What a great project. I wonder how the modern sensors stack up against the military version in times of jitter and drift, that might cause some surprises. Larger sensors have a lot of inertial filtering compared to smaller ones.
I also think that the MTBF target the original had will be vastly exceeded by this replica due to the reduction in component count, but it will probably be more susceptible to bitflips. But you won't be flying that high if you put this on a drone. Please post future updates.
You are doing something really interesting here in terms of discovering scaling of such algorithms with respect to vehicle size and sensor response, I'm super interested in what else you will discover. I'm also not aware of anybody doing something similar in the past.
I also think that the MTBF target the original had will be vastly exceeded by this replica due to the reduction in component count, but it will probably be more susceptible to bitflips. But you won't be flying that high if you put this on a drone. Please post future updates.
This is experimental,
We need to address MEMS vs mechanical gyro noise profile
Gain constants in the nav kernel were tuned empirically for the original sensor suite.
Then we got SEU yes :)
noted, though flying at 30m AGL as the Viggen did in low-level attack mode keeps the cosmic ray exposure manageable
Will write a handover doc for Saab, they can do the rest of the dev :)
Yes will post updates
(we run two other drone projects, uas-phone-swarm for iOS and Android
There is a public blog post on it. Patreon/Formerlab
Modern smartphones more powerful than the ck37 tech)
https://www.patreon.com/posts/152446275
Public access
https://www.saab.com/newsroom/stories/2017/december/datasaab...
PDF:
http://www.datasaab.se/Papers/Articles/Viggenck37.pdf
Now we got it on a %50-100 FPGA board.
Half the size of a credit card. I guess the original computer onboard was 12-15 kgs