@ortac, I’d be interested to know where your fetish for Kalman filters for the attitude problem comes from, and how exactly you do the linearization.
Peter wrote:
All you can do is the nonsense with the pendulous vanes
Why should this be nonsense? And why should correct startup require you to not move?
As long as the average of the acceleration vector is zero, the mechanism starts up correctly. And the average of the acceleration had better be zero, otherwise your plane would disintegrate quite quickly. Unlike some glass, I’ve never heard a mechanical gyro topple on a rolling takeoff…
Peter wrote:
But with modern avionics there is an opportunity to fix this properly
If by “fix this properly” you mean introduce new and not so uncommon failure modes (like iced up pitot → no attitude information), I’d disagree.
If by “fix this properly” you mean introduce new and not so uncommon failure modes (like iced up pitot → no attitude information), I’d disagree.
I agree it has been done badly.
People get to rely so much on the one big piece of “glass” in front of them that if they suddenly get a big cross across it, they are pretty well lost. My friend had four of these on his Aspen EFD1000, in IMC. The transition to the little back up instruments is going to be hard work.
Peter wrote:
People get to rely so much on the one big piece of “glass” in front of them that if they suddenly get a big cross across it, they are pretty well lost. My friend had four of these on his Aspen EFD1000, in IMC. The transition to the little back up instruments is going to be hard work.
That’s the one advantage with using a G1000 with a KAP140 rather than the integrated autopilot. The autopilot uses a dedicated turn coordinator so if you lose the attitude in the EFIS, it will still keep the aircraft wings level and at the same altitude while you sort things out.
What is a worry in all this is this nasty sequence of events:
pitot heat failure => major loss of avionics and an autopilot failure
It is bad enough with the King KI256-based autopilots where you have
vac pump failure => KI256 failure => autopilot failure
but at least you don’t lose anything else.
Is there any way to have two pitot tubes, with any of these “glass” avionics?
Fetish? If you say so.
The specific mathematical solution is not relevant to the conversation anyway if we are just talking about conceptually how the various information available should be best exploited.
As Peter said, new technology capabilities may have just moved the goal posts in terms of what level of integrity is acceptable.
Personally I think the design in question is fine. It provides a high degree of integrity and a virtual guarantee that if an attitude is being displayed, you can trust it. Unlike old technology where you can never be sure if you should trust an AI when looking at it.
No doubt there have been instances of G1000 attitude failures (red crosses) but I doubt there have ever been G1000 false attitude indications.
And if you get the red crosses, you look at the standby AI. Better than realising that you just followed a vacuum AI into an unusual attitude and now have to use a turn coordinator and the ASI to get out of it.
So where is the issue?
A failure if the AHRS is very, very rare.
I have actually not heard of one in the Cirrus/Avidyne fleet. What DOES fail sometimes is the PFD itself, for example the backlighting. But the AHRS will keep working in that case. Also the (blind mounted) TC can fail., which is needed as a reference. But it’s failure will only produce a CAS message “TC Fail”, but it will not disconnect the A/P.
ortac wrote:
The specific mathematical solution is not relevant to the conversation anyway if we are just talking about conceptually how the various information available should be best exploited.
Well it helps to find out what the background of the poster is…
ortac wrote:
Personally I think the design in question is fine. It provides a high degree of integrity and a virtual guarantee that if an attitude is being displayed, you can trust it. Unlike old technology where you can never be sure if you should trust an AI when looking at it.
That sounds like straight out of a marketing brochure. Would you care to enlighten us why you think it provides “a high degree of integrity” and how high high is?
Flyer59 wrote:
A failure if the AHRS is very, very rare.
How rare is “very very rare”? Of our 4 less than 4 year old C172SP with G1000, the flux gate of one already failed. Granted not AHRS, but another component that “should not fail”.
I think I answered your question in a previous post.
With gyros+accelerometers only, how would you detect a miscompare of data between the two in the event of one of them being faulty? How would you identify the faulty one?
Adding speed and rate of turn data as a third source of information (whether GPS or IAS+magnetic heading) allows you identify discrepancies that you couldn’t otherwise.
It is trivial to think of failure scenarios where a mechanical AI would give false indications. But can you identify a scenario where gyros, accelerometers and GPS/airdata all agree, but are wrong, resulting in a false indication? Ignoring for now aerobatic attitudes and the possibility of software algorithm faults which is a different argument.
Peter wrote:
Is there any way to have two pitot tubes, with any of these “glass” avionics?
Of course. Each air data computer can have it’s own. You flew in a TBM850, that one has two IIRC.
Of course. Each air data computer can have it’s own. You flew in a TBM850, that one has two IIRC.
As does the Mirage, Meridian, Mustang and most high performance G1000 aircraft intended for IFR.
It is trivial to think of failure scenarios where a mechanical AI would give false indications.
One if the FSI simulator emergencies was using the mechanical backups to decide which ADC was providing false information. Initially fly attitude and power then find which one seems wrong while knowing that the left pitot static system feeds the Left ADC and the standby. Same applies to attitude. You have three independent sources. Follow two.
