Problem Description
On a Sinumerik System 3 controlled CNC, the main spindle runs at approximately 10% above the programmed setpoint. With a 1000 rpm setpoint, the actual speed stabilizes near 1100 rpm. After several seconds the controller drops the spindle to a stop and raises the KB alarm. The fault is intermittent: under some conditions the speed settles to the correct value and the alarm does not appear, but on the majority of cutting programs the same over-speed condition reappears within a few seconds. At low rpm, an audible mechanical noise is present and intensifies during deceleration.
This pattern — a consistent positive speed offset, intermittent latching, and a contour-following alarm — is characteristic of an analog speed-loop calibration fault, not a mechanical failure of the spindle itself. This article walks through the diagnostic flow, the most likely root causes, the calibration procedure, and the verification steps required to return the machine to production.
System Architecture: Sinumerik System 3 Spindle Drive
The Sinumerik System 3 family (System 3, System 3T, System 3TT) was Siemens' second-generation CNC platform, in service from the late 1970s through the early 1990s. The spindle subsystem in nearly all System 3 installations is a fully analog drive commanded by a DC voltage reference from the NC:
- NC setpoint output: ±10 V analog reference (10 V = max spindle speed for the configured axis)
- Drive module: 6SC61xx or 6RA-series analog thyristor DC drive, or a 6SE-series converter for AC spindle
- Spindle motor: 1PH (AC) or 1GG / 1HU (DC) series
- Speed feedback: DC tacho generator on the spindle, typically 60 V / 1000 rpm, mounted either on the motor rear shaft or on a spindle-encoder end
The NC reads the actual spindle speed from the same tacho (or from a pulse generator on the spindle) and uses it for both display and closed-loop functions such as rigid tapping, threading, and contour monitoring.
Understanding the "KB" Alarm
In Sinumerik System 3, the KB alarm (Kontur-Bahnüberwachung, contour-path monitoring) is the position-following error alarm for the spindle. It is raised when the actual spindle angular position deviates from the commanded position by more than the configured threshold. The fault window is tight in three operating modes:
- Threading (G33 / G34): the spindle angular position must lock to the Z-axis feed; any speed error becomes an angular error that accumulates cycle by cycle.
- Synchronized spindle actions: rigid tap, oriented spindle stop, or any G-code with spindle-synchronized axes.
- Feed-per-revolution (G95): feed is derived from actual spindle speed; if speed drifts above setpoint, the actual feed exceeds the programmed feed, and the axes' following-error monitors can also trip.
A 10% over-speed in a rigid-tap cycle drives the Z-axis into the tap faster than the chip load can retract, generating a per-revolution feed error that exceeds the contour tolerance. The KB alarm then stops the spindle. This is fully consistent with the symptom that the alarm is intermittent: simple G97 (constant-rpm) moves may complete before the error window expires, but synchronized moves fail almost immediately.
Root Cause Analysis
A consistent 10% positive offset between setpoint and actual speed, on an analog drive with DC tacho feedback, has a small, well-defined set of root causes. Use the following table to triage before touching any adjustment:
| # | Cause | Likelihood | Distinguishing Sign |
|---|---|---|---|
| 1 | Tacho feedback trim pot misadjusted (loop gain too high) | High | Speed offset is stable and proportional to setpoint across the range |
| 2 | Tacho generator output drift (magnet aging, brush wear) | High | Offset is constant in V / 1000 rpm regardless of commanded speed |
| 3 | NC setpoint reference scaling (MD) misadjusted | Medium | Speed display in NC matches analog output of drive; fault is upstream |
| 4 | Tacho cable open / intermittent (one brush lifted, partial strand break) | Medium | Offset is irregular and can be RPM-dependent |
| 5 | Drive current limit set too high; speed loop saturated | Low | Speed climbs past setpoint only on load transients, not idle |
| 6 | Mechanical belt or coupling slip | Low | Drift only under load; idle speed correct |
| 7 | Tacho polarity reversed (drive sees negative feedback) | Low | Drive runs away or oscillates, not a steady +10% |
The reported 10% offset — proportional across the speed range, combined with a noise symptom at low rpm — points strongly to a tacho feedback calibration issue compounded by brush / commutator wear on the DC tacho itself.
Diagnostic Procedure
Perform the following measurements with the spindle free to rotate and the machine in a safe state. Lock out the spindle's drive-enable circuit, and use a properly grounded, isolated differential oscilloscope or a high-impedance (10 MΩ) DMM.
Step 1 — Confirm the setpoint reference
- Command a low spindle speed (e.g. 100 rpm) and a high spindle speed (e.g. 3000 rpm) via M03 in MDI mode.
- At the NC's analog output terminals (backplane of the NC, ±10 V reference), measure the DC voltage with respect to analog ground.
- Verify that the voltage scales linearly: 1000 rpm setpoint on a 3000 rpm max-rpm axis should produce +3.33 V (i.e. 10 V × 1000 / 3000).
Expected at ±10V = max_rpm axis:
100 rpm -> 0.333 V
500 rpm -> 1.667 V
1000 rpm -> 3.333 V
3000 rpm -> 10.000 V
If the measured setpoint is correct, the fault is downstream of the NC — in the drive, motor, or tacho loop.
Step 2 — Measure the tacho voltage
Measure the DC voltage directly at the tacho terminals on the drive end of the cable (not at the NC end, to avoid cable-drop effects). A standard industrial tacho is rated 60 V / 1000 rpm (Siemens 1GV series, or compatible).
Expected tacho voltage at 60 V / 1000 rpm:
100 rpm -> 6.00 V
500 rpm -> 30.0 V
1000 rpm -> 60.0 V
3000 rpm -> 180 V
If the tacho produces the expected voltage at the measured actual rpm (e.g. 1000 rpm → 60 V), the tacho is healthy and the calibration must be done in the drive. If the tacho is producing 66 V at 1000 rpm (i.e. 10% high), the tacho itself is over-spec — either replace the tacho or recalibrate its trim network.
Step 3 — Isolate the fault by load test
With the drive running in test mode (current loop only, speed loop disabled if supported) or under light cut, command a fixed setpoint and measure:
- No-load actual speed vs. setpoint
- Loaded actual speed (light cut, e.g. facing pass) vs. setpoint
If no-load speed is 10% high and loaded speed drops back toward setpoint, the drive is saturating the speed loop — the I-term is over-driving, and current limit is set too high. If both no-load and loaded speeds are 10% high, the offset is in the reference scaling of the drive, not in the current loop.
Step 4 — Listen to the tacho
The reported low-rpm noise that intensifies during stopping is a classic tacho brush / commutator signature. On a commutator tach, the brushes ride on a copper commutator with discrete bars; as the commutator wears, individual bar transitions produce a small AC ripple superimposed on the DC output. At low rpm, this ripple has a long period and is audible as a whine. During deceleration, the commutator pattern shifts relative to brush contact and the noise modulates.
This noise is not the direct cause of the 10% offset, but it confirms the tacho is mechanically aged and is the prime suspect for replacement.
Solution: Tacho Feedback Trim and Hardware Repair
5.1 Recalibrate the speed-loop trim pot
On the 6SC61xx spindle drive, the tacho feedback voltage enters a summing junction against the NC setpoint. The loop error drives a current reference, which drives the thyristor bridge. A multi-turn potentiometer on the drive PCB (typically labelled Rxx for "Tacho-Teiler" / tacho divider or P1 for "n-soll / n-ist" gain) sets the tacho voltage scaling.
Procedure:
- Lock and tag the machine. Isolate the spindle drive from mains at the disconnect.
- Wait the manufacturer's discharge time — 5 minutes minimum for 6SC6100 and 6RA drives; the capacitor bank retains dangerous voltage.
- Open the drive cabinet. Locate the spindle drive module and identify the regulator card.
- Identify the n-ist (actual speed) trim pot — usually a 10- or 20-turn blue / white Bourns pot on the regulator card. Reference the drive's documentation for the exact location.
- With the spindle safely uncoupled from the workpiece (no cutting tool engaged), command a mid-range setpoint (e.g. 50% of max rpm).
- Using an isolated screwdriver, turn the trim pot counter-clockwise in small increments (typically 1/8 turn at a time) until the actual speed matches the setpoint as displayed on the NC screen.
- Verify across the full speed range: 10%, 50%, 100% of max rpm.
5.2 Adjust the NC setpoint scaling (alternative)
If the drive has no accessible trim pot, the calibration can be pushed upstream into the NC. In Sinumerik System 3, the spindle setpoint scaling is a machine-data parameter; the exact MD number depends on the System 3 sub-version (3, 3T, 3TT) and the drive type (6SC6100, 6SC6101, 6RA21, 6RA22). Always consult the Siemens Industry Online Support portal for the as-installed configuration before modifying MD.
The scaling relationship is:
n_actual_drive [V] = n_setpoint_NC [V] × (K_drive / K_tacho)
To correct a +10% error, the setpoint gain K_drive must be reduced by a factor of 1 / 1.10 = 0.909, or the tacho feedback gain K_tacho reduced by the same factor.
5.3 Replace the tacho generator
If the tacho itself is producing over-spec voltage (Step 2 showed 66 V at 1000 rpm on a 60 V / 1000 rpm tacho), the tacho magnet has weakened or the brush / commutator interface is degraded. Replace the tacho with a Siemens-rebuilt or new unit of the same type:
- 1GV series for older spindles
- 1FU series for newer spindles with integral encoder
5.4 Inspect the tacho cable and connectors
Open the tacho junction box (usually on the spindle head or motor terminal box). Check for:
- Tightened terminals — loose terminals read as a higher source impedance, which interacts with the tacho's output impedance and can read high.
- Oxidized or green copper conductors.
- Shield integrity — a broken shield picks up drive-side noise that can offset the tacho reading.
- Continuity of both tacho brushes — a partial brush lift gives an RPM-dependent error.
A common, often-missed cause is a single broken strand in a multi-strand tacho conductor; the resulting resistance increase makes the tacho read 5–15% high depending on the drive's input impedance.
Diagnostic Flowchart
Verification
After the trim-pot adjustment or tacho replacement, perform the following verification tests before returning the machine to production.
Test 1 — Static speed accuracy
Command the following speeds in MDI and record both the NC display and the independent reference (hand-held optical or mechanical tach on the spindle nose):
| Setpoint (rpm) | Expected (rpm) | Acceptable Tolerance |
|---|---|---|
| 50 | 50 | ±2 rpm |
| 100 | 100 | ±3 rpm |
| 500 | 500 | ±10 rpm |
| 1000 | 1000 | ±15 rpm |
| 3000 | 3000 | ±30 rpm |
Test 2 — Repeat the threading / tap program
Run the program that originally triggered the KB alarm — typically a G33 threading pass or a rigid-tap cycle. The KB alarm should not appear.
Test 3 — Load response
Run a facing pass at 80% of programmed feed and 50% of programmed rpm. The actual speed should remain within ±2% of setpoint under steady-state cutting load.
Test 4 — Low-rpm noise check
Run at 10 rpm and 50 rpm for 30 seconds each. Listen for the commutator-ripple noise. If still present, the tacho brushes are worn — schedule replacement, but the machine may continue in service until spare parts arrive.
KB Alarm Recovery Procedure
After the underlying speed error is corrected, the KB alarm typically self-clears on the next NC reset. If it latches:
- Select NC reset on the operator panel.
- If the alarm persists, cycle the spindle drive enable (M03 then M05).
- If still present, power-cycle the NC (after backing up parameters) — this clears latched contour errors in older System 3 software.
- If the alarm returns within 30 seconds of any spindle command, the speed error is not actually resolved — re-measure tacho and setpoint voltages per Steps 1–2.
Common Field Errors to Avoid
- Adjusting current limit to fix a speed error. A speed error is a voltage-loop problem; current-limit changes will not correct a 10% offset and can mask the fault.
- Replacing the drive module when the tacho is bad. A new drive on a degraded tacho will still show the offset.
- Confusing KB alarm with a drive-side fault. The KB alarm in System 3 is the NC's contour monitor reacting to a real, measurable speed error, not a drive hardware fault.
- Ignoring the low-rpm noise. It is a leading indicator of tacho failure; ignoring it converts a 30-minute calibration into a multi-day tacho-replacement outage.
- Modifying NC MD without recording the original value. Always write down the factory MD before changing spindle-scaling parameters; an undocumented change can be impossible to reverse on a 30-year-old control.
Preventive Maintenance Schedule
For Sinumerik System 3 spindles with analog DC tacho feedback:
| Interval | Action |
|---|---|
| Monthly | Listen for tacho noise at low rpm; record actual vs setpoint at 500 rpm |
| Quarterly | Measure tacho voltage at 1000 rpm; compare to last quarter (drift > 5% → schedule replacement) |
| Annually | Inspect tacho brushes, clean commutator, check cable / connector integrity |
| Every 2–3 years | Replace tacho as preventive measure on heavily used spindles |
| At every NC software reload | Verify spindle MD are at documented values |
Safety Considerations
Sinumerik System 3 spindle drives are open-frame thyristor converters with exposed DC bus bars at 500–700 VDC. Maintenance work on the drive or the spindle itself must follow these minimum practices:
- Lockout / tagout at the main disconnect. Verify zero energy at the line side.
- Wait the full manufacturer-specified discharge time (5 minutes minimum for 6SC6100 / 6RA drives) before opening the cabinet.
- Verify zero energy at the DC bus with a properly rated (1000 VDC minimum) voltmeter.
- Use isolated tools, one-hand rule, and a second person present for live measurement.
- Spindle must be free to rotate during speed tests; ensure no tool or workpiece is engaged.
- For any work near the spindle, secure the spindle against inadvertent rotation (mechanical brake, if fitted, or a wooden block against a non-critical surface for testing only).
Cross-Reference to Other Siemens Platforms
The same diagnostic flow applies, with hardware differences, to:
- Sinumerik System 8 / Sinumerik 810 / 820 — similar analog drive, but parameter changes go through the NC's machine-data interface; trim-pot locations differ.
- Sinumerik 840C — early digital drive (SIMODRIVE 611), uses encoder feedback rather than tacho; the symptom would be a following error in the drive, not a steady offset.
- Sinumerik 840D with SIMODRIVE 611 digital — fully digital drive, no analog tacho; the equivalent fault would be an encoder count error reported as axis following error, not a steady proportional offset.
If the controller has been retrofitted or upgraded, confirm the current drive type before assuming the tacho architecture described here applies. The original Siemens Industry Online Support portal remains the canonical source for parameter and alarm references on legacy Sinumerik systems.
FAQ
What does the "KB" alarm mean on a Sinumerik System 3?
KB (Kontur-Bahnüberwachung) is the contour-path monitoring alarm. It is raised when the actual spindle position deviates from the commanded position by more than the configured threshold, typically during threading (G33), rigid tapping, or feed-per-revolution (G95) moves. A persistent 10% spindle over-speed will cause a position error large enough to trip the alarm within a few seconds.
Why is the actual spindle speed 10% higher than the setpoint on a Sinumerik System 3?
The most common cause is a tacho feedback trim or scaling error in the analog spindle drive (6SC6100 / 6RA series). Because the speed loop is closed by comparing the NC's setpoint voltage against the tacho's DC feedback voltage, any scaling error in the tacho path appears directly as a speed offset. A 10% offset is consistent with a misadjusted trim pot, a tacho whose magnet has weakened below its rated V / 1000 rpm, or a high-resistance joint in the tacho cable.
How do I measure the tacho voltage on a Sinumerik System 3 spindle?
Connect a high-impedance DC voltmeter (10 MΩ or higher) across the two tacho terminals inside the drive cabinet or at the drive's tacho input. With the spindle running at a known setpoint, measure the DC voltage. A standard Siemens 1GV tacho is rated 60 V per 1000 rpm; at 1000 rpm you should read approximately 60 V. If you read 66 V at 1000 rpm, the tacho itself is producing over-spec voltage, which is one of the root causes of a 10% over-speed.
Can I clear the KB alarm without fixing the underlying speed error?
You can clear the alarm with an NC reset or a power cycle, but it will return within seconds on the next threading or rigid-tap program. The KB alarm is a consequence, not a cause. Always correct the speed-loop calibration or replace the tacho first, then clear the alarm.
Is the low-rpm noise on the spindle related to the speed error?
Likely yes. A commutator-ripple whine that increases during deceleration is the classic signature of worn tacho brushes or a dirty commutator. The brushes and commutator do not directly affect average tacho voltage in a major way, but they do indicate that the tacho is mechanically degraded. A degraded tacho can produce both an offset (from internal resistance variation) and noise (from brush contact bounce), and is a strong candidate for scheduled replacement.
Can I adjust the speed by changing NC machine data instead of opening the drive?
Yes, on System 3 variants that expose the spindle setpoint-scaling MD, the offset can be moved into the NC. However, always record the factory MD value before changing it, and verify the change with an independent tachometer. Pushing the calibration upstream is a workaround, not a fix — a degraded tacho will continue to drift and eventually exceed the MD adjustment range.