesterday about flying without a GPS lock. I haven’t been able to recreate the issue in the simulator and I’ve just discovered the report that got me worried was from -rc1 and there have been enormous numbers of fixes since -rc1.