Replies to This Discussion

DeveloperPermalink Reply by Randy on July 19, 2012 at 5:27pm

Jeff,

     Rate D will be a input based upon the acceleration of the yaw.  So the higher the acceleration, the higher the counter acting force from the engines will be.  In theory a very good thing but because there is noise in the sensors, the acceleration values can jump around a lot and that could cause you troubles.

     By the way, you should be able to turn on logging of the PID values by going into the CLI, "logs", "enable PID" and then when you're tuning one of the yaw parameters with CH6, you'll find at the same time that the PID values are logged to the dataflash.

     One last thing, 2.6.1 has a much improved yaw controller (implemented by Jason and Robert Lefebvre).

• ▶ Reply

Permalink Reply by Jeff Smith on July 19, 2012 at 9:34pm

Awesome, thank you for the quick reply. I read up some on PID loops and tuning but for the most part everything I found eventually lead to "unless you buy our tuning software you are guessing" LOL. I do think I get the theory of the PID reasonably well but would lack in trying to explain it to someone else. I'm not sure it is as one to one with what is happening in APM as one would suppose.

I have the tail fairly happy with low P values, found that the best thing was to first zero out all the yaw settings and get the mechanical setup of the tail to keep the heli from rotating aggressively, then tune Rate Yaw P, then Stab Yaw P.

I still can't see any difference with the "I" values though,  zero them out all seems fine, I run them up high, all seems fine...    Can you shed light on the "I" term considering a traditional heli?

I will update the firmware as soon as it's up on the Wiki site... (gotta tease with an update that isn't here yet :-)

• ▶ Reply

DeveloperPermalink Reply by R_Lefebvre on July 20, 2012 at 6:38am

Yes, you can live with all the I's being zero, but it will cause problems with the old yaw controller, particularly when you are trying to do a slow, precise yaw movement.  Sometimes it will go the wrong way. And yes, it required very precise rudder servo setup.

The new controller completely fixes all that.  Still, you should always use some Yaw_Rate_I, I recommend a number around 0.010 to 0.040.  Yaw_Stab_I should be zero, and in fact I advocate it's complete removal from the program to avoid confusion.  It just shouldn't be there, as there is no sense to it.

Yaw_Rate_I basically acts exactly like an automatic yaw servo trim.  It trims for zero yaw automatically, and in fact, completely eliminates the need to worry much about your mechanical setup.

Yaw_Rate_P is very much like gyro gain on a traditional heli.

Yaw_Stab_P can also have an effect like gyro gain as well, making matters confusing.  It also controls the maximum yaw rate in a roundabout way.  I think this part of the code needs some improvement.

• ▶ Reply

Permalink Reply by Jeff Smith on July 20, 2012 at 10:00am

I think possibly the reason I was not seeing a change with the "I" term was because I did the mechanical trim on the tail first so there was no additional trim required for proper tail behavior...

If I understand PID loops correctly I should be able to increase the "P" term value and use the "D" term to help avoid an oscillation?

I realy think logging my PID tuning is going to be the key to getting all the loops dialed in. I have a weekend camping with just the wife and heli's so by Monday I'll either be insane or much better at this LoL!

My current goal is to read through the entire 130 page monster traditional heli thread. So far I'm on page 26 and have learned more than the wiki ever taught me :-)

Great thanks be to the pioneers who sacrificed heli's so I could start with a fly able product!!!

• ▶ Reply

DeveloperPermalink Reply by R_Lefebvre on July 20, 2012 at 10:27am

Yeah... I dunno if reading the monster thread is much use anymore.  So much of that is outdated.

Yes, D term can reduce oscillation with higher P terms, but you always have to watch out for gyro noise.  

Well, the reason why I term wasn't needed is because of the outer P term.  Yaw heading, is basically the integration of rate.  So the Yaw_Stab_P acts like an I term on rate.  So what happens is if your servo isn't adjusted, it will rotate a bit, until it has, maybe a 5° error which you don't really notice.  That 5° error will multiply by the StabP, and feed down to the rate PID.  Basically, the StabP will be requesting a persistent yaw rate.  It all balances out so that the persistent yaw rate being commanded (which it's not actually achieving!) is enough to stop the rotation.

But it's super sloppy, particularly with the way the yaw controller worked before.  Many of us would try to yaw slowly to the right, but it would go left instead!  Or it would jump to the left before starting to turn right... it was maddening.

But yes, the only reason you weren't upset was because of the mechanical setup you did, which is good, but not necessary anymore.  You should be able to ballpark-it, using best-practices when setting up, but you should absolutely not have to keep landing and adjusting linkages!

• ▶ Reply

Permalink Reply by Derek Ernewein on July 20, 2012 at 5:27am

I found the Wikipedia article on PID controllers very useful in understanding how they work and what each value does. You can skip the history part but try d the whole thing. This information is important in order to get your multirotor to fly the best that it can and get your moneys worth out of your investment.
http://en.wikipedia.org/wiki/PID_controller

• ▶ Reply

Permalink Reply by Flo on July 26, 2012 at 2:10pm

Hi,

i have problems settiing up my small quad for Yaw.

When flying fast it doesn´t  hold the angle.

Can you give me more details about the parameters.

Is there a difference between Yaw  and Pitch/Roll.
To my experince both Values - Stab and Rate are used in Arco Mode. Is this correct?

What is the differnce between those two P Values?

And what are the I-max values for?

regards

Florian

• ▶ Reply

Permalink Reply by Jeff Smith on July 27, 2012 at 9:05am

Florian, roll is rotation on the Y axis (if you put you hand out and go thumbs up/ thumbs down), pitch is rotation on the X axis (fingers pointing to the ceiling/floor), yaw is rotation on the Z axis (parade wave). Make sense?

Stab yaw (stability) is how fast the APM trys to correct the quad itself (
I.e. turning towards a new waypoint), Rate yaw is when you tell it to turn.

The I term is like a trim that applies itself (I.e. the wind is blowing to the left, without "I" the quad is blown left, then corrects, over and over. With "I" the qual learns to lean right to cancel out the effect of the wind. IMAX is how far you are willing to let it lean over. If the wind stops too much IMAX and you can't correct the lean to the right quick enough to avoid a crash.

• ▶ Reply

Permalink Reply by Flo on July 27, 2012 at 12:31pm

Hi Jeff,

thx for the explanation.

Is Yaw Stab also used in Acro Mode? (i think there is no waypoint or stabalisation...?)

At the moment i have problems flying fast in Acro mode.

Making a U-Turn  - Yaw axis isn´t stable enough after the turn.

Yesterday i tried the new 2.7 Firmware in the backyard. When i give a huge Yaw input from RC the quad (in Acro and Stabalize Mode) didn´t stays leveled. It raises one of the front arms.

Do you have an idea how to correct this?

• ▶ Reply

Permalink Reply by Jeff Smith on July 27, 2012 at 2:01pm

Have you tuned the yaw rate P yet? Sounds like the P is too low or the I is too high. If when you stop turning left the quad keeps turning left (stick centered) the quad continues left I would say its the I value...

I would guess that the raising of one arm in a pirouette would indicate a motor imbalance, try lowering the max trim for that channel.

P.s. you may have better luck with quad specific questions in the multi rotor section, in this section we deal with traditional heli's ;-)

• ▶ Reply

Permalink Reply by Flo on July 30, 2012 at 3:09pm

Hi Jeff,

thx for your help.

I tried the new 2.7 Firmware, lowerd the I term an check CG.

Sometimes it continues left after a quick or long stick input. But it´s getting much better.

 

 

 

• ▶ Reply