SAS for unmanned probe not targeting orbital vectors

[MadJackal]

I recently launched a ship carrying multiple unmanned probes. The main ship is able to target orbital vectors without issue, but the probes will not do this once separated. Here is a screenshot to give some idea of what's happening.

Here I have selected the prograde vector in the SAS menu. The probe has power, RCS thrusters and monopropellant, but will not face prograde without manual repositioning. Instead it "jiggles" slightly in what can best be described as a less fuel efficient alternative to stability assist. I have not found a solution to this so far and any help would be greatly appreciated.

[Mikki]

1 hour ago, MadJackal said:

I recently launched a ship carrying multiple unmanned probes. The main ship is able to target orbital vectors without issue, but the probes will not do this once separated. Here is a screenshot to give some idea of what's happening.

...
Here I have selected the prograde vector in the SAS menu. The probe has power, RCS thrusters and monopropellant, but will not face prograde without manual repositioning. Instead it "jiggles" slightly in what can best be described as a less fuel efficient alternative to stability assist. I have not found a solution to this so far and any help would be greatly appreciated.

@MadJackal

IIRC this probecore lacks a inbuildt Gyroscope, attach the smallest SAS module and it should work.

Edited December 24, 2015 by Mikki

[Plusck]

7 minutes ago, Mikki said:

@MadJackal

IIRC this probecore lacks a inbuildt Gyroscope, attach the smallest SAS module and it should work.

Surely the core can just use RCS without needing a gyroscope, no?

 

However, I see the RCS thrusters are clipped into the solar panels. Could that be the source of the problem?

[swjr-swis]

2 hours ago, Plusck said:

Surely the core can just use RCS without needing a gyroscope, no?

 

However, I see the RCS thrusters are clipped into the solar panels. Could that be the source of the problem?

I've never had trouble with clipped RCS units or panels.

The probe core you've used is not visible in that picture. It must be clipped. Is it under the klaw? The klaw is notorious for having a very obnoxious collission mesh that will make it jitter like a volcano about to burst, enough to easily overwhelm a small probe's attitude control.

If the Klaw is clipped and jumpy: it's possible at times to 'settle' the jittery klaw by arming or disarming. Once it's settled the probe should be easier to control.

Edited December 24, 2015 by swjr-swis

[MadJackal]

6 hours ago, swjr-swis said:

I've never had trouble with clipped RCS units or panels.

The probe core you've used is not visible in that picture. It must be clipped. Is it under the klaw? The klaw is notorious for having a very obnoxious collission mesh that will make it jitter like a volcano about to burst, enough to easily overwhelm a small probe's attitude control.

If the Klaw is clipped and jumpy: it's possible at times to 'settle' the jittery klaw by arming or disarming. Once it's settled the probe should be easier to control.

Here is a better look at the attachment of the probe core.

Klaw was attached directly to the green "joint" and was not offset afterwards. There is no visible jitter except from RCS when I try to target an orbital vector.

 

9 hours ago, Plusck said:

Surely the core can just use RCS without needing a gyroscope, no?

 

However, I see the RCS thrusters are clipped into the solar panels. Could that be the source of the problem?

The clipped RCS thrusters do not seem to be a problem when I am maneuvering manually. I also don't think that would affect the non-clipped thrusters.

 

EDIT: I also just remembered that the main ship has an avionics hub attached.

Is it possible that the game is confused by the sudden absence of this part?

Edited December 24, 2015 by MadJackal
New information

[Mikki]

On 24.12.2015 at 2:37 PM, Mikki said:

@MadJackal

IIRC this probecore lacks a inbuildt Gyroscope, attach the smallest SAS module and it should work.

The answer is this, it lacks a gyroscope... 

[MadJackal]

14 hours ago, Mikki said:

The answer is this, it lacks a gyroscope... 

Launched again with a SAS module on the probe, this was a partial fix. The probe now rotates toward vectors but it still wastes monopropellant with jittering and stops about 10-15 degrees short full of alignment.

 

EDIT: Relaunched a 3rd time without avionics hub on main ship, problems still persist.

Edited December 26, 2015 by MadJackal

[swjr-swis]

2 hours ago, MadJackal said:

Launched again with a SAS module on the probe, this was a partial fix. The probe now rotates toward vectors but it still wastes monopropellant with jittering and stops about 10-15 degrees short full of alignment.

 

EDIT: Relaunched a 3rd time without avionics hub on main ship, problems still persist.

A few more things to try:

• The SAS reaction wheel you added should really be all you need for attitude control. When all you need is rotate or steady the probe, try leaving RCS off. Enable RCS only when you need to translate/thrust along the axes.
• Even then, before engaging RCS, hit CAPS LOCK. This will toggle 'fine control', which means the RCS units will thrust at only 10% of the normal. This will diminish the jittering caused by the RCS and save monoprop. (While on fine control, hold the RCS keys a bit longer whenever you need 100% thrust)
• In the VAB, check the CoM and the RCS unit placement. Especially with a single ring of RCS units, you need to place them spot on to the CoM, or it will add a torque that makes RCS have trouble to keep control and direction.

[RexKramer]

On December 24, 2015 at 5:57 AM, MadJackal said:

I recently launched a ship carrying multiple unmanned probes. The main ship is able to target orbital vectors without issue, but the probes will not do this once separated. Here is a screenshot to give some idea of what's happening.

Here I have selected the prograde vector in the SAS menu. The probe has power, RCS thrusters and monopropellant, but will not face prograde without manual repositioning. Instead it "jiggles" slightly in what can best be described as a less fuel efficient alternative to stability assist. I have not found a solution to this so far and any help would be greatly appreciated.

There is a known issue, which results in jitteriness and craft pointing not exactly at the selected vector.  It seems to affect small light vessels the most, especially ships which may have more SAS or RCS power than they really need.  

To the OP, I don't think you are making any errors in the construction of your probes- the issue is not limited to specific parts, or clipping, or even the Klaw.  Here's a couple of possible workarounds I've used-

- Reduce the amount of rotational force relative to the mass of the vessel.  For SAS, use the probe cores which provide the least SAS torque, as this is not tweak able in stock (yet).  I don't use RCS for rotation, but if it is possible to reduce the max thrust of RCS ports (like you can with engines), that may help.

- Conversely, increase the mass of your probe.  This isn't a great suggestion, you worked hard conserving mass after all.

- Don't use the vector hold modes, instead just use the basic SAS attitude hold.  The basic SAS mode doesn't seem to be affected by the issue.

Finally, I don't use mods at all.  I'm curious if other users can comment if mods exist which reduce or eliminate the problem.  I'm thinking even MechJeb might do the trick.

For your exact probe, here's what I would try: Get rid of the SAS unit you added.  Replace the small probe core with one of the slightly thicker ones, which have minimal SAS force, and use SAS for alignment, not the RCS.

[MadJackal]

10 hours ago, swjr-swis said:

A few more things to try:

• The SAS reaction wheel you added should really be all you need for attitude control. When all you need is rotate or steady the probe, try leaving RCS off. Enable RCS only when you need to translate/thrust along the axes.
• Even then, before engaging RCS, hit CAPS LOCK. This will toggle 'fine control', which means the RCS units will thrust at only 10% of the normal. This will diminish the jittering caused by the RCS and save monoprop. (While on fine control, hold the RCS keys a bit longer whenever you need 100% thrust)
• In the VAB, check the CoM and the RCS unit placement. Especially with a single ring of RCS units, you need to place them spot on to the CoM, or it will add a torque that makes RCS have trouble to keep control and direction.

Thanks for the suggestions; in regard to RCS placement, they were already dead-on center of mass for the probe at 50% monopropellant. I tried your other two suggestions, fine control seems to have no effect on RCS use, and using only gyro rotation eliminates the jitter but worsens the probe's ability to target vectors correctly. Also going back to your idea about the klaw I relaunched with it unattached and that had no effect.

[MadJackal]

7 hours ago, RexKramer said:

There is a known issue, which results in jitteriness and craft pointing not exactly at the selected vector.  It seems to affect small light vessels the most, especially ships which may have more SAS or RCS power than they really need.  

To the OP, I don't think you are making any errors in the construction of your probes- the issue is not limited to specific parts, or clipping, or even the Klaw.  Here's a couple of possible workarounds I've used-

- Reduce the amount of rotational horse relative to the mass of the vessel.  For SAS, use the probe cores which provide the least SAS torque, as this is not tweak able in stock (yet).  I don't use RCS for rotation, but if it is possible to reduce the max thrust of RCS ports (like you can with engines), that may help.

- Conversely, increase the mass of your probe.  This isn't a great suggestion, you worked hard conserving mass after all.

- Don't use the vector hold modes, instead just use the basic SAS attitude hold.  The basic SAS mode doesn't seem to be affected by the issue.

Finally, I don't use mods at all.  I'm curious if other users can comment if mods exist which reduce or eliminate the problem.  I'm thinking even MechJeb might do the trick.

For your exact probe, here's what I would try: Get rid of the SAS unit you added.  Replace the small probe core with one of the slightly thicker ones, which have minimal SAS horse, and use SAS for alignment, not the RCS.

This sounds almost exactly like what I have been experiencing. Good to know it is a known issue but I'm a little bummed that it is not fixable in vanilla. I would likewise be interested to know if any mods fix this

[RexKramer]

19 hours ago, MadJackal said:

This sounds almost exactly like what I have been experiencing. Good to know it is a known issue but I'm a little bummed that it is not fixable in vanilla. I would likewise be interested to know if any mods fix this

Unfortunately, the issue has been around for a while (at least 0.90, probably earlier).  The process of the autopilot (SAS or ASAS) orienting vessels generally works fairly well in KSP.  It turns out the process of determining which inputs are necessary to achieve the desired orientation is somewhat complicated.  Google Proportional Intergral Derivitave (PID), that's the mathematical basis for autopilot controllers.

In the real world, controllers are optimized for specific applications, for example different aircraft have differing efficiency in responding to various control inputs.  I'm not sure how exactly KSP handles this, my guess is a one size fits all controller is used.  Some errors which players have noticed with the stock KSP autopilot include:

- The craft jitteriness and inability to track accurately with light ships that you have noticed.  I've mentioned a few workarounds earlier in this thread.  Additionally, you may notice a higher resource drain (electric or RCS) when this is occurring.

- Overshooting the desired attitude.  This happens with heavier ships as well, in fact it is probably more annoying with heavier ships than lighter ships.  Ideally, an autopilot would overshoot little if any.

- Ships with more rotational torque in one axis than another produce annoying results.  Two easy ways of encountering this are have more RCS on one axis than another (pitch vs yaw), and more SAS in one axis than the other, which is easy to see with the Mk2 cockpit (which has a lot of pitch torque, but less yaw torque).  This one can cause some really bizarre hunting around for the right attitude.

Until the jitteriness issue is resolved, the best solution is to use the least amount of SAS (or RCS) you can on light probes. Unfortunately, for extremely light probes (and yours looks extremely light indeed), even the smallest SAS unit, found in the basic HECS and OCTO cores will still cause ships to spas out.  The only solution I have found in stock is to just use the basic Attitude Hold, not the vector hold modes.  You can still use the vector modes to initially spin your ship to the correct vector, but then switch back to the basic mode to avoid the mentioned problems.

The issue is known to the developers.  Creating a one size fits all controller for KSP is difficult, the range of possible mass, rotational inertia, and rotational force combinations is pretty large.  Those of us who played KSP before the handy vector hold modes were introduced are still pretty thankful we have them at all.  Good luck!

Edited December 27, 2015 by RexKramer

[MadJackal]

4 hours ago, RexKramer said:

Those of us who played KSP before the handy vector hold modes were introduced are still pretty thankful we have them at all.  Good luck!

Yeah I've been playing on and off since around .21, I remember thinking the orbital vectors were a godsend. I guess the only reason I haven't had this problem before is because I usually like to build things big.

Edited December 28, 2015 by MadJackal

[swjr-swis]

On 12/27/2015 at 3:01 AM, MadJackal said:

I would likewise be interested to know if any mods fix this

MechJeb comes with a SmartA.S.S. module which attempts to do SAS better. Additionally, it is tweakable in a few ways the stock SAS modules are not. I am not sure if it helps in this distinct case, but it's worth a try for its myriad of tools and info screens (and even when installed, it's all optional to use or to ignore, depending on how much assistance you want/need).

Speaking of tweakables: Tweakable Everything is also one to consider, because it adds tweakables to stock that allow to adjust how much torque force you wish reaction wheels to use, I think on a 0-100% scale. Might help too.