Fumes

  • *
  • 20
Negative Tau
« on: February 11, 2017, 03:55:48 pm »
After a Autotune I can see that I have a negative value in Settings/SystemIdentSetting/Tau.
It is −3,77117014

Is that normal?

Also i have those values in Data Objects/SystemIdentState/Noise
Roll 0,86450666
Pitch 1,00162148
Yaw 7,74460125

Is that normal?

The reason I'm asking is because I've upgraded motors and escs to more powerful ones and it (the quad) is really sensitive in terms of PIDs.
On the old setup i was able to swtich batteries for example, (heavier/lighter) without any noticeable change in oscilations.
Now its on its tippytoes all the time. I've also changed from OneShot125 to MultiShot which could change the performance a bit.

The PIDs from the Autotune is significally lower compared to the previous setup also.

Is that normal?

Sorry for all the basic questions but I really want to understand this.
« Last Edit: February 11, 2017, 04:01:42 pm by Fumes »

Re: Negative Tau
« Reply #1 on: February 11, 2017, 09:11:40 pm »
Tau is based on the time delay from sensor / command sent to reaction measured.  The measured delay is 'e' to the power of Tau.

e to the −3,77117014 is 0,023025105 seconds, so it take about 1/43rd second to respond.  That is an average answer and perfectly normal.

Those noise values are very low.  That's good!  It looks like you spend a lot of time to make sure your motor / prop adapter / prop combos are balanced.  :)  Some of my cheap motor / plastic prop quads are as high as 1000 (that is bad) and still fly OK.

Fumes

  • *
  • 20
Re: Negative Tau
« Reply #2 on: February 11, 2017, 11:06:46 pm »
Thank you for taking the time to respond to this.   :)

Then i wonder why my stable PIDs are much lower with the current setup?
Is it okay with lower PIDs as long as it "feels" locked in or should i chase higher values by tuning?
The new build AutoTune produces PIDs that are oscilating a little bit, so by turning down the knob with SmoothQuick just a pinch I came up with these new values.

I have attached pictures of my old and new PIDs as I can't see how to export the numbers in a good way to display them here.

Re: Negative Tau
« Reply #3 on: February 14, 2017, 07:04:42 pm »
Your outer loop values are higher in the new version...

Are the oscillations only in ATtitude mode?  The rest of this post assumes that your problems are with ATtitude mode.

You must fly the tuning procedure with a set of PIDs that do not oscillate.  In a few cases it may be that there is an invisible oscillation that needs to be removed before tuning.  Sluggish PIDs are OK, but twitchy ones are not.  You will not get consistent tunes with PIDs that are too high.

I think there is an issue, especially with fairly powerful machines.  Basically ATitude mode uses Rate mode.  I recall that when flying ATitude mode, the maximum roll rate is artificially limited and the math assumes it is not.  The math assumes that ATitude mode rolls as fast as possible and you move the stick slowly if you want it to roll slowly.  I recall that the fix is to up the "Max Rate Limit (all modes)" to a very high value like say 1500 and retune.  It wouldn't hurt to up it to a lesser value, at least at first, because this is not well tested.

If you want your ATitude mode to  be rate limited, then after "correctly" running AT, disable AT, and manually reduce your Outer Loop PIDs

Post back if that fixes it for you so I can advise others.  :)

Fumes

  • *
  • 20
Re: Negative Tau
« Reply #4 on: February 14, 2017, 07:48:43 pm »
My problems are both with Attitude and Rate but Attitude is worse.
I have now re-run AT with the sugested settings. I have only touched Max Rate Limit (all modes). Everything else in that bank is as it was before. I have also posted the values from that bank (#1). I've also posted the new numbers from Bank #3 which is my default destination for AT.

It did not produce much better values unfortunately. The numbers posted here and in the previous post is the "turned down" values via SmoothQuick.

I have a Sparky 2 FC. I read somewhere that it is sensitive of vibrations. More than the old cc3d for example. Maybe the new motors are disturbing the gyro and gives lower PIDs?

Fumes

  • *
  • 20
Re: Negative Tau
« Reply #5 on: February 14, 2017, 07:57:40 pm »
I forgot to post the PIDs from bank #1

Re: Negative Tau
« Reply #6 on: February 15, 2017, 08:40:28 pm »
If your SystemIdentState.Noise values with new motors (only valid after AT and without powering off) are below say 100 then I think you don't have a vibration problem.

I think that the general rule is to run AT with it the least powerful / heaviest configuration and it will not oscillate in other configurations.

I mainly use Sparky2 FC which has the same gyros etc as Nano (MPU9250).  The CC3D and Revo use MPU6000.  As far as I know, the 9250 doesn't have worse vibration problems and I fly Sparky2 a lot and tested the AT code with it while AT was under development.

I have seen some quads that wouldn't tune well, but I only ported the code from TauLabs/dRonin.  I have seen GyroTau (Stab Expert page) cause oscillations with AT.

The last thing I can suggest is that you fly the tuning with a low set of PIDs.  Whatever you have that works well, multiply all numbers by 0.6 and tune with that if it is flyable.  It will be slow to respond and "not well connected".

Fumes

  • *
  • 20
Re: Negative Tau
« Reply #7 on: February 15, 2017, 09:05:45 pm »
I'll sort it out eventually.  :)
My SystemIdentState.Noise is always between 1-5 units so vibrations are out of the question probably.
To be superclear i'll ask one more time;
Is it okay with the lower PIDs as long as it "feels" locked in or should i always chase higher values by tuning?

I don't have much experience with quads (I only have flown my own zmr250 china clone) and I don't really know how it should "feel" when it's very well tuned.
I have pretty much experience from rc-helis though, and the flybarless models translates pretty well i presume.

Re: Negative Tau
« Reply #8 on: February 16, 2017, 01:11:03 am »
Perfect PIDs react as quickly as it is possible (for a PID controller) to react.  In ATitude mode that means it gets to the commanded bank angle as quickly as possible.  It feels very connected and does exactly what you ask it to.

PIDs that are too high oscillate.

PIDs that are way too low feel sloppy (like driving a car with a lot of loose play in the steering wheel), loose the correct pitch angle in fast forward flight, wobble a lot in descents.

In ATtiude, if you are running less than max "max rate in all modes" then you aren't using perfect PIDs anyway.  It would take a good amount of PID reduction to get past that issue.  Imagine a sports car that has a perfectly tuned suspension.  Now put a shock absorber on the steering wheel so you can't move it quickly.  You would have to detune the suspension a lot to be able to notice a difference with the steering wheel slowed down.

I think where you would first notice the PIDs being too low is in vertical descents wobbling a little more and small pitch changes in fast forward flight doing strange things.  You might be able to notice a 10% PID change there before you could tell from the feel.