Compassless flight plus GSF - more questions than answers :) state that to enable it, one should disable compass use.

1: does it mean this otherwise useful method, is of no help for bad compass data due to installation error, calibration error, or electromagnetic fields?

2: is it an always-active safeguard ? or only active once ?

3: would it handle or help in this event: Crash during takeoff due to bad yaw reset (/electromagnetic interference) ?

4: which version of ArduPlane was the first to have it ?
5: which version of Copter will have it ? (or is it in master now?)

AFAIK only master has it.

Thank you.
I hope somebody can enlighten me on the details about how it operates.

Sure, here is all the available information:


  1. it is always there as a fallback when using a compass if the mag innovations get too high and exceed the ekf mag gate limit…but that is an “under the hood” operation that the non-advanced user does not need to know about…therefore, its not something I would add to the vehicle wiki…a dev might add it in the developer section…
  2. always there…switches in or out as needed…just like many of the EKF recovery mechanisms…its not a quick thing however…innovations have to be outside limits for a bit for the switch and it takes a while to switch back if mag innovations recover (they usually do not, since its either bad cal or dead compass)
  3. sorry, cant answer that…Randy might
    4/5. notice that this page is not in the normal wiki…its in the UPCOMING FEATURES section…master only stuff…when it will appear in a stable release is unknown…

@hwurzburg this helped a lot to further understand GSF. I still have a query, what happens when GPS is the source for yaw (moving Baseline), none of the EK3_SRCn_YAW is set to GSF and any one of the GPS goes bad.
Will GSF correct the heading?

@tridge would have to answer that… I expect not since GSF uses GPS

it should do, yes, as long as one of the GPS modules can still provide a good position/velocity, but not immediately, it would wait for a timeout. I haven’t tested fallback to GSF when using GPS yaw though, so should be tested in SITL