Dev Call Nov 1, 2021

Issues & Pull Request

ESP32 Sub Modules




Attendee count (max): 22

UTC2259 - Fail to arm if in RC_Calibration by IamPete1 · Pull Request #18514 · ArduPilot/ardupilot · GitHub

  • Don’t arm in rc calibration
  • It works this way in the documentation but no GCS actually implements this standard
    • QGC is supposed to do it but it doesn’t seem to?
  • Messages (common) · MAVLink Developer Guide
  • Need something on the GCS mailing list
  • Recommended procedure is take props off
    • And possibly safety switch

UTC0002 - Plane: loiter altitude control like FBWB by IamPete1 · Pull Request #18624 · ArduPilot/ardupilot · GitHub

  • Make loiter altitude control like fbwb
  • Checks if stick mixing is enabled first
  • Flight option bit to disable
  • If pitch trim is outside dead zone you could end up climbing/descending
  • Default to on or off?
  • Roll radius out as well when stick-mixed that way?
  • Terrain following?
    • Cruise does terrain following, fbwb does not
    • This should do terrain following
  • Buzz: kinda feels a bit sub-optimal to me. Is there a way it could learn trim centres…?

UTC0009 - Plane: split transition in to separate class for SLT, tailsitter and tiltrotor by IamPete1 · Pull Request #18697 · ArduPilot/ardupilot · GitHub

  • Splitting up the different quadplane types
  • Real-life tail-sitter testing from Marco and tridge
    • slt and tailsitter work nicely
    • Tiltrototor not flown yet IRL
      • Griffin flown in RealFlight
      • 788 bytes

UTC0014 -

  • ADC scaling / testing not working
  • Vservo - bec on servo rail and make sure reported voltage is right
  • Tridge will test
    • Works
  • Need to get MdB to approve

UTC0017 - AP_Param: Class Conversion Function by hendjoshsr71 · Pull Request #18994 · ArduPilot/ardupilot · GitHub

  • Convert parameters from all classes
  • Tested with eedup
  • Merged!

UTC0019 - AP_Airspeed move into AP_Vehicle by hendjoshsr71 · Pull Request #18838 · ArduPilot/ardupilot · GitHub

  • Ap_airspeed up to vehicle
  • Vehicle-in-header problems
    • Tried to add it to the whitelist?
  • Could hide the parameter trees if we felt like it
  • Priority of running the scheduler tasks means we’re leaving stuff in the vehicles ATM
  • Don’t want EKF to consume airspeed data
  • Need to have it off by default so we don’t mix airspeed up with barometers!
  • Need to not use zero-offset in table
    • Top-level-key can be stuffed
    • Ends up with a conflict

UTC0032 -

  • Add airspeed support to Copter
  • Synthetic is producing length of airspeed vector
    • Full vs x
    • Pitot don’t just produce x component
      • Roughly flat over 15 to 20 degrees
    • AOA up by 1 it will actually get aoa-0 equivalent
  • Copter can do wind-speed with drag
    • With airspeed it should probably show drag-coef values in VFR_HUD
  • [10:36 AM] Peter Hall: Copter people will probably be expecting the XY airspeed vector.
  • [10:37 AM] Peter Hall: Otherwise you might end up with the X airspeed going down as you speed up. Because you pitch down to go faster.
  • Will test this this weekend
  • 3 or 4k on smallest build
  • Matekh743 copter is down to 14kB
    • Lost 256kB off H743
    • Miscalculating free flash
      • Storage sectors weren’t taken into account
      • Thought we had 240kB free.
        • Wrong!
  • -Os rather than -O2 as first port of call
  • [10:45 AM] Peter Hall: Makes sense to remove frame types that have more motors than the boards have PWM out.
  • [10:46 AM] Peter Hall: Lots of those ‘betaflight’ boards only have 5 or 6 PWM outs.
  • Need to check for space
  • He’ss run --build

UTC0040 -

  • Add c++ and lua extensions for vscode
  • Ide will detect and offer install of add-ons
    • E.g. astyle add-on with our astylerc
  • Great Britain for English!
  • Merged!
  • Sid’s written an extension to expand stuff
  • VSCode is apparently much better than Eclipse now-adays
  • Two pages in Wiki doesn’t help
    • Some mixed in with debugging on linux!

UTC0051 -

  • Mavproxy mcast and autosysid improvements
  • Mcast links not appearing in “link list”?
    • Works in master

UTC0103 -

  • Summing battery by mask
  • Parameters reorganised in smbus battery stuff:
    • Share a table pointer and staggered index
    • Smbus parameters will be broken older master to newer master
      • Won’t affect stable branches
    • We have simulators for maxell and rotoye
    • This is a nice structure
      • Need to check for conflicting IDs
  • merged!

UTC0113 -

  • Nooploop use base class uart
  • Someone is having trouble on the forum with it
    • EKF-related rather than driver issue
  • Noop-loop uart pointer probably broken in master
    • Uart not initialised

UTC0119 -

UTC0127 -

  • Y4 frame support
  • Is this just a quad?
    • Y has cross-axis coupling
  • Why?
    • Because it looks cool?
    • Hard 3d stuff
      • Feel is better according to some fliers?
  • A-Tails are in there as a frame type

UTC0136 -

  • Fix failure to return mode on short fs action
  • Ending up in circle mode mean we don’t revert
  • NeedsTesting
  • Follow-up PR to do more complete fix

UTC0141 - git submodules

  • Github smacked us - git:// no longer works
  • 11:42 AM] Peter Hall: If we hold of for a while they might fix it again…
  • [11:43 AM] Peter Hall: Or back down on the breaking change.
  • Completely blind-sided by github
  • :44 AM] Peter Barker: It looks like they’ve actually screwed up!
  • [11:45 AM] Peter Barker: i.e. this shouldn’t have happened until Jan 2022
  • [11:45 AM] Peter Hall: Yeah, that was my read.
  • [11:45 AM] Peter Hall: Maybe they pressed the wrong button.
  • [11:45 AM] rmackay9: ah, wait there’s a Nov 2nd date…
  • Maybe we should use some git magic to pull the submodules as needed?
    • So if you configure and don’t have a submodule it grabs the relevant modules?
  • Timing - all of this happened right on the call
  • Vendors pointing to their own remote might also be broken
  • Probably won’t fix the uavcan → dronecan change
  • [11:56 AM] Peter Hall: This is only temporary for now, we don’t need to rush out a fix in the next few hours.
  • [11:56 AM] Peter Hall: We can take a few weeks to do it nicely.
  • [11:56 AM] Peter Barker: They’re not saying for how long, right?
  • [11:57 AM] Peter Hall: We’ll also run several short brownouts on this date.
  • [11:58 AM] Peter Hall: So should only be today
  • [11:58 AM] Peter Barker: Bloody hell. Did they really think this through in the context of CI?

UTC0200 - Plane update

  • 4.1.3 coming
    • Scripting matrix frame class
    • Fix rangefinder double-correction
    • Vtol land when close to landing point to not do an approach
    • Suppress rate d term
      • “Ground mode”
      • Rate d-term causes control surfaces to bounce around
      • Disarmed + not flying suppresses d-term in roll/pitch rate controllers and halves the d-term
    • And perhaps more from the discuss forums

UTC0202 -

  • Randy’s been away for a week, a little behind
  • Beta testing for 4.1.1 rc1
  • Three issues have come up in 4.1.1
    • “Reports requiring investigation”
    • Missing IMU on MatekH743
      • Power issue?
    • Fport disconnection after 25 minutes?
      • In 4.1.0 as well so not necessarily a blocker
      • Originally reported in Plane
      • Set RC protocols to f-port only?
      • Archer RS
        • Tridge has one of these
      • fport 2 is at firmware version 1
      • fport 1 is at firmware version 2
      • Completely different protocols
    • Internal error doing temperature calibration
    • QGC can’t download parameters over wifi
      • Peter can’t reproduce
  • Lightware serial lidars seem to be problematic - slow rumble from users
    • E.g. Repowering required to get them to work?
    • Specify a new enum entry to force a protocol version?

UTC0216 - controls update

  • SCurve silly value check
  • Jerk limit really high?
    • Code assumes you can achieve that
      • Navigation can move really slowly
    • Fix coming
    • Limit max jerk to what the aircraft can achieve
      • Attitude controller has acceleration limits
      • Roll angle is closely related to acceleration
      • Roll rate is closely related to the jerk
      • your roll acceleration is closely related to your jounce
      • If we know the jounce we can calculate the jerk, and that’s what we pass into scruve
      • If jerk is large and jounce is small we assume it will take a long time to achieve jerk level
    • So a new check for unobtainable values added
    • Docile navigation
    • 4.1.2 fix

UTC0221 - close