My quad keeps getting newer.

[makattack]

I've build two mini quads now, an old FPV250/plastic & glass framed mini quad from HobbyKing with afro 20A ESC's and multistar 1806 motors and a flip32 FC. It was my first quad, and I pretty much used it as a learning platform. I built a newer, cleaner version using trimax's Diamondback G10 based frame that I bought for FTFF2015 (but ultimately didn't attend). That one was using newer gear and a PDB which made it cleaner/easier to setup.

Just for giggles, I decided to blow off the dust on the first mini quad (the FPV250) and update the CF software along with the ESC's. They were running SimonK but I now have the latest BLHeli firmware on it, and the flip32 has CF 1.14.1 (despite the known bugs).

The changes were great enough between the old version I had and the current version that I decided to tune it from scratch. First I checked to see that it flew with the default PIDs, and it did. It was super stable, but not particularly agile. This is stock setup with the now default luxfloat/default PID's. The blheli setup was also pretty much stock except for the throttle range calibration (done in blhelisuite). Other than that, CF was also setup with the proper min/max throttle, and I enabled oneshot125.

While I expected it to fly well stock, I was surprised at my manual LOS tuning effort. I don't have an openlogger blackbox connected, as I like to use UART1 for the bluetooth modem, and I use SBUS so UART2 is assigned to that.

Anyway, with luxfloat, when I started to manually tune, I set the roll/pitch I and D values to 10 (values that seemed low, but would still let it fly) and kept the P at the default (40?). I setup the RX channels to perform the inflight tuning, and when I started increasing the P value, it just kept going up to 200 before I stopped trying to increase P. Still no oscillations and it's definitely more responsive.

I feel like I can keep increasing the P value even more, but it just seems wrong to me.

Has anyone experienced this with the above software and settings combination on a miniquad size? I know it's hard to compare, but this sort of behavior really has me confused. I know the scaling/values used for PID tuning have changed over time, but with the previous version of CF I had (1.8 I believe -- yes, I don't fly multirotors much), I had found that luxfloat didn't like really high P values. Of course, I also was running simonk firmware as well without the damped light / oneshot implementation.

Anyway, just thought I'd see what others are seeing for top end P values in similarly equipped quads.

 

[PsyBorg]

Did you set your rates before tuning? If rates are low then I believe the software looks at it like you are setting up an AP quad and want super stable flight. I am not familiar with the latest changes but the 4, 6 9 (40, 60, 90) depending on what software is used and where they use the decimal places for accuracy rule is pretty common across the board for tighter racing / free style quads.

I have my roll and pitch rates set to 900 degrees per second and yaw at 1000. I know Kiss settings are different from clean flight so you may have to use the online calculator to see how many degrees per second you are actually spinning in clean flight. You need to find the rates you are comfortable using before you start tuning or its all kind of wasted effort.

 

[makattack]

Ah, thanks PsyBorg! I'll look into that. I left the rates at default as with the older software, I didn't need to adjust the rates just to tune. In fact, I used to tune, then set the rates. Amazing how things change.

 

[PsyBorg]

Take a look at this when you go thru the process again. This is what made tuning finally click for me.

 

[makattack]

Ah yes, I've seen skitzo's tuning video. I believe he's also a KISS user, and I just learned that KISS is also a fork of MultiWii, so I guess it's not all that different!

Was reading up a summary of RC Rates & Roll/Pitch Rates from this article:
https://oscarliang.com/rc-roll-pitch-yaw-rate-cleanflight/

Coupled with the CF documentation:
https://github.com/cleanflight/cleanflight/blob/master/docs/PID tuning.md

Leads me to think that actually tuning luxfloat with stock rates should be ok. I looked at the release notes for CF and did see that at one point, all the values changed such that the defaults would match the MW23/rewrite PID controllers. The version I last had, the lux controller would never fly with the same PID values without shaking the quad to death (severe oscillations)

That sort of explains why I'm seeing a high value of P... I think I need to push the envelope a bit more. I only charged up two batteries, and because I really wasn't believing what I was seeing, I basically used one to get to the P=200, I=10, D=10 setting, and the other just flying around to see that it could actually fly. I'll push it up more to see how far I can go.

I'm going to push the PID tuning first because according to the CF docs, the RC rate and Pitch/Roll rates only affect stick sensitivity with LUX. I'm basically trying to push the P value up until I see oscillations, then back down a bit and push up D to smooth out more, and I until it holds angles in wind.

I believe what's happened is that by:
* updating my esc's to blheli 14.8 which have damped light.
* enabling oneshot125
* Upgrading the FC to CF 1.14.1 with the updated LUX controller (since 1.13 according to the release notes)

I just have to expect a set of very different tune values for LUX. I somehow had imagined that maybe they setup the PID controller such that unwanted oscillations won't happen any more with some auto damping cut off feature. I know they added a GTune feature, which I didn't enable, but thought maybe it was in play somehow.