Dev Call Apr 13, 2020

Issues & Pull Requests




1 Like

Attendee count (max): 23

UTC1102 -

  • Two-stage PR
    • Disable EK2
    • This PR will change it to 3 magically
  • Downside is that if a user has fancy EKF parameters set then the vehicle may not be flyable
  • If we don’t do this then moving backwards and forwards will be very painful
    • E.g. having both enabled at the same time
      • And running out of memory/CPU
  • Do this on the basis that few people actually do set any EK2 parameters
  • Setup defaults to new thing we want rather than doing runtime sets?
    • Wouldn’t catch people who have set things to EKF2
  • This doesn’t change the default to be EKF3 on boards that can support both
  • Most people on 1MB boards would be running EK2
  • This specifically checks for someone to move from EKF2 to EKF3
  • Not supporting EK2 may break Vision for some people on 1MB boards
  • [9:09 AM] (Channel) MdB: _ALT_SOURCE is probably the most played with EKF related tuning (straightforward conversion in that special case)
  • This PR fixes a specific bug where people will be flying on DCM

UTC1115 - Henry’s PR - turn off EK2 on 1MB boards?

UTC1119 -

  • AP_Scheduler debug enhancements
  • Tridge has a different idea for this
    • Make it available via ftp
  • Close or mark for rework

UTC1121 -

  • Add support for ProfiLEDS
  • Want Randy to look at this
  • A lot of stuff to go back into a stable release
  • ChibiOS bits end up, like master
  • Tridge isn’t keen on putting this into Plane stable
  • LUA stuff is nice and working
  • No Pong in stable as the stick input stuff isn’t backported
  • Brings back some parameter conversion
    • Can’t upgrade to 4.0 from 3.4 / 3.5 after this
      • Ran out of servo functions
    • Change in upgradeability
  • Tridge would prefer to be getting onto 4.1 sooner
  • Needs to be tested with d-shot
  • Non-essential feature ordinarily pushed to 4.1
    • Cool marketing thing
    • Extended beta already planned for 4.0.4
  • Will need to check flash size
    • or to check sizes

UTC1129 -

  • Derive RPM from harmonic notch filter
  • Expose values in RPM driver
  • Appears over mavlink telem
  • Andy used Srxl driver to expose RPM values onto transmitter
  • Harmonic notch would ordinarily be a consumer of RPM source
    • If you don’t have an RPM sensor then the FFT can measure an RPM
  • Need to select where you want RPM to come from
  • FFT is a set of spectral strengths that we could do a peak search on to find an RPM
    • The harmonic notch takes the FFT data and does that peak search
    • Needs to a module that processes the FFT data to determine the fundamental frequency
      • That’s currently in the HarmonicNotch
        • Could move this into the FFT?
          • So it then gets used by the RPM and the Harmonic notch
  • Would be nice to get RPM from FFT without actually having a Harmonic Notch filter
  • FFT for label?
  • merged!

UTC1139 -

  • Srxl support
  • IOMCU needs testing
  • Did we verify that the telem is in the right directory?
  • AP_RC_Telemetry
    • Move the others later
  • Needs to be flight-tested on frsky
  • Yaapu verified fport but should verify the others

UTC1145 -

  • Add support for custom rotations
  • Size blew out
    • Probably now depends on a new library
    • Ensure it is disabled for AP_Periph etc etc
    • Shouldn’t have parameters either
    • Minimise features….

UTC1147 -

  • Double-notch harmonic filter
  • Flown and works fine
  • Leonard did design
  • Two narrower notches
    • Less group delay
    • Squashes the noise
  • Notch filter is an axe cut in the spectrum
    • Good for helis!
    • Multirotors have multiple sources
      • Noise growth on each size
    • Double-notch puts two axe-cuts right next to each other which overlap
      • More of a bucket style of response
      • Significant broadening of attenuation size of notch filter
      • Narrower, so same group delay
      • If the board can handle the additional processing they can improve the quality of their harmonic notch filtering
      • How much extra CPU?
        • Watch sys_status.load
          • Double vs single?
          • Should be twice the cost of the existing single notch
          • Particularly on existing F4
          • PM.Load SYS_STATUS.Load
          • Flown on F4 with long fft enabled and this works fine
          • Hitting limits on CPU and memory
            • Even on capable boards
          • OA hit limits because background thread was being starved of resources
          • Comments on methods to indicate cost is good
          • Should we have a health check on the Scheduler?
            • Or some sort of reporting
          • Could already be making this available via SYS_STATUS
            • Have to worry about the loop-scaling we do
              • Multiply by slowdown

UTC1158 -

  • Flash storage changes for H7 support
  • Has been tested
  • Can test the three different types of flash now in SITL
  • Fixes an issue with people getting internal errors wrapping sectors
  • SPI timeout or corruption of IMU data
  • About a second to erase 1 128k sector
  • We’ll never do two in a row
  • Our watchdog interval is 2 seconds
  • merged!

UTC1159 -

  • Add UAVCAN support into SITL
  • Tridge needs to look at this
  • Why double-precision arithmetic all of a sudden?
  • Testing and work and review
  • The can-serial-passthrough thing should be pulled out into a separate PR
  • PR should really only contain changes in SITL

UTC0004 -

  • Needs to address existing comments - in particular the check-and-forward thing

UTC0007 -

  • Already closed
  • Real bug if you have two compasses
  • When we finish calibrating there’s a 0.8 second pause which will annoy the EKF
  • Second compass can have the wrong attitude for the remainder of its calibration
  • Will be revisited!

UTC0008 -

  • Changing blprotocol.cpp
    • No response from author yet
  • Needs testing

UTC0010 -

  • MdB to be thoroughly prodded

UTC0010 -

  • Peter still to review

UTC0011 -

  • EK2 and EK3 operation without good compass data
  • Needs a rebase
  • Depended on optional fallback to compass

UTC0012 -

  • @param and @sys
  • MdB hit “submit” on his review without looking too hard at his old comments
  • Delicate because seeking has to get exactly the same data
  • Pack_param has all of the logic for packing parameter into
    • Should it return a boolean?
    • Needs to fill to end of buffer?!
    • Next read starts with those bytes?!
  • Aiming to put all knowledge of packing into one function
  • Tridge would like to get this one in as he’d like to start building on it

UTC0013 -

  • Can be merged after testing with large missions
    • And valgrinded

UTC0022 -

  • Added ability to simulate it
  • Scaling factor added
    • Needed for T265
    • Camera extrinsics (“camera form”)
    • Sensor startup if you are close to things it seems to get scale wrong
    • Probably a firmware bug on T265?
    • Not just a vision system
      • Uses accelerometer
      • Inertial system
      • So shouldn’t have a scale error
  • pp: so yes I experienced about 20% compared to the LidarLite on the Z axis
  • merged!

UTC0026 -

  • First attempt to mount might fail
  • Write out on log open
  • Tridge thinks we can preserve the use of Log_Write

UTC0034 -

  • Oops,

UTC0036 -

  • Hott broken in master?
  • Timing of uarts etc etc
  • Need to get things fixed before adding more features

UTC0035 -

  • AP_NavEKF2 clear error message
  • We don’t use our normal pattern here of asking the subsystem to fill in a string as part of the prearm_checks call
  • Randy will comment on it pointing back at the Copter code
  • Should look at doing a prearm_checks(char*buf, uint8_t len) call on EKF

UTC0042 -

  • Reset-time-ms
  • Already discussed with tridge and Paul
  • EK2 isn’t handling resets correctly ATM
  • Maybe we should forget about this and getting external nav into EKF3
  • Probably a simple fix

UTC0044 -

  • We will be moving away from Ardu to Rover
  • No reason to do all the vehicle in the same day
  • Merged!

UTC0051 -

  • Merged

UTC0052 -

  • Merged!

UTC0053 -

  • get/set an entire AP_Mission item
  • Maps into mission-item-int structure
  • Looks like mavlink message
  • Get-item / set-item
  • Needs code-style fixes
  • Could be refactored to share more code into the missionitemprotocol
  • Retpacket stuff needs to be looked at
  • Should be OK for locking purposes with the scheduler lock
  • MdB will be going off and looking again at this

UTC0101 - Plane?

  • No, lots of system-level stuff
  • Syncing with Copter could be problematic with profiled stuff

UTC0102 - Copter and Rover

  • non-GPS-navigation development
  • This is delaying 4.0.4
  • Loiter-turns broken
  • Circle in auto mode, rangefinder terrain following etc
  • Circle in automode didn’t point to center of circle
  • Rajat raised and patched the issue

UTC0103 -

  • Fixes neopixel sync issue

UTC0105 - Rover

  • Growing calls for 4.1 release for Rover
  • Moving-baseline stuff

UTC0106 - Close

1 Like