AC 2.0.38 - the good the bad and the ugly

Took my Arducopter out for spin today.

  • Stock standard Arducopterwith my I2C Sonar mods
  • 2.0.38 Software
  • Calm weather
  • Standard PIDs and setup

Good:

Stabilise and Simple modes work very well - no issues

Alt Hold works very well most of the time - more on this later

Loiter works very well - most of the time

This code is good and getting better every time I fly - Big thanks to Jason and others for all the hard work

 

Bad:

RTL still seems way to aggressive - thunders back towards launch location and climbs hard. Lesson here is to move away from launch point before selecting RTL and set a lower RTL altitude before flight.

Alt_Hold is great when in Sonar range but when in pure Baro mode there is an interaction between throttle and Baro which sets up a pulsing behaviour in the throttles. The pulsing is no where near as bad as the pulsing we had in sonar Alt_Hold before the sonar fixes.

3690882704?profile=originalIn this image the red trace is throttle, green is sonar and blue is Baro. This is all in Alt_Hold mode. Note how on the left when flying in sonar range the throttle is nice and steady. When the Arducopter climbs to 8 metres or so note how the throttle is pulsing. This behaviour may be solved by filtering the Baro data to smooth its response or having different PID values at different heights.

Logs do not line up properly with 2.0.38 and planner 1.0.49, wrong values in wrong columns.

When in Alt_Hold or Loiter the response from using the throttle stick on the radio to climb or descend is slow and I usually over shoot. Seems like too much 'dead band' or something similar. I move the stick nothing happens, move some more, still nothing, then a smidge more and whoosh your climbing or plummeting. (climbing is the more favourable!) I 'bounced' gently twice today. I also noted that loiter mode will not descend below a minimum level for me of about 30cm or so - if I take it down gently. This may be an artefact of my i2C sonar or something else.

My sonar data has a delay in registering fast ascents or descents. I need to check if that is my fault (in my i2C sonar code) or something else. You can see the green trace in the graphs below 'ramps' up or down during rapid vertical movement, sometimes lagging the baro trace.

Loiter works reasonably well but could do with some PID tuning. I note that in the unreleased 2.0.39 code the Loiter PID values have been adjusted.  When I engage Loiter the arducopter starts swaying back and forth in one direction, then this becomes a small circle or oval - less than 3 meters across today. I don't think it is a GPS error or GPS accuracy issue due to the way the movement is sinusoidal. It looks like a PID tuning issue

3690882663?profile=originalIn this graph the red trace is long_err, the green trace is lat_err

 

Ugly:

I finished flight two today with a flip and crash at low altitude while flying in loiter mode

No idea why - could be a hardware issue - I still have bullet connectors on my motors. Just a broken prop today - nothing to cry about. Was only 5 mins into a full battery - voltage was good (11.55V) when checked afterwards. It could be software related - in the log the 'long_error' value spikes and then plummets just before the flip. Nothing conclusive.

Here it is at normal speed and slow motion

 

Logs & Graphs:

6Aug2011Flight1.log

3690882711?profile=original

6Aug2011Flight2.log

3690882575?profile=original

 

07Aug2011.flight2.log

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • It got even more 'ugly' today

    After being chased of my nice soft test location by an over officious park ranger I had to fly somewhere less forgiving.

    The flight began with a good stable flight in Stabilise and then Alt_Hold.

    I landed, then took off again for another flight, within seconds I had a flip and crash.

    Ugly this time - two props, two arms, two motors? Time for major surgery and a possible trip to the jdrones webstore. I'll pull it completely apart and rebuild it with all the lessons learnt since it was first constructed - soldered ESCs, ESCs out on the arms - Jani style, gel mount for APM board, long landing gear legs, LED lighting etc.

    Again there was no warning or notice of anything going wrong, just the slightest rocking motion and bamm over she goes. Logs and video are attached.

    If anyone can have a look at the logs for me and tell me what happens around line 33250 i'd really appreciate it. I cant see any cause of the crash in the logs but my understanding is limited.

    Crash Log File

     

  • I have also experienced a flip two days ago. I am on 2.0.38 and fly in X - config. First attempt with the new version were quite successful. During a second attempt the copter fell out of the sky rolling over its left side. Altitude was about 2m, no sonar attached, flight mode was stabilize. Unfortunately, there was a little pond nearby and it dove right into it. Even the copter was in the water for about 90s, two motors were still running once I pulled it out. After trying it off, everything seems to work fine and I am able to fly it. I could not get the log file however. I don't think I had a hardware issue since all my cables are soldered. I am a little worried to fly it again, since I cannot determine the cause of the flip.

  • Here is some video of the Loiter oscillations (at 8x speed)

     

     

     

  • 3D Robotics

    Very helpful report--many thanks. 2.0.39 will probably be released this weekend with a bunch of changes, which should hopefully address some of these issues. We're getting there!

This reply was deleted.

Activity

Hiroki Tanaka liked Hiroki Tanaka's profile
Oct 13
Derrick Davies liked lisa TDrones's profile
Sep 23
More…