CNC Servo Vibration, Overshoot and Noise: A Phenomenon-Based Troubleshooting Table-CNCShare

Q: For high-frequency squeal, should I lower the gain or use a filter?

Use a filter first, then raise the gain, rather than simply lowering the gain. First confirm it is resonance: if the vibration only appears in a certain speed band or the frequency is fixed, it is mechanical resonance. Merely lowering the gain is escapism because the vibration reduces but rigidity goes with it and accuracy worsens. The correct order is: measure the center frequency and bandwidth with SERVO GUIDE or a frequency counter rather than by feel, select the filter by frequency level and set the correct axis, remove the resonance point, then raise the speed gain 2021, then re-measure and repeat. If all filters are used and it still vibrates, stop tuning and investigate the machine.

Q: Does a quadrant protrusion mean I should raise the gain?

Not necessarily, it is two-sided. A quadrant protrusion can indicate the inertia ratio is too low, or it can be a backlash compensation problem, and the two are handled differently. If the whole circle is bad, the inertia ratio is too low, so raise 2021. If there is a step only at the reversal point, 2048 is too small, so add 50 each time. Do not blindly add gain the moment you see a protrusion; first confirm whether it is the whole circle being bad or a step only at the reversal point.

When someone reports “the machine shakes,” that’s often all they say — just two words. But high-frequency squeal, low-frequency flutter, start/stop overshoot, and machining-surface ripple are four completely different problems, with completely different parameters and handling paths. Get the direction wrong and you tune it worse and worse.

This article is organized as “phenomenon → classification → troubleshooting,” not as a list of parameter numbers. You only need to judge which one you’re facing, then follow the corresponding branch.

Safety note: Servo parameters directly affect motion stability. You must complete a parameter backup before adjusting, and verify under non-production, low-speed, single-axis conditions. When resonance is severe the axis may vibrate violently and even crash into the limit; if anything looks abnormal, stop immediately and roll back to the previous parameter set.

1. First Distinguish the 4 Kinds of “Shake”

Why classify first: take “gain” as the example —

High-frequency vibration usually means the gain is relatively too high (or a resonance point exists) → lower the gain or add filtering.
Low-frequency flutter / quadrant protrusion usually means the gain is too low → raising it appropriately improves things.

Get the direction backwards and it gets worse the more you tune. So the first step is always to judge which one you’re facing.

2. High-Frequency Vibration / Squeal: Resonance and Filters

First confirm whether it is resonance

Criterion: is the vibration strongly correlated with a particular speed or position? If it only makes noise in a certain speed band, or the frequency is fixed, you can basically judge it as mechanical resonance — the mechanical system has a natural frequency being excited by the servo.

At this point merely lowering the gain is escapism: dropping the gain does reduce the vibration, but the rigidity goes with it and machining accuracy gets worse. The correct approach is to use a filter to cut the resonance point first.

How to use the filter (per BEIJING-FANUC “Servo Tuning Guidance”)

Step 1: Find the center frequency and bandwidth of the resonance point. The official instructions clearly state that the center frequency and bandwidth of the vibration must be detected by some method (e.g. SERVO GUIDE software or a frequency counter). Estimating by feel does not work.

Step 2: Select the filter by frequency level. There are 4 HRV filter groups. Usage points:

The system provides 4 filter groups; if there are two or more resonance points, they can be combined.
When setting parameters you must select the corresponding axis — this is the step most easily missed.
The damping value is commonly set to 5% or 10% (per some references); a smaller value gives greater suppression.

Step 3: After removing the resonance point, raise the gain again. The official procedure reads: after the resonance point is eliminated, a higher speed gain (parameter 2021) can be set. Then the frequency response must be measured again, repeating until the requirement is met. This is a “filter → raise gain → re-measure” cycle, not a one-shot fix.

Step 4: When to give up tuning. The official instructions also give a clear stop signal: if vibration cannot be eliminated after using all the filters, carry out a mechanical-factor investigation.

What “good frequency response” means

The official criteria (for reference when you have SERVO GUIDE or similar tools):

The response bandwidth (the 0 dB region on the amplitude-frequency curve) must be wide enough — wider is better — achieved mainly by adjusting position-loop gain (1825) and speed-loop gain (2021).
After using the HRV filter, the amplitude at the high-frequency resonance frequency should be below −10 dB.

Notch filter (some references mention 2205 / 2212)

Some references list 2205, 2212 etc. as “notch filter” related parameters. The specific numbers and availability of such parameters are strongly dependent on the system model and software version, and the correspondence differs between references. My suggestion: follow the filter function names actually displayed on your machine’s servo parameter screen; don’t hard-fill numbers from references. If filter parameters are filled in wrong, they can actually amplify the vibration.

3. Low-Frequency Flutter: Inertia Ratio and Speed Loop Mismatch

Phenomenon: a duller, larger-amplitude back-and-forth sway, or sway during acceleration/deceleration.

The direction for this class is the opposite of the previous one — low-frequency problems are often not “gain too high” but the inertia ratio setting not matching reality. Troubleshooting order:

Confirm the actual inertia ratio. You can measure it manually or use the system’s automatic measurement function, then compare with the set value. A set value either too large or too small will cause oscillation.
Adjust the speed-loop related parameters per the official sequence (see FANUC Servo Gain 2021 / 1825 Adjustment): start at 128, decrement 128 → 64 → 0 when vibrating.
If still unstable, only then does it involve decrementing the position-loop gain.

Conversion relationship (official formula): speed-loop gain (%) = (1 + load inertia ratio / 256) × 100. This formula helps you judge what level the current setting is at.

Caution on parameter numbers: the official 0i-D start-up manual writes 2021 = load inertia ratio (LDINT), but references disagree on 2020 vs 2021. Follow the name shown on the parameter screen.

4. Start/Stop Overshoot: Acceleration/Deceleration and Backlash Acceleration

Phenomenon: overshoots past the target at start or stop then comes back; a step appears at the reversal position.

Two main directions:

① Acceleration/deceleration time — extending the accel/decel time reduces mechanical shock, suitable for heavy-load conditions. The specific parameter number varies by system model; follow your machine’s manual. This article does not list uncertain numbers.
② Backlash acceleration (2048) — handles the shape at the reversal instant: protrusion at reversal → add 50 each time; depression → subtract 50 each time.

Note where to read: look at the shape at the reversal instant, not the steady-state segment. The handling directions for protrusion and depression are opposite — do it reversed and it gets worse the more you tune.

5. Machining Surface Ripple: Quadrant Protrusion and Backlash

Phenomenon: a regular pattern on the machined surface; a protrusion or step appears at the quadrant when machining an arc.

The two-sided nature of quadrant protrusion (key)

A quadrant protrusion can mean either “inertia ratio too low” or “backlash compensation problem,” and the two are handled in different directions:

If the whole circle is bad → the inertia ratio is too low; raise 2021.
If there is a step only at the reversal point → 2048 is too small; add 50 each time.

Don’t blindly add gain the moment you see a protrusion — first confirm whether it is “the whole circle is bad” or “there is a step only at the reversal point.”

Backlash: electrical compensation vs mechanical gap

Two things must be distinguished:

The backlash compensation parameter is electrical compensation — it can only compensate, not eliminate.
Mechanical backlash (the gap between screw and nut) physically exists.

If the mechanical gap itself is too large (screw wear, loose bearing, loose coupling), no amount of parameter tuning can recover it — you must fix the machine first.

6. Should HRV Control Be Enabled?

HRV (High Response Vector) is FANUC’s servo control technology used to raise current-loop response speed and improve rigidity and tracking precision.

What it solves

Raises servo-control rigidity and tracking precision, suitable for high-precision contour machining.
Some references mention HRV control can eliminate low-frequency vibration on low-rigidity machines.
Spindle application: for frequent start/stop, enabling HRV improves dynamic response.

HRV1 / HRV2 / HRV3 / HRV4

The current-loop response becomes progressively faster. The higher the level, the faster the response — and the higher the requirement on the machine.

Enabling method (version differences are large)

Claims online are not unified, for example:

Some references write No.2013 #0 (HRV3) = 1.
Some references write enabling HRV3 requires 11240 #1 = 11240 #4 = 1.

These two correspond to different systems/configurations; don’t mix them. My suggestion: HRV enabling involves servo function bits and options — first confirm your machine has the corresponding option; follow your own machine’s servo parameter screen and manual; don’t copy bit numbers from the internet.

Risks of enabling HRV

The higher the level, the higher the requirement on mechanical rigidity. If the machine itself has resonance or insufficient rigidity, blindly raising the response may amplify the vibration.
Not “the higher the better” — choose according to the machine’s actual mechanical characteristics.
Changing HRV is a functional change, different from simply adjusting gain, with a broader impact.
Back up before changing; verify axis by axis after changing.

In one sentence: HRV is not a “vibration-kill switch.” It raises response capability, not automatically eliminates mechanical defects. When the machine has problems, enabling a higher HRV level may make them more obvious.

7. Self-Diagnosis Reference Table (phenomenon → ranked candidate causes)

Follow this table: first find the phenomenon you observe, then work through the candidate causes in order (ordered “common to rare, cheap to expensive”). Usage notes:

Work from top to bottom, verifying after each change; don’t change multiple parameters at once — otherwise you won’t know which one took effect, and you can’t roll back.
Record the before-value after changing, for easy rollback.
The same phenomenon may have several causes stacked; after handling the first one, if the problem remains, continue down the list.

8. When to Stop Tuning and Go Check the Machine

This is the most important section of the article. If vibration cannot be eliminated after using all the filters, carry out a mechanical-factor investigation. Once the filters are used up and it still vibrates, stop spending time on parameters.

Mechanical items to check (cheap to expensive)

Guide lubrication — poor lubrication is an extremely common cause, and the lowest cost.
Coupling — loose, worn, or installed off-center.
Ball screw — backlash, wear, loose bearing housing.
Bearing — abnormal noise, play.
Mechanical rigidity — the structure itself lacks rigidity; parameters cannot fundamentally solve it.
Tool and cutting parameters — the ripple may not be a servo problem at all.

Can servo vibration be solved with damping pads?

Not recommended as a solution. Adding damping pads may make the vibration “sound a bit smaller,” but:

It does not remove the excitation source — the resonance point is still there.
It may mask the real fault, letting the mechanical problem keep deteriorating.

The correct order remains: first confirm mechanical state → targeted repair → then fine-tune with parameters/filters.

Closing

Solidify this flow:

Classify first: high-frequency squeal / low-frequency flutter / start-stop overshoot / machining ripple — four classes, four directions; get it backwards and it gets worse the more you tune.
For high-frequency resonance, filter first then raise gain — don’t just lower gain; if the filters are used up and it still vibrates, stop and check the machine.
Quadrant protrusion is two-sided: whole circle bad = inertia ratio too low; only at reversal = 2048 too small.
HRV is not a vibration-kill switch; it raises response, and may amplify problems when the machine is defective.
Troubleshooting order: mechanical → feedback and grounding → parameters.
Back up before tuning; verify after each change; record the before-value.

Disclaimer: This article is compiled from the BEIJING-FANUC “Servo Tuning Guidance,” the FANUC official 0i-D start-up manual, and public technical materials. Parameter numbers, filter configuration, and function bits vary by system model (0i-C/D/F, 16i/18i/21i/30i etc.) and software version; follow your machine’s official manual and parameter screen display. Servo parameters directly affect motion stability — complete a backup before adjusting and verify step by step under non-production, low-speed conditions; operations involving safety functions (such as Dual Check Safety) must follow the relevant procedures and authorization requirements.

Zero-base learning entry: PMC Ladder Video Course From Zero.

Frequently Asked Questions (FAQ)

1. How do I classify servo vibration into the right type?

There are four kinds, and their handling directions differ: high-frequency squeal (usually gain relatively too high, or a resonance point exists), low-frequency flutter (usually the inertia ratio setting does not match reality, i.e. gain effectively too low), start/stop overshoot (acceleration/deceleration and backlash acceleration 2048), and machining-surface ripple (quadrant protrusion and mechanical backlash). First judge which one you face based on the phenomenon, then follow that branch; getting the direction backwards makes it worse the more you tune.

2. For high-frequency squeal, should I lower the gain or use a filter?

Use a filter first, then raise the gain — do not simply lower the gain. First confirm it is resonance: if the vibration only appears in a certain speed band or the frequency is fixed, it is mechanical resonance. Merely lowering the gain is escapism: the vibration reduces but rigidity goes with it and accuracy worsens. Correct order: measure the center frequency and bandwidth (with SERVO GUIDE or a frequency counter, not by feel), select the filter by frequency level and set the correct axis, remove the resonance point, then raise the speed gain (2021) — then re-measure, and repeat. If all filters are used and it still vibrates, stop tuning and investigate the machine.

3. Does a quadrant protrusion mean I should raise the gain?

Not necessarily — it is two-sided. A quadrant protrusion can indicate the inertia ratio is too low, or it can be a backlash-compensation problem, and the two are handled differently. If the whole circle is bad, the inertia ratio is too low, so raise 2021; if there is a step only at the reversal point, 2048 is too small, so add 50 each time. Don’t blindly add gain the moment you see a protrusion — first confirm whether it is “the whole circle is bad” or “a step only at the reversal point.”

4. When should I stop tuning parameters and go check the machine?

If vibration cannot be eliminated after using all available filters, stop tuning and investigate mechanical factors. Check items cheap to expensive: guide lubrication (very common and lowest cost), coupling (loose, worn, off-center), ball screw (backlash, wear, loose bearing housing), bearing (noise, play), mechanical rigidity (structural; parameters cannot solve it), and tool and cutting parameters (the ripple may not be a servo problem). Also, don’t treat damping pads as a solution — they don’t remove the excitation source and may mask the real fault.