Configuring SINAMICS Notch Filters for 840D sl C-Axis Whistling

When a C-axis on a SINUMERIK 840D sl grinding machine starts to whistle after a heavier workpiece is mounted, the symptom is almost always mechanical resonance that has been re-excited by the speed-control loop. The motor itself is not failing; the closed-loop gain is high enough at the new workpiece mass to drive a structural mode into sustained oscillation, and that oscillation shows up acoustically as a high-pitched tone. The correct response is not to back off the gain to the point where the axis loses productivity, but to add a digital filter in the current or speed setpoint path that suppresses the offending frequency.

This article covers the parameter set used to add and tune those filters on a SINAMICS S120 drive controlled from a SINUMERIK 840D sl NC, how to identify the offending frequency when the drive software is older than the auto-tuning revisions, and how to verify the result. The same approach is valid on any SERVO drive object (DO) on a SINAMICS S120, not only the C-axis of a grinder.

1. Problem Description and Symptoms

Symptom pattern seen on the C-axis of the grinder in question:

• A continuous, narrow-band audible whistle, typically 300 Hz to 1500 Hz, that appears as soon as the axis moves and stops within a fraction of a second when the axis stops moving.
• The whistle is louder and starts at a lower speed when a heavier workpiece is fitted, which is the indicator that the resonance is on the mechanical side (workpiece-tool-spindle system), not the motor bearings.
• Position following error and the position-controller output remain stable; the oscillation is in the speed loop, not the position loop.
• The whistle is independent of feedrate, so it is not a spindle-speed artifact (e.g. tooth mesh on a gearbox). It tracks the C-axis motion.

Acoustic emission at a single fixed frequency is the signature of an undamped or under-damped second-order system being excited by a periodic input. In a closed-loop speed controller, the periodic input is the torque-producing current ripple at the speed-controller bandwidth corner. When the speed-controller proportional gain p1460 is increased to handle the heavier load, the open-loop gain at the mechanical resonance frequency crosses 0 dB with insufficient phase margin, and the loop oscillates at that mode.

2. Root Cause Analysis

Three things change when the workpiece mass goes up on a C-axis:

1. Total inertia at the motor shaft increases. The speed controller was originally tuned for a much lower J_total. The mechanical time constant of the axis, Tau_mech = J_total / dM, grows. The integrator time p1462 is sized to the original Tau_mech, so a higher inertia gives more phase lag at the new crossover, which usually forces a lower p1460 if left unchanged.
2. Resonant frequency drops. A heavier tool/workpiece on the same mechanical structure lowers the natural frequency of any compliance in the drivetrain (coupling, belt, encoder mounting, workpiece-fixture interface). The mode that was previously well above the speed-loop bandwidth moves into the bandwidth, and the controller excites it.
3. Loop gain needs to increase. For the same acceleration, torque demand grows, which generally requires a higher p1460. The combination of a higher proportional gain and a resonance that is now inside the bandwidth is what produces the whistle.

Conclusion: the symptom is a speed-loop / structural-mode interaction. The fix is to (a) re-tune the speed loop for the new inertia, and (b) add a notch filter that removes the offending mode from the loop's plant response before the proportional gain acts on it.

3. SINAMICS Speed Control Loop Architecture

The SERVO drive object on SINAMICS S120 has a cascaded structure: position controller (in the NC) feeds the speed setpoint, the speed controller (p1460, p1462) feeds the torque setpoint, and the current controllers close the inner loop. The current setpoint can be filtered by up to four current-setpoint filters before it reaches the current controllers. Those filters are the primary tool for suppressing mechanical resonance.

For the whistling symptom, the filters that matter are p1670...p1686 (current-setpoint filters 1...4). They are digital biquad stages inserted in the torque path. The first filter is enabled by default after current-controller autotuning; the remaining three are off until configured.

4. Key Parameters for the C-Axis Speed Controller

Before adding a filter, document the existing controller state. The drive object for the C-axis is the SERVO DO; its parameters are accessible from the SINUMERIK HMI under Startup > Drive System > Drives > SERVO_xxx > Parameters, or from STARTER/Startdrive through the drive device tree.

For a higher-inertia load, expect p1460 to settle roughly 1.3x to 2.0x the original value once a filter is in place, and p1462 to be 1.5x to 3.0x the original value because the larger J_total yields a longer Tau_mech.

5. Auto-Tuning Functions in SINUMERIK 840D sl

SINUMERIK 840D sl integrates the SINAMICS commissioning into the HMI under Startup > Drive System > Drives > SERVO_xxx > Commissioning > Automatic Controller Setting. The exact menu text and feature set depend on the SINUMERIK software version. The discussion in the source thread notes that SW 2.05 does not have the auto-tuning menu; SW 2.6 and later (NCU software 4.4 SP2 and up) ship with the drive-side auto-tuning functions enabled.

If the controller is at SW 2.6 or later and the SINAMICS firmware is V4.x, run Startup > Drive System > Drives > SERVO_xxx > Commissioning > Automatic Controller Setting > Speed controller optimization (with load). The routine will:

1. Move the C-axis over a small angle (typically ±5°),
2. Inject a pseudo-random binary sequence (PRBS) into the torque setpoint,
3. Capture the resulting speed response,
4. Compute the open-loop frequency response via FFT,
5. Fit a closed-loop PI + notch filter model,
6. Write p1460, p1462, p1670...p1682 directly to the drive.

After auto-tuning, run a controlled motion test with a representative workpiece and listen for whistle. If the resonance is still present, drop the auto-tuned Kp by 20 % and add a second manual notch filter at the residual peak.

6. Manual Filter Design with Notch and Band-Stop Filters

For a controller at SW 2.05 without auto-tuning, the offending resonance frequency must be measured and a current-setpoint filter configured manually. A "notch filter" in the SINAMICS sense is a band-stop filter (type 2 biquad): it passes all frequencies except a narrow band around its center, where the gain drops sharply. The filter is described by four parameters per stage:

• Type: 1 = low-pass PT2, 2 = general 2nd order (band-stop / band-pass / lead-lag), 3 = band-stop, 4 = band-pass, 5 = general 2nd order with high-pass shelf
• Denominator natural frequency f_dd [Hz] — the frequency where the denominator has its pole pair
• Denominator damping d_dd — the damping of the pole pair (lower = sharper notch, higher = broader)
• Numerator natural frequency f_nn [Hz] — the frequency of the zero pair that creates the notch
• Numerator damping d_nn — the damping of the zero pair

For a pure band-stop (notch) at one frequency, set f_nn = f_resonance and d_nn = 0.7 (moderate depth), and set f_dd slightly above f_nn (typically 1.05 to 1.2x f_nn) with d_dd = 0.3 to 0.5. The width of the notch and its depth are then controlled by the d_nn / d_dd ratio.

7. Filter Parameter Set (p1655...p1686) Detailed Reference

There are four current-setpoint filter stages on a SINAMICS S120 SERVO. Each stage is enabled by the corresponding bit in p1656:

For the C-axis whistle, the typical configuration is:

; ---- Current-setpoint filter 1: notch at the measured resonance ----
p1656.0 = 1                  ; enable filter 1 (if not already on)
p1670    = 3                 ; type = band-stop (notch)
p1671    = 480               ; f_dd = 480 Hz  (denominator natural freq, slightly above notch)
p1672    = 0.4               ; d_dd = 0.4
p1673    = 455               ; f_nn = 455 Hz  (numerator natural freq = measured resonance)
p1674    = 0.7               ; d_nn = 0.7

; ---- Current-setpoint filter 2: 2nd notch if a 2nd peak remains ----
p1656.1 = 1
p1675    = 3                 ; type = band-stop
p1676    = 1200              ; f_dd
p1677    = 0.4
p1678    = 1150              ; f_nn (second resonance from coupling)
p1679    = 0.7

; ---- Speed controller re-tune for the heavier workpiece ----
p1460    = 0.85              ; was 0.45; raise Kp after filters are in place
p1462    = 0.018             ; was 0.008; longer Tn for the larger J_total

For a band-stop filter in SINAMICS, the convention is f_nn = f_resonance, f_dd > f_nn. If the auto-tuning tool has written the values, the typical result is f_dd / f_nn between 1.05 and 1.15. Do not invert the relationship; the biquad becomes unstable if f_dd < f_nn and damping is low.

8. Identifying the Resonant Frequency with STARTER / Startdrive

On a controller without auto-tuning, the resonance is identified by injecting a small chirp or PRBS reference into the open speed loop and capturing the response. The procedure:

1. Open STARTER (or Startdrive for newer TIA Portal projects). Connect to the drive device containing the C-axis SERVO DO.
2. Open the trace tool (Trace & Graph Generator). Configure two signals: speed setpoint (r0062) and speed actual value (r0061). Set the trigger to "immediate", sample time 0.5 ms, recording time 4 s.
3. Pulse-inhibit the drive. Set p1120 (ramp-up time) to a small value, e.g. 0.05 s, so a short reference step is fast enough to excite the mode.
4. Enable the drive. Issue a small jog step in the C-axis (e.g. 1° at 1000°/min). Start the trace.
5. Stop the trace. Open the trace in the FFT view (right-click > Signal characteristics > FFT). The peak in the speed actual value spectrum (and the corresponding dip in the coherence function) is the resonance frequency f_resonance.
6. Repeat the test with the heavy workpiece mounted. The peak frequency will drop; this is the value to write into p1673 (numerator natural frequency).

An alternative measurement that requires no extra tool: use the online spectrum analysis of the torque setpoint (r1503) at a constant speed. The torque setpoint will show a tonal peak at the resonance frequency while the whistle is audible. This method is less precise than the PRBS / FFT approach but works with any STARTER version.

9. Setpoint and Actual-Value Filters as Alternatives

If the resonance is below ~200 Hz, a band-stop filter is the wrong tool; the speed-loop bandwidth typically needs to be 2x to 3x the resonance frequency, and a notch that low would also notch out useful control signal. In that case, the alternative is to add a low-pass (PT2) on the speed actual value or the speed setpoint:

• p1130 (setpoint PT2 low-pass): Reduces the controller's excitation of the mode, but also adds phase lag and degrades contour accuracy. Use only if p1460 is the limiting factor and the axis is not in a tight contouring cut.
• p1413 / p1414 (actual value low-pass): Removes the high-frequency component from the actual speed used by the controller. Same phase-lag caveat; do not combine with very low p1462.
• p1441 (actual speed smoothing time constant): A simple first-order low-pass; gentler than p1413/p1414, suitable for encoder noise, not for mode suppression.

For a C-axis grinding machine that does contour-following between the wheel and the workpiece, the cleanest answer is a band-stop filter at the resonance frequency (above 200 Hz), not a low-pass on the actual value.

10. C-Axis Specific Considerations on a Grinding Machine

The C-axis on a cylindrical grinder is unusual in that the workpiece itself is a tuned structure. The first resonance is often the workpiece-fixture system, not the drivetrain. The frequency is then proportional to 1/sqrt(m_workpiece). The original poster's report that the whistle appeared only after a heavier workpiece is consistent with this.

Practical implications for the filter design:

1. Set the filter frequency for the heaviest workpiece the machine will run, not the lightest. With a lighter workpiece the resonance moves up and out of the notch's effective range, which is fine; the filter will simply have no effect at the higher frequency.
2. If the machine grinds a range of workpiece masses, leave filter 1 (notch) on the heaviest-workpiece frequency, and consider adding filter 2 (notch) on a second modal frequency 1.5x to 2x higher. Filter 3 (notch) can be set to a third mode if needed.
3. Verify the C-axis position loop in the NC (MD of the position controller) is not also being excited. A sustained whistle that does not decay when motion stops is a speed-loop mode; a whistle that decays when motion stops is a position-loop mode and is treated by adjusting the NC's Kv factor and possibly the position setpoint smoothing (MD of the NCK).

11. Current and Torque Verification

A question raised in the field report: what current is drawn, and what is the rated current? When evaluating a C-axis filter, the relevant limit is the short-time current capacity of the SERVO drive object, not just the rated motor current. Verify the following before declaring a filter successful:

1. Drive utilization (r0033) stays below 100 % during a representative grinding pass with the new workpiece. If utilization is consistently above 90 %, the new inertia is at the limit of the drive's capability and a filter will not solve the underlying issue.
2. Motor utilization (r0031) does not exceed the rated current of the motor for the duty cycle. The rated current is read from the motor data set and is p0305 [Arms].
3. Torque utilization (r0031) does not exceed the short-time torque limit. The maximum torque utilization over a full grinding cycle should be below the S6 / overload capacity stated on the motor nameplate.
4. DC-link voltage (r0026) does not collapse under heavy load; if it does, the drive is regenerating and the input filter is undersized. A DC-link collapse can also cause current-controller instability, which can sound similar to a resonance whistle but is at a different frequency.

For the current measurement question in the source, the answer depends on whether the value is single-phase or three-phase. SINAMICS r0027 reports total DC-link current; r0031 reports motor torque utilization as a percentage of rated; r0078 reports the actual torque-producing current Iq in [Arms]. The single most useful number for verifying the filter is r0078 versus the load profile, not the line current on the input side.

12. Step-by-Step Commissioning Procedure

The recommended procedure for a 840D sl at any software level, applicable to the C-axis in the source scenario:

1. Baseline readback. Save the current parameter set using the HMI function Startup > Drive System > Drives > SERVO_xxx > Commissioning > Save parameters to CompactFlash. Also export the drive parameter set from STARTER/Startdrive. Do not skip this step.
2. Measure the resonance. Use the FFT-on-trace procedure in section 8. Record the frequency, the Q-factor (peak width at -3 dB), and the peak amplitude in dB. Save the trace file.
3. Configure filter 1 as a band-stop at the measured frequency. Use the parameter values from section 7. Set p1656.0 = 1.
4. Copy RAM to ROM and restart the drive (or do Copy RAM to ROM and re-power the NCU if running). The filter is inactive until the next ramp-up.
5. Run a controlled motion test. Jog the C-axis back and forth at the speed used during grinding, with the heavy workpiece mounted. Listen for the whistle. If still present, the peak was either misidentified or is a higher-order mode.
6. Add filter 2 if needed. Repeat the FFT measurement with filter 1 active. A second peak at a higher frequency indicates a second mode. Configure filter 2 as a band-stop at that frequency (section 7).
7. Re-tune the speed controller. With the filters in place, increase p1460 in 10 % steps and re-run the motion test. Stop when the whistle returns or when the position following error during a contouring pass starts to oscillate. Reduce p1460 by 20 %, leave p1462 at the previous value (Tau_mech-based), and document the final values.
8. Run a full grinding cycle with a real workpiece. Verify the surface finish is acceptable, the following error stays within the programmed tolerance, and the drive utilization is below 100 %.
9. Save the parameter set to CompactFlash and back up the STARTER/Startdrive project.

13. Verification and Acceptance Test

After the filter is configured, the following checks confirm a stable installation:

14. Troubleshooting Matrix

Common failure modes during filter installation, with their diagnostic signature and the corrective action: