More autopilot questions

[Fishbreath]

I don't have any special knowledge, just a few hundred hours in the Ka-50. :P I'm not sure that what I'm describing is how the autopilot would work in an ideal world, or even how it actually works in the real world. To the best of my knowledge, though, I am describing how it does work in the DCS Ka-50.

 

The goal in hold mode is simply to hold the desired attitude. Your definition of angles relative to the horizon will do.

 

The pilot of BS is never ever supposed to do a slight movement of a cyclic to change its pitch/bank without hitting the trimmer first? I find it hard to believe. To navigate, fly a course. For sure. But eg. for a landing I don't agree. I can land spot on with FD. But still, it's easier to do with moving the cyclic around a centered hover/trimmed point without touching the trimmer.

 

It depends on the nature of the change. Short-term changes are okay to do without the trimmer. When I'm in a hover or trimmed for slow flight for an attack, I'll frequently use the rudder pedals to move the Vikhr targeting cue without using the trimmer. I'll also trim for a near-hover before landing, and work the stick to put it down without trimming again. Some of that is because I don't have a force-feedback stick. If I did, I would use the trimmer even while landing, since with force feedback, hitting the trimmer would just reduce to zero the stick forces required to hold the current stick position.

 

For changing pitch or bank longer-term, yes, I pretty much never move the stick without re-trimming. If I had force feedback, about the only time I'd ever make control inputs without trimming would be for yawing for the Vikhrs.

[Hunden Ynk]

Of course I have tried it. I know how it behaves. And again as I told you before. I'm not asking these questions to learn how to fly. I'm asking them to understand the details. I agree fully that the way pitch/bank hold works is somewhat intuitive. But if it works the way I think it does there shouldn't be a jump in case of the trimmer. And there is. So I'm trying to understand.

Sorry, I just mentioned it to illustrate that this might be (and most likely is, from my point of view) a matter of deliberate design - not a bugged system.

[Fishbreath]

Again. This is the part of your explanation I don't get. Why it doesn't have it while:

1. I added very little stick movement. 3-5%?

2. It has enough authority to counter my movement in Hover mode.

 

So the famous 20% of authority means something different for Hold and Hover/Route?

 

No, but the definition of error does. In hold autopilot mode, the autopilot is holding a pitch, bank, and heading in relation to a horizon an infinite distance away. 'Error' is the difference between the current angles and the set angles. Since 3-5% control input only yields a very small error, the autopilot only makes a very small correction—less than the 3-5% you're inputting.

 

In hover mode, the autopilot isn't holding angles as a primary consequence: it's holding a heading, a speed (zero), and a concrete position over the earth gleaned from the navigational systems. 'Error' means offsets from those things.

 

If you're holding a five-degree bank against the hold autopilot, you're adding a fixed error. Your control input will only cause five degrees of bank. If you're holding a five-degree bank against the hover autopilot, you're adding a constantly-increasing error: every second you move a few meters further away from autopilot's set position. Because the error increases—not the angles, as in hold mode, but the distance from the hover position—the amount of control input the autopilot is programmed to use increases.

 

I already did that experiment. I know what you mean. But still for small cyclic movements I don't understand why there isn't enough authority.

This really could have been implemented in a way that stick movemenet without a trim in such case just moves the reference point and AP stabilizes itself around it.

As I touched on, that would have been a lot more expensive and a lot more complicated, and any flight control system that can fly a helicopter is already expensive and complicated. :P

 

That said, it's definitely a great system when it can be done—the F-16's flight control system works a lot like that. When you stop making control inputs, it stabilizes there. It really takes a lot of the work out of flying. It makes me wonder whether the state of the art in helicopters has moved that direction—it would make the single-seat attack helicopter idea more viable.

 

Edit: I want to further emphasize that I could be entirely wrong about everything I've said on the autopilot making proportional responses, and I wouldn't even be particularly surprised, given how complicated the Ka-50's flight control systems are (or perhaps 'must be'). The manner of operation I've described, however, does seem to me to be consistent with the way our Ka-50 works, so it's at least a useful approximation for someone. :P

Edited September 25, 2013 by Fishbreath

[Havner]

In hover mode, the autopilot isn't holding angles as a primary consequence: it's holding a heading, a speed (zero), and a concrete position over the earth gleaned from the navigational systems. 'Error' means offsets from those things.

 

If you're holding a five-degree bank against the hold autopilot, you're adding a fixed error. Your control input will only cause five degrees of bank. If you're holding a five-degree bank against the hover autopilot, you're adding a constantly-increasing error: every second you move a few meters further away from autopilot's set position. Because the error increases—not the angles, as in hold mode, but the distance from the hover position—the amount of control input the autopilot is programmed to use increases.

 

Sorry. I still completely don't get your explenation. We agree on the definition of bank. We agree on the goal of the AP and yet you seem to take the fact that the hold AP can't reach its goal as normal. Not only with big errors. But also with small ones where it does have an authority. This is bug. Plain and simple. Now, I'm not saying that the AP is buggy. But if your explanation were true it would. Only one explanation comes to my mind. That your reasoning can't be true.

 

As I touched on, that would have been a lot more expensive and a lot more complicated, and any flight control system that can fly a helicopter is already expensive and complicated. :P

 

So difficult that I managed to program it myself in a TM warthog script language :P Seriously. On top of the hold AP.

 

About F-16. It has electronics between the stick and plane control surfaces. Fly by wire is purely software. Don't know what is the case with BS. If it's mechanical then it might be difficult. But I don't believe that the software would be too difficult for kamov engineers.

 

 

Anyway. Thank you for a discussion. I got some other thoughts for explaining my questions. And surely some of them came from a different points of view presented by people here. So definitely I learned something.

 

Hopefully maybe at some point someone else will be kind enough to give some more insight :thumbup:

[Fishbreath]

Sorry. I still completely don't get your explenation. We agree on the definition of bank. We agree on the goal of the AP and yet you seem to take the fact that the hold AP can't reach its goal as normal. Not only with big errors. But also with small ones where it does have an authority. This is bug. Plain and simple. Now, I'm not saying that the AP is buggy. But if your explanation were true it would. Only one explanation comes to my mind. That your reasoning can't be true.

 

I'll try one more time before I give up, just in case I still haven't made myself understood (and it's probably my fault :P). First, read this, which is a handy explanation of three components in typical autopilots.

 

A proportional correction algorithm is, it seems to me, the primary mechanism in the Ka-50's autopilot: in hold mode, the amount of control input the autopilot uses is directly proportional to the amount of error from the reference attitude. The error from the reference attitude does not increase when you're holding a fixed attitude offset from the reference attitude, and so the maximum permitted autopilot inputs also do not increase. (It may ramp up to the maximum permitted autopilot inputs for a given error, but I haven't had much Ka-50 stick time in the last few weeks.) The autopilot only has 20% input, and the pilot has 100% input. It's probably a design feature, not a bug, that the pilot has enough input to override the hold autopilot, both so that the pilot can offset around a trimmed setting (for landing or slight turns for weapons employment) and for safety reasons as Flagrum suggested.

 

The autopilot is still responding proportionally in route or hover mode, but the error is different. When you hold a fixed attitude away from the attitude the autopilot chooses, the error increases, because the reference position is a speed and heading. As you hold the nose down, the speed increases and the offset from the reference position increases. As such, the autopilot's proportional response grows larger. It appears that the autopilot is responding with an integral correction, because you're holding a constant control input and the autopilot control input is increasing, but this is only an illusion. The autopilot is responding to the error, which is the integral of the attitude error your control inputs are producing.

 

Saying 'it's bugged' or 'it's wrong' is getting into saying how the autopilot should be, and that's not productive. I seriously doubt this is a DCS bug, and I haven't seen any evidence to suggest that my explanation for the autopilot is incorrect, whether or not you like it. :P

 

About F-16. It has electronics between the stick and plane control surfaces. Fly by wire is purely software. Don't know what is the case with BS. If it's mechanical then it might be difficult. But I don't believe that the software would be too difficult for kamov engineers.

The Ka-50 isn't fly-by-wire—the flight controls are attached directly to the rotor control servos, and the autopilot is also attached directly to the rotor control servos with authority to move them up to 20%.

 

Actually, come to think of it, that's probably the answer to basically all of your concerns, and the reason why there's no auto-trim mode: the pilot's controls and the autopilot both go directly to the rotors without talking to each other, so the autopilot has no way of distinguishing pilot control inputs from other errors. You can do it with the Warthog software because you're (presumably?) using it to operate the trimmer, which serves to pause the autopilot so it isn't trying to correct pilot errors.

 

Edit: for what it's worth, this has been a useful conversation for me, too. I had an instinctual mental model of the Ka-50's autopilot, but after this I have a formal model I can use to explain it to my Shark-flying friends.

Edited September 26, 2013 by Fishbreath