DEV call 17 Apr 2017 4pm PST = 17 Apr 2017 2300 UTC



  • quaternion logging,
  • added thrust vectoring support,
  • oneshot fix,
  • booster motors for multicopters,
  • multi-output for multicopters
  • Airspeed seems coupled with pitch on DCM




Attendee count (max): 27

09:01am - Tom and Maker Faire
Intel has invited ArduPilot to come to Maker Faire in the Bay Area
ArduPilot will be in the intel booth
Same part of Intel that make Joule and Edison
Demo in booth
Vision processing
Quadcopter and rover
May 19->May 21st
Jani is supplying remove-pilot-before-flight tags
Great audience to promote to so good opportunity for schwag
Intel is paying for a few developers to attend
No Intel Aero because it isn’t a DIY thing
DIY is what Maker Faire is about, after all!

09:10am - Randy and Copter
3.5rc5 coming this week
Dataflash-over-mavlink fix

Francisco is handling an issue with servo function output
Servo gimbals may stop working when RC is lost
One-short bugfixes will be mashed together and merged
If RC is lost various things don’t work. Sprayer, landing gear, parachutes, …
Probably came in with servo channel changes
Probably only copter and sub
Output servos were only being set at same rate as input RC frames

Randy is running a little behind with support

09:21am - Rob and mode-change crash
Patch was set against 3.3.2 but failed to come into master
Rob is happy to do the patch but unable to functionally test it

09:23am - RC failsafe bug
Not sure if this has been fixed in master
When we go into receiver failsafe, maybe 5% of the time not all the channels come back simultaneously
Some of the channels come back significantly after other channels
Can cause mode changes!
Motors can stop on heli
Infinite roll/pitch issues….
874 value comes through
frsky receiver outputting bad values on sbus?
Need to make sure it is an s.bus protocol
X-series apparently only do s.bus
Sbus output comes as a single 24-byte frame
Can’t really partially decode it
Spectrum split channels between packets so can be vulnerable
Logic analyzer would work
Could also get a binary capture of the uart and use command-line tool of tridge’s to decode it
Rob is putting checks into heli to make sure input values are within-range
Hard to reproduce
Perhaps debouncing is hiding it
X9e taranis set to no-pulses
Tridge asks for an issue to be created and assigned to him with as much detail as possible
Needs to go into range-test mode, tin foil at both ends, cup hands over receiver
Tridge suggests attenuators
But the antenna is soldered
Rc5 going out later this week

09:36am - Solo-on-Master with Matt
[9:36 AM] (Channel) Matt_Lawrence: Solo before & after RTH to RTL Throttle Output Graph:

Solo before & after RTH to RTL Motor Output Graph:

Solo before and after RTH to RTL photos:

Solo Full Rabbit Mode:

Another user’s Solo showing landing detection works superbly.This would have flipped on a stock solo. Also shows off the AWACS mod!
Green Cube is working really well
There are some nuances with getting it working happily with the companion computer
Companion-computer smart-shot was completely stuffing things up on master
Matt has changed things to just use normal RTL
Was causing issues outside of RTL somehow, too.
Matt pushed things hard and it worked flawlessly
Including in full-rabbit mode
Landing detection in master works really, really well, as the image shows
RTK on Solo means cm-accurate positioning
Still working on reliable way of users loading firmware
Also, if compass-variance creates a red-fault, what should the copter do?
Fall out of sky?
Dronekit-python scripts force a land mode
Vehicles tend to pitch and roll over and descend at constant speed while accelerating……
Maybe companion computer is commanding it poorly?!
Maybe there’s a NaN in the controllers?
If it still has its lean angles it should be able to cope
Battery failsafe allows vehicle to recover and fly home and land itself happily
Maybe the mode change resets a NaN in the controller?
Ceases responding to all control input
Does the CC lock out all controls when doing a failsafe land?

10:00am - Randy and Aero
documentation for intel Aero is in Wiki
Including parameter file

10:01am - BBBlue and AP_Sync
Performance issues mean that this image is currently in limbo

10:04am - tridge and Plane and other things
Logging quaternions now to solve gimbal-lock issues during transition / log analysis
Need to get this stuff documented
Tridge has a GSoC candidate who would like to do documentation for dataflash
Their application wasn’t put in because of various things
Tools need to be able to do conversion to Euler angles for analysis purposes
Thrust vectoring happening
Merged into master
Working in simulator
First test flight is done
Tuning was awful
Learning from scratch on tuning these things
Tridge needs to sit down and chat with Leonard as to how to go about tuning these
Unvectored tailsitter in SITL now
Vectored to come
One-shot fix
It got broken again
And fixed again
Bl-heli one-shot really cares about the refresh rate
During startup ArduCopter was doing motor writes at 59.8Hz
Caused BLHeli to never initialise
Even when it all switched to 400Hz
Must be careful to always send stuff out at 100Hz+

Booster motors
Tricopter with petrol booster motors
Matt R’s idea
10cc petrol booster motor

New photo by Andrew Tridgell
Lots of power for a small vehicle
80% hover throttle in pure-electric
Petrol-motor at half-throttle hover-throttle was 35%
Current draw 30A to 10A
Can add a booster to any multicopter
Very loud tricopter
Multi-output for multicopters
Fix for srv_channel library
Also allows multiple channels for same motor
Could fly an octoquad as a normal quad
Can set SERVOn_FUNCTION multiple times
Multiplier on AP_Motors throttle
Could instead set things up so that electric motors spin at e.g. 10%
Need Leonard’s input
Need to cope with petrol motor conking out
Lots of yaw torque
Tricopter yaw servo was only part-way over
Need to work on the OH&S issues….
Leaning over vehicle to start it….

10:27am - MdB and blowing-into-Pitot-tubes
Pitch-down moment when blowing into Pitot tube
Attitude.pitch is changing
Taking cross-product of gyroscopes and air-speed sensor
Gives you estimate of accelerometer correction
Thus Attitude-estimate change due to airflow
We should continue to use EKF for attitude if it thinks it is good even if it is unhappy with its absolute position estimate

10:32am - Tom and AP_Landing
Tom wants to do a frontend/backend split
Sid also noticed that 3.8 fixed a landing-slope discontinuity bug

10:34am - Assist Angle is saving lots of aircraft now-adays
If you stall or wrong-attitude then quadcopter will save the vehicle

10:36am - Sub with Jacob
Doing release for sub 3.5

wants to tag -rc1
Will work with tridge to get this right
Need to force-push the tag
Which is dangerous
Randy does release notes, version, then tag

That’s probably ArduCopter
Matt + Solo underwater?

10:42am - mavlink and Jacob

Wants to be able to spin based on motor number
May conflict with tridge’s spin-all-motors-in-sequence count
Mavlnk xml documentation doesn’t reflect that
This would require constant input from GCS
Sliders in user interface to spin motors faster/slower

10:48am - decimation of dataflash message
Peter is putting a patch together to reduce the frequency of the bad-sd-card-logging message

10:53am - battery monitoring with MdB
Can see cells in packs going bad!
Patch to reduce smbus traffic
75% failure rate talking to Solo battery
Not present on normal Solo
sending/receiving at same time?
Fix for PL lidar went in
Closer to spec
2-transactions is closer to spec
So change was made in all of AP_HAL
There’s a flag where individual device drivers can change this behaviour
So can revert the global change, perhaps

10:59am - buzzer issues

Unknown tone

11:02am - output limit

Tridge is going to look at this

11:07am - long term Baro drift with Michael
Holger’s PR

MdB needs help as he doesn’t get Holger’s comments
Interpretation of comments will be needed