Dev Call June 15, 2020

Issues & Pull Requests

New log analyzer





1 Like

Are we still using mumble for the dev calls?

Same question as above.

I’m currently on Mumble but I don’t see anyone in the “Weekly devcall”

We are using Discord

1 Like

Thank you on discord now

1 Like

Darn. We’ve moved the devcall to Discord; instructions are on the Wiki.

Where did you find the information which led you to Mumble? Do we need to
update them?

I was also under the impression we left a message on the mumble server ot
encourage people over to Discord. You obviously didn’t see it, so I guess
we might need to do better there… > * Weekly meetings on Mumble Server


I’ve created
to fix that.

Moving to Discord is a relatively recent development; maybe a month now.



Attendee count (max): 29

UTC1100 -

  • Code tridge pointed to was removed
    • Should be much safer
  • Needs code review from Peter / tridge and others
  • Interaction with frsky sensor PR?
    • Might be able to recycle/reuse code
  • Should CI post the size difference somewhere?
    • Complicated
  • How much flash size does this add?
  • Telemetry is so useful generally, so behind minimise features is maybe not appropriate for these
    • Should be behind some ifdefs
    • Henry Wurzburg: its 3.3K more on omnibusf4pro
  • Dynamic allocation and parameters weirdness
    • We do have a way to register parameters
    • Parameters exist all the time…
    • Only enabled when a serial protocol is set
      • Serial protocol 10 is existing s-port protocol
      • There’s no additional gate - transmitter just starts sending new message
        • Initial download might not get parameters if we use a parameter pointer
    • Enable parameter?
  • Needs more testing by community members
    • Put a call out for testers
  • Travis is a little annoyed too

UTC1128 -

  • Dynamic harmonic notch filters
  • Drive sensor frequency of lowest harmonic using fft or rpm or whatever
  • Stack second and tird on top of that
  • What this PR allows you to do is select the centre frequency of all of the notches
  • Fft tracks three peaks
    • Can set sensor frequency for three notches
  • Or use RPM sensor on 4 motors to drive 4 notches
  • Much more precise
  • RPM sensor works really well
  • Side-effect of changing the notch frequency is own source of noise
  • Input has to be smooth
    • Not the case with RPM
      • LPF fixed it
      • Tconst of half the trate
        • 100Hz TRATE worked really well
          • Uses up a huge amount of the available 19200 bandwidth
  • LPF on gyros should come after the notches so the LPF attenuates the filter-induced noise
  • Randy B likes it
  • Write-notch-log-message
    • Deref without checking?
      • Fixed array in inertial sensor
  • Should Leonard look at this?
    • Can be merged if Leonard says yes

UTC1135 -

  • Airmode on switch
  • Arming switch only disarm at zero-throttle
  • Accidental disarming with existing arm/disarm
  • Have another option for only-disarm-at-zero-throttle
    • Or is-landed with Copter?
      • Probably can’t win here as landing detector may not be triggering
  • Current magic-airmode is odd
  • Can be merged once the state has been moved out of ap

UTC1144 -

  • CRSF protocol implementation
  • Tridge has done a first-pass review
    • Some things to change
  • Don’t use packed unless you need it
  • 11-bit-rc-input
    • 16 channels
    • Common structure and common unpacking structure
  • Tridge is sending some hardware for testing

UTC1149 -

  • Andy thinks this is done
  • Video is cool
  • Basti has been testing too
  • Char **values in param metadata?
    • Text version of parameter
  • key/index/group-id to get a number to put in static data?
    • Migrating parameters might break this horribly
    • Check cost of scanning for parameter names instead
    • Cross-vehicle stuff is awkward
  • Tridge and Peter need to look at it

UTC0002 -

  • Merged!
  • Clear fence breaches when fence type disabled

UTC0002 -

  • Can be merged after testing

UTC0005 -

  • Merged
  • Naming is a bit odd on the next-id thing

UTC0009 -

  • Battery stuff, need MdB around to do this properly
  • Split into generic / maxcell
  • Heavy lifting is still in generic
  • Testing done on Maxcell 6-cell and maxcell 12-cell
  • We were making pointless i2c calls
    • Fixed
  • No battery at boot?
    • Battery has to be healthy (main voltage) for 15 seconds
      • Then you must be getting cell voltages
  • Mavlink2 extension rather than taking the minimum
    • Second array of 10?
  • MdB to review

UTC0015 -

  • Fix Plane forward throttle voltage scaling
  • Previously worked if cruise throttle would never hit max
  • This allows you to set things up so full throttle is available at all battery levels
  • Merged!

UTC0018 -

  • Programmable glitch filter
  • Move glitch filter up to cover all rangefinders
  • In backend now
  • Tried a couple of versions
  • Baro filter relies on high sampling rate
  • So used filter from hc-sr04
    • Added parameters
    • Results are posted in PR
  • Do we have two parameters per rangefinder?
    • Yes
  • Filter coeff zero it disables everything
    • Still does too much maths?
  • Glitching gives average rather than no value?
    • Rapidly descending vehicle in landing?
  • Extrapolation may be broken by averaging
    • Status.distance_cm structure is bad in rangefinder
    • Race condition in update_status stuff
  • API change so update_status
    • Stop filling in distance_cm in backends?
  • Lying about having new data when we don’t could be bad
  • The extrapolation is in trouble after this patch
    • Need to know what the LIDAR lag is
      • Do what the EKF does with time horizons and whatnot
  • Rejecting means you jump to using baro
    • Reject time?
  • Copter glitch protection should have worked for Henry
  • Need two data products
    • Raw for EKF
    • Filtered
  • Need to look at Copter and Plane filtering
  • Need to make sure Copter glitch works
    • Then base filter based on Copter
  • Proportional glitch detection based on height
    • % with minimum?

UTC0051 -

  • List was scrubbed
  • Changes made as requested from last week
  • Can be merged once spaces are removed from temp

UTC0055 -

  • Arming check has been inserted
  • Some comments from Pete which haven’t been resolved
  • Tridge wants to do some testing on this one

UTC0058 -

  • Merged

UTC0101 -

  • Merged!

UTC0106 -

  • Can be merged after its tested not to spew to the console

UTC0119 -

  • Merged

UTC0122 -

  • Conflicting
  • Needs someone to talk to it

UTC0123 -

UTC0125 -

  • Revisit of PiccoloCAN parameters
  • Uavcan needs parameters to know which ones to send
    • So there’s precendent
  • Don’t want to add parameters that don’t do anything
  • Rate in hz or ms?
  • Uavcan uses hz

UTC0141 -

  • Limit forward throttle during transition
  • Consider it a takeoff and reuse throttle takeoff parameters
  • Merged!

UTC0143 -

  • Running out of time
  • Tridge has reviewed / not ready

UTC0145 - running overtime

  • Spending a really long time merging PRs
  • Two calls?
    • One for PRs one for other things?
  • Limit it to ¾ of an hour for merging PRs
  • Try to get more merged without bringing to DevCall
  • Two meetings?
    • Eats up a tonne of time
  • Invert so status updates are done first
    • Current structure so European PRs can get merged
  • contractors are onboard
  • Clocking hours

    • Get some of the ones tagged DevCall to be reviewed?
  • Copter and Plane updates first?

    • There were complaints that they just wanted their updates looked at
    • Try to limit the time to 5 minutes/vehicle?

UTC0152 - Copter update

  • 4.0.4rc2 out soon
  • dataflash issue is a potential in-flight crash
    • Don’t turn logging on while flying with flash-based logging
    • long-standing
  • CubeOrange serial bandwidth issue is also long-standing
  • Rc later today

UTC0154 - Plane update

  • New beta about same time as Copter one

UTC0156 - new log-based viewer coming

  • Open-source alternative to
  • Putting it on ArduPilot servers
    • Chrome only for the time being
      • Fix coming

UTC0155 - close

1 Like