Dev Call Jul 24, 2017 2300 UTC


  • 3.5 Updates


  • Updates


  • 3.8 Updates
  • Plane, AP_Landing and flight termination
  • Looking for DF logs with bad airspeed data (and useful logging)


  • P4pro
  • dsm fix
  • canbus,
  • fast servo speeds



Attendee count (max): 25

09:04am - tridge and hardware
PX4 Pro merged!
nice port done by K Lopez
Px4v1 bug fixed by tridge
Preexisting, shown up by this port
Not in Copter 3.5
Probably not in Plane 3.8
DSM Spektrum receivers on PixRacer
Bug introduced when DMA was introduced on SPI bus
DMA was turned off on the Spektrum port
Time-based framing
If no DMA for receive the timing can go off relatively easily
You get incorrect channels on your RC inputs
It was relatively easy to reproduce
Channels have been jiggled around a bit
Extremely timing-sensitive
Request that those with PixRacers try the fix (by flying master)
Even if they’re not flying Spektrum
May also affect S.BUS
No checksum ,timing-based
Decoding is a little more robust
3.5.1 will be released to contain this fix
Usually if there is a DMA channel conflict on NuttX the board doesn’t boot

09:12am - tridge and faster servos on Plane
Belly-sitters need faster servos
Already had these for the lift motors on quadplanes
SERVO_DEF_RATE (or similar) parameter
Problem in Rover where this code is currently triggering all the time

09:19am - Sub and reboots
Should we do that on all vehicles?
We already have low-level code in place that ensures we don’t send short pulses for one-shot but nothing for normal PWM ATM

09:19am - hotplug UAVCAN GPSs
Copying of states between GPS and uavcan can cause issues
Should the instance variable be in the backend?
short -term fix for this is probably needed
Reported issue that uBlox driver is doing redetection, which would be very bad

09:25am - Fixes for Mixer generation
Not a perfection representation
Manual RC mask will be required for complicated setups (e.g. MdB)

09:26am - Plane delayed
These issues have delayed Plane another week

09:26am - MdB and AirSpeed sensors for Plane
Blocked airspeed tubes dataflash logs would be nice!
Want to do blocked-tube-detection
Look at elevator integrator
Should log all synthetic airspeed estimates
Dual airspeed setups would also work to fix this issue

09:28am - termination for Plane
Another flight mode?
Or stage for auto?
Switching into auto and a specific stage seems odd
This is different to “AFS-termination”
More of a pop-the-parachute thing
Maybe it should be hooked in near the parachute code?
But still want to stabilize on the way down
Just stabilise the descent
But you still never exit this
Could be nice if you could get out of this
Spiral landing instead?
Flare near ground
“Landing mode”?
If they haven’t set up for spiral or deep-stall then the mode would be FWBA glide?
Or reject the mode change?
On Plane every mode is possible
On Copter we can reject mode changes
We don’t have spiral landing… yet
Descending loiter and constant airspeed would be relatively quick
Complicated because Plane likes to have paths for missions
So we’d probably end up calling into AP_Landing from outside
CRASH mode?

09:41am - Randy and Copter
3.5 was released
Smooth so far
Big jump in number of support questions on forums
Some crashes to be looked at
But a lot of non-critical things
Possibly smoothest release we’ve done so far
Couple of issues
DSM issue as mentioned earlier
3.5.1rc1 almost ready to go out
Some alpha testers looking at it
“Did it fix my problem?” people
3.5.2 plans
GitHub Projects list
Backports area
A 3.4.6. User brought up some stuff, meaning 3.5.2 might get some features
GPS glitching
GPS glitching won’t trigger EKF failsafe any more!
Need velocity and compass innovations to trigger a failsafe
Master has a change which uses velocity/position and compass innovations
Reporting of glitching
EKF has a glitch flag
We should report this and use it for arming
And possibly for a failsafe
MdB Randy: new MAVlink message that reports all those would be great! (and system accuracy estimates)
EKF special-casing of under-5m can cause issues
Never using compass
Been reduced to 2.5m
Just reduces loophole size
LED Solo changes
Maybe remove the EKF3 GPS wait-forever patches
Confusing users a lot
Buffer-sizing patches in EKF3

09:49am - Rover and Randy
Pierre, Randy and Peter have been working in Rover a bit
Rover-mode patches have gone in
All vehicles eventually should have this
Makes more cleanups apparent
Too many callbacks into the vehicle code
Pierre is working on next few steps
Paul has been working on wheel encoders with Randy
$15 each
One slightly controversial change
Defaulting Rover to always use the EKF
Do we need more logic to make it use EKF?
Randy says it still seems to use DCM even when saying always-use-ekf
Do without DCM on Rover?
Go to HOLD mode if EKF is bad
Bad for boats!
If GPS-only and it is bad
With DCM it can come home
With EKF it won’t come home
Change Navigation so that it can be happy with non-GPS forms of navigation
Randy tried this
But apparently found the wrong place
Randy’s going to have another go at fixing NavEKF rather than supplying the constant

10:03am - issues
MdB and
Telemetry log doesn’t contain raw pressure difference
One absolute and one differential pressure
Are we using these fields incorrectly?
Will fixing this break AntennaTracker?
Fixing one might break the other
Only using absolute, not differential
So no problem
Nobody else seems to use the pressure differential
Comments in mavlink say this should be an airspeed sensor essentially
Our airspeed reporting is pretty poor
PR will come into change the value in this field
Something is better than nothing
Original problem was that we have no way of telling what the protocol being used ever between the receiver and PixHawk

10:17am - Urus
Chats on gitter and Skype
Many people say that we don’t carry dead code and should remove it
A year ago code was let into master
Thinking was that AVR branches might be taken over and maintained
And that the code behind Urus might be PR’d
We want to remove Urus code because we don’t carry dead code
The code isn’t being used ATM
People aren’t flying it
It can be added back later
Forks are always a possibility!
Hiroshi needs another month to put a PR together to make the Urus code in AP master worthwhile
PR will be created to remove Urus code
Hiroshi can comment there

10:24am - PX4FLOW
Latest firmware is from 2014?
Windows10 can’t update these?
DFU mode is a possibility
Philip will play with this

10:29am - S Quilter and GSoC
Almost have a PR ready!
But broke everything today
Randy has offered to test