Thanks for the tips of making LOS rate profile! I found it's more easier with this to get basement tuning done without crash. :) And Betaflight 4.3RC2 now released. Please consider to make a step by step tuning video from ground zero with the new slider and PTB so the beginners could follow. Thank you Brian!
thank you, found this on the discord pinned msgs. I'll try to move both roll and pitch at the same time to get better results, mine are too noisy at the moment on my 6inch
As always great information! Now people hopefully stopping to put filters to the absolute limits and instead getting a really clean signal to noise being able to drive higher gains which is way superiour than getting a millisecond less latency on filter delay. You want a good looking step response, go revisit your filtering, I love it! :-)
Exactly. It’s typically just a matter of better utilizing the DN and or adding slightly more low pass filtering on the gyro, not enough to cost much if any latency. It might even call for looking for revisiting mechanical or electrical issues, but in either case when it’s visible in the step traces it’s usually worse than people think.
Great stuff, thx! Switch off iterm relax as well and crank up thrust linear in BF... under the assumption you're not hitting any saturation the pid controller within the loop only acts linear on an almost linear and decoupled plant (at least below 100 hz, rpm filters and adaptive notch above). The small influence from the magnitude should vanish, thats at least what i saw.
My next vid! Absolutely right. You might have noticed from the logfile name I’ve done this test. Have to think more about this one to properly address it. There’s a subtlety there that most are unaware of
Hello, I have been trying to learn about the adjustment of PIDs for a while and the truth is that I have been lucky to find your videos, it seems to me an incredibly simple and fast method. I have finally been able to adjust some of my quads well. I have the problem when I mount a naked gopro on a soft mount, because then I start to get curves with very strange shapes, especially at high PD. It is as if the first peak was below 1, and then it rises above the line of 1. In this situation I am not at all clear how I should adjust the PIDs, and I would like you to help me. Thank you very much in advance and greetings from Spain.
indeed it is a simple method! well tbh it is conceptually simple, but it certainly can be confusing when there is oscillation or noisy gyro, both of which cause this curves to vary and can confuse the results. I wish I had more time to make more explanatory videos. I do have a tuning service which I have not advertised because Im already swamped but you can find me on the Blackbox log facebook group (Brian White).
Wanted to add a minor point here. If you decide to use a LOS rate profile with low max rate, be cautious not to hit the stick end positions. This could induce abrupt steps in the input. I’ve not tested in detail whether this is really an issue yet but just wanted to throw it out there.
Thanks Brian, a question about this: I watched Chris Rosser and some other tutorials, they say to do some full stick and full 360 deg movements, and in your videos seems like you just did some relatively small movements, compare to their 360deg. So would the result be different between these two ways? Thanks
@@xilinwang4770 if you do full stick, it’ll leaves sparse movement data and throws bad step resp traces into the average (this is why you get this dc offset when you go that approach). My method involves continuous ~200-300deg /s movement stick input. The system is linear enough that the results generalize across all sized stick input.
@@PIDtoolbox okay I get it…but when doing actual flights you’ll have to do full sticks sometimes, so the bad step resp traces still happen right? I’m a little bit confused when I have gotten nice curves when tuning, then after a real flight I checked the curves, those overshoot and oscillation came back again. Should I just leave them or do more tuning based on the new curves? Thanks!
@@xilinwang4770 yes but bad step traces don’t mean bad flight. It means bad data collection is all. you can’t measure step responses from typical fight because there’s typically not enough movement. The RESULTS you get from real flight are not that meaningful because it’s pulling junk from the noise floor, mostly because most typical flying is essentially cruising around hardly moving the sticks then throwing in occasional full stick deflection. The tool can’t gather much useful info from this, and this is why you get the dc offset where the trace never seems to hit the target line. This is not a reflection of bad flight characteristics just a limitation of the methodology. See the algorithm runs a moving window over the data deconvolving the impulse response at each step then integrates those to get the step resp and averages those to get the nice mean step resp trace you see. The problem is during periods of just cruising or with little movement, it throws crappy data into the average. You see, what it’s doing is in effect computing the function that describes the fundamental resp property of the system and this just works best in a controlled environment with continuous controlled input. The beauty is the step resp curve is nearly identical whether you are moving at 50deg/s or 500deg/s. As you approach higher and higher rates there may be some minor deviation from this linear relationship but it’s essentially the same.
Hey man! first of all, THANK YOU! What if you could choose in and out points independently for pitch/roll (and yaw?) That would get around having to do these very specific circular pitch/roll movements shown in the basement tuning example. Doing those manouvers consistently is harder than it looks for most people.
@@PIDtoolbox Hey! Thanks for replying! :) point and click would be AWESOME! A step further would be to be able to drag the bars side to side. What I meant with "independently" is to be able to choose a different in/out point for roll and a different in/out point for pitch, instead of fixed n/out points for all axis. That way you can clean the step response calculation of bad signal to noise ratio parts independently.
@BW, any idea why PTB vs PlasmaTree step-response are massively different for same BBL? PT would tell me i need a little more i term, but PTB would tell me i need much less. EmuFlight in this case. (which obviously has more variables around iTerm [iDecay, SPA, Boosts, etc]) +1👍 very good info in here irregardless of flightware 💪
Just one question (sorry if it is mentioned before) Is there a way to import all the flight takes(logs) in one go from a single BBL file? Right now I have to select the BBL file and import the log 1 then selecting the same BBL file wait it to load and select the 2nd log and so on.
ya, not really. it's trickier than you think getting that working, and even if it was working, the plot legends need labels and these currently come from the file names. It is on my list of things to improve though. what I recommend is making separate distinct files because you're also gonna need this to keep things organized and for the proper labels and legends in the plots.
@@PIDtoolbox With Betaflight 4.3 you can change the slider with the OSD and than I just hit "save and reboot" so I get new files. Btw. will be there are updated tutorial for the new sliders, because PD balance changes now the P value and not the D values. What I tried was to move PD Balance in 0.2 steps and change PD gain so the P value stays the same
@@EuLeRFPV for sure! I'm just kinda waiting till we know for sure what 4.3 will exactly look like. I think the sliders are there now, but still some minor unresolves
How to switch off FF in new BF 4.3 version? Something is changed in params and I cannot decrease FF to zero. It’s very easy to use BF for tuning process as sliders could be changed from OSD. Please try
@@PIDtoolbox all the way left, the slider will be at 0.5, and FF values will not be zero. If we want FF = 0 we should set it manually, not using slider. This point is not clear to me. Do we want FF = 0 ?
@@Tonynux just set FF slider as far left as possible and transition to 1. FF wont be off but minimized which is fine. After you're tuned then set FF as desired.
2:45 yep! The method we use to analyze the approximate step response is base on the assumption that the system is in the stationary state!
Thanks for the tips of making LOS rate profile! I found it's more easier with this to get basement tuning done without crash. :) And Betaflight 4.3RC2 now released. Please consider to make a step by step tuning video from ground zero with the new slider and PTB so the beginners could follow. Thank you Brian!
I have a vid doing step by step basement tune in BF4.3 ua-cam.com/video/cUzNuMVxkBM/v-deo.html
thank you, found this on the discord pinned msgs. I'll try to move both roll and pitch at the same time to get better results, mine are too noisy at the moment on my 6inch
Thank you so much. Could you please elaborate when would you use Y correction in Step Response
Love your tuning skills sir! Would hire again 10/10
As always great information! Now people hopefully stopping to put filters to the absolute limits and instead getting a really clean signal to noise being able to drive higher gains which is way superiour than getting a millisecond less latency on filter delay. You want a good looking step response, go revisit your filtering, I love it! :-)
Exactly. It’s typically just a matter of better utilizing the DN and or adding slightly more low pass filtering on the gyro, not enough to cost much if any latency. It might even call for looking for revisiting mechanical or electrical issues, but in either case when it’s visible in the step traces it’s usually worse than people think.
As usual, very useful and makes sense! Thanks Brian (Voilààààà) :)
Great stuff, thx! Switch off iterm relax as well and crank up thrust linear in BF... under the assumption you're not hitting any saturation the pid controller within the loop only acts linear on an almost linear and decoupled plant (at least below 100 hz, rpm filters and adaptive notch above). The small influence from the magnitude should vanish, thats at least what i saw.
My next vid! Absolutely right. You might have noticed from the logfile name I’ve done this test. Have to think more about this one to properly address it. There’s a subtlety there that most are unaware of
Hello, I have been trying to learn about the adjustment of PIDs for a while and the truth is that I have been lucky to find your videos, it seems to me an incredibly simple and fast method. I have finally been able to adjust some of my quads well. I have the problem when I mount a naked gopro on a soft mount, because then I start to get curves with very strange shapes, especially at high PD. It is as if the first peak was below 1, and then it rises above the line of 1. In this situation I am not at all clear how I should adjust the PIDs, and I would like you to help me. Thank you very much in advance and greetings from Spain.
indeed it is a simple method! well tbh it is conceptually simple, but it certainly can be confusing when there is oscillation or noisy gyro, both of which cause this curves to vary and can confuse the results. I wish I had more time to make more explanatory videos. I do have a tuning service which I have not advertised because Im already swamped but you can find me on the Blackbox log facebook group (Brian White).
@@PIDtoolbox Ok thank you very much, I will contact you through Facebook.
Great tips man.
Excellent
Wanted to add a minor point here. If you decide to use a LOS rate profile with low max rate, be cautious not to hit the stick end positions. This could induce abrupt steps in the input. I’ve not tested in detail whether this is really an issue yet but just wanted to throw it out there.
Thanks Brian, a question about this: I watched Chris Rosser and some other tutorials, they say to do some full stick and full 360 deg movements, and in your videos seems like you just did some relatively small movements, compare to their 360deg. So would the result be different between these two ways? Thanks
@@xilinwang4770 if you do full stick, it’ll leaves sparse movement data and throws bad step resp traces into the average (this is why you get this dc offset when you go that approach). My method involves continuous ~200-300deg /s movement stick input. The system is linear enough that the results generalize across all sized stick input.
@@PIDtoolbox okay I get it…but when doing actual flights you’ll have to do full sticks sometimes, so the bad step resp traces still happen right?
I’m a little bit confused when I have gotten nice curves when tuning, then after a real flight I checked the curves, those overshoot and oscillation came back again. Should I just leave them or do more tuning based on the new curves? Thanks!
@@xilinwang4770 yes but bad step traces don’t mean bad flight. It means bad data collection is all. you can’t measure step responses from typical fight because there’s typically not enough movement. The RESULTS you get from real flight are not that meaningful because it’s pulling junk from the noise floor, mostly because most typical flying is essentially cruising around hardly moving the sticks then throwing in occasional full stick deflection. The tool can’t gather much useful info from this, and this is why you get the dc offset where the trace never seems to hit the target line. This is not a reflection of bad flight characteristics just a limitation of the methodology. See the algorithm runs a moving window over the data deconvolving the impulse response at each step then integrates those to get the step resp and averages those to get the nice mean step resp trace you see. The problem is during periods of just cruising or with little movement, it throws crappy data into the average. You see, what it’s doing is in effect computing the function that describes the fundamental resp property of the system and this just works best in a controlled environment with continuous controlled input. The beauty is the step resp curve is nearly identical whether you are moving at 50deg/s or 500deg/s. As you approach higher and higher rates there may be some minor deviation from this linear relationship but it’s essentially the same.
@@PIDtoolbox got it, thank you very much!
How does the algorithm that converts logs to step response work? Where can i read info about that?
Hey man! first of all, THANK YOU! What if you could choose in and out points independently for pitch/roll (and yaw?) That would get around having to do these very specific circular pitch/roll movements shown in the basement tuning example. Doing those manouvers consistently is harder than it looks for most people.
that is a nice idea. it's a pain plugging in number so a point click thing would be ideal.
@@PIDtoolbox Hey! Thanks for replying! :) point and click would be AWESOME! A step further would be to be able to drag the bars side to side. What I meant with "independently" is to be able to choose a different in/out point for roll and a different in/out point for pitch, instead of fixed n/out points for all axis. That way you can clean the step response calculation of bad signal to noise ratio parts independently.
What about y correction, what is that function works to.?
@BW, any idea why PTB vs PlasmaTree step-response are massively different for same BBL?
PT would tell me i need a little more i term, but PTB would tell me i need much less.
EmuFlight in this case. (which obviously has more variables around iTerm [iDecay, SPA, Boosts, etc])
+1👍 very good info in here irregardless of flightware 💪
hmm, if i enable Y-correction, it's more consistent. maybe i need to search what this is for. :P
One more question, please. It's possible to use pidtoolbox to tunning althold or positionhold as well??? Very thanks!!❤
I intend on deveoping the step resp tool to interpret gps pid control
@@PIDtoolbox Nce!!👍
I have tried everything in this video plus some other things i just cant get it smoother i need help
www.pidtoolbox.com
Flights in acro or stabilize mode? Cant find this info. Thank you.
acro. still not sure stabilize works properly. have to do some tests
I have the same question too . 😀
Just one question (sorry if it is mentioned before)
Is there a way to import all the flight takes(logs) in one go from a single BBL file? Right now I have to select the BBL file and import the log 1 then selecting the same BBL file wait it to load and select the 2nd log and so on.
ya, not really. it's trickier than you think getting that working, and even if it was working, the plot legends need labels and these currently come from the file names. It is on my list of things to improve though. what I recommend is making separate distinct files because you're also gonna need this to keep things organized and for the proper labels and legends in the plots.
@@PIDtoolbox With Betaflight 4.3 you can change the slider with the OSD and than I just hit "save and reboot" so I get new files. Btw. will be there are updated tutorial for the new sliders, because PD balance changes now the P value and not the D values. What I tried was to move PD Balance in 0.2 steps and change PD gain so the P value stays the same
@@EuLeRFPV for sure! I'm just kinda waiting till we know for sure what 4.3 will exactly look like. I think the sliders are there now, but still some minor unresolves
How to switch off FF in new BF 4.3 version? Something is changed in params and I cannot decrease FF to zero. It’s very easy to use BF for tuning process as sliders could be changed from OSD. Please try
To turn off FF and dynamic damping you enable expert mode and push the slider all the way left. This will turn off FF and dynamic D
@@PIDtoolbox all the way left, the slider will be at 0.5, and FF values will not be zero. If we want FF = 0 we should set it manually, not using slider. This point is not clear to me. Do we want FF = 0 ?
@@Tonynux oh sorry I was thinking BF4.3, that's how the sliders will operate in the latest.
@@PIDtoolbox I have BF 4.2.9, so how do I have to set FF? And what about FF transition? 1 or 0 ?
@@Tonynux just set FF slider as far left as possible and transition to 1. FF wont be off but minimized which is fine. After you're tuned then set FF as desired.