Dev Call Nov 9, 2020

Issues & Pull Requests




1 Like

Attendee count (max): 16

UTC1100 -

  • Can’t use --sysid willy-nilly or user’s can’t set their system-id parameters

UTC1107 -

  • Probably not the right fix, but something does need to be done
  • Peter will look at this

UTC1113 -

  • Raising max-scale-factor
  • Wider scale factor because failure to calibrate is painful
  • Merged!

UTC1115 -

  • Don’t fail radio in a single bad frame
  • 4.0.6 airspeed and throttle nudging might be bad
  • Throttle pulsing and whatnot?
    • Long-range transmitters are often only 20Hz
  • 1-100 integer granularity on throttle in Plane is awful
  • For stick mixing:
    • Tridge
    • Peter
  • Against stick mixing
    • Randy
    • MdB
    • Leonard (apparently)
  • Backporting this to Plane stable

UTC1127 -

  • We’ll be breaking APM Planner if we do this
  • Onboard log viewer as an alternative?
    • Open-source
  • EKF has been using instance numbers since December last year
  • Announced instance type in October last year
  • Tridge and MdB think we should merge it
  • Will ask Andy about things
  • Change motor offsets columns rather than normal offset columns
  • Should move offsets into separate message
  • There’s a new tool to give percentage of log used by messages
    • Showed a problem up in EKF logging…
  • Delta-compression using log messages might be coming
    • Array of floats
    • Between neighbouring messages usually only the bottom byte of a float will change
    • If a message makes the ringbuffer then it is guaranteed-to-disk
    • XOR new version vs old version, write delta
    • Might be really good for flash-based logging
  • Breaking stuff is sometimes worthwhile
    • See FTP stuff

UTC1134 -

  • Clear wind estimate on disarm
  • Yes, we do do wind compensation in PosHold
    • Lean angle
  • PR for proper wind estimation in Copter coming
  • Use build server statistics to work out whether users actually want things like poshold

UTC1135 -

  • Log secondary attitude from configured EKF
  • Merged

UTC1159 -

  • Disk-theory gain scaling
  • New parameter for disk loading
  • Can estimate wind flow over surfaces
  • Want test results from real-world flights….
    • Henry? Marco?

UTC0008 -

  • Copter Spin min / spin max / expo in forward flight for planes
    • Decided this is not a good idea.
    • Esp now we have Plane battery scaling

UTC0010 -

  • MdB doesn’t see any other problems with this
  • Need to work out what’s up with the bus number comment
  • Two external buses on Pixhawk4

UTC0017 -

  • Slew position resets
    • But not in modes that understand resets like loiter
  • S-curve stuff might conflict here
  • Leonard will want input here
  • Randy likes this

UTC0024 -

  • Want to get this in this week
  • Peter and tridge are happy with this
  • New Replay
  • Works really well
  • Costs 2.5kB in flash
  • ~4kB of RAM on CubeBlack
  • Does slightly increase CPU costs
  • Accurate Replay is useful for analyzing issues
  • Number of scheduling misses increases
  • /4000 moves from 260 to 300
  • Once this is approved we can get some other stuff merged
  • Would need to work out if CPU cost is OK given the benefits
  • Free mem went up on F405 Wing
  • User visible changes
    • One IMU you only get one parameter set
    • … and we don’t allocate some data structures and cut some code out
  • AP_DAL - data access layer
    • AP_NavEKFs go via this to get data
    • We log data going through the DAL
    • We have a linking test to make sure deps don’t get added
  • GPS affinity fix in place
  • Gps_get_yaw_degrees fix
  • Generic function for calculating a filter
  • Paul has looked at changes to the EKF
    • He’s comfortable with these changes
  • Application of trim offsets is inconsistent
    • Euler vs Quaternions are different!
    • AHRS never rpy from EKF, rather quaternions
    • Shouldn’t be putting trims into the EKF
      • AHRS layer should add it consistently to all of the outputs
    • Rotation matrix body-to-vehicle is passed into the EKF for no good reason
    • Graphing NKF1.roll vs ATT.Roll you’ll get a discrepancy equal to your AHRS trim
      • Delayed horizon
  • Perfcounters removed
  • When do we put these in?
    • This-arvo?
  • How do we know if we’ve broken something?
    • Well, we’re writing something to work out if we’ve broken something
      • Replay can’t help us find the problem….
    • Has been reviewed by several people
  • Review is our best tool
  • Backporting to pre-Python thing
  • Log-file-bufsize can get much larger now
  • Default buffer size on H7 increased to 200kB
    • Usually have 700kB free
  • F7 is 80kB
  • Tridge will look a little more at the CPU usage and then merge
  • Should be able to tune the EKF using this by specifying parameters
  • Change default based on what GPS is reporting?
  • M6 vs F9 have vastly different accuracies
  • Septentrino vs uBlox reported accuracy differences AKA “my GPS doesn’t lie” parameters

UTC0052 -

  • Reliable triggering of an internal error
    • Uploading of 305 waypoints causes internal error?
  • SPI fail
  • When we are disarmed the flash sectors can roll over one to the other
    • Erasing the flash can cause this
    • CPU stops while erasing as instructions can’t be read
  • Already have stuff in master to suppress reports when this happens
  • When your CPU stops you can expect this sort of behaviour
  • 4.0.5 has fixes for this
    • Suppresses
  • Can still have problems when stuff completes simultaneously when the CPU comes back to life
  • Emit a message when flipping sectors?
  • Must not get people used to these internal error messages
    • They often indicate dangerous things happening

UTC0100 -

  • GPS primary stuff
  • Merged!

UTC0102 -

  • Alti have been testing
  • This turns on by default

UTC0106 -

  • Dump all serial on a port to SD card
  • Need to set serial port protocol
  • Still need the open/close file thing

UTC0115 -

  • Backporting fixes to Copter 4.0.x stable
  • We are trying to do critical fixes
  • New sensors?
    • Relatively low risk to add new e.g. rangefinder?
  • And boards?
  • Don’t backport stuff that people haven’t asked for either
  • [12:21 PM] WickedShell: GPS stuff no, if I stuffed something up lets find out in master not in backport…
  • Needs to build on Plane

UTC0025 - Copter update

  • 4.0.5 is out and seem to be OK
  • Simple avoidance is a little too aggressive when trying to stop
    • Not a new thing
  • Trudging forward with 4.1
    • S-curve stuff getting very close now
    • Sub autotests still need to pass
    • Quadplane autotests need to pass
    • Real bugs!
    • Hopefully by week-end we’ll have s-curve PR

UTC0028 - Plane update

  • Bugfix from Pete for another release potentially

UTC0029 - Buttons

  • Do we really want separate parameters for PWM pins?
  • Peter to grab tridge to discuss

UTC0034 - close

1 Like