Autopilot behavior not linear using HDG or NAV

Bug Reporting Hub Aircraft & Systems
, , , , ,
42

About autopilot during approach we can also discuss about dymanic gains that should be available during approach.

About the other topic: The topic should be then more specific: Are we talking about simple airplanes or complex ones? As a rule when I ask for autopilot engage I check what the bank selector is at: Auto, 15, 25?
For now the logic about HDG engagement during a turn is not even clear. Usually when you ENGAGE HDG mode then the airplane will pick up the heading that is shown when bank is less than 3 degrees. It is the whole HDG that needs to be discussed here otherwise it will end up about a patch on a patch.

43

I know some of the planes have a “bank” – reduced bank – button. My favorite is the TBM and it has a max bank (default) of 30 degrees with 15 degrees if the bank button is activated, but that didn’t work until the Dec 22 update.

On the AP, in a recent patch they changed the HGD, ALT, and VS select behavior. You used to be able to set your heading bug to whatever you wanted, then press the HDG button, and when activating the autopilot, it would turn to that heading. Pressing the heading rotary would do HDG sync.

Part of them breaking the AP was making it so that anytime you pressed the HDG bug, it either changed to ROL mode – but didn’t hold the bank angle, it leveled the wings – or it synched the heading and went into HDG mode. So you could no longer do the old “I’m tracking a VOR radial, get a vector from ATC, dial it in, and then hit HDG mode”. Now you have to first hit HDG mode which will synchronize you to the current heading, THEN dial in the vector while in HDG mode is activated.

The applied the same logic to ALT and VS, making IFR flying difficult. To change target altitudes, you first have to hit ALT to sync your altitude and go into ALT mode, THEN dial in the selected altitude. Same with VS – hit the VS button and it will reset the selected VS to zero, then you have to dial in the desired rate.

There is a lot of complaining going on about the flight models (different for each plane but with some commonalities) and the AP, flight planner - GPS - FMS issues, and other stuff.

Asobo is making steady progress on things, but we’ll just have to be patient. Meanwhile, you can still fly old school IFR flights using VORs and non-GPS approach procedures. It’s still a lot of fun with the great weather effects!

44

Instead of grooming part of the backlog based on votes, Aboso should consider basic functionalities. You can go all other the place and touch many modules or then focus on “all” autopilots and then release. To complement what you are saying about HDG, VS and so on, for two men cockpit operation there is a standard phraseology used between monitoring pilot and flying pilot: “Preselect heading 270” (just turn the knob for me to 270 without the mode engaged). “Engage heading”, “Select heading 270”. “Engage heading select 270”.
Anyhow the bank selector makes sense in multi engine operation (OEI SID or Missed approach) or at high altitude.

45

I’m curious about the “dynamic gains that should be available during approach”. What do you mean by that?

And by the way, you actually wrote “dymanic”, which if spelled “dimanic” sounds like a very good word to describe these forums. Manic = alternately euphoric and angry. Di = two or both. Dimanic = both euphoric and angry at the same time. :blush:

46

The ILS is a funnel. When far out, the autopilot aggressively makes corrections to stay on LOC and GS. As you get closer to the antennas you need to reduce the gains of the PID so you do not end up with oscillations. So this is what is called dynamic gains. As consequence, it reduces the amount of bank and large pitch changes during the approach. On the 777 the pitch is also controlled by the LAM (Landing Attitude Modifier) that controls aileron drooping to ensure a stable pitch during the approach. Then it is up to the autopilot sub-modes like FLARE to command nose up, NOSE LET DOWN to land the front wheel, and finally sub-mode roll out. Building a simulation is building a system of systems and it requires a lot of knowledge about all the systems interaction to know exactly what to patch and make it work. For instance, a NOSE LET DOWN feature could be totally hosed by a poor grond effect model and vice versa.

1 Like
47

Ah! I just looked up your profile. I could tell you knew good stuff, but didn’t realize the context! Just this morning I watched Lemione’s YouTube video about the May F35 crash at Eglin. That was all about mistakes in not selecting the correct command mode during the approach, then hitting hard on the main and nose gear, then not understanding the mode the plane would change to, then trying to go around and the plane ignoring his manual inputs.

PIDs! This is a HUGE source of pain to me in the sim, and I started a big discussion over at Avsim on PIDs and the autopilot turn bug Asobo introduced. Discussion of PIDs is on topic in this thread, so we can have at it and most everyone will just tune us out – with a little bit of complaining, probably.

AP TURN ISSUE AND PIDs
So like I said, back in late November, Asobo changed the AP turn behavior (think HDG mode heading change) so that it:

  1. Rolled in proportionally according to the amount of heading change – it took a heading change of more than 90 degrees to achieve anything near 25-30 degrees of bank.

  2. Never stabilized on a constant bank until time for roll-out. Just another way of saying proportional control was in effect.

  3. Proportionally rolled out very, very, very slowly, taking years to go the final 3 degrees or so of heading error, and frequently missing a course / localizer intercept.

SYSTEMS CONTROLLING SYSTEMS – AKA AP CONTROLLING TURNS
Clearly, this was not a problem with the flight model PIDs (or various flight model tables). This had to do with the AP in HDG and NAV mode.

The previous behavior clearly showed that the AP itself was in direct control of the roll in to a max bank, and remaining at that constant bank until time to rollout. The rollout did seem to be PID-based, but only for the last few degrees of heading error. Until then, it seemed to be linear with a constant turn rate. The roll in was a constant acceleration to a max turn rate velocity, which was then held until max bank was achieved. This was the AP system dynamically changing “modes” and choosing between systems and various manual and automated inputs and feedback loops.

After they broke that logic, the behavior clearly was PID all the way, with P and D gains set low, and the I nearly non-existent. I think they did this to try to get rid of the pitch “porpoising” and roll oscillations. In heading, the Integral drives the final few degrees very slowly, but if they increased it, they would get the bad oscillations in roll upon any minor adverse input.

Note that in the sim, there are two parameters that will filter I and D – called “boundary” parameters. For the heading PID, the I boundary is in degrees of heading error, and the D boundary specifies when the Derivative factor starts to take effect (minimum limit). So I is only applied within +/- some degrees of heading error (close to 0 error when P and D go close to 0), and D begins to dampen the oscillations caused by P only above some number of degrees of heading error.

AP PIDs IN THE SYSTEMS.CFG FILE
The AP PID parameters are in the SYSTEMS.CFG file, which is encrypted on all but the base aircraft. Prior to the mid-December patch, the PID parameters had an effect, but after that, the turn behavior is better but the parameters not longer affect anything. The MAX_BANK parameters seem to work now, though, whereas before they didn’t.

So now we again have to await an Asobo fix, and hopefully their opening up more control over the way the autopilot behaves.

Your comments about dynamics open up another can of worms. Elevator effectiveness and lift vector should be proportional to speed. Trim doesn’t work as it should because the aircraft never seems to reach stability in pitch, roll, or yaw – and all of those axes are way too sensitive at speed. When an aircraft is moving quickly through the air, its inertia and aerodynamic design (wing dihedral, etc.) should act as stabilizing forces. They don’t right now.

So there is a lot of work left to do, and right now, we don’t have many ways to “roll our own”. The controls and interfaces are not yet sufficiently implemented or exposed to do much more than get close.

It’s still fun to fly and we can live with it for now. It just makes me even more excited to see what Asobo has up its sleeve with each new update!

48

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.

52

Hello, I’m checking to see if the AP turn rate behavior works more realistically with the latest sim updates - if anyone would like to provide feedback please do. Thank you!

53

This is still very much an issue. Enabling HDG Mode in the 172 and setting the heading bug the plane will bank well past standard rate. As a Real World Pilot with over 1000 hours in 172s that are equipped with a KAP140 it does not behave this way IRL. When you are in heading mode and set a heading it will bank to standard rate.

1 Like
54

Screenshot_1
Screenshot_11667×1127 222 KB

Executing a 180 Degree turn with the KAP 140 in HDG mode should be 1 minute. It is timing at about 45 seconds. Its simply just adding way too much bank which matches both by attitude and turn coordinator.

1 Like
55

I’m incredibly impressed by the simulation of the Garmin suite and autopilot, but this issue is making IFR flying a nightmare.

I use the sim to practice approaches that we don’t often do in real life - like raw data VOR approaches (no GPS assistance). But the Garmin autopilot consistently turns at 25 degrees angle of bank, not a standard rate turn. This is not how the real system works.

Garmin AFC systems use standard rate turn bank angles with a bank angle limit (usually 22 or 25 degrees, depending on the aircraft). Every Garmin AFCS I’ve flown does this, and you will find this standard rate turn and roll command limit in the associated pilot’s guide.

The current fixed bank angle system makes heading mode pretty much unusable under around 160 knots.

This is implemented correctly in X-Plane, so I can’t imagine it would be too difficult to implement in MSFS.

It’s a small change that would make a world of a difference!

56

I have noticed the same thing, and pointed it out on the WTT discord, and have supplied evidence from POHs, Garmin simulators, Xplane, and live flights that the lack of standard rate turns in the autopilot does not model actual behavior.

Feels like they’re still in the denial stage of grief.

1 Like
57

I wonder what’s keeping them from making the change?

58

They think the current behavior is the correct one.

EDIT - they think that standard rate turns are for humans only, and that autopilots should just make sure that they sequence way points properly, and that autopilot behavior varies from equipment to equipment and so modelling correctly is hard.

59

That’s not at all what we said, so hopefully we can clear up any misunderstanding. :slightly_smiling_face:

What we did say is that to the best of our knowledge, per engineering docs, and our references, the standard rate limitation only applies to HDG mode, and the FMS computes turns for GPS/FMS modes using a fixed bank angle. We have not observed in the trainer or in other real world references a sharp decrease in bank angle in those modes to adhere to standard rate under the speed at which it would be necessary. However, we did also say we would double-check our references and investigate the issue.

We also have confirmation from other pilots and documents on different systems (Collins, Honeywell) that they are computed using fixed bank angles only. For example, the PL21 FMS Manual is very explicit about this, and even shows diagrams of various potential wind errors or overshoots. It would be unusual for Garmin to buck the industry for flight path vectorization, as they tend to also use slightly simpler vectorizations than other FMS units. But, it isn’t impossible they’re doing something different there. There’s no regulations that specifically require standard rate for procedures, though there are documents that use standard rate as one of the recommendations for safe area bounds.

We also said that we are aware that HDG mode in the real thing does do standard rate, and that we do not presently have that simulated. It is on our list to simulate, but is low on the priority list as it hasn’t really come up as feedback except one or two times. This would apply to the GFC autopilots, but not necessarily to other autopilots (there are a number that don’t do this).

The message you are referring to was not from a developer but from an active ATP (rated in multiple types) who was drawing from his real world experience with other systems while in FMS/LNAV modes. So, again, maybe Garmin does something different, but it requires much more digging.

Hopefully that helps!

2 Likes
60

Matt, you are correct. I am an A&P specializing in avionics. I currently work in corporate aviation, maintaining Dassault Falcon 900s and Bombardier/Misubishi CRJ-200s, but for many years I worked for an authorized Garmin dealer. I have installed and maintained many GNS-430 and GNS-530 systems on a variety of GA aircraft.

I moved to the corporate world before the advent of the GTN-750.

A Garmin autopilot will indeed use standard rate only for turns in HDG mode, (which it does by monitoring the rate gyro in the turn coordinator), and a fixed bank angle for NAV mode turns.

Only older GA autopilots like the Century III/IV and most STEC systems use standard rate for all turns in both HDG and NAV modes.

The Collins ProLine series, both the older PL4 and the newer PL21, turn using one of two fixed bank angles: 12.5 degrees or 25 degrees. It rolls into and out of turns at a standard roll rate of 3 degrees per second. This suffices for most navigation modes.

However the standard Proline series is not compatible with RNP-AR, as that requires an autopilot capable of using a continuously variable bank angle in turns. In the Collins line, only the PL21 Advanced and Proline Fusion systems have that ability.

Continuously variable bank angles in autopilot turns are a standard feature found only in larger Boeing and Airbus autopilot systems - all of which are capable of adhering to RNP 0.17 limits.

4 Likes
61

Thanks for the clarification both!

As a pilot who has used G1000, G3X, GTN 750/650, GFC5xx and KAP140, I’m pretty sure that in all cases the HDG mode uses standard rate turns. I’m pretty sure that G3X/GTN/GFC also uses standard rate for NAV (and have documented such), but can’t really say for the other combinations.

I agree it doesn’t matter too much for NAV mode, but it does matter quite a bit for HDG mode IMO. Using HDG to fly vectors or enters holds when it banks too sharply causes problems, and I doubt that’s what ATC expects.

2 Likes
62

Hi Matt! What can I do from my side to increase the priority of fixing heading mode on GFC autopilots? It is making flying without LNAV very challenging, which is what I am trying to practice. Thank you!

63

Indeed, this is a real problem and should be elevated to a higher priority and fixed. I’m an ATP corporate pilot and CFI-IAME and recently tried using MSFS to teach one of my students holds recently using the autopilot in heading mode. I was very surprised to see a nearly 30 degree bank angle! Whereas, I should have seen a 3 degree per second turn as done in any GNS430/530 real aircraft I fly in. I hope that you can get this fixed and strive to achieve a level near that of actual aircraft as I’m sure that you audience would expect near perfection as much as possible.

2 Likes