LeSving wrote:
The earth magnetic field is a perfectly good and usable field for orientation in 3D. … What you get with a 3D magnetometer is a vector pointing in the same direction relative to the earth (more or less) all the time, unaffected by accelerations and velocities of the aircraft.I have never understood why Garmin uses GPS as (extra, but necessary by the looks of it) input.
If you put a 3D magnetometer in the earth magnetic field you are perfectly right.
For practical flying, however, it is much more useful to put that device into an airplane and therefore in the local magnetic field at the place of the airplane where you put it – and here things get difficult. In an airplane, unfortunately, there are many other sources of magnetism and therefore the local magnetic field can be quite different from the earth field surrounding it.
Therefore if you want to do something based on a magnetometer within an airplane, you need to compensate for the differences between the local field and the earth field. With the common 1D-magnetometers we have in our planes (aka as “Compass”) we do this by a) mounting them as far as possible away from the electrical system and b) having a deviation card with each magnetometer which is updated from time to time. Such a deviation card is useful when you are only interested in the actual direction but gets quite complex if you measure rate of change with it.
Therefore if you certify a magnetometer for other uses, you need to look at individual types individually and there be very specific about the place of installation (typically as far out at the wing as possible but far enough away from the strobes as well).
Garmin took the business decision to use GPS (which is not installation location independent) to go the easy way to certification on a long AML – accepting the discussed downsides of such a solution. If the had gone the magnetometer way, it would still be only certified for few types.
LeSving wrote:
There is the MD302
I have it. Love the device but would not use it as primary in IMC situations.
It shows exactly the issues you would expect from such a device that has not external magnetic or GPS stabilization: It tends to tilt in long periods of non straight and level flights.
When you e.g. fly in a constant descent for let’s say 30 minutes (which is not that uncommon if you e.g. cruise at FL 180, have a friendly controller and descent straight into your destination at a comfortable 500ft/min), the horizon on the MD302 tilts right. In the first device I had that was very significant (at about 3 deg built per minute) so that I had it exchanged and now it’s more in the ballpark of .5 deg/min tilt but still noticeable over long periods of time.
For standby this is perfectly fine (as I know and just would avoid such long descents), for primary it is not.
Really interesting posts above.
As a data point in our aircraft we have a G500. A very good and certified system but eventually I managed to break its algorithm. In several 100 hours it never missed a beat but then we did land at Courchevel and at the end of the steep part I once got quite slow obviously with a massive pitch up attitude. That made the AHRS quit for a few seconds until we did taxi to the flat part of the apron where it did recover fast.
That’s very interesting. I can do that with the SG102 also but much more with one specimen than another one. On a bench test the former one shows a parameter called ImuRate=2 whereas the latter unit shows 50.
When you e.g. fly in a constant descent for let’s say 30 minutes (which is not that uncommon if you e.g. cruise at FL 180, have a friendly controller and descent straight into your destination at a comfortable 500ft/min), the horizon on the MD302 tilts right. In the first device I had that was very significant (at about 3 deg built per minute) so that I had it exchanged and now it’s more in the ballpark of .5 deg/min tilt but still noticeable over long periods of time.
It is really weird that it converts long term pitch into a roll error. No wonder they can’t demonstrate it is any good for primary (IMC, by implication) use.
I also fail to understand the stuff higher above by using an accelerometer to long term stabilise the pitch/roll solution (which is trivially generated from the pitch and roll rate gyros). A coordinated turn is indistinguishable from straight and level and the gravity vector (as sensed by any accelerometer) will also point directly down at the floor during a correctly coordinated turn.
The apparent direction of the gravity vector will still be “down” in a coordinated turn, but its magnitude will increase.
In very round numbers, for a rate one turn:
the angular velocity ‘omega’ = is 3 degs / second = 1/20 radians / second
the aircraft speed ‘V’ might be 100 knots = 50 metres / second
the centripetal acceleration is omega*V = 2.5 m/s/s
gravitational acceleration is 10 m/s/s
The vectors are at right angles and pythagoras tells us the ‘up’ acceleration increases by some 3%.
That could be measurable, unlike the pitch change in a long descent?
PS a right tilt coming from a down pitch sounds like a gyro ‘correction’.
Edited to use Pythagoras correctly :-o
Peter wrote:
A coordinated turn is indistinguishable from straight and level and the gravity vector (as sensed by any accelerometer) will also point directly down at the floor during a correctly coordinated turn.
As written above the gravity vector will increase but in real life that does not work so well. But you can detect the turn by the yaw rate which works better. If anybody is interested I made a comparison between the G500, my software and reality:
Yes I forgot the gravity vector increases if you do a turn at a constant altitude 
but the increase needs to be handled carefully. For example the SG102 I have on the bench (ADXL321J accelerometer) is showing 0.992 (1.02 for a while after starting up) and sure as hell it isn’t flying. So for a lot of maneuvers the delta G is of a similar order to the normal drift. As regards absolute G, according to this it is under-reading my local gravity by roughly 1% although one could deal with that by measuring the gravity at startup.
Very interesting video, Sebastian. The ADL200 looks pretty good. Are you using some integrated AHRS module? For clarity, I am not looking to manufacture any AI product What I might have a go at one day is making a KI256 emulator which uses ARINC429 pitch/roll as the source.
I should do a video comparison of the SG102 (SN3500 reversionary AI mode) and the KI256.
The Garmin G5 manual says that it will give a valid (but less accurate) attitude indication without air data and GPS. It also initializes pretty much instantly, and in a tailwheel plane shows the correct nose up pitch when just turned on, stationary, on the ground, so it must have some sensing of the gravity vector to do that during startup.
The G5 has also been through several software updates; on early versions of the software, the horizon would often be 5 degrees out in roll (although correct in pitch) until the instrument had been running for about 5 minutes – really only noticeable if your time from startup to takeoff was very short. However, all recent (end of 2016?) and later versions of the software haven’t had this problem.
Malibuflyer wrote:
which is not that uncommon if you e.g. cruise at FL 180,
I can assure you, that is no issue with the P2008. With two persons on board, you are lucky if you manage (have the patience) to climb to 6000 feet
Malibuflyer wrote:
In an airplane, unfortunately, there are many other sources of magnetism and therefore the local magnetic field can be quite different from the earth field surrounding it.
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). The magnetometer itself also has a 3D accelerometer, so it is a 3D magnetometer + 3D accelerometer. The accelerometers are used while calibrating. Calibration is done in flight, setting it in calibration mode and do some turns, similar to what is done on mobile phones, the “8”.
I think a lot can be done with sensors and Kalman type processing. At the same time it is funny how complex and fuzzy all this appear to be. A mechanical AI is one single gyro, and it does the job just fine
Cool video ! Could you explain a bit more how different AHRS systems actually work? What is needed, what is not needed etc in terms of sensors?
shows the correct nose up pitch when just turned on, stationary, on the ground, so it must have some sensing of the gravity vector to do that during startup.
Yes; sensing tilt is easy, with a cheap chip. You get tilt by comparing the three G values. If two of them are identical the tilt is 45 degrees, etc…
The trick is using this to long term correct the rate gyros which give you pitch and roll. It looks like there is no magic solution, and the (correct) assumption that the three G vectors must be 0,0,1 in the long run is evidently not quite usable. Perhaps the coordinated turn scenario (which can produce > 1G permanently) is the problem.
I recall a G500 install in a WW2 Mustang and Garmin had to do a modded version, so it could initialise in the high pitch orientation on the ground.
Actually that is the issue my last post was trying to address.
If your yaw rate sensor detects a rate one turn, and your ‘down’ acceleration increases by 3%, you can compute the ‘true down’ and use that instead of (0,0,1) during the turn.
If the two sensors do not correspond the software will have to do ‘something clever’ to correct :-)
