Fantastic video and great explanation mate. Would be good to see some filter tuning strategy, iterm relax and anti gravity using your method and pid toolbox. Thank you so much and keep well.
I have just concluded my testing using angle mode. Previously I have been using your old system with great success. I was excited to try this. I used a 5 inch quad lightweight with no action cam. This method was a fail/success for me. Utilizing the Feedforward made my quad have a lot of propwash and flew very uncomfortably. However after simply turning off FF the quad flew amazing. I set the I gain just .10 Higher and it was magic. This quad is one of my first and very familiar with its handling. This is the best it has flown ever! So I have concluded so far at least with this quad, That tuning with Angle Mode on Might actually be better. Thanks for the Video it was fun to test this and I think I will try this Angle mode method on my other quads.
thx man , superb education ;) i get outstanding results with the basement method ( just flying outdoors ;) this way will speed up the work a lot , thx for sharing the knowledge
Really great video and very good help.Thanks for that. Unfortunately, I miss an evaluation of the I-term at the end. You talk about it, but some data on different I Gains would have been very nice.
Great work Brian 👍 Would you be able to make a short video to explain how you ended up on the setting for i-Term please? I'm guessing you were able to identify the best case by using the traces, just like FF? Apart from tuning yaw, this is the last part of the tuning puzzle for me 🙂
The reason feed foward is different in angle mode is because of the gyro. If you look at the data at the end of the peaks the gyro drifts away. That's caused by the quad angling itself. If I'm not mistaken. Just a bit more information being processed. That's just my logic take it with a grain of salt. It does make sense though.
This is fantastic! Finally an easy to follow, step by step guide. And it gives superb result! Just now, spent 40 minutes and got a tune for my quad, that i never had even close! Would be even greater if don't skip i-term tuning. And make the same short and concise video about filter tuning. For filters, please take some hard to tune quad, like 7" long range or cinelifter. Where you actually have issues
Tuning filters is pretty individual from rig to rig and there are different strategies to attack noise. Will be hard to explain it all in a single video for general use.
🤩🤩🤩🤩 oh man this vid has some juicy tips! Thats a genius idea tuning in angle. I just built a kabob power pick 23mm 3450kv 4" 4s and did some blackbox tuning last weekend for my first time. I ended up adding a decent amount of FF and jeez louise it tracks my heartbeat now! or my shaky finger😅 I'll try the rc smoothing with a cutoff like you did. Thank you!
Thank you again for this new video. i noticed that the order is now bit different to the other long video and now with the FF step before using the I-Term slider at the end. If this will give a bit better results, i will do the tuning again. i also finished yesterday my flywoo explorer HD build, which ist similar to this build in the video. AOS 7 is next. ;-) i still dont know why these Videos have not that attention they should have. there are so much pilots who hate tuning their quads like me 10+ years. i will post this in our german forum, maybe it will help someone out.
Yea I’m not too strict about the exact order because I always see room for revisiting any aspect of the tune but for most smaller rigs I think this is a better order. For bigger rigs where iterm can be an issue I like to do it before FF sometimes so I can plot it in the step tool without contamination of FF. But whatever works I say. In terms of channel growth, it does take time especially when you don’t put out content very frequently. But I’m happy with how it’s growing. Always appreciate the share. Thanks!
@@PIDtoolbox OK. i will keep this in mind when i tune the bigger AOS 7. when retuning in a different order, the steps will be not so much any more, because the good range got already smaller. its a process. i still learn with every log.
I’ve watched all of your tuning videos and learned a lot! I tried your method on a 5 inch cinematic rig and it does fly a lot better! Are you gonna make a video about filters in the near future? So far I just applied your method with default filters. Haven’t tried reducing them
hi! thanks for awesome video! so simple and easy to follow and clear. you said FF doesnt work in angle mode 2 years ago; but wtih BF 4.5 angle mode is changed i think "earth referenced" now? This can give it an improvement about FF? I am very noob with tuning; just bugged my mind and wanted to ask you. Thanks again!
Thanks for this great informative video. After PD balance is set, I know you mentioned to look at gyro/set point traces when analyzing FF, should we be looking at step traces at all when tuning FF? Lastly when it comes to I term tuning, are we looking at step traces too?
Did you had problems where log file doesnt open? It just doesnt really load up. Shows - loading and never loads up... Seems like file is too big (4mb), but even if i fly a little - it same, gives 4 mb file -_- Black box logging on 1/2 (1600Hz) Gyro Scaled.
Not exactly the same because I term would be changing as well if manipulating PI. That’s not ideal for adjusting P-D ratio. But if you drop the iterm slider such that I is very low or even off, then it’s technically the same.
Hi Brian, thanks for your amazing tool!! a question .. do you suggest to flag the Y correction in step responsive tool? I saw that you don't flag always this option... Thanks!
Hi Brian. Thank you for your job, this video is very helpful. After follow all of your steps got my quad flying really well. I've adjusted the pitch to roll ration (both sliders at the same multiplier) to get the same delay in both axis, now in the step response have a little overshoot in the pitch axis. Does it require a little more Pitch Damping without moving the pitck tracking or lowering the pitch tracking alone?
Only if it’s P related overshoot. I’d bet my money it’s iterm. Drop iterm and test again. If overshoot is gone then PD balance is right. The add iterm back but you’ll need more pid gain and/or FF or reduce iterm a bit on pitch.
@@PIDtoolbox Hi Brian. I lowered the I term multiplier to 0.5 and the problem still occurs. Previously I lowered both pitch:roll ratios that got after adjust they to equilize latency on both axis. Obtained 1.35 on both with your method, now lowered to 1.25. If it's an I term related overshoot then I can't tune it with sliders because I can't adjust I term on one axis alone. Or I have to adjust something else. I would try to move pitch damping alone to 1.35, but D term is way high in pitch right now. Step responses: ibb.co/g75X05M. Sliders: ibb.co/gSvdjJd. Thank you so much!
Hello, thank you so much for this great tool and this tutorial. Some things are not clear to me: when i test for the damping d gains the master multiplier must remain at 1? then i have to test the master multiplier having already set the correct damping d gains? thank you
Everything should be held constant during any of these sets of tests except for the thing that you’re manipulating so in that case it’s Damping. But the master multiplier need not be at 1.0 for these tests and in fact depending on the rig sometimes it’s better to have the master multiplier start a little bit higher or a little bit lower depending on the authority. For example on an under powered cinelifter the default master multiplier of 1.0 is a bit low and may lead to ambiguous PD bal results. Re the other part of your question if I understand it, yes when you get PD balance right then I typically start increasing MM, and it doesn’t matter if you repeat MM1.0. Replication won’t hurt anything
So I've went through filter tuning and this PID tuning. I think it went pretty good and got great results. Really good and easy to follow guide. However now I feel like the motors are a little bit to warm for my liking, not hot but to warm. What is the best way to get cooler motors without doing everything from the start. More filtering? Less Damping? Does this ruin my pid tune?
Properly tuned PIDs will not make motors hot. If you are getting how motors and you are confident it is either PID or filter-related, then it would arise from one of 2 possible things: (1) excessive PID gains and/or excessive filtering (causing excessive filter delay) are causing low freq feedback oscillation. (2) not enough filtering is resulting in noisy dterm, which causes a noisy pidsum, which in turn causes a noisy motor signal feeding into the ESCs. Both of these are easily observed right in the raw gyro and dterm traces. If neither of these possibilities, then motors will also get how if the rig is over weight and the motor rpm is really high just to get the thing off the ground (as with many cinelifters). So going through the PTB basement tuning steps is not complete if you see (1) or (2) above. you have to know where to look to determine those.
@@PIDtoolbox Thanks for the fast and detailed response. Yes my quad is pretty much on the heavier side even for a 7inch. Need quite some throttle to lift of the ground. Motors aren't really hot, just warm. I am still able to touch them without feeling pain. Just would like a little bit more headroom in case of a crash and having to fly back with a bent prop. I will try your suggestion. Thanks again
Hello sr i have a strange one here...when iam tunning ff (stick response) y move the slider but then when i check it with the toolbox, the graph is the same no matter in wich position i put the slider (stick response) dont know what iam doing wrong. Thanks
Make sure you don’t have leftover logfiles stuck in main folder. On mac it’s really a problem if you don’t unlock logfiles. It’ll keep reading the wrong logfile
@@PIDtoolbox i don't use mac but i will try that. And an "autofactor" of 90 could be the problem too? Thanks for your time it's really valuable what You do. Really thanks
How do you manage to get such smooth graphs? No matter how many logs I write - I almost always get very wavy lines that make it extremely hard to tell the difference. this is one of the reasons I prefer to use blackbox viewer and manual log analytics.
Nice video! I don't get it at 7:22 - 8:00. You tried to set the latency ratio the same, but while adjusting, the result in your example even got bigger (from 2ms to 4ms), which doesn't make sense to me.
Not sure I understand what you mean. But, I didn’t show the results after pitch was adjusted (just forgot to show this, but it’s right there in that data). I was comparing two different points in that plot just to show what values I would pick.
@@PIDtoolbox Aah that explains. I thought the picture while explaining was the one. I wanted to ask about a trouble I have with setpoint response tool in PIDToolBox - While my I term is 0, the setpoint looks smooth, but when I set the I term over 0 , it overshoots. It's like this on all the logs. Is that normal? Thanks!
@@tomasspetersons4158 iterm is tricky to understand. When pid error is low (in other words when the error is mostly already minimized with P and D), then iterm has a very wide tuning window such that even very extreme values will show little iterm modulation (especially when using sufficient FF). But when there is still a decent amount of error then iterm accumulates higher and adds significantly to the pid sum. In that case you’ll also see it shape the step response as a sort of slower prolonged overshoot. Key there is to add more FF (and/or more PD gain) to close the error gap. Now the minor issue then is you can’t really interpret the step response when FF is added (it tends to exaggerate overshoot), so you just analyze set point and gyro traces at that point. Maybe one day I will resolve this issue but for now the step resp is really only useful for optimizing PD ratio and PD gains and maybe PI balance if you choose to go that route
@@PIDtoolbox Great answer! Thank you! So I did tune P and D gain with I term 0. And then added I term and started tuning FF till gyro and setpoint were in sync. But I thought, maybe a good way would be getting good P:D balance and then starting tuning FF. Then getting good response, and then maybe adding I term? Anyway, you solved my confusion. I feel like going with raising P and D gains would be a good road for the freestyle quads, but for racing, would be too much for the motors. Hope there will be some kind of a good method for tuning racing rigs. Thank you!
@@tomasspetersons4158 I had the same doubt, I adjusted the ratio of P and D and it was perfect, when I started adding the I an "overshot" started to appear even with very low I values.
Hello, I am a faithful follower of your method to adjust pids and the truth is that I love the system. But I have a doubt and I wanted to ask you to see what is better. If I want to adjust pids of a drone that usually carries a gopro, which is better? adjust the pids without the camera and then raise the multiplier a bit or adjust the pids directly with the camera on top?. I use a softmount for a gopro 5 and the fit results are a bit confusing.
great explanation, thankyou :) but you didnt show the step response tool graph when incrasing ff and i.. does the curve remain flat to Y or it jumps with some overshoot? (witch is what i usually experience when i put ff and I to more than 0.2 on sliders)
The step resp tool is not recommended for FF. It tends to over estimate overshoot that’s not apparent in the traces. This is also why I didn’t plot iterm there either. Once FF is introduced you can really only properly evaluate it direct from traces
@@PIDtoolbox ❤️ that's a really great news... I was following Your previous tutorial where FF and I term where evaluated on step response tool.. And i couldn't get anough far from 0.3.. Thanks 🙏
I have a question, I have a costum system where I need to process the output, for this I process it as CSV. Now is there a way to insert a CSV into PIDtoolbox without running it through the blackbox_decode first? Thanks in advance!
@@PIDtoolbox Ok, thank you very much for the quick response! Though I have another one, could it be that PIDtoolbox expects some response from blackbox_decode? (Side info I am now trying to integrate my code inbetween the PIDtoolbox and blackbox_decode, but now the toolbox opens the loading bar and then stops. I've tried looking through the source code a little but i didn't find anything)
Really appreciate your videos, Brian! I tried this method on a Squirt I have been having issues with. It all balances well in Angle mode, but when I switch it to Acro, the motors and stack get very hot, and the motor noise changes dramatically. It's fine again when I switch back to angle. Any suggestions? Filtering? Thanks for your brillaint content. I too would love to see a tuning showdown between you, Mark and Chris! ☺✌
I been wondering over the last few days whether it'd work the same in angle mode... I literally almost asked yesterday too. But figured it was a dumb thing to ask 😂 I guess not
With this method, the tune is based on the step response on the lower end of the setpoint range. Does it generalize to higher rates like 1000deg/s if you tuned at 200deg/s max?
Generally speaking yes it does. This is because the system is relatively linear within the range of settings we use. And where it is nonlinear there are dynamic parameters that are there to help out (like TPA ‘D’ at 0.65, iterm_relax_cutoff, FF, all of which help the maintenance of set point to gyro relationship at the extremes).
@@PIDtoolbox thanks for the answer. I think it makes sense. Most of the time we fly in the center stick position so that's the most import part for have tuned well. One follow up question, does an indoor tune deal well with external factors like wind? Indoors when the setpoint is zero there isn't much pushing the quad around but outdoors this is quite different.
@@avsaase yes in fact because it tends towards higher PIDs because you’re able to push the master multiplier higher in a controlled setting. So once you reach the upper tuning bound there it’s recommended to step it back a bit to a more safe range. But you’d never otherwise know what the upper bound of the tuning window is without this method. If you are not gaining any response latency at higher gains (or you’re beginning to see oscillation) then there is no benefit to going higher. No more reduction in response latency indicates that the motors have given all they can so settings higher than that won’t buy you any extra control over external forces, at least from the stand point of PIDs. You still have ESC settings which can change things. But the output of the resp latency shown in Ptb step plots gives you a good sense of whether you’re at the top and where you might want to set things in that overall tuning window
@@PIDtoolbox awesome. I'll give it a try today. I'm tuning a 7 inch quad in a smaller room than your basement so it'll definitely help doing this in angle mode.
Hey Brian very impressive i must say. ive got a problem with my new quad its an apex5 and the pids showing in pid toobox are different to when i click the same file and check pids by looking in blackbox info logs. and also when i change my pids in betaflight they dont correspond to the pids displaying in my dji googgles . its bloody driving my crazy. any ideas?
Hello, I wanted to ask you about the I term. Once the P-D and FF ratios are adjusted, when I start to increase the I term, the graphs lose their shape and are excessively overshoot, which forces me to work with very low values of I term. Is that normal?
brian, about the FF, after i get the highest MM and then start to add FF then my step response will showed massive overshoot.. did you ignore the overshoot?
No. The step resp tool exaggerates the overshoot caused by FF. You should not interpret those plots after you add FF. That’s why I did it direct in the traces. Recall there was no overshoot there.
@@TERBANGFPV correct. or do iterm before FF. I have no hard rules about this. But technically I think you can still look at iterm effects in step resp with FF. You'd just ignore the initial be FF-related overshoot and look for the slower iterm related overshoot. I think it can still show up there. But the truth is with enough FF, iterm never get over
averaging for 50hz rate is not recommended. I think it causes a lot of delay to do 2 or 3pt averaging when the samples are only coming in 20ms apart (50hz). Higher rx rates call for it
@@PIDtoolbox Thank you for the reply! It would be too easy to think training mode is a more "natural" substitute to acro than angle for testing, but I don't know any details either, it was only my first thought when I read the video title. Of course you can configure the max angle for training mode, but no idea how the clamping works exactly
What a nonsense. Using the PTB now, open files, some wait cursor appears, nothing happens, even let it run for 30 minutes. Happed to BFL or CSV files. Besides that, the UI is horrible. For example, who using using radio buttons for check boxes instead? Pff, maybe it's a good thing no public releases are being made and it's now behind a paywall.
Here's me thinking the video title gives it all away, but there were so many little nuggets of advice hidden within. 👌
I second that. Thank you so much, Brian! This video made my day on so many levels. I am next tuning my micro longrange now. 👍🏻
It was raining on the weekend, so I tried this PID tuning.
It worked better than other tuning methods.
Thank you for telling me!
You are awesome! Best PID tuning video + software there is! Saving this video for all future uses.
I’m so grateful for your work Brian. This has absolutely changed my experience of the hobby for the better. Thank you.
Man I have a 4" that has been giving me hell. I tried tuning with this method and now she's flies on rails. Thanks!!!!
It's a beautiful and extremely useful tool for PID tuning. This is what I was looking for. Definitely supporting now.
That set point tracking is amazing holy moly can't wait to try this myself
Fantastic video and great explanation mate. Would be good to see some filter tuning strategy, iterm relax and anti gravity using your method and pid toolbox. Thank you so much and keep well.
I have just concluded my testing using angle mode. Previously I have been using your old system with great success. I was excited to try this. I used a 5 inch quad lightweight with no action cam. This method was a fail/success for me. Utilizing the Feedforward made my quad have a lot of propwash and flew very uncomfortably. However after simply turning off FF the quad flew amazing. I set the I gain just .10 Higher and it was magic. This quad is one of my first and very familiar with its handling. This is the best it has flown ever!
So I have concluded so far at least with this quad, That tuning with Angle Mode on Might actually be better. Thanks for the Video it was fun to test this and I think I will try this Angle mode method on my other quads.
thx man , superb education ;) i get outstanding results with the basement method ( just flying outdoors ;) this way will speed up the work a lot , thx for sharing the knowledge
Top tip I may try this method on the big fella. I’ve been getting great success with my 5” s
Best educational video about pid tunning!
Great tutorial, thank you, my 5" now flies beautifully :)
That was a cool explanation! Thank you!
This is great, thanks! I Tried tuning my 5" inside on ACRO the other day, but that was way too sketchy. Gonna give this a try.
I love your videos. Thanks for the work because this really helped me tune for the first time with Blackbox data
Really great video and very good help.Thanks for that.
Unfortunately, I miss an evaluation of the I-term at the end.
You talk about it, but some data on different I Gains would have been very nice.
Great work Brian 👍
Would you be able to make a short video to explain how you ended up on the setting for i-Term please? I'm guessing you were able to identify the best case by using the traces, just like FF?
Apart from tuning yaw, this is the last part of the tuning puzzle for me 🙂
Thanks for everything ✌🏼
wow. perfect tune now even if you can't fly well!
Fricken. Amazing. I might actually tune more now. **tip of the hat**
The reason feed foward is different in angle mode is because of the gyro. If you look at the data at the end of the peaks the gyro drifts away. That's caused by the quad angling itself. If I'm not mistaken. Just a bit more information being processed. That's just my logic take it with a grain of salt. It does make sense though.
This is fantastic! Finally an easy to follow, step by step guide. And it gives superb result! Just now, spent 40 minutes and got a tune for my quad, that i never had even close! Would be even greater if don't skip i-term tuning. And make the same short and concise video about filter tuning. For filters, please take some hard to tune quad, like 7" long range or cinelifter. Where you actually have issues
Tuning filters is pretty individual from rig to rig and there are different strategies to attack noise. Will be hard to explain it all in a single video for general use.
🤩🤩🤩🤩 oh man this vid has some juicy tips! Thats a genius idea tuning in angle. I just built a kabob power pick 23mm 3450kv 4" 4s and did some blackbox tuning last weekend for my first time. I ended up adding a decent amount of FF and jeez louise it tracks my heartbeat now! or my shaky finger😅 I'll try the rc smoothing with a cutoff like you did.
Thank you!
Feedforward in angle code has been rewritten so it behaves almost like expo to allow smooth and snappy. Great for whoops
Thanks for sharing bro
Thank you again for this new video. i noticed that the order is now bit different to the other long video and now with the FF step before using the I-Term slider at the end. If this will give a bit better results, i will do the tuning again. i also finished yesterday my flywoo explorer HD build, which ist similar to this build in the video. AOS 7 is next. ;-) i still dont know why these Videos have not that attention they should have. there are so much pilots who hate tuning their quads like me 10+ years. i will post this in our german forum, maybe it will help someone out.
Yea I’m not too strict about the exact order because I always see room for revisiting any aspect of the tune but for most smaller rigs I think this is a better order. For bigger rigs where iterm can be an issue I like to do it before FF sometimes so I can plot it in the step tool without contamination of FF. But whatever works I say.
In terms of channel growth, it does take time especially when you don’t put out content very frequently. But I’m happy with how it’s growing. Always appreciate the share. Thanks!
@@PIDtoolbox OK. i will keep this in mind when i tune the bigger AOS 7. when retuning in a different order, the steps will be not so much any more, because the good range got already smaller. its a process. i still learn with every log.
Now I just need a basement
😆
Great video Brian..
Thank you for making PID tuning so simple
I’ve watched all of your tuning videos and learned a lot! I tried your method on a 5 inch cinematic rig and it does fly a lot better! Are you gonna make a video about filters in the near future? So far I just applied your method with default filters. Haven’t tried reducing them
Will do more on filtering for sure. I’m just really slow getting vids out.
Exactly how you want to tune your quad if you want it to fly custom made :)
Beautiful 😁
Thanks Mark!
hi! thanks for awesome video! so simple and easy to follow and clear. you said FF doesnt work in angle mode 2 years ago; but wtih BF 4.5 angle mode is changed i think "earth referenced" now? This can give it an improvement about FF? I am very noob with tuning; just bugged my mind and wanted to ask you. Thanks again!
Thanks for this great informative video. After PD balance is set, I know you mentioned to look at gyro/set point traces when analyzing FF, should we be looking at step traces at all when tuning FF? Lastly when it comes to I term tuning, are we looking at step traces too?
Hey Brian, can u cover about dynamic damping.. are we doestn need dynamic damping or it's not so important?
Should i tune filters first ? Thanks
@PIDtoolbox what version is the one in the video? I don't get the the small toolbar like the one in the video 10:14 I'm running latest version
'ms to Hz' is called 'Period' in latest version, and it's a button in the control panel
My graph always returns to 1.05 using this method. D gain from 0.2 to 1.8 . What does this mean?
Did you had problems where log file doesnt open? It just doesnt really load up. Shows - loading and never loads up... Seems like file is too big (4mb), but even if i fly a little - it same, gives 4 mb file -_- Black box logging on 1/2 (1600Hz) Gyro Scaled.
Hello! One question.
For the PD balance, it is the same to move the damping slider (D gain) or to move the Tracking slider(PI Gain)?
Thank You!
Not exactly the same because I term would be changing as well if manipulating PI. That’s not ideal for adjusting P-D ratio. But if you drop the iterm slider such that I is very low or even off, then it’s technically the same.
Thank you Brian! So are you using 15 for RC smoothing on all your Crossfire rigs?
No. Usually ~20hz for freestyle but when I’m looking for smooth, 15hz or even 10hz can really round out the edges. Rates alone won’t always do it
Next time I watch this I won't be baked.. 👍
Is this still relevant in the order of process ?
Hi Brian, thanks for your amazing tool!! a question .. do you suggest to flag the Y correction in step responsive tool? I saw that you don't flag always this option... Thanks!
Not sure I understand what you mean by “flag it”. It should be used only in special circumstances
thanks.. in this video at time 2:52 the Y correction is flagged … from 3:40 the flag is OFF
@@FPVTUNING that is the check box that turns it on or off
Amazing! What is the benefit of increasing I-term in flight?
Drift and control during sustained sharp turns
Hi Brian. Thank you for your job, this video is very helpful. After follow all of your steps got my quad flying really well. I've adjusted the pitch to roll ration (both sliders at the same multiplier) to get the same delay in both axis, now in the step response have a little overshoot in the pitch axis. Does it require a little more Pitch Damping without moving the pitck tracking or lowering the pitch tracking alone?
Only if it’s P related overshoot. I’d bet my money it’s iterm. Drop iterm and test again. If overshoot is gone then PD balance is right. The add iterm back but you’ll need more pid gain and/or FF or reduce iterm a bit on pitch.
@@PIDtoolbox the problem is that can't adjust I term just for pitch, has to do that for all axis. Will try it tomorrow. Thanks Brian!
@@PIDtoolbox Hi Brian. I lowered the I term multiplier to 0.5 and the problem still occurs. Previously I lowered both pitch:roll ratios that got after adjust they to equilize latency on both axis. Obtained 1.35 on both with your method, now lowered to 1.25. If it's an I term related overshoot then I can't tune it with sliders because I can't adjust I term on one axis alone. Or I have to adjust something else. I would try to move pitch damping alone to 1.35, but D term is way high in pitch right now. Step responses: ibb.co/g75X05M. Sliders: ibb.co/gSvdjJd. Thank you so much!
And what about pid balance between damping and tracking as chris rosser says? D term must be 2.5 more times than p term? Thanks for all this?
That’s achieved with damping. Order of operations: P-D ratio, P-I ratio, MM, FF
Dterm 2.5x more than Pterm? No. That’s what the P-D bal tests are for. To optimize P to D
Hello,
thank you so much for this great tool and this tutorial.
Some things are not clear to me:
when i test for the damping d gains the master multiplier must remain at 1?
then i have to test the master multiplier having already set the correct damping d gains? thank you
Everything should be held constant during any of these sets of tests except for the thing that you’re manipulating so in that case it’s Damping. But the master multiplier need not be at 1.0 for these tests and in fact depending on the rig sometimes it’s better to have the master multiplier start a little bit higher or a little bit lower depending on the authority. For example on an under powered cinelifter the default master multiplier of 1.0 is a bit low and may lead to ambiguous PD bal results.
Re the other part of your question if I understand it, yes when you get PD balance right then I typically start increasing MM, and it doesn’t matter if you repeat MM1.0. Replication won’t hurt anything
What about D-Max ?
Nice Video, quick question, where did the "ms to Hz" button go in v0.55?
It’s the check box called “period” now
How about acro trainer mode?
Hello, if I have tuned the quad based on this tutorial and in theory it's fine, I don't need to add dynamic damping?
In most cases yes
Okay. Tuning went well. But. Do you bring dampening back up after tuning or you leave at 0?
Leave it at 0
What does pd balance test stand for is it the damping slider or pitch damping? Thanks!
Yes, damping changes D, while holding P constant, hence it changes the P-to-D balance
So I've went through filter tuning and this PID tuning. I think it went pretty good and got great results. Really good and easy to follow guide. However now I feel like the motors are a little bit to warm for my liking, not hot but to warm. What is the best way to get cooler motors without doing everything from the start. More filtering? Less Damping? Does this ruin my pid tune?
Properly tuned PIDs will not make motors hot. If you are getting how motors and you are confident it is either PID or filter-related, then it would arise from one of 2 possible things: (1) excessive PID gains and/or excessive filtering (causing excessive filter delay) are causing low freq feedback oscillation. (2) not enough filtering is resulting in noisy dterm, which causes a noisy pidsum, which in turn causes a noisy motor signal feeding into the ESCs. Both of these are easily observed right in the raw gyro and dterm traces. If neither of these possibilities, then motors will also get how if the rig is over weight and the motor rpm is really high just to get the thing off the ground (as with many cinelifters). So going through the PTB basement tuning steps is not complete if you see (1) or (2) above. you have to know where to look to determine those.
@@PIDtoolbox Thanks for the fast and detailed response. Yes my quad is pretty much on the heavier side even for a 7inch. Need quite some throttle to lift of the ground. Motors aren't really hot, just warm. I am still able to touch them without feeling pain. Just would like a little bit more headroom in case of a crash and having to fly back with a bent prop. I will try your suggestion. Thanks again
Hello sr i have a strange one here...when iam tunning ff (stick response) y move the slider but then when i check it with the toolbox, the graph is the same no matter in wich position i put the slider (stick response) dont know what iam doing wrong. Thanks
Make sure you don’t have leftover logfiles stuck in main folder. On mac it’s really a problem if you don’t unlock logfiles. It’ll keep reading the wrong logfile
@@PIDtoolbox i don't use mac but i will try that. And an "autofactor" of 90 could be the problem too? Thanks for your time it's really valuable what You do. Really thanks
@@gonzafpv2843 or right, old version.
@@PIDtoolbox a high value of autofactor (90, stock is 30) it has nothing to do with it? Thanks
How do you manage to get such smooth graphs? No matter how many logs I write - I almost always get very wavy lines that make it extremely hard to tell the difference. this is one of the reasons I prefer to use blackbox viewer and manual log analytics.
Low freq broadband noise causes noisy step resp traces
I have to ask what the filter settings are on this quad? Great video!
No RPM because the ESC is blheli s. Instead using 3xDNs, and single gyro lpf at 150Hz. Dterm default i think
Nice video! I don't get it at 7:22 - 8:00. You tried to set the latency ratio the same, but while adjusting, the result in your example even got bigger (from 2ms to 4ms), which doesn't make sense to me.
Not sure I understand what you mean. But, I didn’t show the results after pitch was adjusted (just forgot to show this, but it’s right there in that data). I was comparing two different points in that plot just to show what values I would pick.
@@PIDtoolbox Aah that explains. I thought the picture while explaining was the one.
I wanted to ask about a trouble I have with setpoint response tool in PIDToolBox - While my I term is 0, the setpoint looks smooth, but when I set the I term over 0 , it overshoots. It's like this on all the logs. Is that normal? Thanks!
@@tomasspetersons4158 iterm is tricky to understand. When pid error is low (in other words when the error is mostly already minimized with P and D), then iterm has a very wide tuning window such that even very extreme values will show little iterm modulation (especially when using sufficient FF). But when there is still a decent amount of error then iterm accumulates higher and adds significantly to the pid sum. In that case you’ll also see it shape the step response as a sort of slower prolonged overshoot. Key there is to add more FF (and/or more PD gain) to close the error gap. Now the minor issue then is you can’t really interpret the step response when FF is added (it tends to exaggerate overshoot), so you just analyze set point and gyro traces at that point. Maybe one day I will resolve this issue but for now the step resp is really only useful for optimizing PD ratio and PD gains and maybe PI balance if you choose to go that route
@@PIDtoolbox Great answer! Thank you! So I did tune P and D gain with I term 0. And then added I term and started tuning FF till gyro and setpoint were in sync.
But I thought, maybe a good way would be getting good P:D balance and then starting tuning FF. Then getting good response, and then maybe adding I term?
Anyway, you solved my confusion. I feel like going with raising P and D gains would be a good road for the freestyle quads, but for racing, would be too much for the motors. Hope there will be some kind of a good method for tuning racing rigs. Thank you!
@@tomasspetersons4158 I had the same doubt, I adjusted the ratio of P and D and it was perfect, when I started adding the I an "overshot" started to appear even with very low I values.
Hello, I am a faithful follower of your method to adjust pids and the truth is that I love the system. But I have a doubt and I wanted to ask you to see what is better. If I want to adjust pids of a drone that usually carries a gopro, which is better? adjust the pids without the camera and then raise the multiplier a bit or adjust the pids directly with the camera on top?. I use a softmount for a gopro 5 and the fit results are a bit confusing.
the gopro will usually only affect the pitch axis so either method will work but you'd only be boosting pitch, not MM.
@@PIDtoolbox ok, thank you very much.
great explanation, thankyou :)
but you didnt show the step response tool graph when incrasing ff and i.. does the curve remain flat to Y or it jumps with some overshoot? (witch is what i usually experience when i put ff and I to more than 0.2 on sliders)
The step resp tool is not recommended for FF. It tends to over estimate overshoot that’s not apparent in the traces. This is also why I didn’t plot iterm there either. Once FF is introduced you can really only properly evaluate it direct from traces
@@PIDtoolbox ❤️ that's a really great news... I was following Your previous tutorial where FF and I term where evaluated on step response tool.. And i couldn't get anough far from 0.3..
Thanks 🙏
I have a question, I have a costum system where I need to process the output, for this I process it as CSV. Now is there a way to insert a CSV into PIDtoolbox without running it through the blackbox_decode first? Thanks in advance!
Short answer is unfortunately no
@@PIDtoolbox Ok, thank you very much for the quick response! Though I have another one, could it be that PIDtoolbox expects some response from blackbox_decode? (Side info I am now trying to integrate my code inbetween the PIDtoolbox and blackbox_decode, but now the toolbox opens the loading bar and then stops. I've tried looking through the source code a little but i didn't find anything)
@@Martin-zo8lz yes it does
Really appreciate your videos, Brian! I tried this method on a Squirt I have been having issues with. It all balances well in Angle mode, but when I switch it to Acro, the motors and stack get very hot, and the motor noise changes dramatically. It's fine again when I switch back to angle. Any suggestions? Filtering? Thanks for your brillaint content. I too would love to see a tuning showdown between you, Mark and Chris! ☺✌
Sounds like a filtering issue.
I been wondering over the last few days whether it'd work the same in angle mode... I literally almost asked yesterday too. But figured it was a dumb thing to ask 😂 I guess not
Me too. I’ve actually thought about this eons ago but never got around to formal testing it till now
when is 4.4 update comming? tried to tune but 4.4 looks different than this.. (and i'm a beginner in tunning)
There’s no difference in the sliders for 4.3 or 4.4
With this method, the tune is based on the step response on the lower end of the setpoint range. Does it generalize to higher rates like 1000deg/s if you tuned at 200deg/s max?
Generally speaking yes it does. This is because the system is relatively linear within the range of settings we use. And where it is nonlinear there are dynamic parameters that are there to help out (like TPA ‘D’ at 0.65, iterm_relax_cutoff, FF, all of which help the maintenance of set point to gyro relationship at the extremes).
@@PIDtoolbox thanks for the answer. I think it makes sense. Most of the time we fly in the center stick position so that's the most import part for have tuned well. One follow up question, does an indoor tune deal well with external factors like wind? Indoors when the setpoint is zero there isn't much pushing the quad around but outdoors this is quite different.
@@avsaase yes in fact because it tends towards higher PIDs because you’re able to push the master multiplier higher in a controlled setting. So once you reach the upper tuning bound there it’s recommended to step it back a bit to a more safe range. But you’d never otherwise know what the upper bound of the tuning window is without this method. If you are not gaining any response latency at higher gains (or you’re beginning to see oscillation) then there is no benefit to going higher. No more reduction in response latency indicates that the motors have given all they can so settings higher than that won’t buy you any extra control over external forces, at least from the stand point of PIDs. You still have ESC settings which can change things. But the output of the resp latency shown in Ptb step plots gives you a good sense of whether you’re at the top and where you might want to set things in that overall tuning window
@@PIDtoolbox awesome. I'll give it a try today. I'm tuning a 7 inch quad in a smaller room than your basement so it'll definitely help doing this in angle mode.
Hey Brian very impressive i must say. ive got a problem with my new quad its an apex5 and the pids showing in pid toobox are different to when i click the same file and check pids by looking in blackbox info logs. and also when i change my pids in betaflight they dont correspond to the pids displaying in my dji googgles . its bloody driving my crazy. any ideas?
Send me the logfile
@@PIDtoolbox no probs, ive just sent one to your facebook. thanks
Hello, I wanted to ask you about the I term. Once the P-D and FF ratios are adjusted, when I start to increase the I term, the graphs lose their shape and are excessively overshoot, which forces me to work with very low values of I term. Is that normal?
I have a vid about iterm. Check playlist
Should you add Y correction ?
Not really. I should remove it
@@PIDtoolbox I noticed I can get results that seem over dampened until I enable it. Then everything looks nice afterwards. Not sure what that means
Very cool...does it work also with BF 4.2.x?
Absolutely. Nothing different there wrt angle mode
@@PIDtoolbox thk u so much 🙂
brian, about the FF, after i get the highest MM and then start to add FF then my step response will showed massive overshoot.. did you ignore the overshoot?
No. The step resp tool exaggerates the overshoot caused by FF. You should not interpret those plots after you add FF. That’s why I did it direct in the traces. Recall there was no overshoot there.
@@PIDtoolbox okay, but then we cant tune i term likr before, i mean we cant see when the iterm starts to add overshoot.
@@TERBANGFPV correct. or do iterm before FF. I have no hard rules about this. But technically I think you can still look at iterm effects in step resp with FF. You'd just ignore the initial be FF-related overshoot and look for the slower iterm related overshoot. I think it can still show up there. But the truth is with enough FF, iterm never get over
why not using averaging in feedforward, does it because using 50Hz link, or the averaging is not much benefit from it.
averaging for 50hz rate is not recommended. I think it causes a lot of delay to do 2 or 3pt averaging when the samples are only coming in 20ms apart (50hz). Higher rx rates call for it
@@PIDtoolbox ok noted sir,,,thank you verymuch
does this method work with training mode too?
Not sure
I think it should actually. Might also be a nice way to do things but I’m not sure how it caps rotation rate.
@@PIDtoolbox Thank you for the reply! It would be too easy to think training mode is a more "natural" substitute to acro than angle for testing, but I don't know any details either, it was only my first thought when I read the video title. Of course you can configure the max angle for training mode, but no idea how the clamping works exactly
I can't believe I watched this entire video of Brian playing around with his P and his D
ive come to the conclusion i'm just an idiot. Been trying to figure this shit out for 6 years and still dont understand anything
Any chance this is ever going to utilize AI? It's just too difficult to understand?
6/
What a nonsense. Using the PTB now, open files, some wait cursor appears, nothing happens, even let it run for 30 minutes. Happed to BFL or CSV files. Besides that, the UI is horrible. For example, who using using radio buttons for check boxes instead? Pff, maybe it's a good thing no public releases are being made and it's now behind a paywall.