I got rejected from the engineering program at most of the schools I applied to because my math scores sucked, but now I feel like I am getting a much more fun education from doing this drone thing and learning from you Sam! Thank you
Interesting presentation. I've only watched the bit about FC mounting so far but wanted to show where I ended up after so much testing. It's kinda hard to figure out if that 'drum' effect you bring up actually happens or not and if it does, are the frequencies meaningful or do they just get chopped out by an LPF. Anyway, the latest design I have for 5" has a pretty unique structure to it in order to help manage these specific issues you're talking about. It seems to do a pretty good job of it but again, hard to tell with only a dozen or so builds. Also, the vibrational/resonance issue isn't the main reason why I personally hate stack screws integrated with arms. It just makes it a total PITA to deal with when you need to replace an arm or deal with really anything at all. See this image for the structure of the Prototype 5: ibb.co/chYBhjQ
I've been trying to tune a 7" for months and just last week was able to make progress. It was just as you describe - a resonant peak at 78Hz - which meant I couldn't reduce filtering, and so couldn't raise PIDs without all sorts of extra oscillation (the bumpy road look). I didn't try a notch, but did remade the bracket holding the GoPro to be thicker and stiffer, and the peak has dropped away almost completely. Still have some more tuning to do, but I've been able to remove a lot of gyro filtering and push P and D up. Fixing mechanical issues really seems critical! I didn't know the camera mount could make so much difference, as a lot of quads seem to have the camera strapped on with no particular system to prevent resonance. Where I live there is almost always some wind so I need to get P and D quite high to prevent bobbles. Hopefully I'm on the right track now.
Thank you so much man! First time seeing your channel and I really appreciate the calm and peaceful dialog with it all. Informative and relaxing content while I'm contemplating our next shendrones build.. thanks again!
Hey Sam. Thank you for this great video. Your channel is one of a kind. I am still struggling with a resonance spike at 35Hz. Should I noticed that I also have the same spike at the same frequency in the Pid D term. Does this mean that I should apply my notch filter on the Pid D term?
You only need a notch filter on the gyro. The d-term is merely reacting to the p-term, attempting to damping the oscillation. The d-term always mimics the p-term, in fact it is always slightly ahead of the p-term in an attempt to "predict the future". This is why you see the d-term going crazy when you have resonance, because 1) it mimics the p-term/motors and 2) it is more sensitive than P and I. The D stands for derivative, that same kind that is used to attempt to predict the stock market (stonks). I've never used a notch filter on the d-term ever before with any kind of improvement yet, but that's not to say it it's not worth trying out for funsies.
This video is amazing, thanks for making it... The phase delay stuff was something i was thinking about so glad you talked about it... My last step is going up in the wind and i usually have to push pids an extra 20 in windy environments, although im just working on heavier longrange rigs for now... might not be as bad on x8's Ive recently designed a 10" frame with flight controller mounted through the bottom plate and not the arms, but i made top/bottom plate 4mm thick hopefully that helps remove some of that resonance you talked about
Would you ever want to intentionally join the camera AND drone resonance together intentionally so you’d only have to apply a single notch filter for both?
Every time I've encountered camera payload resonance it's just one peak usually on the pitch axis. As such, by applying a gyro notch filter at that frequency, you eliminate the resonance. Resonance occurs because the camera's natural frequency and the drone's natural frequency are close enough together to excite each other. The two systems together cause one resonance frequency and thus you need just one notch filter. Unless of course you are getting a resonance on the roll axis as well (which is possible). In that case you would need two notch filters. At that point though, design changes must be made. As an addendum, there are ways to nullify resonance before it even starts (mechanically) without having to resort to notch filters. This is some secret sauce though that will be revealed by my drone designs when PerformanceFPV launches.
Thanks Sam - really nice walk through ! It's noticeable how much time you spend with this topic ! BTW, do you have looked into gyroflow for stabilization ? If you are interested, would be great to get some "real camera samples" to work on. We basically just need video + Betaflight Blackbox recorded simultaneously and need to make a lens preset based on a workflow you how from steadyXP.
Hi, I haven't used gyroflow yet before, but I am part of the FB group and loosely follow the topic. Right now my focus, and the only thing I have time for, is improving the drone platform itself as much as possible. I've used SteadXP in the past and made that video, but I am hopeful that gyroflow can one day surpass SteadXP and be able to process high bitrate cinema codecs.
@@strawhatsamsounds like a good plan - Elvin does a great job with gyroflow and braw/prores are on the list - as well as hardware to log more efficient/easy. 😉✌️
I got rejected from the engineering program at most of the schools I applied to because my math scores sucked, but now I feel like I am getting a much more fun education from doing this drone thing and learning from you Sam! Thank you
Do you have cinelifter frame for sale?
Interesting presentation. I've only watched the bit about FC mounting so far but wanted to show where I ended up after so much testing. It's kinda hard to figure out if that 'drum' effect you bring up actually happens or not and if it does, are the frequencies meaningful or do they just get chopped out by an LPF. Anyway, the latest design I have for 5" has a pretty unique structure to it in order to help manage these specific issues you're talking about. It seems to do a pretty good job of it but again, hard to tell with only a dozen or so builds. Also, the vibrational/resonance issue isn't the main reason why I personally hate stack screws integrated with arms. It just makes it a total PITA to deal with when you need to replace an arm or deal with really anything at all. See this image for the structure of the Prototype 5: ibb.co/chYBhjQ
I've been trying to tune a 7" for months and just last week was able to make progress. It was just as you describe - a resonant peak at 78Hz - which meant I couldn't reduce filtering, and so couldn't raise PIDs without all sorts of extra oscillation (the bumpy road look). I didn't try a notch, but did remade the bracket holding the GoPro to be thicker and stiffer, and the peak has dropped away almost completely. Still have some more tuning to do, but I've been able to remove a lot of gyro filtering and push P and D up. Fixing mechanical issues really seems critical! I didn't know the camera mount could make so much difference, as a lot of quads seem to have the camera strapped on with no particular system to prevent resonance.
Where I live there is almost always some wind so I need to get P and D quite high to prevent bobbles. Hopefully I'm on the right track now.
Thanks, man. Very good info material. Be sure to by a good mic to record your voice.
Thank you so much man! First time seeing your channel and I really appreciate the calm and peaceful dialog with it all. Informative and relaxing content while I'm contemplating our next shendrones build.. thanks again!
Hey Sam. Thank you for this great video. Your channel is one of a kind. I am still struggling with a resonance spike at 35Hz. Should I noticed that I also have the same spike at the same frequency in the Pid D term. Does this mean that I should apply my notch filter on the Pid D term?
You only need a notch filter on the gyro. The d-term is merely reacting to the p-term, attempting to damping the oscillation. The d-term always mimics the p-term, in fact it is always slightly ahead of the p-term in an attempt to "predict the future". This is why you see the d-term going crazy when you have resonance, because 1) it mimics the p-term/motors and 2) it is more sensitive than P and I. The D stands for derivative, that same kind that is used to attempt to predict the stock market (stonks).
I've never used a notch filter on the d-term ever before with any kind of improvement yet, but that's not to say it it's not worth trying out for funsies.
What a legend you are
This video is amazing, thanks for making it... The phase delay stuff was something i was thinking about so glad you talked about it... My last step is going up in the wind and i usually have to push pids an extra 20 in windy environments, although im just working on heavier longrange rigs for now... might not be as bad on x8's
Ive recently designed a 10" frame with flight controller mounted through the bottom plate and not the arms, but i made top/bottom plate 4mm thick hopefully that helps remove some of that resonance you talked about
You're welcome! Yes with sandwich plates as thick as 4mm and with the edge of the FC as close as possible to the arms, I think you'll be fine.
I don’t own a cinelifter, yet, but this video was really helpful to me! Subbed ;)
When dealing with this rigs always use ear protection. Be kind to your ears so they do not ring for the rest of your life.
Great informative video
Puts light on some issues with latest designs.
Great Content Bro - Thanks for sharing
Thanks for the vid bud
Very nice video, Good job!
Would you ever want to intentionally join the camera AND drone resonance together intentionally so you’d only have to apply a single notch filter for both?
Every time I've encountered camera payload resonance it's just one peak usually on the pitch axis. As such, by applying a gyro notch filter at that frequency, you eliminate the resonance. Resonance occurs because the camera's natural frequency and the drone's natural frequency are close enough together to excite each other. The two systems together cause one resonance frequency and thus you need just one notch filter. Unless of course you are getting a resonance on the roll axis as well (which is possible). In that case you would need two notch filters. At that point though, design changes must be made.
As an addendum, there are ways to nullify resonance before it even starts (mechanically) without having to resort to notch filters. This is some secret sauce though that will be revealed by my drone designs when PerformanceFPV launches.
@@strawhatsam excellent! Stoked to see what’s next! Thanks for sharing your knowledge in this video
Thanks Sam - really nice walk through ! It's noticeable how much time you spend with this topic !
BTW, do you have looked into gyroflow for stabilization ? If you are interested, would be great to get some "real camera samples" to work on. We basically just need video + Betaflight Blackbox recorded simultaneously and need to make a lens preset based on a workflow you how from steadyXP.
Hi, I haven't used gyroflow yet before, but I am part of the FB group and loosely follow the topic. Right now my focus, and the only thing I have time for, is improving the drone platform itself as much as possible. I've used SteadXP in the past and made that video, but I am hopeful that gyroflow can one day surpass SteadXP and be able to process high bitrate cinema codecs.
@@strawhatsamsounds like a good plan - Elvin does a great job with gyroflow and braw/prores are on the list - as well as hardware to log more efficient/easy.
😉✌️
What I took away from this video:
what is that? = “que es eso”
Also- with resonance, it DOES magically go away