How to fix propwash in Betaflight
Propwash is the oscillation you get when the quad flies back through its own disturbed air, mostly on dives, split-S, hard turns and throttle chops. You reduce it by letting D work harder without letting noise through: clean up the noise floor first (RPM filter, less filter delay), then adjust the P:D balance, then D max and I-term relax. Propwash can be reduced, not removed.
What is propwash?
The props push air down. When you descend, turn hard or chop the throttle, the quad drops into that churned air and the props lose grip on it. Thrust becomes uneven for a moment, the quad gets knocked about, and the flight controller has to catch it.
In the air it looks and feels like a fast shake in the video, strongest at the bottom of a dive or on the way out of a split-S, and it fades as you reach clean air again.
In a blackbox log it is energy at roughly 20 to 100 Hz in the gyro and the D-term, concentrated in the moments you descend or chop the throttle. It is real motion of the quad, so it shows in the filtered gyro and in the unfiltered gyro alike. Motor noise and frame resonance look different: they sit at higher frequencies, and motor noise moves with throttle. See blackbox logging for tuning for how to record one, and how to read a step response for what the log tells you.
Why don’t filters fix propwash?
Filters remove noise, which is vibration that the quad is not actually doing. Propwash is movement the quad really makes, at frequencies far below any sensible filter cutoff. A filter that reaches down to 20 to 100 Hz would only delay your gyro signal, and a delayed signal makes the controller react late, which makes propwash worse.
The filters matter in a different way. D is the setting that fights propwash, and D amplifies noise. The cleaner the signal reaching D, the more D you can run before the motors get hot. That is why the noise floor comes first. The filter tuning guide covers that side.
What helps, in order
Change one thing at a time, in small steps. After each change, arm on the ground, hover for about 20 seconds, land and touch the motors before you fly hard.
1. Get the noise floor down so D can work
- RPM filter. With bidirectional DShot on (
dshot_bidir, defaultOFF), the RPM filter removes motor noise precisely and with little delay. That lets you open the D-term lowpass filters and run more D. See the RPM filter setup guide. - Less filter delay. Every lowpass filter delays the signal. If the log shows low, flat noise above 100 Hz, raise the cutoffs a step. D-term lowpass 1 defaults to 75 to 150 Hz dynamic (
dterm_lpf1_dyn_min_hz75,dterm_lpf1_dyn_max_hz150) and lowpass 2 to 150 Hz (dterm_lpf2_static_hz) in every version this guide covers. Do not remove filtering altogether.
2. Adjust the P:D balance
Propwash gets worse when P is high compared with D, because P keeps pushing while D is not damping. Raise D in small steps, or lower P. Raising D warms the motors, so check them after every flight. The defaults are P 45 and D 30 on roll, P 47 and D 34 on pitch.
D has two names depending on your version, so check the setting-name converter:
| What | 4.3, 4.4, 4.5 | 2025.12, 2026.6 |
|---|---|---|
| Base D (roll, default 30) | d_min_roll |
d_roll |
| Peak D (roll, default 40) | d_roll |
d_max_roll |
3. D max and D max advance
D max raises D above its base value when the quad moves fast or the gyro changes quickly, so you get more damping in the moments propwash happens, while the motors stay cooler the rest of the time. Raising peak D can help propwash if base D is already where the motors allow.
Two related settings exist: d_max_gain (default 37 up to 2025.12, 0 in 2026.6) and d_max_advance (default 20 up to 2025.12, 35 in 2026.6). Check what your version defaults to before you change them. A peak D equal to base D means D max does nothing.
4. I-term relax
I-term relax holds back I-term build-up during fast stick moves. It defaults to iterm_relax="RP", iterm_relax_type="SETPOINT", iterm_relax_cutoff=15. If the quad rolls out of a dive and I-term has wound up, the recovery gets slow and sloppy. The default is a good place to stay; see the feedforward, I-term and TPA guide for when to move it.
5. Dynamic idle
dyn_idle_min_rpm (default 0, which is off) keeps the motors from dropping to a very low speed on throttle chops, so the props keep some grip on the air. It is a motor setting and needs bidirectional DShot. It is mentioned here because it affects propwash, and it is not part of PID tuning.
6. Flying technique
Propwash is worst when you chop the throttle to zero in a dive. Keep some throttle in a split-S, and roll out of a dive earlier than feels natural. No setting replaces this.
Fix these first, tuning can’t
Bent or damaged props, a loose flight stack, loose motor screws and worn bearings all make propwash look much worse than it is. Sort those out before you spend time on settings.
Typical order of changes
| Step | Change | Check |
|---|---|---|
| 1 | RPM filter on, filters tidy | Motor temperature after a hover |
| 2 | Raise filter cutoffs if the noise floor is low | D-term noise in the log |
| 3 | More D, or less P, in a small step | Motors still cool |
| 4 | Peak D and D max advance | Propwash after a dive |
| 5 | I-term relax, only if recovery is slow | Behaviour out of split-S |
How do I test propwash?
Fly the same short sequence before and after each change: a few split-S and a few dives with a throttle chop at the bottom, on the same battery and the same props. Compare the same moves in the same log view. Wind and a fresh battery against a tired one both change the result, so keep conditions the same. The 20 to 100 Hz energy in the D-term and gyro should drop, and the shake in the video should be shorter.
What can I realistically expect?
Less shake, a shorter shake, and a quad that recovers faster out of dives. Propwash cannot be removed, because the air really is disturbed. Frame, props and motor choice set a floor that tuning cannot get below. If more D only makes the motors hot, stop there.
How FPV PID Tune handles it
The test pack includes split-S and punch-outs, which give the analysis the propwash evidence it needs. The app reads your log in the browser, measures the 20 to 100 Hz energy, and treats it as a PID problem rather than a filter problem. An AI picks values, and a validator holds every change inside fixed safety limits, with small steps per round. Every value is read back after Apply, and any mismatch is undone automatically.
