Why a 5-inch racer tune destroys a 5 kg quad

The problem

You build something big. Ten inch, fifteen inch, a lifter with a camera under it, maybe twenty-two inches of carbon and a payload. You flash mainline Betaflight or iNAV, because that is what mainline firmware is for, and you get the default tune.

That default was written for a 250 g five-inch racer. Not "roughly derived from" — written for. It assumes an aircraft that can change its rotation rate almost instantly, that has eight times its own weight in thrust available, and whose structure is stiff enough to be treated as a rigid body. Your aircraft is none of those things.

So it oscillates. Or it does something worse: it holds together in a hover, then flips on the first hard input, or it cooks its motors in ninety seconds of gentle flying. And here is the part that makes it a genuine problem rather than a nuisance — there is nowhere authoritative to go for a better starting point. The stock defaults are built around small aircraft, and the further you get from a five-inch quad the thinner the official guidance becomes. So people trade numbers on forums: someone's 10-inch, someone else's X8, someone's 5 kg mapping rig with completely different arms. You paste them in, and the aircraft oscillates differently.

Nobody tells you why, so you cannot reason about it, so you are reduced to guessing with expensive hardware. This article is the why.

Everything that changes when the aircraft gets big

Five things change. They are not independent, and they all push the same direction: the aircraft gets slower, and your control loop has to get slower with it, or it will tear itself apart.

Rotational inertia

Moment of inertia scales roughly as mass times radius squared. Both of those numbers go up on a big build, and the radius term is squared.

Take a 250 g racer with motors about 110 mm from centre, and a 5 kg platform with motors 350 mm out. The mass is twenty times larger and the radius is roughly three times larger — the radius contributes a factor of about ten on its own. Multiply those and the aircraft's resistance to angular acceleration is up by a couple of orders of magnitude. This is not a scaling-up of the same problem. It is a different problem.

The consequence people anticipate is that it is slow to start rotating. The consequence that actually hurts is that it is slow to stop. Once several kilograms of aircraft is rotating, all that angular momentum has to be taken back out by the motors, and they have to start doing it early. A controller that waits until the error is obvious before it brakes has already lost.

Natural frequency

A heavy aircraft oscillates slowly. There is no literal spring in a multirotor's attitude dynamics — the aircraft is a free body, and it accelerates at torque divided by inertia. The spring, such as it is, is the P gain: it pulls the aircraft back toward the attitude you asked for, and the aircraft's inertia resists. Raise the inertia while leaving the gains alone and the resulting oscillation slows down.

That has a consequence worth holding on to: a control-loop oscillation moves when you change the gains. Raise P and it speeds up. A structural resonance in the frame does not move at all, because its frequency is set by the stiffness and mass of the carbon, not by anything in the flight controller. That difference is the cleanest way to tell the two apart when a big aircraft starts to wobble.

So the wobble you get on a heavy quad is not the fast buzz of a five-inch. It is a slow wallow — a few hertz, something you can count with your eyes. Anyone who has learned to diagnose oscillation on small quads is calibrated for the wrong band entirely and will mistake it for something else.

And this is the dangerous part. On a racer, the frequencies where the control loop gets into trouble sit far above the frequencies at which the aircraft actually manoeuvres — the failure lives in its own separate region, up where you can filter it. On a heavy aircraft the loop's trouble band has come down to sit right on top of the band where the aircraft is genuinely flying. There is no clean separation left. An instability at 3 Hz is not a cosmetic buzz you tolerate; it is the aircraft doing something structural at the exact rate you are trying to steer it, and that is why heavy quads do not shake themselves apart gently — they diverge and come down.

Motor and prop response time

This is the single most important difference, and it is the one people skip.

A large prop is a heavy disc of plastic or carbon on the end of a motor with a big rotor, and it has its own rotational inertia. When the controller asks a motor for more thrust, the motor does not deliver thrust — it delivers torque, which accelerates the prop, which eventually spins fast enough to produce the thrust that was requested. On a 5-inch prop that transient is fast enough to ignore. On a 15-inch prop it is not: commanded RPM and actual RPM visibly diverge, and thrust arrives measurably late.

Late is the operative word. Lag in a feedback loop is what makes the loop unstable. The controller sees the error, commands a correction, the correction does not arrive, so the controller — dutifully — commands more. By the time the thrust finally shows up, the aircraft has moved on and the correction is now pushing the wrong way. That is the definition of a phase margin you do not have, and it is exactly how a stable gain becomes an oscillating one with no change to the number.

This is why a heavy aircraft cannot run the fast, twitchy gains that make a racer feel connected. The racer's loop can be aggressive because its actuators are effectively instantaneous. Yours are not.

Thrust-to-weight and control authority

A racer runs 8:1 or better. A heavy platform might have 2:1 — and less than that once it is loaded.

Rotation is produced by differential thrust: some motors go up, some go down. That differential has to fit in the gap between where you are hovering and where the motors run out. If you hover at 30% throttle, there is an enormous amount of room above you. If you hover at 70%, there is almost none, and the first serious correction the controller asks for saturates the motors — and a saturated motor cannot correct any further. The controller has, at that moment, no authority left in that axis. The mixer clips, the axes bleed into each other, the I term winds up on an error it cannot fix, and everything you observe afterwards is downstream of that.

So: aim to hover at or below half throttle. If you hover at 70%, the tune is not your problem. Your aircraft is over-propped, over-loaded or under-powered, and no PID number rescues an aircraft with no headroom.

Frame flex

Long carbon arms with heavy motors on the end are springs. There is no way around this — it is what a cantilevered beam with a mass on the tip is.

That structure has a resonance of its own, and the gyro is bolted to the middle of it. So the gyro is not measuring the aircraft's rigid-body motion; it is measuring rigid-body motion plus the arms flexing. A control loop fast enough to chase that resonance will feed it — it will push the motors at exactly the frequency the arms want to move at, and the arms will oblige.

This is a difference in kind, not degree. The rigid-body assumption underneath every quad PID controller you can flash is simply not true on your aircraft, and it is why big machines sometimes oscillate in ways that no gain change fixes. If you find a resonance that follows the airframe rather than the tune, the answer is mechanical — stiffer arms, better motor mounts, better soft mounting under the flight controller — not another 10% off D.

What that means for each PID term

Directions only. Read the full explanation of what each term responds to first if any of this is new.

Term	Generally moves	Why
P	Relatively higher	There is far more inertia to overcome. A gain that produces a snappy correction on 250 g barely moves 5 kg. But you will hit the oscillation ceiling earlier in frequency terms, and it will be a slow wallow rather than a buzz.
I	Matters much more	A heavy aircraft has persistent forces to hold against: an off-centre payload, a CG that shifts as fuel or battery drains, real wind loading on a big airframe. Only I can supply a permanent correction. Under-I on a heavy platform reads as a slow, boaty drift the aircraft never finishes correcting.
D	Proportionate, not racer-high	This is the trap. Racers run high D because their loops are fast and clean. On a slow, laggy, flexible airframe, high D amplifies gyro noise and excites the frame resonance, and it does it through motors that respond too slowly to actually damp anything. You want enough D to stop P ringing, and not one unit more.
Feedforward	Worth more than you think	It gets the motors moving on your stick rather than waiting for the error, which is exactly the head start a laggy actuator needs.
Filtering	Gentler	The loop is already lagging badly from prop inertia. Every filter is more delay on top, and you have far less phase margin to spend than a racer does. Solve noise mechanically wherever you can, and lean on RPM filtering, which costs the least delay for what it removes.

Why you cannot copy someone else's numbers

Look back at that list. Inertia, natural frequency, prop response time, thrust-to-weight, frame stiffness. Every one of them is a property of a specific aircraft: its arm length, its prop diameter, its motor rotor mass, its all-up weight today with today's payload, and how stiff its particular arms happen to be.

Two 5 kg quads can differ in every one of those. A short-armed X8 with small props and a long-armed flat X with 18-inch props are not the same control problem, and they are not close. The number that flies one of them will oscillate — or flip — the other. "5 kg" tells you almost nothing.

So this article contains no PID values, and that is deliberate. Publishing numbers for an airframe we have not measured would be irresponsible: they would be wrong for your aircraft, you would have no way of knowing why they were wrong, and you would be finding out at altitude over something expensive. What we can give you is the reasoning, and the reasoning is the thing that transfers.

How to actually approach a heavy tune

1. Fix the airframe first. Arm bolts torqued, motor mounts solid, props balanced, no cracked carbon. On a heavy build, mechanical noise is not a nuisance — it is the constraint that determines every gain you will be allowed to use.
2. Check your thrust-to-weight before you tune anything. Hover it, read the throttle. Above about 50% and you should be changing hardware, not gains.
3. Start low and conservative, everywhere. Come at a heavy tune from underneath. An under-gained heavy quad is mushy and unpleasant and lands fine. An over-gained one diverges at 3 Hz and does not.
4. One term, one axis, one change. Roll and pitch have genuinely different inertias on most big airframes — much more so than on a racer, where the difference is small enough to ignore. Tune them separately.
5. Fly the same manoeuvre every time. A gentle hover, a slow forward pass, one moderate bank. Not a freestyle line. You are looking for the onset of trouble, not performance.
6. Read the log, not your feelings. Look at setpoint versus gyro, and look at the motor outputs. If the motors are clipping, nothing else in the log means what you think it means.
7. Then, and only then, push a term up until you see the first hint of oscillation, and back off. That is where the aircraft lives. Find the wall gently, deliberately, with plenty of altitude, and then stay away from it.

The things that will hurt you specifically on a heavy build

• You get one chance at a first hover. A racer with a bad tune bounces off the grass and you laugh. A 5 kg aircraft with a bad tune destroys itself, and possibly whatever is nearby. There is no cheap failure mode. Every habit in this article exists because of that asymmetry.
• Failures are slow, which reads as safe, and is not. A 3 Hz divergence looks survivable for the first two cycles. It is not. It is growing, and it has enough energy behind it to arrive somewhere very quickly.
• Motor heat is not cosmetic. Big motors carry big currents, and a D-term oscillation on a heavy build will destroy windings rather than merely annoying you.
• The tune drifts with payload. Add a camera, change the battery, mount a gimbal, and you have changed the inertia tensor and the CG. Your tune was a property of the aircraft you tuned, and that is not this aircraft any more. Re-check after any significant change.
• Do not tune it in wind. A big airframe presents a big surface. You will spend an evening chasing a gust.
• Set rates the airframe can actually deliver. Commanding a rotation rate a heavy aircraft physically cannot achieve is a fast route to a saturated, mushy, unpredictable machine — see rates on a heavy aircraft. Lower rates are not a compromise. They are honesty.