Your system has worked better for me than any others out there. I have been able to get my quads flying the best they have ever by following your procedures. Only I needed some minor adjustments from the goggles out in the field, but only the pitch gain and yaw pids. Thank You for this great video!
I personally don’t use it but feel free. Once you know what range of damping effect you have on your particular rig, then you’re in a better position to know how best to set D min vs D max. You just have to know what you want to get out of a dynamic D. You could for example set your D max from step response results then set D min to a slightly under dampened state to get a snappier/faster response latency, ideal for racers for example. Or if you want a D boost kind of effect then set D min at the value you obtain from PTB and D max as a D boost. Just depends what you want and why
Should we always enable Y correction even if feedforward is set to 0? Also when I add any sort of I Gain it overshoots and messes up my graphs. Is that normal or do I have to readjust the P to D balance once I set the I gain??
Awesome guide, finally I am able to tune my quads properly. Thanks a lot! One silly question I still have. Tuned an AOS 3.5" and an FlowX Mini 3". After adjusting D Gains and Master Multiplier, the curve looks perfect. Adjusting I gains makes it look worse. Is is a Problem when I ends up on 0? Added 0.2 I gain but it does not look as clean as 0
Typically this would not happen if the error gap is reduced. Usually once FF is brought into the picture it will close this gap more and iterm won’t do that. The conundrum is you can’t examine the step resp with FF so you just have to analyze the raw SP and gyro traces to confirm the correct amount of iterm. There is a vid after this one that describes this. Check those out
I am in the exact same position. My lines looks almost perfect untill i add I Gain and it overshoots. I noticed you adjust Drift Wobble - I Gain before the master slider is that the correct order? I can't lie, i've watched a video from someone else using this method and they said to tune I Gain after Master Slider.
First of all, thanks for all this! Why yaw its not being tuned? I want to tune a cinewhoop for indoor flying i use lot of yaw beacause of low tilt camera (real state) should i worry abojt yaw tunning?
Yes yaw should be tuned. It’s just not part of the BF slider system to treat it as a separate parameter. To optimize yaw requires removing it from the slider control and inputting manual inputs. It’s also tricky to wobble pitch without having the copter jump in altitude so indoors is kinda out of the question.
@@gonzafpv2843 well there’s no need for D on yaw. So you have P and I. So PI balance and PI gain is all you would test. For PI balance I’d manipulate I until you get slow bounce back. Then add PI gain holding the ratio constant until there’s no measurable decrease in latency. Something like that might work
Hi Brian, i noticed a completely different resoult between the basement los flight logs and the effective flight with sharp manouvers, full stick flips and rolls.. in the first one i managed to flat down the curve perfectly but in the actual flight i have massiveovershoot in the step response tool. is it normal? thank you for your time :)
in the actual flight i added FF.. could it be the reason for the overshoot? and if so, in that case shuould i leave it at 0?.. because as i improove it overshoot more and more
@@fefpv5001 yes it's most likely from FF. But also it depends on what you're doing between the full stick deflections. if there's a lot of pauses between a series of moves then the step resp will still tend to extract data that will go into the average, and this data is usually crappy. but it's most likely FF changing things
Really good stuff! I'd love to see a Rosser, Spatz, PIDToolBox tune-out. Each of you tune the exact same builds (three different copters of course) using your methods and then see how similar the values end up being.
i just wanna say BIG THANKS to you. with this video i was able to tune my 3´´ Tadpole to smooth flying and got the oszillations out of the video. it has little to almost no propwash left, but on normal cruise, fast climb and fast forward sinking the horizon looks really stable in goggle dvr. Video quality is better, more details, even almost in the dark. hoovering is also absolutely smooth and stable. i will upload tag my videos so everbody can see the improvement. BF 4.3 is also a big improvement. i fly since 2005, but never had so much fun, tuning was 17 years a PAIN.
Thanks for the good video, but I don't know what I am doing wrong, I have all the settings like you and my graphs are horrible crazy going up and down on my brand new Apex 3" build. Can't read anything out of this...
Hi!I just bought some batteries for my chimera7. Should I pid the quad with the heaviest battery or the lightest? I’m using 6s1p 4000 mah and 6s2p 8000mah, is the difference in weight too high to have only one pid set up?
Very helpful how you take us through this method step by step and show the changes is more then useful . I just don’t understand how your channel isn’t getting way more traffic. Technique method and implementation is just first rate. Many thanks you are my go to from now on. Hope to see more soon. Best bar none. 👊 Keep them coming. Maybe some more black box insight also.
when i tune the master slider i get more and more an overshoot. pd und pi i make it like you (curves loos identicaly) so that i have no overshoot. after masterslider it reach earlier and earlier the setpoint but with massiv overshoot. 7inch heavy ciopter mit 4s2p liion pack
keep FF off, and Iterm low during MM tests. Then add iterm, and finally FF, but do not use step resp to analyze effect of FF. that has to be done from traces directly
I don't have the skill (or guts) to do as sharp a stick movements as you do in this video. My plots therefore look less sharp. Is that still useful data or do I need to have sharper movements?
it's not necessarily why your plots don't look as 'sharp' as you say. but if the plots are unreliable then it could help to do longer tests. if you dont know whether the data is reliable, here's a test. do the same test with same settings twice. compare plots. it they are very different, then your data is unreliable. then do the 2 tests again, but for a longer period of time (more data). then compare. it should get more reliable with more data, and then you can be more confident in the results
Good tutorial! It looks like it can be completely automated by Betaflight itself. Its algorithmically possible to choose best curve from toolbox in realtime I guess. Did you've proposed this idea to BF developers yet? I see it like one take (single lipo) process - like arm, enable "autotune mode", lift off, start shaking your drone in all directions and BF will adjust D gain based on response curve in realtime. Then I and MM, done. Right?
Not quite as simple as you think but certainly doable. There are many edge cases that make it less than simple though. Also this is my project. BF have different interests but we do talk about it once in a while. There’s just so many other things focused on atm
@@PIDtoolbox I understand. Specially after math lab package was downloaded and installed 😀 Thank you for your work! First time I see real math behind PID tuning process and it easy to follow.
Hello, do you think I can do the LOS test flights in angle mode? Or it must be done in acro? I feel more confident doing it in angle if there's no counter indication about it. Thanks
TBH I am not absolutely sure. Some people have done that and it seems to work but I have not confirmed whether it’s correct for reasons I don’t yet understand. I’d say give it a shot and possibly compare to one example in acro and see. I need to do these tests to confirm but haven’t got round to it
Is it right that if you have more filtering applied you can raise PID Values higher by loosing latency. So there is a balance between filtering and PID? So whats your Suggestion for Filter Setup before Basement Tuning? Filter Setup before or after Basement Tuning?
not exactly. More filtering, more latency, and more latency means you might not be able to push PID gains as high. It's counter intuitive. But basically, too much filter latency will limit how high you can push PID gains because it'll generate feedback oscillation and potential flyaway; similarly too little filtering will also limit PID gains because the gyro is too noisy and you'll get hot motors, and potential flyaway because of too much noise going into the pidsum.
@@andreua8779 no, you should not use the step resp to tune FF. All you do is examine the effect on sp-gyro tracking, from the raw traces. You should also examine the quality of the F-term while you’re at it, to be sure your FF settings are working properly
I imagine this process would be very similar using Kiss?. Although Kiss doesn’t use sliders etc manually increasing D with a set P to find the balance and then the same for I would that get things somewhere close? Any advise for Kiss users wanting to use PidToolBox in this way?
@@PIDtoolbox any reports on the logs not loading properly? When I try to trim the logs I get the cross hair first appear to select the start of the trim, then when try to click on the end of the trim it’s like doesn’t let do anything. When trying to open step response files they kinda just want to buffer and says insufficient data? Ever heard of anything similar before?
Thanks for sharing this video! Your process is very methodical and makes sense. Being an Aerospace Engineer I appreciate your well thought out approach. Do you have any recommendations for filter settings before tuning the PID gains?
Thanks so much for putting this together! Never really understood or been good at tuning so having a step by step really helps! I did the angle mode flights and on my roll axis, I get a bit of over shoot with damping at 0.6 and 0.8 but the lines are smooth. As soon as I set damping up to 1.0, 1.2 and 1.4, the line on my roll axis isn’t very smooth, it gets squiggly just before the flat part. Any ideas what might be causing this? Thanks!
Few questions: 1) You start out with a arbitrary P number? It does not matter at what value you start as it gets adjusted later with the master multiplier? 2) What is Y correction? 3) Can you link those PID sliders to a knob on your radio (adjustments) to quickly tune? 4) If you have the opportunity, I guess it doesn't hurt to do actual flights instead of basements hovers, right? 5) Peak time gets worse when latency drops (file 4,5)
Great video, clear and simple. This seems like a great way to tune and I can't wait to try it. Can you link to the video where you describe what moves to put the quad through for the basement tune? I haven't been able to find it on your channel yet.
Took me half a day to tune my thicc, and the HD video is really good, no stab needed ! Big thanks you for your work, the soft and your explication. I think the next step will be a video to understand the filter, but I think filtering is really different between each drone.
Just tried this with a cinelifter and wow it works great! Thanks for everything! Can you use this same method for yaw? Just hover and do a lot of yaw moves to set P and I?
So complete and interesting about new sliders process. Interesting part about PI (remove I first, then adding it) on step response. Thanks a lot, as usual, very informative and makes sense :)
At 2:22 I can see you have tuned the Flywoo Explorer. I am about to tune my Explorer4 (1. gen) and my Rekon5 (6S version) because they should handle the weight of 4S Li-Ion packs. Do you have any suggestions how to tackle these tunings for quads with low power to weight ratio? With the Rekon5 I still had plenty of overshoot at PDbal 2.0. BF sliders need to go further ;) N4V1G4T0R also noticed to tune differently for quads with assymetrical axis, like these dead cat frames. I watched your video about tuning the 10" Li-Ion. But any caveat would be welcome.
One option is to start with PI gains set higher before starting with the damping slider so that you have more room to go when using the MM. that’s probably what I’d do there. I should do a vid basement tuning my ultra lightweight long range quad. Hmm
But what about Dynamic Damping? You just turn it back to 1.0 default? or at 0? Thre's no need in tuning it? Thanks! Super interesting and clear everything!!
I do the same kind of basement tuning and I use Horizon mode, I do a series of sharp roll left and right and pitch forward and backward movements. Do you use Acro mode for your tuning ? The mode shouldn't affect the tuning right ,?
I love that concept of basement tuning. Thank you so much! Just a quick question about disableing FF. Isn't it necessary to set FF Transition to 1 anymore, because the Stick Response slider reaches all the way down to zero in BF 4.3?
No, when the FF slider is put all the way to 0, FF is off. That’s all you need to do the tests. Then when done is set FF back to defaults and adjust jitter and smoothing as desired using the “tooltips” “?” There for info on how best to set this
Hi Brian, If I plan to use dynamic idle, do you suggest to tune it after the pids or before going into tge process you showed? My guess is that since it improves motors responsiveness it should be active when tuning but I'd realky like to hear your thought
latency in the step resp graphs is the latency at 0.5 response (50%). Similar to a median since the curves are kinda like a cumulative freq plot. But yea, a time constant at 63% would also be practical
Very clear and detailed, exactly what i was look for. A question, do you really leave iterm relax on for the tuning? When do you know you need to increase or decrease it?
Great point. I’ve left iterm relax as is for the moment under the assumption that it’s mostly not being triggered with these smaller moves. So the iterm estimate we get here should be decent for moderate stick moves. When you do fast full stick the default relax takes care of any windup. You could certainly turn relax off for these tests if there was some concern that it’s already having some effect on these moves.
At the end of your tune video you did not mention what you would do with the Dynamic Damping? You have it at 0. I set mine at .5 and I liked the way your tune proccess worked for me. Also what about filter tuneing and does this affect your procedure at all? This would be after you tune the filters, is that correct?
great question, I've responded to that one elsewhere here. but will copy paste my response here: "I personally don’t use it but feel free. Use of dynamic damping (which is essentially making D term dynamic with amount of stick) is not widely understood I dont think. You really need to know why you want to use it. If you wish to use it, then once you know what range of damping effect you have on your particular rig, then you’re in a better position to know how best to set D min vs D max. You just have to know what specifically you want to get out of a dynamic D. You could for example set your D max from step response results then set D min to a slightly under dampened state to get a snappier/faster response latency, ideal for racers for example. Or if you want a D boost kind of effect then set D min at the value you obtain from PTB and D max as a D boost. Just depends what you want and why"
What does it mean, if you can't seem to tame the overshoot with D? is that due to esc settings?(48khz?) too much P already? too much filtering?(pretty much stock / rpm on)?
depends. Is overshoot there when Iterm is off? Sometimes adding some more gain will reduce it, if the authority is really low. So if you're testing P-D balance but both are fairly low, it might help to try the tests starting with higher P. Ive also seen that lower ESC pwm can help. but Im speculating on what your specific issue is. only way to know for sure Id have to see the logfile.
As I'm now using Inav I'm taking this new slider method and making an Excel Spreadsheet to replicate the steps (preserve the ratios). The first attempt went pretty well , but now This explanation makes it much clearer why I was getting confuses. Simple things like dropping the I out, so obvious in hindsight. Your videos are a good pace and you explain the logic of what you do and why. Its refreshing. I've watched a few 4.3 videos and this has been the best so far.
great! I was planning a couple of quick vids on how to convert this method to INAV and EMU. Same concepts for sure. I am starting to think that maybe PID gain should come before Iterm though. This might seem odd, but on bigger rigs, if P starts low, the wobble can lead to under tuning Iterm if done before gains, whereas when gains come back in, error goes down naturally, and then often you can afford more iterm. so there's a kind of nonlinearity sometimes on iterm that doesn't hold as nicely as the P to D balance. Nonetheless, this should work as is for anything < 7 inch
@@PIDtoolbox yeah I'm using to really understand this not just follow a bouncing ball, I've had to go right back and understand filters etc as from what you have helped me see is the impact that can have is detrimental to the rest of the tune. ( By default mine are severely over filtered on Inav. As a new person to this level I'm still cautious to know what's safe and whats not so taking baby steps . I think that max d and d filter is almost a step/balance on its own delay/noise ratio loving your videos really really helpful