I had a problem with my new quad today. 
When I arm it the quad just goes almost full throttle and shoots up.
I tried everything:

• Check motor orders, spin direction, props
  
• Board orientation
  
• Change ESC protocols
  

But the only thing that fixed it was disabling 32KHz sampling rate option, and stick with 8KHz or lower looptime, Despite this is the Kakute F4 and should be capable with 32KHz...
While I am looking for the answer I thought I would share that info here in case someone run into the same problem and wondering why.

I've read somewhere that it's a commom problem with the Kakute and although advertised, it can't quite handle 32/32 (I'm guessing your CPI usage is pretty high?) Have you tried 32/16? Take off on arming is usually a sign of noisy motors or too much frame flex causing d term feedback. If you've got D on PT1 maybe change it to biquad and see if that makes a difference but heavier filtering sort of defeats the purpose of higher rates. Maybe someone with more knowledge of filtering can chime in. I will add though that I was just fiddling around with a new build with the Radix FC and when I tried 32/32 I had to lower the PIDs almost 50% before it was even flyable

So the problem turned out to the noise!

When the gyro have 32KHz sampling mode enabled, it becomes so noisy even when it was just sitting on the floor, it caused the quad to freak out when armed.
The fix was to to add some more filtering for the gyro.

set gyro_stage2_lowpass_hz = 150

The quad flies beautifully now with 16K/8K.

---

(13-Apr-2018, 07:34 PM)RENOV8R Wrote: I've read somewhere that it's a commom problem with the Kakute and although advertised, it can't quite handle 32/32 (I'm guessing your CPI usage is pretty high?) Have you tried 32/16? Take off on arming is usually a sign of noisy motors or too much frame flex causing d term feedback.  If you've got D on PT1 maybe change it to biquad and see if that makes a difference but heavier filtering sort of defeats the purpose of higher rates.  Maybe someone with more knowledge of filtering can chime in.  I will add though that I was just fiddling around with a new build with the Radix FC and when I tried 32/32 I had to lower the PIDs almost 50% before it was even flyable

The problem was unrelated to looptime, as I even tried 8K/8K and the problem was still there. see my post above for fix