Filters and PID tuing in Betaflight

How I tune filters and PIDs in Betaflight using PIDtoolbox pro (a musthave, current v.0.93)

radio

• ADC Filter OFF
• global functions -> GF1 ON Lua Script bfbkgd for time sync from radio’s RTC to the betaflight flight controller

quad

• must have a capacitor
• must have new props installed, all screws tightened, no dangling antennas or XT60
• stack bolts must be secured to the frame with nuts first

ESC

• ESC tuning
• https://esc-configurator.com or https://am32.ca, or locally
• set PWM frequency to 48 kHz or 96khz for more thrust (and less braking force) and less overlap with harmonics
• set type to 2S+ if it is not a 1S AIO, disable temperature protection
• set motor timing to 15 degrees for better performance and cooler motors (go back to the default 22.5 deg in case if there are any desyncs) for a 5 inch, set to default for smaller quads
• motor settings in am32 must be set close to the actual ones. manual protocol selection DSHOT

Betaflight preparation

• backup previous config (dump and diff all showdefaults), flash the latest version of Betaflight using online app or locally
• load RC_LINK profile for the corresponding refresh rate. less smoothing will result in hotter motors when the tune is pushed to extremes。 lowering the factor to like 20 at ELRS150Hz gives me stepped setpoint curve on sharp moves like a roll.

  set rc_smoothing_auto_factor = 30
  set rc_smoothing_auto_factor_throttle = 30
  

• enable bidirectional dshot, set correct dynamic idle value is based on the prop size/pitch, here and here
• set PID authority to 100% from 50% default (do not change this after PID tuning).

  set pidsum_limit = 1000
  set pidsum_limit_yaw = 1000
  

• if MCU has enough compute power, set ESCs to DSHOT600, set pidloop frequency to set pid_process_denom = 1 (disable gyro low pass 2 if theres not a lot of noise above 500Hz) or to the highest frequency possible. pidloop below 2khz will disable dynamic notches, be aware.
• motor_poles must be set correctly (usually 12 for small motors and 14 for large motors)
• set blackbox to 2 kHz, debug fft_freq
• tune for lighter weight. quad tuned for heavier weight will fly away if lighter, especially in AIRMODE with pidsum_limit 100%

PID and filter tuning process

• Brian White and Joshua Bardwell https://www.youtube.com/watch?v=FsJNHI2HWlg
• tuning workflow https://www.youtube.com/watch?v=sqT4MACi3d8
• tuning workflow https://www.youtube.com/watch?v=ehvQm8Rqrzk
• tuning workflow https://www.youtube.com/watch?v=4lZ8BY_9KBs
• PID fundamentals 1 (must see) https://www.youtube.com/watch?v=tBI2TjePeWA
• PID fundamentals 2 (must see) https://www.youtube.com/watch?v=Ylb8mYi91DI
• PID fundamentals 3 (must see) https://www.youtube.com/watch?v=inOe7J6l2lU
• if needed, set linear rates 150/150, angle strength 50, angle limit 30
• for spectral analyzer tool only (step response tool ignores any movement less than 20 degrees per second): trim logs each time to remove takeoff and landing
• the process:
  • on the default PIDs, set ff and dmax to zero, iterm gain to 0.2 (also: lower MM for larger rigs, bump MM for underpowered quads)
  • record hover and wiggle, use both for filter adjustment
  • use PIDtoolbox to adjust the filters. apply the least amount of filtering (otherwise: high delay -> propwash): dynamic notches, filter weights and q to filter most of the noise while keeping the delay low, balance filter strength and delay caused by the filters. ideally keep the noise under -40 db (the noise is mostly meaningless below -30). leave gyro lowpass 2 on if there is noise after 500hz. there must be no broadband noise. broadband noise comes from electrical issues like missing capacitor or bad gyro/FC/AIO. try flashing ESCs from 24khz to 48khz or 96khz to isolate the problem. try changing props to a different model to lower the noise around the motor bands. dterm lowpass filter tuning
    • adjust

      set rpm_filter_weights = 100,100,100
      set rpm_filter_fade_range_hz = 50
      set rpm_filter_q = 500
      

  • in angle mode, record wiggle with dterm gain 0.6-1.8 step 0.2, in step response tool find the best curve without overshoot (to have a headroom for mm). compare roll and pitch latency and adjust pitch:roll balance (increase pitch damping/tracking if there is more delay on pitch, until the delay on pitch and roll is about the same)
  • in angle mode, record wiggle with master multiplier 0.6-1.8 step 0.2, choose the lowest latency before oscillations begin in step response tool
  • (optional) in angle mode, record iterm gain 0.5-2 step 0.5, choose the best curve without overshoot or oscillations in step response tool. lower iterm relax cutoff values reduce iterm windup for underpowered quads.
  • in acro mode, record ff gain 0.5-2 step 0.5, choose the closest following curve without overshoot, lowest latency (shortest period) between setpoint and gyro in the main window plots.
  • iterm https://www.youtube.com/watch?v=Sq_DFjmvVDE

(optional) use blackbox explorer to adjust filters

• https://www.youtube.com/watch?v=E3s5XYk3M74
• UAVTech’s workspace for the blackbox viewer is here

after tuning

• set the anti-gravity to 5, voltage sag compensation (depends on vbat_warning_cell_voltage), thrust linearization
• set motor_output_limit = 95 for a 5-inch to protect the ESCs
• set igains to 0.5-1.0. the lager the quad is, the lower the igains (5 inch to 1.0)
• do hover, slow rampup punchouts, forward flight, test for propwash (split-s, sharp 180 turns, dives), throttle chops to test antigravity (nose dives - bump up, throbbles - lower), rolls, flips, then analyze in PIDtoolbox
• watch for hot motors or overshoot on sharp moves in logs

problems

• about the washout problem with ducted frames, when quad not descending with lowered throttle. also, use props for ducted frames, gemfan D series
• hot motors, 1 2 . Too Much D-Term, Too Little Filtering, or Both. d-term oscillations caused by too much d gain. hot motors might be caused by RC smoothing being too low for the tune.
• wobbles
• propwash - increase master multiplier, increase dgain, decrease dterm filtering (dterm filter slider to the right), increase dynamic idle. also better throttle control and lower pitch props. UAV Tech - lower delay. UAV Tech - P/D balance. set pidloop frequency higher. the theory of propwash
• insufficient filtering coupled with pidsum_limit = 1000 and/or AIRMODE can cause a flyaway
• 13inch filters
• frame resonance noise - lower filters, lower mm
• low freq frame noise
• if there is a slow bounceback after a roll or a flip, lower I term gain
• if the noise floor is high on a heavily loaded quad, try higher PWM frequency (48khz -> 96khz) to get around
• high noise around motor frequency harmonics might be caused by fake or low quality props
• big quads require lowering iterm gains https://www.youtube.com/watch?v=M7mcUf05JmY
• if DSHOT_TELEM error pops up from time to time when arming, set motor_pwm_protocol = DSHOT300
• tinywhoop loses orientation after setting throttle to zero in acro but keeps steady if to do the same in air mode:

  set pidsum_limit = 1000
  set pidsum_limit_yaw = 1000
  

other tools for log analysis

• https://github.com/KoffeinFlummi/bucksaw
• PIDtoolbox fork using Octave https://github.com/dzikus/PIDscope
• https://skypulse.ua/pidpulse/

betaflight autotune

• autotune can be used to evaluate the current tune https://www.youtube.com/watch?v=uKX9W5skYJQ

build 2026 firmware with -DUSE_CHIRP, enable chirp injection (mode 55) on a switch

set debug_mode = CHIRP

• do: full throttle, rolls, flips. do chirps for each axis until the message end appears on the osd

references

• https://oscarliang.com/fpv-drone-tuning/
• https://oscarliang.com/pid/
• https://www.youtube.com/playlist?list=PLycwhvl4h5nNbqRvcVB8y6fSPAgMbBo8p
• https://www.youtube.com/playlist?list=PLFPBjpbd5xKQAzyblBStGKYtNdP9FCpRe