This is a discussion re the ArduCopter-3.1 AutoTune feature.  A place for people to report successes and failures, ask questions and get answers.

The AutoTune wiki page is here.

 

Views: 74207

Reply to This

Replies to This Discussion

David, is the the Disco Pro or the Disco?

When you say half what you expected did it at least improve things or did it make it worse?

Ed,

I can't agree with you there, reducing the filtering (if you have very low/no vibrations) does increase the performance, especially in the wind, this is even more evident on large birds (like my 1300mm Octo). Thats kind of like saying turning the gain down on the sonar if you have a clean signal won't reduce performance........bit of blanket response don't you think? The setting is there for a reason.....

What this is doing is "smoothing" out the sensor readings and technically reducing how often "new" sensor readings (or should I say how much variance is allowed in a given time frame) are being fed into the control loop, the lower the number the more "smoothing" applied making the entire control loop slower/less accurate but less likely to react to a sudden "outlier" value (in this case coming from vibrations). If the "hertz" are stated correctly you are taking a reading from the sensors which it is around 100hz if I am not mistaken (in actual fact I think the accelerometers can output a higher hz, but lets not worry about that for now and just work with nice round figures) and then chopping the sensor outputs up and basically taking the "average" (smoothing) reading without out major outliers over 5hz and giving this to the control loop. So instead of taking a reading at 100hz you are actually taking a modified reading at 20hz (1/5th of the sensor data rate). Of course filtering sensor values reduces performance, there is a delay induced from the filtering and less actual value changes a second being made available to be used in calculations (in this example 5 times less data), but it also makes other things substantially more critical, especially vibrations when it comes to the accelerometers. The trade off here is how much of a difference does decreasing filtering make compared to how hard it is to setup a frame to get a "clean" set of values to begin with and how many people will be able to achieve low enough noise (vibrations) to take advantage of less filtering.  

If you remember back to 2.9 when this first came in, default was 42hz but many people had issues due to a lack of adequate vibration dampening to the FC, to fix this issue they could increase the filtering by reducing it to 20hz at the cost of performance or remove/isolate the vibrations from the FC. This then became the default (20hz) in in 3.01 and I would hazard a guess this is due to a bunch of complaints due to users not getting the vibrations low enough to take advantage of less filtering.

In the end I think what you should of said is "for most people decreasing the filtering won't provide much of a tangible increase in performance, it is assumed that most people won't adjust this value and this assumption is applied to auto tuning". 

Randy, what did you guys do, I ran the autotune and my TBS Disco Pro flies like a dream.  Dare I say better than any multicopter I have ever owned and that includes my 2 NAZA-Ms and my APM 2.5

Thanks man, you guys are the best!!!

Below are my PIDS (Flown on a TBS Disco Pro, 4200 mAh 4S)

Raph

This is running a 3s 3000mha Lipo

David

Edgar

This is my own built based on a drawing published by Max Levine back in January.

Before Autotune it felt real solid bit of a oscillation in roll but after Autotune the best way it could be described as soft.

In fast forward flight if you center the stick it would pitch up and down several times before settling into a hover.

David Ardis

Thanks, I went ahead and tried, it have found mine better after auto tune at least with accuracy, but my time was limited as it was getting dark and battery beeping.  I tried a fast forward flight but not a really long one, ill try more after the new year and I can give a better result.

Thanks for the reply, Ed,

I was almost done auto tuning when my quad fell out of the sky! motors just stopped really not sure what happened

I've attached my log , im thinking that looking at it at line 15063 it had finished the tune however it just started to drop and starting at line 15144 i can see ctun is reading my alt is dropping quickly till im on the ground at witch point it detects the crash at the same time i change back to stabilize mode

it all happened so fast im not sure what to think but id really appreciate it someone would be able to give my log a look over and let me know if there's anything there i should be worried about.

Attachments:

Question, are the autotune results repeatable? also am I best to attach my gimbal etc ie have it fully loaded when preforming the test?

After autotune it works like a "Dream Machine"!!!

4S 390KV 17X5.5" 1700gr quad.

Thank you all developers....

Well done!!!

Can someone tell me what to look for in a flight log to figure out why autotune is not working?  I arm the apm2.6 take it to a hove about 20' up and 20' out.  Flip Flight mode to alt-hold and then flip the chan 7 switch set to autotune.  1 our of 5 times the quad actually starts the routine on the roll axis.  When I looked at the log it showed maybe 30 lines for Loop 0 and then Loop 1 and never got to the pitch axis (confirmed by visual during flight and log).  There were no ERR messages in the log. 

Maybe a better way to ask is, where can I see why the quad is not entering autotune and why it is stopping in the middle?  I see MODE messages in the log where I go from stabilize to alt-hold to loiter and back.  Perhaps I don't have the right level of logging enable?  I will time tomorrow to fly it again so I want to make sure that I have the right logs enabled and also that I know what I should be looking for.  I have 915mhz telem radios too but did not have the groundstation turned on.  Should I?

Does Auto Tune adjust the trims?

no ...

Reply to Discussion

RSS

Groups

Season Two of the Trust Time Trial (T3) Contest 
A list of all T3 contests is here. The current round, the Vertical Horizontal one, is here

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service