Checkout https://westphae.github.io/stratux/ – he has an open-source ahrs that plugs back on the stratux
LeSving wrote:
Yes. MGL has found a way around that one. The (3D) magnetometer is placed typically in the wing tip (as far away from ferrite material and magnetic fields as possible).
My point was more that if you use GPS it is (largely) agnostic to where you install it while if you use magnetometers it is not (e.g. must not be too far out the wing in most planes as there typically is the strobe and its driver with a magnetic field).
That leads to the situation, that a GPS based system is comparatively simple to get certified for a large fleet of types on an AML in a single step. For a magnetometer based system you need evaluate installation spots type by type (and sometimes variant by variant and than certify per type (including specific installation instructions per type).
wleferrand wrote:
Checkout https://westphae.github.io/stratux/ – he has an open-source ahrs that plugs back on the stratux
Interesting 
(I doubt I will dive that deep though )
Malibuflyer wrote:
That leads to the situation, that a GPS based system is comparatively simple to get certified for a large fleet of types on an AML in a single step. For a magnetometer based system you need evaluate installation spots type by type (and sometimes variant by variant and than certify per type (including specific installation instructions per type).
That’s probably true. And that means a GPS is the only cost effective way to go for certification as a primary A. I would call that more an extension of the GPS than “true” AI though.
The situation is that a certified AI, for practical/economical reasons, has to be based on GPS as input. This also means there has to be a backup, and that backup cannot be based on GPS as input ? Or it wouldn’t be a true backup ? What are the requirements as backup when the MD302 can be used as a backup?
It’s easier with non certified stuff
@MalibuFlyer on the MD302, do you have the magnetometer? I have it and I had huge drift issues; we’ve disconnected it and it’s now rock solid.
wleferrand wrote:
Checkout https://westphae.github.io/stratux/ – he has an open-source ahrs that plugs back on the stratux
I’m pretty sure that is long since merged into stratus proper – https://github.com/cyoung/stratux
All About AHRS in the stratux wiki
Interesting. They seem to use a GPS in a really primitive way though.
It seems that a long term roll error is fixable easily. If you see say 10 seconds of constant GPS track, you assume wings-level flight, and zero the roll angle. Not right away, but just nudge the roll value towards zero every time you see such sustained periods.
That is a reasonable approach for roll fixing, unless the plane is upside down, but (a) upside down flight as a result of loss of control will probably never produce a straight GPS track 
and (b) types which have an engine installation capable of sustained inverted flight are rare, and anyway you can document that limitation in the POH.
But you can’t do that with the pitch angle because most normal planes cruise at about 2.5-3.0 degrees UP pitch. OTOH I don’t see why there should be short term pitch errors (due to aircraft acceleration/deceleration) if the pitch rate gyro is any good at all. One has to assume you aren’t using totally sh1tty gyros.
Also you can’t fix yaw errors, because one can fly a plane with a heavy foot on the rudder, permanently (and there is no solution to that afaics, short of a very accurate magnetometer and GPS track and TAS etc). I don’t know how the SG102 does this; more than likely any yaw indicated when you stick a boot into one of the pedals will gradually dissipate
They also use accelerometers in the stratux. This will give you the “ball”, and sort out lots of other stuff.
Tell me how a ball in a curved glass tube can show yaw
I wrote “ball”, not yaw
Peter wrote:
Tell me how a ball in a curved glass tube can show yaw
It won’t, but the “ball” (an electronic one) being off centre, with a zero yaw rate, will correctly indicate that the plane is flying a bit sideways.
The basic conditions with a ball are:
ball centred and no heading change – you must be flying straight with wings level.
ball off to the left and no heading change – you must be uncoordinated with too much right rudder.
ball off to the right and no heading change – you must be uncoordinated with too much left rudder.
ball centred and heading change – you must be in a coordinated banked turn.
I suspect this is how the G5 can come up with a good enough roll solution even with the air data and GPS failed as the sensors required for the “electronic ball” can be used as part of the calculation to correct for the drift in its gyros. (The G5’s ball appears to be done purely by sensing the gravity vector).
