An SR22 with the G1000 and GFC700 does not have the supposedly brushless servos?
YES, it does, sorry. I was not aware of this so i had to consult the maintenance manual (see above) The GFC700 uses the GA80/GA81 servos. The DFC90 uses the existing S-TEC servos, made by Globe.
@Snowbirdxx how did you arrive at your opinion? It seems to contradict what I know about these problems.
It is true that Kalman filters, also known as linear quadratic estimation, are extensively used to solve the navigation problem, i.e. to combine multiple navigation sensors, such as satellite pseudo ranges (GNSS receivers), or VOR, DME, and INS measurements (FMS systems) of possibly very different accuracy / reliability, into a single coherent position and velocity vector. With the exception of very few GNSS receivers that use direct least squares solution, all GNSS receivers and likely all FMS systems use a Kalman filter at their core. This is btw also why many GNSS receivers take ages to arrive at an initial position fix if they are kept in motion. They assume an initial velocity vector of zero, which causes the solution to diverge faster than the sensor input can make it converge.
Now people have certainly tried to apply Kalman filters to the attitude problem. I still don’t see why velocity input should be mandatory. After all, the attitude problem is concerned about your 3 rotational degrees of freedom, and only very indirectly (via vehicle dynamics) to the position vector or its derivatives (velocity, acceleration). But the attitude problem is highly nonlinear. Sure you can linearise around the current state, and this is likely going to work if all you ever do is move around more or less straight and level and with little acceleration, but is almost certain to fall over (topple) if you ever spin or anything like that.
But even if Kalman filters were used, it would be relatively easy to deal with a degrading sensor, after all that’s the whole point of Kalman filters. You can input your (a priori) knowledge about the accuracy of the sensor in the appropriate position of the covariance matrix Q_k, and you get what the Kalman filter thinks about this sensor (a posteriori) out of P_k|k, so you even get a failure indication out of the filter.
It is much more natural to solve the attitude problem in the SO ( 3 ) rotation group. That makes it robust no matter how your attitude changes, and avoids Gimbal Lock
Snowbirdxx wrote:
If that fails, GPS speed could take over, but AFAIK this is not certifiable
Yet this is exactly what Garmin has done in the G1000…
From personal experience: the DFC90 will not “dive down” when losing airspeed indication in cruise flight.
That’s interesting! Are you sure the GFC700 can use the GPS speed? Since they have the wind and the altitude.. they could (can?) even correct that GPS ground speed …?
I think it comes down that you really have to know what these systems are doing. Be it a sudden disconnect or a pitch down to “regain speed”, once you know what can happen you will be prepared.
For me this will result in a) always having Pitot heat on (although i have before in IMC) … and b) handfly with FD in icing conditions …
Boscomantico: it did in my case even that bad that the 2 of us sitting up front hit our head against the ceiling with the airspeed rising sharply. I agree that the DFC90 is a great AP and much, much better and more stable than the STEC55, but …
From personal experience: the DFC90 will not “dive down” when losing airspeed indication in cruise flight.
@ Bosco:
Was that an iced up pitot aswell? Can you explain what happened? I would have expected it to first announce “UNDERSPEED” .. and then, approaching the “stall speed” lower the nose. But in my experience it will gently lower the nose in such conditions. But i have no experience with Airspeed = 0
The wind calculation needs IAS GS HDG TRK so you can’t use wind to correct an unreliable GS.
@aeroplus: I don’t doubt it did what it did in your case. Mine doesn’t.
@flyer59: nothing happens. I tried it once on purpose (i.e. not switch on pitot heat), and a second time I forgot.
From my experience, airspeed indication will not decrease gradually. It will work and then, the next moment, it will be gone.
The wind calculation needs IAS GS HDG TRK
Of course … sorry for beeing stupid.
Is it possible that other parts of the airframe were iced up and that the autotrim brought the plane out of trim and that it started to dive when the A/P disconnected?
But whatever it was (I’ll ask Avidyne later) it shows, again, that you cannot trust any A/P in icing conditions. MUCH bigger planes with autopilots that cost xx times as much were brought down in ice when they were flown on A/P … The main problem ist that you will not realize what the autotrim system does if, let’s say the tail gets iced up …
