Step Response turns "whether the aircraft follows the hand" into a comparable curve: the horizontal axis is the time after the command, the vertical axis is the response ratio of the gyro relative to the setpoint, and the ideal steady state is close to 1. Rise time, overshoot, oscillation, and convergence allow you to quantify P:D balance, overall control effort, steady-state locking of I, and FeedForward's startup response. The 18 ms, 1.5x overshoot, and 160 ms oscillation in the video are just a log example and are not numbers that all aircraft should aim for.
The goal of PID tuning is for gyro to follow setpoint
The pilot uses the sticks to command a target angular velocity, or setpoint, while the gyro records the aircraft’s actual angular velocity. PID tuning is not about obtaining an attractive set of numbers. It aims to make these two curves match with reasonable delay and stabilize after a maneuver. Step Response normalizes many commands of different amplitudes to compare how the aircraft catches up after a step command.
The horizontal axis shows milliseconds after the command. The vertical axis can be understood intuitively as the gyro/setpoint ratio. Initially the aircraft has not moved, so response is near 0. It then rises toward 1. A value above 1 means actual angular velocity has overshot the command. Repeated crossings of 1 indicate continuing oscillation; finally approaching 1 represents steady-state tracking.
Consider rise time, overshoot, and settling together
The video’s example reaches the target at about 18 ms but overshoots to roughly 1.5, then oscillates for about 160 ms before settling. Looking only at how quickly it reaches the target would miss serious overshoot. Conversely, pursuing zero overshoot alone may make response very sluggish. Real optimization balances rise time, peak value, oscillation count, and settling time for the aircraft and flying style.
A curve is only as trustworthy as its samples. Each axis needs several clear command changes. Tiny movements, collisions, stalls, or data noise may distort the result. Roll, Pitch, and Yaw differ in mechanical behavior and sample count; one axis’s curve cannot establish conclusions for the other two.
Relate the curve to P, D, I, and FeedForward
The video’s sequence is to find P:D balance first, reducing overshoot and oscillation. Once the ratio is stable, increase P and D together while retaining it to raise overall control strength. Then use I to improve final locking at 1, and finally FeedForward to shorten the initial response to stick movement. This order reduces confusion from changing several parameters together.
No graph can determine gains on its own. Excessive P can cause oscillation; excessive D amplifies noise and raises motor temperature; excessive I may produce slow oscillation; excessive FeedForward makes inputs abrupt. Make one small change at a time, repeat the same maneuvers, and record Blackbox data, flight feel, noise, and temperature together.
Resolve mechanical and sensor problems before optimizing
The video ends by emphasizing that excessive aircraft noise and gyro faults must be ruled out before PID tuning. Loose arms, damaged propellers, bent shafts, bearings, soft mounting, or unsuitable filtering can contaminate gyro data. Increasing D may then merely turn a mechanical problem into motor heat. Inspect the spectrum and raw gyro data before analyzing Step Response.
At the time of recording, the demonstrated Analyze PID tool required login and was free. Its service availability, algorithm, and interface may have changed. Whatever tool you use, understand its axes, normalization, and sample filtering. Do not treat an automatic chart as a one-click route to the best PID values. Consult current official documentation for Betaflight sliders and fields too.
Operation steps
- Understand the Step Response axes
The horizontal axis is time and the vertical axis is normalized response. At 1, gyro has reached setpoint; above 1, it has overshot.
- Check the samples and coverage of all three axes
Ensure Roll, Pitch, and Yaw each have enough clear maneuvers. Exclude samples with collisions, stalls, or obvious noise.
- Measure rise time, overshoot, and settling time
Do not look only at the earliest target arrival. Also record the peak, number of target crossings, and time needed to settle.
- Adjust P:D balance first
Judge the relative ratio from overshoot and damping. Make a small change, then sample again with the same maneuvers.
- Increase overall P/D while preserving the ratio
Increase only in small steps while noise, oscillation, and temperature remain safe. Immediately roll back if high-frequency abnormalities or noticeable heating appear; do not keep approaching a destructive limit.
- Use I and FeedForward to finish settling and response tuning
I improves steady-state locking; FeedForward shortens the initial response to a command. Adjust and retest them separately.
- Return to raw-noise and gyro checks
If the curve is abnormal, rule out mechanical vibration, sensor faults, and filtering problems before continuing PID optimization.
FAQ
Is reaching 1 faster always better in Step Response?
Not necessarily. A fast rise with large overshoot and prolonged oscillation still gives poor control. Consider both the peak and settling.
What does 1.5 on the vertical axis mean?
In the video’s normalized explanation, it means actual angular velocity reaches about 150% of the command: substantial overshoot, not a recommended target.
Does a shaky curve always mean incorrect PID gains?
Not necessarily. Insufficient samples, mechanical vibration, gyro faults, or filtering problems can contaminate the result. Check the raw logs and hardware first.
Can the tool automatically produce the best PID settings?
No. Analysis tools provide a quantitative view. Final decisions still require sample quality, Blackbox signals, temperature, flying style, and safe test flights.
Full timeline transcript
Transcripts are arranged according to video time, making it easy to quickly locate the explanation content. Transcript language: Simplified Chinese.
Hello everyone, I am QuadRank. Last time we learned how PID works. Today we will use that knowledge to look at PID tuning. Ultimately, tuning is an optimization problem. I remember simple optimization exercises from high school: a company buys 10 trucks, each earns a certain amount per year, there is an annual depreciation rate, and maintenance costs change with age—
—then the question asks in which year selling the trucks would maximize net profit. PID tuning is also an optimization problem, but differs slightly from that example. Net profit is the objective there. What is the objective in PID tuning? If you watched the previous videos, you can answer immediately: two curves, the gyro curve and the setpoint curve.
We want the curves to match as closely as possible: actual aircraft angular velocity should follow the angular velocity commanded by the sticks. In the previous video’s analogy, Nobita’s position should stay close to Shizuka’s. You might wonder whether there is a way to quantify how well the two curves match. Conveniently—
—there is: the step-response curve. How do we read it? The horizontal axis is time, here from 0 milliseconds to 500 milliseconds. What does the vertical axis represent? Recall the previous analogy: Nobita wants to catch Shizuka and keep his position aligned with hers.
Think of Shizuka’s position as the reference, always at 1.0. At 0 milliseconds, Nobita starts at 0. Over time, under the PID algorithm, he approaches Shizuka’s position at 1.0. He may overshoot, then oscillate, before finally settling at 1.0.
That is the analogy. What does the vertical axis actually represent? The ratio of gyro to setpoint. Ideally it should be 1, right? But aircraft response always has some delay. Initially it is 0, then after ten-odd or twenty-odd milliseconds it eventually approaches 1. Let us practice reading the curve with a log. I have opened one here and will click Analyze PID.
Note that this button requires login and is currently free, though I cannot promise it will remain free in the future. If you want to use it, try it soon. Clicking generates three graphs for the Roll, Pitch, and Yaw axes. We will look at one axis for this demonstration. The horizontal axis is time and the vertical axis is gyro divided by setpoint.
At about 18 milliseconds, actual angular velocity catches the commanded velocity. Suppose setpoint commands rotation at 100 degrees per second. At around 17 milliseconds, the aircraft has gradually risen from 0 degrees per second to 100 degrees per second. But it overshoots because this aircraft is not tuned well.
How far does it overshoot? To 1.5. A value of 1.5 means 150 degrees per second. We asked it to rotate at 100 degrees per second, but it reached 150 degrees per second. It realizes it is going too fast and comes back, but overshoots the other way—to perhaps seventy or eighty degrees per second, around eighty. It is now too slow, so it goes back again.
Eventually it settles near 100 degrees per second. What is our PID-tuning objective? First, make overshoot as small as possible. Second, make the oscillation as short as possible. How long does it oscillate here? Around 160 milliseconds, which is quite a long time.
Another goal is to reach 1.0 as quickly as possible without overshoot. Reaching 1.0 quickly and avoiding overshoot can conflict, so we need a tradeoff. Finally, we use I to keep it locked as closely as possible around 1.0. Understanding the step-response curve gives us a basis for PID tuning.
Here is the usual tuning sequence. First find the best P:D ratio to reduce or eliminate overshoot. Once that ratio is found, increase P and D together while maintaining it. When the motors become a little hot to touch or obvious oscillation appears, reduce them somewhat. In Betaflight 4.2, the slider can increase or decrease P and D together while preserving their ratio.
PID limits depend on many factors, which we can discuss in detail later. The third step is increasing I until the curve settles well at one. During the first and second steps, keep I relatively low because excessive I can also cause oscillation, especially when P and D are low. If P and D are already high, this effect is less pronounced. The fourth step is to use—
—Betaflight FeedForward to shorten the time taken to reach 1, making the curve steeper. Finally, before PID tuning, ensure the aircraft does not have excessive noise and the gyro has no problems. Friends and I have encountered faulty gyros. We may make a video to share those experiences later.
That is all for today. Next time we will use what we have learned in these videos to tune one or two aircraft in practice. Stay tuned, and thank you for liking, saving, and sharing. I am QuadRank; see you next time.
