Sid Refactored to give more space in the DroneCAN array.
Merged!
Andrew Good to backport to 4.6, but via a beta that will get adequate testing.
UTC0710
A It’s unconditionally built, there are no ifdefs. But it’s not getting autodetected.
Would be nice to see a flight log, along with another IMU for comparison.
The clang build fails with errors.
Peter Are we okay with them claiming this IMU for their board? A They provide reasonably good FC boards. P ARK also provide support for this sensor, via a PR. S Why do they integrate with an FPGA? The chip is SPI. A Because it lacks a FIFO. S There are data-ready pins, there are cheaper ways to do this.
UTC0721
A Okay, let’s approve this. R I don’t think it’s such a popular option. It might be broken.
Merged!
UTC0723
Merged!
UTC0724
P It’s very similar to LIS3MDL, but I understand it’s hard to share code with OwnPtr.
Merged!
UTC0727
A hal_bootloader_build is probably a better alternative to declaring a static. P Is AP_FILESYSTEM_FATFS_ENABLED true in bootloader? It shouldn’t be.
We probably need to explicitly disable it in the bootloader hwdef file.
A That will allow us to not use the static keyword.
Merged!
UTC0739
Matt Still a WIP. P It would be nice to add an autotest or two. M It’s a bit hard, because you need the motorboat frame in an autotest to show the bug. Ris_destination_valid may not be always true, can you double-check? P You can use customize_sitl_commandline() to pass custom frames.
UTC0752
P Weirdly for CubeBlack we save almost 1kB! A If a DroneCAN compass works still in CubeBlack, then let’s call it good.
UTC0758
A Let’s double-check that gdb debugging still works on ChibiOS.
Also, let’s fence the asserts behind debug builds.
UTC0800
P Keep in mind that sometimes SITL code runs also in FC boards when doing Sim On Hardware. A An increasing frame_num doesn’t guarantee that we are in lockstep.
You can’t do actual lockstep against RealFlight.
Extrapolating that much (up to 100x) might not be a very good idea. Perhaps we need to account for the RF time jitter directly within EKF_TYPE 10.
Best to consult what we do for extrapolation in X-Plane.
We were also doing this as well for RealFlight in master, but this PR removes this call.
Andy Yes, the idea is to base the extrapolation based on the gyro timings.
Also, the extrapolation is not in the future, but in the past. We’re extrapolating to gradually reach the latest measurement.
A Are these micro-steps done in burst? Andy Yes. A In this case this will probably not work well in slow machines.
At least in my machine pausing in gdb works, and the EKF innovations are good!
Cygwin doesn’t seem to work, the control packets don’t seem to come through.
Unfortunately gdb doesn’t play nice with Cygwin.
UTC0839
P Leonard will probably want to review this. Andy Actually he has! He suggested changes to the API.
UTC0853
P We’ll either have to stop introducing new compasses or drop old ones and their support. The space requirements are getting untenable.