Jittery SAS on small craft

[Claw]

The issue with stock SAS is in a few layers. The stock SAS does use a PID controller. However, it doesn't have a dynamic or manual tuner, which is a problem when one considers the huge range of potential craft that the SAS has to deal with.

For small craft (low moment of inertia) and high control-ability (lots of torque, rcs, and/or flight controls), the stock SAS has a really hard time dealing with keeping up on a moving target.

Also, the constantly moving target causes some inherent issues with the pilot maneuvers (prograde, etc) vs. the "stability assist." So the SAS ends up in a situation where the P constant is too high for small ships with lots of authority, and the target is moving. So the target moves slightly and the SAS overshoots the target right off the bat. (In addition to some slightly different logic.)

As for the single Mk1 capsule can't even point at the markers, there's an issue that's akin to divide by zero. This basically causes the SAS to "blow up" it's calculations, leading to a failure to point the right way.

 

My StockBugFixPlus add-on does include some fixes for the RSAS. It dampens down the PID values a bit and includes a basic dynamic tuner. If you activate the Plus features, it also gives access to a manual tuner. This unfortunately doesn't fix everything though, because there are some other issues with how the stock SAS detects overshoots and how it tries to compensate for high torque.

[KerbonautInTraining]

The best thing to do with big ships is to turn off SAS, slowly turn yourself towards the node, stop your rotation very close to the node, then turn on SAS and switch it to the node you're pointing at.

E: Wait. This topic has "small ships" in the title. 

Edited March 21, 2016 by KerbonautInTraining

[Racescort666]

On March 21, 2016 at 6:51 AM, michelcolman said:

Also, it's not just SAS. Manually controlling a small pod can be next to impossible even with caps lock on. Instead of just reaction wheels on/off, we should have a slider controlling their torque so we can turn it way down if the craft has become much lighter.

I have always assumed that setting up a joystick would help fix this but I've yet to try it. 

[Stoney3K]

Does the stock SAS have some way to measure the amount of control authority that is actually on a craft? If there is a measure for that, the PID loop can be easily tuned to allow consistent SAS action across all ship designs. In principle, the control loop's P value should be inversely proportional to the amount of authority on a craft. (In lay man's terms: Small wheels need a big push on the stick, huge wheels need a tiny push to move the same amount)

Latency is also an issue. Oscillations don't exist unless there is some measure of latency in the SAS control system, which is easiest to notice if you have a craft with both flight control surfaces and reaction wheels on it. The control surfaces are slower than the torque wheels, causing the two to fight each other for control and the craft to wobble along a single axis during atmospheric flight.

[michelcolman]

Even if it doesn't have a direct measurement of control authority, it should be able to detect right away that a control input is havig a huge effect, and reduce its inputs accordingly. But that's not the only problem: even on craft with more limited control input, it still does a very poor job. For example, while facing about 10 or 20 degrees off retrograde and clicking prograde, you would expect the indicator to start moving directly away from the retrograde symbol. Instead, it starts to move in some other direction, often perpendicular to where it needs to go, then starts arcing towards prograde, still accelerating until it passes prograde (not through it, but 10 or 20 degrees to the side of it), then starts slowing down and arcing back to prograde again, usually going through it on the second pass but overshooting it again, coming back and overshooting a third time, etc. It can take minutes to finally stabilize, especially on ships with underpowered reaction wheels. It just keeps accelerating until it's really close to the target and there's no way it can stop in time.

[John FX]

I used to find this a problem and then I started using Persistentrotation and RWSaturatable and kept down the torque on my smaller probes which got rid of 90% of the issue.

Of course then I installed RO which meant I have barely enough torque and that is with reaction wheels that saturate.

It would be good for stock reaction wheels to actually settle when assigned a heading rather than jittering.

A reaction wheel that had a maximum radials/second would be good then whatever size your craft is it would only spin up to a certain speed. It would address many sins.