I've been tuning a 1S 3" toothpick with a ~70g AUW on 11000KV 1303 motors.
I eliminated bounceback by removing dynamic damping and wiggling the master slider (ended up at 1.8), and then I tried to reduce bounceback on snap rolls and flips by moving P&I. Moving it below 1 immediately made flying worse so I went up in increments of 0.05. Up until 1.5 the bounceback has been reducing, but it got worse again at 1.55 and worsened more at 1.6.
What else can I do to completely stop the quad from bouncing at the end of snap maneuvers?
#PID tuning - cannot erase bounceback
1 messages · Page 1 of 1 (latest)
@austere radish I was told you're an absolute wizard! When you got time, can you pitch some tips at me? Because I'm at a total loss here
My guess is he's gonna ask for BB log and DVR 
Ein Moment
That's the best I could get it to fly
That's the blackbox, not sure which flights are included because the DVR doesn't capture OSD
This is the tune that gave me the least bounceback
I was fiddling with the P&I gains slider, move it in any direction away from 1.5 and it flies worse
XD 😁
@rose shell I will look into that later (after 21 o'clock CEST)
Appreciated!
Can you make me a BB (and DVR) with stock PIDs and filters?
It's quite unflyable with stock pids and filters >.>
But I'll try
I crashed it three times just because of how uncontrollable it was stock
Uh okay, uhm that's not good
Than please only stock PIDs
What firmware version?
4.3
Stock PIDs are the problem, it only becomes somewhat good when I boost the master slider significantly
The D gain is too small by default
It's 3am for me, I'll be able to record more tomorrow (around midday-ish my time?)
Okay, I can work with that
I will first investigate the current logs
Reason for 0 Feed forward?
Chris Rosser
XD
The reason for my question: if you play with feed forward you can solve unflyable quads
Yeah, I just follow his guide for tuning Betaflight 4.3. It worked for other two toothpicks, this one is a little tougher and it is also the lightest of the three
Because ff bypasses some things in the pid loop, so we can tune D and I and resolve missing reactions or so with ff
I will look into the logs
Maybe I can see stuff already there
~70g, 3". The first one was 3.5", 99g, also flew like shit initially but I could tune it to near-perfection following Rosser's guide, another was 150g and flew near perfectly already with stock PIDs, tuning that one was a breeze
So I guess the default tune does not handle the large-prop-low-weight combination
Tbh I suspect a frame problem (I always do XD )
I can see most of this with a stock bb flight
Sir
That's why I asked XD
My frame is perfect x3
Like frame, motors, props, esc
Looks sexy!
I don't like the colors, but that's preference XD
I could not find that canopy in a different colour, was out of stock everywhere
Print it 🤪
First picture, is a wire near the gyro?
Perfect
Reason i picked the canopy is because it's entirely self-contained
I will come back when I analyzed the logs
(the VTX was flipped upside down since taking that picture, the antenna is now on the other side and away from the FC)
I have tuned three other quads successfully >.>
There are configurations, that are not tunable
I guess
Because of some combined stuff nobody can fix 😦
But this is the cleanest, most solid build I've made so far, so I was surprised a little
Perfect parts, perfect build, together doesn't work
Yeh
Yeah, I will look
Maybe I see a big tell
I guess this involves a lot of back and forth
If everything fails, I'll just bring back FF and see how it goes, otherwise with the current tune it's perfectly flyable unless I max out the sticks with my rates
Oh what rates?
XD
One wild guess, it. And handle that rates
But I can see if that's the case in the logs
Hmm
Alright. I'll be fast asleep in 2 hours so I'll come back and read in the morning
Thank you for looking into it
If you can, try it with max 900
Sure
I had a 3" which had a slight unfixable weight balance problem
This only flew perfect with lower rates
Interesting, good to know. I think you can see it in DVR, if I don't quite max out the sticks it tracks smoothly
Yes I saw
lets do this
@rose shell sadly i must say, this BB is pretty much useless. No rolls, flips or full throttle punches
have you seen this?
orange is feedforward
in this example, you dont have enough I
without any realy logs to look at: increase I a lot and increase D a little
please provide a new BB with flips, rolls and full throttle punches 🙂
filter wise, you can even decrease filters 😄
this is FCKING clean bro
as you already guessed correctly 😄 i dont see a mechanical issue
Going to grab a better log today. Strange because I was erasing it specifically to record for the step response tool, but I guess I got carried away
That's the best visualisation a pilot could hope for
Naturally it rains just as I'm about to leave the house
Could not get the log, the field where I fly is one huge puddle right now. Will put this on hold till next week
Okay 🙂 ping me please, when you have new data
we have another problem
yaw is fine
pitch and roll not
- your P and I is way to high
- your D is too low
2 can be better, when reducing P and I
since its a BIIIIG overshoot AND the line is above 1 => reduce P by quite a margin
reduce I a little
increase D a little
Dampening: 1.1
Tracking: 1.4
Pitch Dampening: 1.1 - 1.15
Master: 1.7
this would be my recommended changes for now
the scale normaly ends at 1.75 which means you overshoot by 75% in deg/s
you are close to 3!
gyro is clean af
d-term is acceptable
but you have some d-term problems (slight)
which my because you have such high P and I terms
Appreciate the input! I'll play with the logs and the step response tool whan I have another field day, the visualisation of how the three terms affect the tracking completed the picture in my head. I'll report the results once done
If you recall the stabilization above 1, it the reason. Your FC fights to match deg/s but never can, because your PIDs are to cranked
This is why I don't love, cranking the masters slider so hard first. You increase everything together
I do it too, than reduce it by 50% so in your case 1.4. than I look at logs and fix each term one by one until its perfect with a little overshoot (1.05 or so)
@austere radish For your perusal! Just got out for another tuning flight, the second arm has all the flips and rolls
yeah but acceptable
So from what I see I need lower I and way more D
P is fine I guess?
Maybe slightly less P
Setting D max to 70 and D to 50
show your current settings please
They're displayed next to the graphs
P 45, I 80, D max 40, D 30
Could use even more D
you should look at the D-Term spectral before doingt hat
this is fine so far (old logs)
P and I are somewhat perfect
go and decrease I a LITTLE! And increase D
like 5
Roll: D = 75, Dmax = 60
Pitch: D = 90, Dmax = 62
i would aim for a little overshoot at the start
general undershoots results in a "laggy" feel
overshoot by max 1.05-1.1 feels realy snappy
Alright, hooking up a new battery, bb in 5 minutes
i will go to sleep now 😄 sorry
but present some Data, always look at d-term noise
and check your motor temps after torture
Alright. I think I can arrive at decent values on my own
And these motors are always cold
increasing D ALWAYS increased motors temps
perfect! 🙂 if not - i am here xD
@austere radish D:
Almost perfect, but now it feels more sluggish and there's some bounceback that feels... almost lazy
I stop the turn and it moves me smoothly back 4 degrees or so over the time of about half a second
Hmm
This is the darker red on the step response tool
Same but no feedforward
More P, less I and even more D?
Adding 2 I to the yaw, otherwise looks perfect
I am on the right track!
It started raining, but before it did I ended up with P 100, I 30, and D in the ballpark of 100, and that flew a lot fucking better than it did last week or at any point today. And motors are still barely warm.
Pretty sure I have achieved the PID tuning zen
I'm dialing it in more tomorrow
Log from the last flight, as I don't know if I'm going to fly again today
This looks almost perfect!
After I tuned via math, I tuned a little just by feeling.. Because I'm the end that only matters
I will look at them log later 👍
I'm heading back right now to try and tighten the P a little more, and get the good FF values
Perfect!
You can play with ff boost as well
This will decrease the ff reaction time, but not the amplitude
Speaking of perfect,
Awesome 🎉🎉🎉🎉
I'll leave the yaw be
I would to
Those PIDs are monstrous though
Yaw is always a problem child
If motors are soundings good and are only warm 🤷
3.5 anti gravity XD I always put that at 10
I actually don't know what it does :D
Try it, it becomes more floaty
Hmm, but I don't really want that
Basically when you let off throttle, it defines how fast the quad falls out of the air
You still can with high anti grav
But yeah, that's preference and different for each quad
I am glad you finally got the PIDs perfect 🎉
Hm?
When the quad is inverted, how does anti-gravity change its behaviour?
Oh, good question
I assume it does it's thing too
Since the FC knows it's inverted
Hmm
But don't quote me on that
If you never played with that, leave it
It's so helpful on 3" quads
Airmode and anti grav helps letting you feel more like a 5" XD with the agility of a 3"
For now, go and fly
Don't get logs, just fly like 10 packs
Yeah
Ah
So, I would say leave it too
Ok, so it flies AMAZINGLY
This looks perfect! Locked in, smooth and jitter free! Not even a little propwash.
As I like Chris rosser so much. I learned tuning from him XD I feel guys methods are too aggressive and lead to stuff that can't be fixed from that point on
So you need to start over ...
I like my method, which just brings the nature of the pid loop to the person. So they understand what they are tuning
That part is missing from rosser imho
His method works well for ~500g 5" quads that are mechanically stable :D
Dunno XD I tune like I told you
And to a lesser degree for builds with a similar vibe
And you not don't need me anymore XD that's the best part
I mean, his method worked for all my toothpicks so far except this one where it just completely fell apart
That one gif, almost everytime gets the click in the person's head
Yep, it's literally the only thing most people would need to understand PIDs