Wow, that's a great video! Impressive flying!

Great job with the video, L0stB1t. That's inspiring. I'm still working up to that level of acrobatics.

Saw this on YT when it dropped ... sick flying man!!!

Could you perhaps share a raw video feed with your settings? 

This quad is shipped with 0.9 on filters sliders. 
In my case, anything above 1.0 results in strong oscilations (especially mid-throttle). Not sure how did you managed to reduce filtering by that much 

I start to believe that these tiny arms are way too thin for these motors?

Hi Philip

My videos are all unstabilized, I don't get much problems with oscillations in the video.

Did you enable the RPM filter? That's probably a must to push those filter sliders up.

As I commented in #18, I do still notice quite a lot of oscillations in the blackbox logs. With PIDToolbox I discovered that there's some kind of resonance around 220 Hz, and at the harmonic of 440 Hz. I'm currently experimenting with setting fixed notch filters to crush those frequencies, and it seems to work really well but I need to do some more testing to find the best balance between filter delays and noise. By the way I noticed that Betaflight 4.3 will come with up to 5 dynamic notch filters, which would be excellent to automatically crush the noisiest frequencies!

I don't think the source of the problem is with the frame. If I spin up a single motor in the Betaflight Configurator, I get some clear resonance right around 34% throttle. When I firmly hold the corresponding arm, the resonance is still there. So I don't think the frame is the problem because my hand would absorb the vibrations before they get transfered to the rest of the frame, right? I guess it's just the motors that are not the best in quality. If there's anyone who has put different motors on this kwad, please let us know.

Here are some PIDtoolbox screens to clarify the noise I'm seeing. You'll find full resolution versions in the attachments.


The first one is from the "Early Spring Flight" video. I had the Gyro filter slider at 1.7 and the D Term slider at 1.5. I had the RPM Harmonics Number set to 2 and the Dynamic Notch range from 100 Hz to 350 Hz.

• The first graph shows all noise as seen by the gyro. I've marked the motor noise, you can see the frequency goes up as the throttle increases.
  
• The second graph shows the filtered gyro noise. You can clearly see that the RPM filter is doing its job because there's basically zero noise in the area I marked in the first graph. There's still the higher frequency noise that starts around 400 Hz at 20% throttle and goes up to 1000 Hz at 50% throttle. I think this is the third harmonic of the motors, so setting that number to 2 in the filters was a mistake, I've changed it back to the default of 3. Furthermore there's a thin line at about 220 Hz which stays there for the entire throttle range.
  
• The third graph shows all D Term noise. It basically amplifies all bad noise and clearly putting the D Term slider at 1.5 is too optimistic.
  

The second one is from a more recent test flight where I had the Gyro filter slider at 2.0 and the D Term slider at 1.5. The RPM Harmonics Number was set to the default of 3 and the Dynamic Notch range from 150 Hz to 500 Hz. So less filtering in the PT1 lowpass filters, and more in the notches.

On top of that I had Gyro Notch Filter 1 at center 220, cutoff 200. Gyro Notch Filter 2 at center 440, cutoff 400. And D Term Notch filter the same as Gyro Notch Filter 1.

• The gyro in the first graph shows pretty much the same noise as before. That's to be expected because the hardware is still the same.
  
• The filtered gyro noise however is much cleaner than before! There's practically zero noise above 100 Hz, only the useful movement data below 100 Hz remains.
  
• The D Term noise is also a lot cleaner as a result of the gyro filters and the extra D Term notch at 220 Hz.
  

I understand that all this is probably a bit complex for most people who prefer to stick to the basic sliders, but in Betaflight 4.3 it will probably be easy to set up additional dynamic notches, which should jump right on the correct frequencies so you don't have to analyse these things in the graphs yourself.

I just had a lunch break test flight where I tried other combinations of filters. If I learn anything interesting from the blackbox results I'll summarize it.

Hey L0stB1t, great job. Your log looks fantastic. I have still the original settings with BF 4.1 and so far only one flight (due to the bad weather in my area). It felt good but I from the beginning considered to update to BF 4.2 for the RPM filtering. But I was hesitant as there is no real good tuning based on 4.2 existing.

Can you share the diff for your recent tuning (the second screen)? This would be fantastic.

Hi MelBourbon

Since I don't know your level of tuning knowledge I'll try to explain as much as possible. I don't want to throw a complete diff at you because some of the settings could be dangerous if you don't understand what they do.

First of all before you upgrade to Betaflight 4.2, make sure you saved your "diff all" to a file. Flashing is pretty straightforward, just select the "MAMBAF722_I2C" version. After flashing, apply the defaults. Betaflight 4.1 and 4.2 configs are fairly compatible so you can probably apply all config lines from your saved diff starting from the "feature" lines and below.

Here are some settings that I recommend and should be totally safe:


Next, make sure to enable blackbox logging. At 2 kHz the 16 MB chip fills up in about 4 minutes, so just a single battery pack. I often set up 3 different PID profiles, fly each of them with aggressive moves (flips, propwash turns...) for 1 minute and for the next battery packs just enjoy flying with the profile which felt/sounded best. At home I compare the differences with PIDtoolbox.




Regarding the filtering, I recommend to just use the GUI. Since I'm not sure that your or all 6S Roma F5's have the same oscillations as mine I recommend to keep the Gyro sliders at 1.8, just in case:



In the diff it should look like this:


If you're sure that the 220 and 440 Hz notches do their job, you may want to disable the Gyro lowpass filters entirely, it works great for me:



It looks like this in the diff:




Now, regarding the PID tune itself, I'm currently happy with the following settings. Note that this is with Gemfan 51466 propellers, the best props for this kwad in my opinion. I found the provided HQ S5 a bit  weak. Gemfan 51433 were better but not a lot. HQ P3 (PB &J) are way too weak. Gemfan 51477 are slightly too heavy (but awesome if you want speed).

Even though I found this tune experimentally, it completely matches UAV Tech's preset for a 5" Freestyle kwad.

You don't have to change the I Term Relax Cutoff and you can leave Dynamic Idle Value empty. More on this below.



In the config:





The following settings are a bit more experimental. I'm not entirely sure myself if they really make a positive difference. These are all per-profile settings, so you may want to apply them for each profile.




Finally, don't forget to:




So, I hope this helps? I'd love to get feedback or have a look at anyone's Roma blackbox traces. Happy flying!

Here's my DVR recording of the lunch break test flight I did yesterday. It's with all lowpass Gyro filters disabled, works great! Read the video description for more details

Hi L0stB1t, this is amazing. Thanks. I have less experience with tuning but know the constraints in regards to applying diff etc. But so far I did tuning only by "feeling" or watching the DVR. Not used the Blackbox so far. But this is my goal for this year. So your details you provided are a great start for me as the Roma F5 is my newest quad. I will update and try to have a fly soon. We are having freaking cold temperatures here in Germany. So flying is no fun currently :-).

One question. If the Blackbox is full this will be no problem for flying the second or third pack, right? It will only not record further, right?

(20-Mar-2021, 01:03 AM)L0stB1t Wrote: Here's my DVR recording of the lunch break test flight I did yesterday. It's with all lowpass Gyro filters disabled, works great! Read the video description for more details

Omg, that's buttery smoooth! I'm not even near to that point with my tune   but going slowly forward and learning...
I have Mamba405US FC (DJI version) so this is probably the reason why I can't go as high with the filter sliders as you do (or not, will find out).
I'm messing up with blackbox second day, trying to figure out what's going on with my quad. So far I can see that the highest oscilations kick in 1300-1450 throttle range (mostly on D-gains).

Your findings are very helpful. Thanks!

(20-Mar-2021, 08:26 AM)MelBourbon Wrote: (...)

One question. If the Blackbox is full this will be no problem for flying the second or third pack, right? It will only not record further, right?

Yes, You can fly without any issues with full blackbox

(20-Mar-2021, 11:32 AM)philip100 Wrote: Yes, You can fly without any issues with full blackbox  

Also, in the Modes tab of Betaflight Configurator you can set up a switch to erase the Blackbox flash.  In the configurator it says it takes ">30s", but I bet it's much faster when you only have 16 MB of flash.  (I haven't tried it.)

(20-Mar-2021, 05:01 PM)SJChannel Wrote: Also, in the Modes tab of Betaflight Configurator you can set up a switch to erase the Blackbox flash.  In the configurator it says it takes ">30s", but I bet it's much faster when you only have 16 MB of flash.  (I haven't tried it.)

Is it also possible by using OSD. Because I don't want to use a channel for that as I use a Tango 2 which has already limited switches.

I applied the setup from L0stB1t and had 2 different PID profiles which I wanted to test (one was the one from L0stB1t). But unfortunately I crashed after 30 seconds or so because I was to aggressive instead of get used to the quad first. So I guess my Blackbox log will be not much helpful. Need to clean the mud from the motors first and will give it another try. But the first 30 seconds felt very good ;-)

(20-Mar-2021, 10:18 PM)MelBourbon Wrote: Is it also possible by using OSD. Because I don't want to use a channel for that as I use a Tango 2 which has already limited switches.

I applied the setup from L0stB1t and had 2 different PID profiles which I wanted to test (one was the one from L0stB1t). But unfortunately I crashed after 30 seconds or so because I was to aggressive instead of get used to the quad first. So I guess my Blackbox log will be not much helpful. Need to clean the mud from the motors first and will give it another try. But the first 30 seconds felt very good ;-)

My quad that uses a 16 MB flash chip for Blackbox has a DJI Air Unit on it, so I don't think I can do it with the OSD, unfortunately.  I'll have to use a channel for it when I try it.

In the Blackbox Explorer, you can use the "i" (for "in") and "o" (for "out") keys to ignore your takeoff and your crash.  Then you may be able to get some useful insights from what remains of the flight.

So, I did a analysis of my short blackbox and filtering in spectral analyzer looks pretty good, similar to the one from L0stB1t. Unfortunately my flight was with another PID profile which I wanted to check (found on youtube) - see below and attached. 

   

But what I noticed is that I have wavy lines of my PID D on roll and pitch in cases where not much stick movement was done (see below). What does this mean and has this any effect. I did not noticed something. The gyro data fits very well with the set point.