Are you on Xbox, Steam or Microsoft Store version?
Answer: Steam
Are you using Developer Mode or made changes in it?
Answer: No
Brief description of the issue:
The autopilot logics are still completely wrong. When autopilot is engaged and giving inputs on the roll axis, the aircraft follows the roll input (ignores the AP) while the AP is steering the in-sim yoke fully in opposite direction (without aircraft response). This is obviously completely wrong. The flight control surfaces are directly connected to the yoke and both you and the AP are manipulating the same set of controls.
In real life you can’t steer “through” the autopilot servo’s, you physically CAN’T move the controls when the autopilot is engaged, the force exerted on the controls is sensed and above a certain threshold the autopilot disengages. The autopilot does disengage in MSFS above a certain control input threshold, this is correct. However, below this threshold all manual control input must be ignored.
Furthermore, the autopilot should disengage upon stall warning activation and not try to maintain pitch / altitude / IAS / vertical speed (whatever mode is active), trim full nose up and pull the aircraft into a deep stall or spin. The autopilot should also disengage upon electrical trim input. All of these logics are still missing. The autopilot should not pull the wings off, as can be seen in the video below.
Provide Screenshot(s)/video(s) of the issue encountered:
In this video a slight roll input is given to the left with AP engaged:
Detailed steps to reproduce the issue encountered:
Engage autopilot, give inputs to the roll axis, observe yoke moving in opposite direction to the roll input given while the aircraft follows the manual roll input and ignores autopilot.
@anon50268670 I fully support your efforts to get this explained and sorted, but I am not holding my breath through 2022 !!
I will however add, there are parameters in the AUTOPILOT part of the system.cfg file, that in many planes, are not being used correctly, as developers have blindly followed default Asobo planes, that surprisingly have them set wrong !!
Like a parameter, that if set correctly, DOES make the AP capture the current VS when it is turned on, as required by GA APs like the KAP140.
The issues you mention are system specific. In many (most? all?) light aircraft overpowering the autopilot is possible and is a checklist item before every flight, overpowering the autopilot does not in and of itself disconnect it, and stall warning does not disconnect it (how would you connect a stall warning reed to an autopilot?).
Agree on the visual aspect that the yoke should follow pilot inputs when the autopilot is overpowered. Also agree manual electric trim input should disconnect it (since I said that I bet there is some autopilot in the world where that is not true either), but that one could be tricky to implement since the sim does not distinguish between manual and electric trim inputs.
On small GA planes you could overrule the AP when it remains engaged for whatever fault indeed, still it should disengage when applying any meaningful input, before steering “through” the autopilot. On most aircraft the AP disconnects upon stall warning activation, maybe not the Cessna with its whistle, but aircraft with electronic stall vanes certainly! I don’t remember how it works on the C172 G1000, I’m sure it will disconnect a certain way before stall (stall speed + x kts for example). On aircraft with electrical trim, the AP disconnects when trim input is given. By the way, we don’t have any of the early “very basic” autopilots in the sim, we are talking mainly about G1000 and higher. Those are already quite advanced autopilots (not just a simple wing leveler).
At the very least the deadzones needs to be increased when on autopilot to simulate some “resistance”. Still everything I have flown so far can’t easily be overruled, from basic wing levelers to the most advanced autopilots. On anything bigger than a Cessna you won’t be able to get the controls to move at all below disconnect threshold.
There must be some kind of autopilot “envelope” established determined by control surface deflection, autopilot trim limit, control force limits, g-limit, angle of attack, speed and pitch / bank limits. Any or all should be available to kick the autopilot offline depending on how complex the system is.
Maybe what is really needed here is SDK support for different types of autopilot functionality? Booleans in the autopilot section of systems.cfg would be one way: “disconnect at stall warning”, “disconnect at control input”, and “disconnect at trim input” would be a start. That way we could have realistic behavior in most aircraft. An envelope like you suggest would be even better, but that needs to come with booleans or some other way to ignore certain envelope limits for autopilots that do not monitor them.
I certainly agree that it takes substantial force to override the autopilot and that is a significant discrepancy in the sim. In real life you would definitely feel it if you try to override the autopilot, in the sim it feels just like normal flying, and that is wrong. Your suggestion to have increased dead zones seems like a great one.
My understanding of the autopilot servos in light aircraft is that they have a clutch that breaks out at sufficient control force, but the clutch reengages when that force is removed and the autopilot itself does not disconnect. That is certainly the case for the KAP 140 and the S-TEC 55X. For the G1000 (I haven’t flown it in real life) all I find in manuals is warnings that overriding the pitch servo can result in the pitch trim running full opposite, not that it disconnects. Even the G3000 has that warning. I see the G3000 in the TBM 930 disconnects on stall warning, I cannot find a G1000 installation that does that (really went down the avionics manual rabbit hole now!).
If you say you can disconnect the G1000 autopilot with manual mechanical trim input and you have tried it I believe you. My big concern there is if MSFS were to implement that behavior it would break autopilots like the S-TEC Twenty/Thirty/Forty family, which has a pitch servo but no pitch trim servo and relies on manual mechanical trim input during autopilot operation. That is not an academic issue, since A2A have said the Comanche 250 will be their first MSFS offering, and their Comanche has that type of autopilot.
When you add Force Feedback Joysticks & Yokes , (like the Microsoft Sidewinder II Joystick ) into the equation that adds another whole dimension, and range of possible more accurate simulation.
Every autopilot with a trim axis has that problem, the autopilot does not feel a difference between you pushing / pulling on the controls or the elevator being out of trim. I flew C172 G1000 before, don’t remember a lot from that time. Trim wheel input likely doesn’t disengage the AP, trim rocker switches definitively will. I get where the miscommunication comes from, with manual trim inputs I mean the trim rocker switches, not the trim wheel. My bad…
When I gently steer through the autopilot in roll mode (to make small course adjustments), if I try to straighten out, the autopilot steers me back in the opposite direction (so I am unable to make small course adjustments without manually dialing in the heading).
You shouldn’t be able to steer through the autopilot in the first place. The one and only mode which allows this is called Control Wheel Steering (CWS) or on some other aircraft called Touch Control Steering (TCS). I don’t believe this is implemented in MSFS. When the autopilot is ON you shouldn’t be able to move the controls at all, if you give enough input the AP should disengage though. Either you are in control, or the AP is, but not both at the same time. In ROLL mode the autopilot maintains the current roll not heading, I assume you mean HDG mode.
I did understand much of that from your initial post (if it seems like I didn’t). Thanks for the info about CWS and TCS. I guess that’s what I really need implemented to do what I want.
I didn’t have HDG mode turned on when the autopilot steered back in the opposite direction. When I stopped turning left, the aircraft turned and banked right.
Hey, the 747 autopilot is also broken. Especially the approach mode, the aircraft plunges well below the glideslope (speed is fine btw, I’m not stalling or anything) with crazy rate of descent (-4000ft/min).
I was confused by the above. It makes sense you don’t have to dial the heading when you are not in heading mode . The problem is that if ROLL mode engages during a 30 degree bank turn to the left and you overrule the autopilot and steer to the right, upon letting go of the controls the autopilot still wants to return to this 30 degree bank to the left. In real life you don’t have this problem as the AP will automatically disengage before you even apply sufficient input to actually move the controls. But this is exactly what CWS/TCS is for. Using CWS/TCS the autopilot resynchs and maintains the new bank upon releasing the CWS/TCS button.
I was just as confused with how the autopilot was behaving. I thought ROLL mode was supposed to keep the wings level, but it basically makes me crash instead
Oh believe me, I’ve encountered many unstable situations , but in this case it seems that I was just using the AP incorrectly, not realizing that I wasn’t supposed to be steering-through it. Thank you both for the tips.
Dear @N6722C , I have posted a new KAP140 mod at Releases · FS2020-USER-TESTER/KAP140-MOD-PACKAGE (github.com) . Maybe you saw this in my post in Third Party Aircraft? I would appreciate if you could test this mod and let me know if the trim run-away can be duplicated. My tests so far are made with Optica, 172-Steam, and Twin Otter. I do not own the BN-2. I note your mention of slew mode - this has not yet been tested. My plan is to fix the disply logic and use the instrumentation in this mod to get a list of core autopilot bugs to report to MS team.
. The problem is that if ROLL mode engages during a 30 degree bank turn to the left and you overrule the autopilot and steer to the right, upon letting go of the controls the autopilot still wants to return to this 30 degree bank to the left. In real life you don’t have this problem as the AP will automatically disengage before you even apply sufficient input to actually move the controls. But this is exactly what CWS/TCS is for. Using CWS/TCS the autopilot resynchs and maintains the new bank upon releasing the CWS/TCS button.
, but in this case it seems that I was just using the AP incorrectly, not realizing that I wasn’t supposed to be steering-through it. Thank you both for the tips.