Why does my quadcopter shake?
Shaking is four completely different faults wearing the same costume. Identify which one you have in about a minute, then fix that one.
The symptom
The aircraft is not flying smoothly. Somewhere between "the video looks slightly buzzy" and "it is visibly juddering in the air," and you want it to stop.
The problem is that "shaking" is four unrelated faults with the same outward appearance. Fixing the wrong one costs you an afternoon and usually makes the aircraft worse: the standard reflex is to lower gains and pile on filtering, which does hide the symptom, at the price of a mushy, drifty aircraft with a slow, dangerous PID loop. So identify the fault first.
Check this in 60 seconds
Do these three before you change a single setting.
- Hold the quad in your hand, arm it, and bring the throttle up slowly (props off if you value your fingers; you will lose the vibration signature, but you will keep the fingers). Does the buzzing start at a specific throttle, or is it there the moment the motors spin?
- Spin each motor by hand. Any grinding, notchiness or side-to-side play in a bearing? Any wobble in a prop or bent shaft?
- Look at the gyro trace, not the aircraft. In Betaflight, open the Configurator's Motors tab, or better, pull one Blackbox log. You cannot tell a 300 Hz vibration from a 20 Hz oscillation with your eyes, and those two have opposite fixes.
That third point is the whole article. Everything below is a way of reading the trace.
Which of the four is it?
| What you see | Frequency | Almost always |
|---|---|---|
| Slow wallow, a visible "breathing" wobble, worse in fast forward flight | low — you can count it by eye | P too high (or too low — see below) |
| Fast buzz that appears under throttle, motors get hot, audible whine | high — you hear it, you cannot see it | D too high — the classic |
| Buzz proportional to throttle, present even hovering, motors cool | tracks RPM | Mechanical — prop, bearing, bent shaft |
| Sharp juddering only when descending or in your own downwash | irregular | Prop wash — not a fault |
Resist the urge to memorise exact frequency bands. The numbers quoted around the internet — a few tens of hertz for a P oscillation, a few hundred for a D oscillation — are ballparks for a 5-inch quad, and they shift downward substantially on a heavy airframe, which has far more rotational inertia and a correspondingly lower natural frequency. What transfers between airframes is the ordering: gain oscillations are slow enough to see, D-term oscillation is fast enough to only hear, and mechanical vibration tracks RPM. Read your own log rather than somebody else's table.
1. Low-frequency oscillation — the gains
A slow, rocking wobble you can count with your eyes is the P term fighting the aircraft's inertia. On a heavy platform this is the one you meet first, because a 5 kg airframe has a much lower natural frequency than a 250 g racer, and the default tune was written for the racer.
Counter-intuitively, this can also mean P too low: an under-gained aircraft does not hold attitude, so it sags and gets corrected, sags and gets corrected. The tell is the phase — if the wobble grows when you push the sticks and settles when you centre them, that is too much gain. If the aircraft feels vague and never quite arrives at the attitude you asked for, it is too little.
Fix: change P in steps of about 10%, one axis at a time, and re-fly the same manoeuvre each time. Do not touch three things between test flights — you learn nothing.
2. High-frequency oscillation — the D term
This is the one people mean 80% of the time. D is a differentiator, and differentiators amplify noise; the higher your D, the more gyro noise the FC turns into motor commands. Those motor commands feed back into the frame as vibration, which the gyro sees, which D amplifies again. That loop is what heats the motors.
Hot motors after a gentle flight are the diagnostic. A motor that is uncomfortably hot to hold, after a hover that demanded no real work from it, is a motor being commanded to oscillate hundreds of times a second. It is the single most reliable sign that D is too high — or that filtering has collapsed.
Fix, in order:
- Lower D by 10–15% and re-fly. If the buzz and the heat go away, you found it.
- Check your RPM filter is actually working (bidirectional DShot, motor poles set correctly). RPM filtering is the single biggest thing you can do for a noisy build, and it fails silently.
- Only then reach for the low-pass filters — and understand that every hertz of extra filtering is a millisecond of extra delay in the loop. Filtering is not free. It is a loan against your control authority.
3. Mechanical vibration — the frame, not the firmware
If the buzz scales cleanly with RPM and is there the instant the motors spin, no amount of tuning will help you. Something is physically out of balance.
In descending order of likelihood: a chipped or bent prop; a prop with a manufacturing imbalance (cheap props vary wildly); a motor bell that is out of true after a crash; a bad bearing; a bent motor shaft; a loose arm bolt; the FC's soft mounts hardened, torn, or over-tightened so they no longer isolate anything.
The test is brutal and effective: swap all four props for a fresh set. It takes two minutes and eliminates the most common cause outright. If the buzz survives four fresh props, start swapping motors between arms and see whether the vibration follows the motor.
4. Prop wash — this is not a fault
Juddering that happens only when you descend into your own downwash, or when you hard-brake, is the aircraft flying through turbulent air of its own making. Every multirotor does this. It can be reduced with a good tune (this is what D_min, dynamic idle and a healthy D term are for) but it cannot be eliminated, and chasing it to zero will wreck an otherwise good tune.
If your quad only shakes in descents, you do not have a problem. You have a quadcopter.
How to know you actually fixed it
The bar is not "it feels better." Vibration diagnosis is exactly where feel lies to you, because you want the last change to have worked.
- Motor temperature after a two-minute hover is back to warm rather than hot.
- The gyro trace at hover is visibly cleaner in the log — not just "the aircraft seems okay."
- The fix survives a different flight. Aggressive manoeuvres, full throttle, a descent. Faults that only show up under load are the ones that bite.
Why this is worth simulating first
Every step above is a hypothesis followed by a test flight. On a 5 kg airframe, each of those test flights costs a battery cycle, twenty minutes, and a small non-zero chance of destroying several hundred euros of hardware while you deliberately drive a control loop toward instability.
That is the exact loop a simulator collapses. Run the same firmware against a physics model of your build, wind D up until it screams, and you will have seen the failure mode — in the trace and in the motor commands — before you ever fly it. Then you know what you are looking at when it happens for real.