Practical combat of servo PID parameter adjustment: bid farewell to jitter and get started quickly

Will the sudden trembling of the servo make your heart suddenly tremble? Don't worry, in fact the distress signal it sends out is that the PID parameters are not adjusted properly. The three values ​​of proportion, integral and differential are like the "three primary colors of character" of the steering gear. Only by properly blending them can it become docile, docile and obedient.

Let’s take a look at some of the most common situations:

If the proportional gain (P) is too large, the servo will tremble continuously like drinking three cups of highly extracted espresso, and will swing back and forth to the left and right during positioning operations.

The integration time is too small and the response of the servo is not sensitive. The action of pointing to the target is done slowly, as if you are walking in the middle of a quagmire, step by step, slowly.

If the differential time is too large, the movements will become nervous, shaking at the slightest touch, and even standing quietly becomes an unattainable luxury.

There was once a student who was in the laboratory in the spring and was crazy about the shaking car servo. He tried the so-called "universal parameters" on the Internet, but the result was that the more he debugged, the worse the situation became. Then he finally understood: PID is not magic, but actually a cause-and-effect dialogue, that is, what value you give, the expression of the servo will be returned. For example, when the ratio (P) is increased from 0.3 to 0.6, the response speed does speed up, but high-frequency vibrations in the corners also begin to appear. The measured data is: for every 0.1 increase in P, the static error can be reduced by 15%, but the overshoot will jump to 8%-12%. See, this is what dialectics sounds like.

So where exactly do you start? Please accept the first adjustment method, which is keyword integration: first P, then I, then D. Don’t be bothered when isolating variables.

Set I to maximum (or 0), set D to 0, and only let P work.

Slowly and gradually increase P until the servo begins to vibrate slightly with constant amplitude, and then adjust P back by 20% to remove the punctuation mark.

At this time, the servo is like a toddler - it will shake, but it will not fall. Perfect!

Why does it have to be this way?The reason is that the integral (I) will have the characteristic of "holding a grudge". It will accumulate past errors. Once the jitter caused by the proportion (P) is remembered by the integral (I), the jitter will become more and more violent.. The differential (D) acts as a "prophet" and implements braking operations in advance based on the current trend of changes. If you reverse the order and adjust the differential (D) first, it's like putting skates on a baby, and the whole logic will become confusing.

Following what the student experienced, after setting P to 0.5 and fixing it, he started adding points. At first, the value of I was 1.0 (unit is second). The servo seemed to be suffering from procrastination, always missing 2 degrees before it could reach the designated position. In desperation, he lowered the value of I to 0.3 with difficulty, and something unexpected happened: not only did the servo quickly lock on the target, but it was also as quiet as a cat taking a nap.However - when the value of I is reduced to 0.1, the rebound generated after the overshoot brings about a new smile-shaped curve.. You see, every parameter has an optimal value range.

The target is proof: There are 200 sets of adjustment records from a certain robot community, which show that 67% of the jitter problems are caused by the P value deviating from the optimal range by more than 20%, and 54% of the slow response is due to the I value being set too large, that is, more than 1.2 seconds. The numbers don’t lie, and neither does the servo.

At this time, we enter the second circle of the upward spiral, that is, the advanced intervention stage. After the basic movements are in a stable state, try to use these techniques.

For asymmetric P, forward motion is given different proportional values, and reverse motion is also given different proportional values. For example, when going up, use P equal to 0.6, and when going down, use P equal to 0.4. It will be more sensible to fight gravity in this way.

Integral limit range setting: It plays a role in preventing the integral from reaching a saturated state. Set an upper limit for the accumulated number of I to prevent the servo from continuing to accumulate crazy points when it is stuck. When released, it will not rush directly into the sky like an ejection seat.

When performing differential operation, only the feedback signal is differentiated and does not respond to sudden changes in the target value. In this way, when your command jumps, it will not cause the servo to produce a "fright cramp" phenomenon.

What will be discovered by careful readers is that this is actually telling the steering gear: "I understand your physical limits." The mechanical structure has its own preferred parameter climate, one of which is a plastic gear, the other is a metal gear, and the high-torque and high-speed types have different expressions of attitude and running performance, which are divided into different types. For the same numerical situation where P is equal to 0.8, the small steering gear will show a gentle state like the spring breeze gently blowing on the face, while the large steering gear will show a strong situation similar to that when a typhoon passes by.

The second key tip (incorporating keywords into it): Advanced style tip – let the data speak for itself, rather than relying on intuition.Take a piece of paper and record the three indicators after each change: time of arrival, angle of overshoot, and amplitude of jitter when stable.. Then compare them just like tasting tea.

It's time for FAQ Q/A - the first sentence of each answer directly gives the conclusion, within fifty words.

Q: The servo makes a hissing sound when it is stationary, how to solve it?

A: Reduce the differential (D) value. If D is too high, the noise signal will be amplified, causing the servo to overcompensate for micro-vibration. Adjusting it below 0.1 can generally eliminate it.

Q: The response is too slow, but what should I do if it starts to shake after increasing P?

First, increase the differential (D) to suppress jitter. D can provide damping, which in some cases allows you to increase P to 1.5 times the original without causing oscillation.。

Q: I need the servo to maintain accuracy even when it is under load. Which one should I adjust?

A: Expand the integral (I) time. It should be noted that what is mentioned here is "increasing" the value (for example, increasing from 0.3 to 0.8), which makes the role of the integral weaker and avoids the situation where the accumulated error overshoots when the load changes.

Q: The position is always different by a fixed angle, like there is a dead zone?

A: Check whether the points item is turned on. Integral (I) is specifically used to eliminate static errors. If you give it a value other than zero (such as 0.5 seconds), it can gradually eliminate the residual error.

Q: After adjusting the parameters, does it work if I change the power supply?

A: Adjust the comparison ratio (P) slightly again.If the voltage drops, the output power of the servo will decrease, which has the same effect as a smaller P value.. Increasing the P value by ten to twenty percent can restore the original state.

Do you still remember that student in spring? At the beginning of summer, the car he owned was already able to turn smoothly with a full glass of water. The conclusion he finally came to was extremely straightforward. PID is not a metaphysics, but a cycle of observation, recording, and observation. Every time you make a parameter change, you must ask yourself whether the "emotion" displayed by the servo is irritability, laziness, or panic, and then guide it very gently.

Suggestions for action (please follow them directly):

1. Prepare a record sheet listing P, I, D and description of the phenomenon.

2. Start with pure proportion, find the critical jitter point, and then return to 80%.

3. Add points and try from small to large until the static error disappears and there is no slow crawl.

4. Finally add the differential, slowly increase from 0, and stop immediately when you feel the movement is "moisturized".

5. Only change one parameter at a time! Don't move both at the same time.

The steering gear is speechless, but every angle and every vibration is recording a log for you. If you understand it, you will find that there is a unique world within the three parameters. Now, go to the lab, close the door, pick up a screwdriver, and listen to the steady song of that little motor.