Your quad flies fine at 30% throttle, then turns into an angry washing machine at 60%. You add D-term. Now it overshoots on flips. You back off D, and the mid-throttle wobble comes back. This isn’t a PID problem — it’s a filter problem. The noise from your motors at mid-RPM is hitting a frequency your lowpass filters aren’t catching, and the PID loop is amplifying the garbage. Fix the filter, and the wobble disappears without touching a single PID value.
How Betaflight Filters Actually Work
Betaflight’s filtering chain runs in this order: gyro raw data → gyro lowpass 1 → gyro lowpass 2 → dynamic notch → D-term lowpass 1 → D-term lowpass 2 → PID controller. Each stage removes noise above its cutoff frequency, but every filter also adds delay. The art of filter tuning is removing enough noise for clean flight without adding so much delay that the quad feels sluggish.
Gyro Lowpass Filters
The gyro lowpass filters clean the raw gyroscope data before it reaches the PID loop. A noisy gyro signal means noisy P and D calculations — the quad twitches because the FC thinks it’s moving when it isn’t.
Lowpass type options:
– PT1: The simplest filter. 6 dB/octave rolloff. Minimal delay (~0.5 ms at 200 Hz cutoff). Use when noise is low and you want maximum responsiveness.
– BIQUAD: 12 dB/octave rolloff. Moderate delay (~1 ms at 200 Hz cutoff). The default for most builds. Good balance of noise rejection and responsiveness.
– PT2: Two cascaded PT1 filters. 12 dB/octave rolloff. Similar to BIQUAD but with different phase characteristics.
– PT3: Three cascaded PT1 filters. 18 dB/octave rolloff. Heavy filtering, significant delay. Only for very noisy builds.
Cutoff frequency: Start at 200 Hz for 5-inch builds with properly balanced props. Drop to 150 Hz if you see mid-throttle oscillations. Go up to 250 Hz if the quad feels mushy. Each 50 Hz change is significant — don’t jump 100 Hz at a time.
Dynamic Notch Filter
The dynamic notch tracks the motor RPM peak in the frequency spectrum and places a narrow notch filter directly on that frequency. This is the single most important filter for eliminating mid-throttle oscillations because motor noise peaks at the RPM of the spinning props, which changes with throttle.
The dynamic notch settings that matter:
– Dynamic notch range: Low = 80 Hz, High = 500 Hz. The notch will track any noise peak between these frequencies.
– Dynamic notch width: 0% = extremely narrow, 50% = wide. Start at 0% and increase only if the notch isn’t catching the full width of the noise peak.
– Dynamic notch Q: 100-250. Higher Q = narrower but more precise notch. Lower Q = wider but removes more adjacent frequencies.
D-Term Lowpass Filters
D-term amplifies noise because it’s a derivative — tiny gyro jitter becomes large D-term spikes. The D-term lowpass filters are more aggressive than gyro filters by default because D-term needs cleaner data.
- D-term lowpass 1: PT1 at 150 Hz on most default tunes. This is the first line of defense against D-term noise.
- D-term lowpass 2: BIQUAD at 120 Hz on most default tunes. Heavy filtering with more delay.
- D-term lowpass 2 type: Can be set independently from gyro lowpass 2.
Filter Settings Reference Table
| Filter Stage | Default Type | Default Cutoff | Effect of Lower Cutoff | Effect of Higher Cutoff |
|---|---|---|---|---|
| Gyro LPF 1 | BIQUAD | 200 Hz | Less noise, more delay | More responsive, risk of noise |
| Gyro LPF 2 | PT1 | 250 Hz | Cleaner gyro data | Less filtering delay |
| Dynamic Notch | N/A | 80-500 Hz | Catches motor noise | Automatic tracking |
| D-Term LPF 1 | PT1 | 150 Hz | Less D-term spikes | More responsive D |
| D-Term LPF 2 | BIQUAD | 120 Hz | Cleaner D, more delay | Sharper D response |
| RPM Filter Harmonics | 3 | N/A | Catches 2nd/3rd motor harmonic | More CPU usage |
Filter Tuning Mistakes That Waste Hours
Mistake 1: Lowering D-term filter cutoff before fixing mechanical noise. A wobbly prop, a bent motor shaft, or a loose arm creates broadband vibration that no filter can fully remove. If your blackbox log shows a solid noise peak at motor RPM that the dynamic notch isn’t catching, check your props before touching filters. I’ve spent an afternoon tuning filters around a chipped propeller that a 30-second inspection would have caught.
Mistake 2: Running both gyro lowpass filters at the same cutoff with different types. BIQUAD at 200 Hz plus PT1 at 200 Hz is two filters at the same frequency adding delay without adding meaningful noise rejection. Stagger them — BIQUAD at 200 Hz and PT1 at 250 Hz, or PT1 at 180 Hz and BIQUAD at 250 Hz. The staggered cutoff frequencies create a broader noise rejection zone without a double delay penalty at one frequency.
Mistake 3: Disabling the dynamic notch for “lower latency.” The dynamic notch adds approximately 1 ms of delay and removes the biggest noise source in the system. Disabling it to save 1 ms while adding 5 ms of D-term filtering delay to compensate is a net loss. Keep the dynamic notch enabled and tune your static lowpass filters around it.
Mistake 4: Setting RPM filter harmonics to 1. RPM filtering targets the exact motor frequency and its harmonics. Setting harmonics to 1 only filters the fundamental frequency, leaving the second and third harmonics (which are often stronger in noise amplitude) unfiltered. Set harmonics to 3 as a baseline.
Mistake 5: Copying filter settings from another pilot’s build. Filters depend on frame resonance, motor balance, prop condition, and build stiffness. A tune that works on a stiff 5-inch Armattan frame will be wrong on a flexible 3-inch toothpick. Start from Betaflight defaults for your build type and tune from there.
⚠️ Regulatory Notice: The 2026 FAA and EASA regulations include noise emission standards for certain categories of UAS that indirectly relate to filter tuning. An improperly filtered quad that produces excessive vibration-induced noise may exceed local noise ordinances in residential flying areas. The CAAC (China) 2026 update to UAS noise regulations introduces maximum sound pressure level (SPL) limits at 10 meters for different weight classes — vibration-induced noise from unfiltered motor harmonics contributes to measured SPL. Proper filtering is both a performance and compliance concern.
Our guide to Betaflight RPM filtering covers the RPM filter setup that pairs with the static lowpass filters in this article — the two filtering systems work together.
The blackbox log analysis guide shows you how to read gyro scopes to identify exactly which noise frequencies need filtering and which are already handled.
Clean filtering starts with clean hardware. The uavmodel F7 flight controller uses the ICM-42688-P gyro with hardware lowpass filtering that removes high-frequency noise before it reaches the MCU — your software filters have less work to do, which means less delay and a tighter-feeling quad.