APM 2.5 PIDs - May I be cheeky?

[Hands0n]

May I be ever so impertinent and ask for a bit of cheat-help? 

Below are the currently working PID in my APM 2.5 quad-rotor Firefly 450 build.  I have, so far, only adjusted the Rate Roll and Rate Pitch (locked together) to establish a very stable Stability mode.  Following advice on my other thread on the build I may need to do a bit more work with D and I, but it really is rather locked in right now so I';m going to use that as my current baseline.

So, going forward, my next big problem are the various "auto" modes that very obviously need tuning. For example, Loiter is very erratic, the motors hunting furiously, altitude all over the place. So much so that it actually feels like it is going to bu99er off like my other 450 did one dark night!  I don';t feel that I can trust it to any of its "auto" modes at all right now.

Thinking, then, that I need to do the same with Rate Loiter to smooth that one out.

I have not yet touched the top row of P values - Stabilize Roll, Pitch, Yaw and Loiter PID. I';m not sure what they do, they look like they are duplicating the other values below, but surely cannot be.

The cheeky bit
Could I ask one of you to post up a pic of your current APM 2.5 Arducopter PIDs panel please.  I';d like to get a bearing/reference on what my [almost] out of the box settings are compared to one that has been set up to fly well. 

I';ve read the Wiki thoroughly, but that last fly away has dented my APM confidence a little 

Thanks awfully 

[Tony Campbell]

The problem I think you';ll find mate, is that even if anyone posts their PID';s, they wont help you much at all. PIDS are different for every multirotor. Weight, frame type, motors ( size, kv, and manufacturer), battery voltage and even esc';s all have a huge bearing on the pid settings. What would be handy to know from a forum member who is the dogs danglies with pid tuning, is what you could change to help your particular symtoms. Sorry I can';t help you with actual tuning advice because I';m not that great at it to be honest. :embarrassed, but I';d really like to see what people suggest, because I';m trying to tune althold and loiter myself. One thing they will ask you straight away is, have you eliminated vibration as much as you possibly can using the flight logs. Vibes causes major issues in althold / loiter.

[Hands0n]

Cheers Tony - I take your points completely. And I was actually quite loathe to ask the [noob] question in the first place. I see it asked all over the place, and for all the wrong reasons.  But what I';m trying to do is to see what the typical broad set up is in terms of what someone has set in the other values baseline from their basic Rate Roll and Rate Pitch PID. 

So for example, my Rate Roll and Pitch P is 0.0562 - on that basis I';m interested in what someone else';s Rate Loiter is in comparison to their own Rate Roll and Pitch P.  As an exercise in understanding it may help, or it may be completely useless.  Its just another angle I';m trying to understand the complexities of PID as applied to APM - it is generally much much simpler in other FCs such as MultiWii.

I am quite happy with the Rate Roll and Pitch Roll now. The aircraft is very [acceptably] stable in both Stabilise and Acro modes.   I want to get that same stability in Loiter before I move on to RTL and, ultimately, Auto. 

As I said, I am being a tad cheeky  ::)

[Marty McFly]

its very hard to explain,but i just use gut feelig if its to lose increase pids ,etc to thight reduce,i no not a good exlplanation,also very difficult to have pids good for both stabilised etc and accro so you need to have a happy medium,marty. :scotland:

[dazza22]

I also think you have vibration problems .
I have 2 setup with APM 2.5. I did use Moongel on the APM to reduce vibration
but  have change to  http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=26457  .I use double sided tape on the foam (both sides ) and stick APM to frame ,so far works great !
Anyway hope you sove your issues .

[Hands0n]

Thanks dazza22 but I do have very stable Stabilise and Acro modes with the Rate Roll and Rate Pitch P values.  Completely satisfactory. It is not as if I cannot get it stable, I really have done so [for Stabilise and Acro]. 

The guidance is now to adjust Loiter and that is where I am right now. It is Loiter [only] that is behaving erratically, just like Stabilise was when I first flew with default values.

Here's an example: the default setting for Rate Roll P is 0.14. This is optimized for the standard Jdrones/3DR quadcopter with 850 motors and a 10″ prop. If, however, you're using the bigger 880 motors and a 12″ prop, which are much more powerful, that Rate Roll P will be too aggressive and your quad will seem quite unstable. So in that case, lower your Rate Roll P to 0.1 or 0.09, and you should see much more stable flight.

http://copter.ardupilot.com/wiki/tuning/

So considering that I am using 950 motors and 10" prop I do indeed need to reduce the setting from the default 0.14 which I ended up using 0.0562. And that makes complete sense.

Where I am not entirely settled with is what area of value to set for the Rate Loiter P.  Instinct says to match it to Rate Roll/Pitch to start with, that is 0.0562 also. But I wonder if that is too soft. So I';m trying to see if there is a ratio that I can use as a starting point.

For example - the default settings are:
Rate Roll / Pitch P = 0.1500  I = 0.1000
Rate Loiter P = 0.5000 I = 0.2500

So, thinking in ratio terms, my current (and very satisfactorily working)
Rate Roll / Pitch P = 0.0562 I = 0.0256
Suggests a Rate Loiter of P = 0.1854 I = 0.0640

I suppose I am going to have to try it out for real - but I';m wondering if that is going to be too low for the APM to handle windy conditions (by whatever definition windy is).

What I';m after doing is to understand a working set of ratios for a well flying machine. My intention was to then see if I can extrapolate that into the necessary unique settings for my own model.  Its the ratios I';m hoping to get to, not the actual values themselves.

[HowDoIfly]

I';m certainly not the person to advise on PID';s (still learning myself) but what I do know (from experience) is what Tony Campbell states.

For instance, when I built my F550 (clone) hex I was looking for starting points in PID tuning and following Peter Kings videos, used the Stab_P he finds works well for his hex';s. I nearly had kittens when I tested it and found my hex behaved better tuning the opposite way. So make of that what you will.

[rickp]

So here';s the thing - I';m finding that most people don';t need to touch anything other than the rate PIDs. The only PID I alter on both my reptile and hexa is the RATE_P for pitch and roll. Thats it. THR_MID is around 35% on both of them, so I have to really reduce RATE_P (0.06 on quad, 0.10 on the hexa).
Everything else is standard.
So tell us more about your frame, motors and cell count etc....

[Hands0n]

Thanks Rickp. My Rate_P are 0.0562. The frame is a Firefly 450 (DJI 450 clone), 950KV motors, 10x4.5 props and 3S battery.

I';m confident of no vibration issue simply because of the stability I have in Stabilise and Acro after tuning Rate_P down to its current level that makes logical sense given my particular build.

It is now that I turn to work on the other PID that I am trying to identify whether our not a ratio exists. Again, logically, one should.  You may be lucky that you';ve not had to adjust the other values but all the advice I read from the Arducopter wiki states a sequence of adjustment beyond just Rate_P.

With the current settings, a solid 3D GPS lock and a protected Barometer the Loiter is unsteady in terms of motor speed as it was initially for Stabilise and Acro. So evidently I do need to work on the automatic PID.

Time is not my friend this week so I was really hoping to get a head start before practical work on the quad this weekend. I have a set of calculated estimates to commence with. Its just better to see previous example for analysis purposes, hence my original question here :-)

Sent from my C6603 using Tapatalk 4

[rickp]

I wouldn';t be sure that you have no vibration problems just ';cos stab mode is fine... Run a test with the IMU logging to be sure...

[Tony Campbell]

Although I have my doubts, the thought that there may be a compatible ratio between the stab and loiter settings is an interesting train of thought. I';m interested to see if it works out or not. Oh, by the way....... just because your vibes are low enough to have a decent stabalise, the same may not be true for loiter or alt hold. The proof of the pudding, would be if you could take a snapshot of your logs from  a brief stab flight and let us see. I can empathise with your lack of time, but at the end of the day, 10 minutes doing that, may save you hours in the end. Good luck with it, and be sure to post your findings anyway, because I really am interested to see how your theory pans out

[Hands0n]

I wouldn';t be sure that you have no vibration problems just ';cos stab mode is fine... Run a test with the IMU logging to be sure...

That is on my list to do ...  But my view is based on the observation that with the default PID the quad was very unstable, motors hunting, difficult to hover.  With the PIDs set it is now nailed to the sky     The remaining "auto" PID require adjustment from default - according to absolutely everything I read on setting APMs up - and that hasn';t been done.  Which, to my logic, is absolutely to be expected.    I have zero expectation of the Auto modes to be stable given that they have their own dedicated PID. 

Taking my reasoning to the next stage - my view is that if I had vibration [or resonance] issues then I would not have been able to get the Rate_P to stabilise the aircraft at all. The default behaviours would persist as the ACCs would not have been able to send a coherent level measurement into the APMs PID logic.   You cannot PID-adjust your way out of such a fundamental root cause as vibration/resonance. I';m the living proof of that LOL 

In terms of logging - I will grab a set and post on here as soon as I can.  After my night-time fly-away I am in no disposition to attempt darkness flights, even for short duration while there is a prospect of the APM repeating what its predecessor did    So its looking like Saturday at the earliest.  I';ll take a flight with the settings as they are in the pic in the OP above.  Then after a series of controlled adjustments.

[ericb]

Hi

I think the Craigs ( Jumpy07 ) guide should be of help......lets you understand what they do and how you need to adjust to resolve :

http://www.multi-rotor.co.uk/index.php?topic=1385.0

[Hands0n]

@ericb - thanks, I am using Jump07';s splendid guide coupled with the Arducopter Wiki.    What I am attempting to do in this specific instance is  to explore a bit of analysis over the entire process (its what I do for my beer vouchers).    I believe that, in the case of the APM 2.5';s particular (peculiar?) array of PID settings that there is a synergy that has not been explained anywhere else. Hence my current focus on the potential to discover a "rule of ratio" that may help others.

For sure, I can sit in the field and bang away at the array of PIDs until I get it right.  But that is absolutely not what I am trying to do here  ::)  Had the APM 2.5 been like all the other FCs I';ve worked with thus far I';d not bother - it is the level of complexity that has stirred me 

Of course, it may very well be that I am on a bit of a quest for the holy grail. Is it that such ratios do not truly exist.

So am I barking?      I';m not quite ready to admit defeat yet 

[Marty McFly]

Danny I like the path your takeing,theres so much to the APM,im always tinkering,it flies well now but i still alter things,Marty. :scotland:

[dazza22]

hi again could you post your flight logs they will help find the issue.
cheers

[Hands0n]

I will indeed do that. Earliest will be Saturday. Then I';ll set to adjusting the Loiter PIDs

Sent from my C6603 using Tapatalk 4

[Hands0n]

Blimmin rain! 

So while I wait for a parting of the clouds I thought that I';d re-do my Compassmot and set up IMU logging in anticipation of getting a few minutes out in the back garden to record a log or two of the flight of the Phoenix (geddit?).  The latter is yet to occur, LiPo are charged and ready, I';m pottering around making myself look busy so Mrs T doesn';t get me into anything major that will cause me to miss the event.  We';ve had a couple of interludes in the wet stuff but the LiPo was on charge at the time. Bah!

So, donning my Compassmot gloves (I';m sexy and I know it  )
[attach=1]

I did three runs sequentially, a few moments apart, and have kept with the final run for now.

Run 1
[attach=2]

Run 2
[attach=3]

Run 3
[attach=4]

So, according to the Arducopter website I am in the "grey zone" where the various auto modes (Loiter, RTL and Auto [waypoint mode]) should work for some, not for others       I do have an external Magno board that I picked up a little while back specifically for the APM.  For now I think I';ll fly with these results and see how it goes. Once I have the rest of the rig sorted (only the auto modes now) I';ll get into refining the build.

Check the % of interference displayed.  If it is less than 30% then your compass interference is acceptable and you should see good Loiter, RTL and AUTO performance.  If it is 31% ~ 60% then the interference is in the "grey zone" where it may be ok (some users are fine, some are not).  If it is higher than 60% you should try moving your APM/PX further up and away from the sources of interference or consider purchasing an external compass or GPS+compass module.

http://copter.ardupilot.com/wiki/ac_compasssetupupadvanced/

Now rain ... kindly do one 

[Hands0n]

Finally, the rain eased off enough for a short few minutes of Stabilise flight in the back garden.  So the log file is now attached.  I would be grateful for any feedback on what you are seeing in there that I should be paying attention to in terms of the ACCs (or anything else for that matter). 

[Hands0n]

Any thoughts on those logs? Is it worth much in refining the vibration dampening down?



Sent from my C6603 using Tapatalk 4

[HowDoIfly]

Your vibes look similar to mine which I understand are too high so maybe look to reduce them?

Your pitch and roll tracks very nicely to the inputs though I';m not sure if yaw is good? (Still a noob at reading logs myself). Compass offsets are within spec though.

That';s about all I can tell.

[Hands0n]

@HowDoIfly - Thanks for the second opinion, it is really appreciated    I';m new to reading and making a good interpretation of them, but work with this kind of forensic data weekly. So it is nice to see empirical data relating to the flight.

According to the Arducopter wiki the vibes are in a "grey" area that may work for some and not for others!

The Pitch and Roll P have been adjusted, significantly downwards from default.  None of the others have, yet.   Although I am quite okay with Yaw at the moment, that may change later.

It is dreadfully windy out there today, so continuing the refinement of PID is proving "interesting"   ::) (Altitude Hold and Loiter are next).

I';ve tried inserting some cut down ear plugs for enhanced dampening, but it does not seem to have made any real difference. At least it hasn';t worsened. I may have to re-mount completely.

Edit: Log files for today attached. Very bouncy out there with strong gusty winds.  Trying not to lose another day.

[Marty McFly]

I have my APM  mounted on moongel,also to get loiter to work make sure your multi hovers at center sticl,Marty, :scotland:

[Hands0n]

Thanks Marty - in the still air the quad hovers mid stick.  I need to put some Expo in the throttle to help with vertical control, can';t see how in the Turnigy 9X or APM yet (not a major priority).

I was thinking of doubling up with the mounting pad material until I can get something more absorbent of vibes.  My APM ios in a 3DR Ardupilot case.

[Hands0n]

Hmm, I';ve removed the ear plug material as I think it added to the vibration transfer into the APM. The quad felt better and more controllable in today';s strong winds.

I managed to inadvertently test "Simple Stabilise" mode  when not noticing that I';d knocked a switch.  That gives good confidence in the Compass, pull back on the Pitch and it made a beeline for the point where I was standing when it lifted off.

Still not enough confidence to flip it into Alt Hold mode in the back garden.  If tomorrow is less windy I';ll go find a more open spot to do that, Loiter and RTL (the three remaining modes to set up).

Pic of the final flight today - it was far too windy to attempt much more.  Those massive Z spikes are when the wind snatched the quad out of the air! 

[Marty McFly]

Yes that graph does not look bad,i get spikes when i land and take off,Marty. :scotland:

[Hands0n]

@Marty - I';ve ordered a pot of Moongel from Amazon http://www.amazon.co.uk/Moongel-Damper-Pads-Pot-4/dp/B0002L52BE/ref=sr_1_1?ie=UTF8&qid=1377465507&sr=8-1&keywords=moongel

So, I think I know the stuff - as it refers to those sticky spiders we used to chuck at windows and watch them "walk" down the glass.   But I am now wondering how you actually mount the FC to these. They can only be tacky. So what additional restraint do you use to keep the FC on the moongel and aircraft itself?

[Marty McFly]

What i didid was cut eight square bits,super glued 2 together then super glued a pair on each corner,then supergluded it the frame,Marty, :scotland:

[Hands0n]

Aha, cool. That makes sense. I was not thinking of a "permanent" fixing. But then again, as the APM is in a 3DR APM case it is removable for servicing anyway   

I';m finding some good reading on the vibration characteristics dampening that is needed

[attach=1]

[Hands0n]

Well thats another tick in the box    Today I managed to shoot down my "Ratio Theory".  Which is good - the logic was not verifiable in practise and I';m left with the conclusion that there is no specific ratio to apply between Rate_P Pitch & Roll and the other "auto" PIDs.  That distraction can now be filed away.

Alt Hold
My current nemesis is the Alt Hold upon which much of the rest of the auto modes relies upon entirely.  As soon as I flip into Alt Hold the quad attempts to hold its altitude, the motors then surge and it flies vertically upwards at an alarming rate.  The only recovery is to knock it out of Alt Hold, back into Stabilise or Acro, and bring it back down nearer the ground manually.  The aircraft is completely unable to manage its own altitude.

I am beginning to settle on Vibration as being the root cause behind this.  Yes, yes, I know, you told me so  ::)   But the evidence left sufficient margin in my mind to discount this ahead of any PID adjustment given that the default were not suited to my particular build. Plus that the downward adjustment of the Rate PID for Pitch and Roll were so completely effective.

So today, following the accumulation of various resource I linked Pin 6 into the APM to one of my Turnigy 9X potentiometers and set it to work to (a) re-adjust the Rate PID for Pitch & Roll and then turn attention to the Alt Hold.

Pitch & Roll have been improved even further using this method to literally "dial in" the values and I have now settled on
P = 0.11980 (previously 0.0526)
I = 0.0390 (previously 0.0256)
D = 0-.0030 (previously 0.0040
The quad now flies in Acro and Stabilise completely locked in and controllable like not before.

I';ve been everywhere with the Alt Hold PID, nothing makes any difference whatsoever. As soon as Alt Hold mode is selected the quad tries to compensate and then accelerates vertically until the mode is deselected. It does not stop at any altitude I am comfortable letting it get to!

Next Steps
A tub of Moongel arrived from Amazon late in the day, by which time I was seriously considering digging a hole in the garden and putting the APM out of its (and my) misery        But I was all APM';d out, so it was time to not lose the entire day off that I';d taken. 

Next time I';m on it I';ll add in the Moongel and take some log readings to compare with the previous (earlier in this thread).  The aim will be to get the vibrations as low as I possibly can with the Moongel and maybe a bit of added weight to the APM';s casing to help deaden it';s responsiveness to the vibes.

Watch this space.  I';ll crack this if it kills me