Micro tuning techniques - different than larger quads


#61

Well its dusk here and the stellar low light performance of the vm275t will be making me wait till tomorrow to test filters off and d gain cut in half. I will say this before anyone gets excited following this too closely … This is not going to produce more than a 1% improvement in what a great tune can do. I just feel like tracking it down … you know - for science. Lol


#62

now please teach me about how to use that thing that makes the pids soften with throttle gain…


#63

I’m going to buy a couple microsd slot breakout boards and glue one to one of my kk100 8.5mm brushed quads. I want to see what they look like black box wise.


#64

Access through the cli… just type dump and you’ll see it in there. anti_gravity_gain = 1

Bardwell says zero in his video but that was just a mistake. He is correct on how it works and I can feel it do its thing when it kicks in. I had to set mine up to 7 to hide my weak motor. (Set anti_gravity_gain = 7 … enter and type save). It’s still not perfect and it only kicks in on a hard punch … so no gentle throttle ups for me on acro recovery - I have to punch it or the weak motor pitches me back.


#65

@NotFastEnuf - In your opinion - Is there anything (setting) in BF that could help me tame the sudden increase in altitude when I hairpin or whip around 180 degrees, or am I just a lousy pilot that can’t control throttle along with the rudder? - hahaha

BTW - I’m really appreciating your input on this subject. I use your (and you seniors’) insight because I have nothing or nobody to compare to, here. Funny the only “quad” you might see here is the Phantom and it’s usually hovering and filming. My quad’s camera faces forward, not down - hehehe


#66

That used to be the yaw D term, which eventually became yaw jump prevention, then disappeared off the configurator in latest versions. I think there is still something in the cli you can set though. I’ll check. It will soften your entry into and out of yaw maneuvers, which accumulates error on the I term and causes a small bounce back correction. It’s a trade off.


#67

its physics.
go fast, lift front edge of wing, quad lifts.
Practice 180* los and you an see whats happening and how to correct it.
I still catch myself popping up on steep fast banks.
Drop that throttle.
keep practicing with your battery time…
“Here, take a :cookie:. I promise, by the time you’re done eating it, you’ll feel right as rain.”


#68

thanks mommy… :slight_smile:


#69

Quad stuck 50ft up in tree, quad recovered. 2 year old around my leg & 5 mo old in my arms. Might cash in for the day and try again tomorrow… feeling like I used up all available good luck.

An old wheel bearing race on a rope - couldn’t have thrown it 1 ft higher. Need to build a potato gun…


#70

So I thought I’d follow up with you guys on the effect of filters - specifically D term filtering. The last few days I’ve been flying with all D term filters completely disabled to force the action of the derivative more in sync with P. The end result to get the craft back in tune after disabling these specific filters was 2 fold. First it did take a smaller D value to smooth out the craft - about 20%. Second, it drastically narrowed the window of acceptable D values. Ie… The tuning window narrowed to less than 5 points where with filtering in place there is closer to a 15 or 20 points range that doesn’t produce bad behavior. Below the sweet spot - P oscillations, and above - larger slower period worbels. So my recommendation is to leave default filtering in place - I did not feel any edge in performance getting the tune locked in without it.


#71

I’d also like to add something about D setpoint weighting and transition values. Many of you may not understand what these do, and how they affect your tuning process. Cause they do… and alot… and if you know what to do with them - they are freaking awesome!

So first let’s translate the official definitions into something we can all understand. If you want to read what the experts have to say - head over to the betaflight wiki. If you want to read what a hack like me has to say… read on

Second, (and later when more time permits) I will talk about what to do while tuning and as a last step to tuning so that your micro is stable in more situations - and feels more the way you like. This feature changes the ways we identified earlier in the thread for what to look for as we adjust values and how we test out tunes. And if you are scratch tuning like we talked about above on default values - you are under tuning.

So definitions in my own words.
D setpoint weight closer to zero softens the craft’s initial response to your stick inputs. Closer to 1 and the fc is gonna push as hard as the pid loop allows to eliminate any error difference between your stick command and where the craft actually is as it begins to respond to your command. This is getting more sharp on initial response to stick commands. Approaching 2 and it’s getting razor’s edge sharp. I personally like .85 which feels “crisp but smooth”… default is 1.

D Setpoint Transition controls how your craft stops moving at the end of a maneuver. …but wait … We’ve been using D coefficients to tame bounce back at the end of a flip or roll - well that’s pretty close to a thing of the past. Default value is .2 and this is an aggressively soft stop to maneuvers hiding bounce back in even the worst of tunes.

So what does this all mean… changing the D coefficient used to be the last step of polishing the feel of a pid tune. Now we can “tune up” the craft to be more stable to external forces like wind and prop wash - and use setpoint weight and transition to “tune down” a more comfortable and butter smooth stick response. If you guys are into this I can rant some more on what I’m doing now that is WAY different than before.


#72

Rant away!


#73

Well I’ve been flying pack after pack just tweaking the tune today. Back story - I had it flying perfect to my stick commands - and taking prop wash pretty well - but a slight breeze reminded me I was just flying a micro. So I wanted to fix that. What is happening is that specifically d setpoint transition at .2 just hides an under tuned quad but is dang near perfect to fly at. So we tune till it feels good to our stick commands in flips and rolls and we stop. The setpoint transition is doing all the work to reign in our under tune to smooth - and a breeze comes along and shows you how bad the tune really is.

Here is what I’m doing that’s different. I’m not flipping or rolling looking for bounce back or sharp crisp feelings anymore - that is being totally controlled by setpoint weight and transition. I’m just flying up high where there is more turbulent air and jacking up P till the big bounces get tighter and tighter and a good P ringing oscillation is just about to stick. Then I do slow decents into my prop wash looking for the same thing. Now I’ve got a craft that doesn’t bounce in the wind but is coming in and out of high frequency vibration as it holds its line much better. Then I crank up D till that goes away. With a setpoint weight of .85 (smooth stick entry) and the setpoint transition of .2 (smooth stick exit) - I go bust a flip or roll or slam into a 180 corner and they are butter smooth starting and stopping. Essentially, I used to tune PIDs to the crafts reaction to my sticks supplemented by expo to smooth it out.
Now… After getting a decent tune to fly on
-I up tune PIDs to the crafts reaction to external forces (easy on the I)
-I tune up antigravity gain to cancel thrust imbalances instead of cranking up I
-Tune setpoint weight and transition to give me the stick feel i want.
-Tune superrates to give me the rotation rate I want
-Tune rcrate and expo to change the shape of the super rates curve

The thinking behind P,I, & D relative to external forces is as follows. D responds the fastest - if you are trying to get small unwanted motions out that move very fast - use D. P is the strongest - if you are trying to get rid of big unwanted motions - use more P (even if it creates some small fast unwanted motion as a result- you can raise D to smooth those ) I responds the slowest - it usually takes a half a second to get into the game. It’s not going to fix big bounces or vibrations - it’s too slow - but it will lock in a wandering line over a distance. It’s got about as much authority as the rudder on a boat in a no wake zone. Just about ineffective for small corrections along the way but still very necessary to get you where you need to go over a distance.


#74

What I want to do next is raise d setpoint weight and transition to 1. This will be reading and calculating from pure error. No stick softening at the beginning end of a manuver. This MAY create a scenario where we can fully use the old way of flipping and rolling to get all the way up to a max tune. It will feel overly mechanical or robotic but should fly with absolutely zero error from sticks or external forces. Then go back and down tune setpoint weight and transition to make it feel butter smooth. This may or may not be a better approach. … I’m gonna try and see.

Side note… ever hit the gas too hard in a big horsepower rear wheel drive muscle car in a corner and have it spin out all the way around… drop setpoint transition down to .1 from .2 … your micro will do the same thing in a turn. Lol its hilarious. In other words … don’t stray far below .2 for transition. LOL


#75

And finally I will say that I previously had to raise antigravity gain to 7 to tame my thrust imbalance. In only the very very hardest of 180 turns where I was pulsing throttle to control altitude climb - this was winding up my I term and giving me one big unwanted jolt down or to the side (about a half a second into the turn - classic I behavior). Since I’ve been able to up tune to external forces - I need much less of this now and higher P gains are doing more of the work that previously multiplying I was doing with antigravity. Next flight will be turning this back down to stop the wind up.

All done rant…

Oh … and my micro doesn’t feel like a micro anymore. He’s got his big boy props on now!


#76

Dude, I’m really enjoying the rant and reading what you are writing. I’m just doing my best to hang on and soak it all in. I like to hear about your adventure in tuning a brushed quad to fly like a brushless micro :slight_smile:

That’s the goal I’ve been trying to obtain since I started tuning, but my approach is far less presice and only deals with the adjusting P I and D. I feel like I am getting lucky when inputting numbers and getting really good tunes…

You on the other hand, are a mechical mastermind of PID tuning quadcopters. Reading your thoughts on the subject is very intriguing and helps open my mind to understanding the bigger picture. Thanks for ranting.


#77

Thanks bro, I appreciate that. You were my inspiration to dive into this, at least I think it was you, when you mentioned something about flying in wind. The next time I flew after that was a little windy and what had previously felt like a great tune - was bouncing all over the place. Well not any more - this is now the best flying quad I have ever owned… and is just a cheapo 8.5mm motor (with horrible thrust imbalance) build but at a good low auw of 55 grams. Plus I went for the 65mm props to up its speed which I usually avoid cause they are so much harder to get flying smooth in prop wash. I wanted this to be a challenge and really evaluate a new procedure. It was a really strange tuning session to just cruise around in the wind and do slow decents into prop wash. I mixed in a few high speed 180’s too for good measure. When i had PID set to manage those conditions, i tweaked the stick feel, and then hit some acro. It was flawless, predictable, and just locked in. I’m loving this thing right now and I have never had such great flights in my front yard flying space as I have tonight. I think I smiled for like 2 hours after. It’s just phenomenal how much more you can do in the air when the tune is solid and predictable and you feel confident you can put it anywhere that you want to. Funny thing, I was a hair away from not even building this as I recently got the brushless bug. I couldn’t be happier that it’s in the air now.


#78

Just for reference… tuning it to feel good to flips, rolls, 180 turns gave me p values in the 80 to 90 range and corresponding d values close to the same knocked most prop wash down pretty good. (I’ve always been a firm believer in high D values for any craft without active braking brushless or of course brushed). This flew great in dead calm air but was garbage in wind. Then I decided to ignore how it felt to my stick inputs and just tune out its behavior to external forces like wind. All the bounce cruising around in wind turned into tight P oscillations around P values of 115 to 130. At 130 for pitch P, it had even almost reined in my thrust imbalance on throttle punches so I stuck to the high side. Then D was upped to the 115 to 120 range to smooth out P oscillation cruising and then up to the 130 to 140 range to knock back the remaining oscillation in the hardest of 180 slide turns.

Insane high values you say… sure maybe. But darn if they aren’t doing the job. I don’t see the affect of wind anymore. And an average pilot like myself was given new found confidence to do things I have NEVER been able to do in my flying space repeatedly pack after pack. Plans with my bro fell through to get my fatsharks back this weekend. I’m really considering priming in an eachine prodvr and hacking it into my hood so I can share this with you guys. Whenever I can dvr again, I will do some side by side comparisons for you with the old tune and the new one just so you can evaluate the improvements for yourself.


#79

I think a kingkong 100/15kv/afunta(/sand… or /a bent shaft.) tuning session is going to happen tomorrow, I like flying them but this sounds really nice! If time consuming.
Once will do both of them though, which is a good start!


#80

Nice. Keep it simple. Start with your tune where it’s at. Assign roll and pitch P values to a three position switch using the adjustments tab. Go up high and find bumpy air. Increase P till you trade a bumpy ride for low magnitude but high frequency P oscillations. Then crank D up next to tame that down. It’ll work. Try my setpoint weight value of .85 and transition of .2 also when you’re done. Let me know what you think of those.