Fantastic testing, Mark! Pretty interesting results! 😃 But it's the first time I see a field of football fields... 😂 Oh, ok... Soccer. 😝 Anyway, stay safe there with your family! 🖖😊
These tests aren't really helpful. The pt1 filter has different cutoffs at lower pidloop rate. So your dterm filtering differs each time you change you pidloop rate, which can more than account for what you think the pidloop rate changes are doing.
@@Quick-Flash As you know, it is: K = (2 * pi() * cutoff) / loop rate PT1 = previous sample + K * (current sample - previous sample) Based on what I see, the PT1 seems to title out a little more delay at higher sampling rates, not less. i.imgur.com/4RhJkxg.gif
"Hey, i did this test, if you want go know what it actualy means, you have to go to my patrion and give me some of your money" While using open source software on the fc, and then open souce software to analize the data. Entrepreneur.
If your sampling at 8k and running pid at lower rates the science of signal processing shows only downsides to lower rates. The only real downsides to 32k are more cpu load and our sensor (aka the gyro) may see more noise. If the noise seen at 32k was the same as the noise at 8k then there is 0 benefit to 8k except for cpu load.
@@uavtech but if your living in 8k land then you get no advantage with dropping your pidloop rate when the gyro is sampled faster. Only advantage is less cpu load.
No. Our control frequencies stop at 100hz. So 200hz for control. For accurate noise capture, which is never higher than 900hz, 2k logging will suffice.
There are controll techniques that do this, but the assumption generaly is that the enviroment doesn't realy change. This is probaply not true enough for quads. I also doubt how well this will work for propwhash which is chaotic and its own research field.
![Seungmin "그렇게, 천천히, 우리(As we are)" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/kAzmhLHePqU/mqdefault.jpg)
Great testing! Glad you brought this up again. Love to see the new data now that BF4.3 has so much better control over looptime errors
Let me download the latest.
@@uavtech classic. +1
Fantastic testing, Mark! Pretty interesting results! 😃
But it's the first time I see a field of football fields... 😂
Oh, ok... Soccer. 😝
Anyway, stay safe there with your family! 🖖😊
Is there a change in feel between the different loop frequencies. A higher loop rate feels more responsive?
No, I don't see how that is possible.
thanks! so interesting!
Appears same amount of prop wash but where less on pitch axis more on roll axis and vise versa
Great topic, work and explanation. Thank you
Why you don't test with fix 1 factor (like logging) or dshot 600 will it perform better ?
Fly better for which?
@@uavtech it mean fix the 2 value and compare with one by one , just because more variable = more inconsistencies
These tests aren't really helpful. The pt1 filter has different cutoffs at lower pidloop rate. So your dterm filtering differs each time you change you pidloop rate, which can more than account for what you think the pidloop rate changes are doing.
So a 100 Hz PT1 at 8k isn't the same as a 100 Hz PT1 at 2k? Can you elaborate?
Nah. The calcs compensate for that (loop rate) when converting Cutoff to K factor.
@@l0stb1t correct, higher loop rates are more accurate though to the cutoff you set.
@@uavtech the calcs are approximations that are faster but not as accurate as the true answer.
@@Quick-Flash
As you know, it is:
K = (2 * pi() * cutoff) / loop rate
PT1 = previous sample + K * (current sample - previous sample)
Based on what I see, the PT1 seems to title out a little more delay at higher sampling rates, not less. i.imgur.com/4RhJkxg.gif
"Hey, i did this test, if you want go know what it actualy means, you have to go to my patrion and give me some of your money"
While using open source software on the fc, and then open souce software to analize the data. Entrepreneur.
HD and logs in this video. :-)
If your sampling at 8k and running pid at lower rates the science of signal processing shows only downsides to lower rates. The only real downsides to 32k are more cpu load and our sensor (aka the gyro) may see more noise. If the noise seen at 32k was the same as the noise at 8k then there is 0 benefit to 8k except for cpu load.
noise at 8k is far less than at 32k though. We crossed that bridge a long time ago.
@@uavtech but if your living in 8k land then you get no advantage with dropping your pidloop rate when the gyro is sampled faster. Only advantage is less cpu load.
@@Quick-Flash yeah, agreed.
More is better. Diminishing returns, sure. More is still better.
It is, but also misses/delays tasks. More on that this weekend
Doesn't nyquist mean you should log at 2x the fastest rate???
Nyquist means log at 2x the max frequency you want to examine.
@@uavtech so to do a comprehensive comparison you would want to log at 16khz to ensure nothing got missed.
No. Our control frequencies stop at 100hz. So 200hz for control. For accurate noise capture, which is never higher than 900hz, 2k logging will suffice.
@@uavtech ok, I understand, so your working assumption for this video is "nothing of interest above 900hz".
Not an assumption, but yes.
8k is the obvious answer because of aliasing going down from 8k
i don't see aliasing. You are assuming there is noise above 500hz. But is filed our before down-sampling.
why can't these flight controllers just teach themselves the correct settings? there's no lack of processing power.
unfortunately there is a lack of processing power, especially on F4's
Why can't cars just drive themselves already?
There are controll techniques that do this, but the assumption generaly is that the enviroment doesn't realy change. This is probaply not true enough for quads. I also doubt how well this will work for propwhash which is chaotic and its own research field.
auto bot infusion tune program .
Good testing! F1 will strongly disagree 😂
But F1 honestly does not fly very well at all (throbbles and loss of cross axis control in sharp stick moves) sooo... 🤷♂️
you so cool
What is your conclusion? Could you re think your video format. Not interesting and then again I don't think it was informative