Servo Motor Chain Drive vs Timing Belt: CNC Rotary Table Retrofit

1. Problem Overview: Coupling a Servo Motor to a CNC Rotary Table

Retrofitting a manual or hydraulic rotary table to servo-driven CNC control requires selecting a mechanical coupling between the servo motor shaft and the table's input shaft. The most common candidates are:

• Synchronous (timing) belt – typically HTD, RPP, or PD profile with molded or metal sprockets.
• Roller chain – ANSI #25, #35, or #40 double-strand chain with corresponding sprockets.
• Spur or helical gear set – direct gear mesh or anti-backlash split gear pair.
• Direct coupling or gearhead – zero-backlash planetary reducers mounted to the motor face.

Each option trades off stiffness, backlash, resonance, cost, envelope size, and maintenance access. For a 30" Pratt & Whitney (P&W) style rotary table driven off a Siemens 1FT/1FK-series servo recovered from an 1980s Cincinnati machining center, the constraints are tight:

• Center distance: approximately 4-1/2" between motor and jackshaft.
• Maximum belt or chain width envelope: roughly 1" nominal, with 1-1/4" possible if flange guidance is removed on one sprocket.
• Existing reduction: 4:1 jackshaft to worm.
• Pre-existing worm/wheel backlash: approximately 1 arc-minute at the table.
• Service access: poor – splitting the rotary table from its base is required to inspect tension.
• Application: 5th-axis trunnion on a horizontal machining center, low duty cycle, semi-finish profiling.

These constraints eliminate most gearhead options (insufficient room for an inline planetary reducer) and push the decision toward belt or chain. The remainder of this article works through the engineering analysis, anti-backlash options, and verification steps that determine which coupling survives in service.

2. Coupling Method Comparison

Roller chain is mechanically robust and tolerant of contamination, but the pitch impulses excite torsional modes in the motor/load system and inject velocity ripple into the encoder feedback. Timing belts attenuate these impulses through the elastomer body and provide a cleaner, lower-noise coupling. Gears deliver the stiffest path but are unforgiving of center-distance error and require precision machined sprockets.

3. Roller Chain Drive Engineering Analysis

ANSI roller chain (#25, #35, #40) is regularly used on industrial servo indexers, packaging machinery, and conveyor servos, but rarely on high-bandwidth CNC axes. The reasons are measurable.

3.1 Chordal Action and Velocity Ripple

As the chain engages the sprocket, it does not wrap around a perfect circle – it forms a polygon. The radial distance from the sprocket center to the chain pin varies by approximately:

Δr = r · (1 − cos(180°/N))

where N is the number of teeth. For a 21-tooth sprocket, Δr is about 0.71% of the pitch radius. The chain enters and leaves the sprocket with a sinusoidal radial velocity variation at the tooth-meshing frequency, equal to the shaft rotation frequency multiplied by the tooth count.

For a Siemens 1FK7060 motor at 3000 rpm with a 21T sprocket, the chordal ripple frequency is:

fchordal = 3000/60 × 21 = 1050 Hz

This frequency often falls inside the velocity-loop bandwidth of a typical servo (100–500 Hz for current loop, 20–80 Hz for velocity loop), and depending on phase relationship, it can be amplified by the controller integral action. The result is a velocity-following error that oscillates at the chordal frequency – visible as a periodic pattern on machined circular profiles.

3.2 Tooth-Engagement Impact

Each chain pin drives into the sprocket tooth with a small impact. The impulse magnitude depends on chain pitch, sprocket quality, tension, and lubrication. This generates:

• Acoustic noise at the tooth-meshing frequency and its harmonics.
• Torsional vibration transmitted back into the motor shaft.
• Encoder noise – particularly on incremental encoders mounted on the motor rear, where the vibration couples through the bearing.

When the encoder is mounted to the motor housing (resolver or integrated encoder), the vibration path is the shortest possible. The track density of a typical 20-bit Siemens absolute encoder (~1.3 million counts/rev) is high enough that sub-arcsecond disturbances are detectable on the feedback.

3.3 Backlash in Chain Drives

Chain slack accumulates as a function of span length and the number of articulation points. For a 4-1/2" center distance with #25 chain, the expected linear slack is roughly 0.5–1.0 mm before tensioning. A properly tensioned chain will have less than 0.2 mm of linear slack, which translates to angular backlash at the driven sprocket of:

θbacklash (rad) ≈ Δs / rdriven

For a 21T driven sprocket (#25, pitch radius ∼33 mm), 0.2 mm slack is about 6 milliradians, or 0.34°. After the 4:1 jackshaft reduction, this becomes 0.085° at the worm, or roughly 5 arc-minutes at the table. Combined with the existing 1 arc-minute worm backlash, the total would be about 6 arc-minutes. The source case notes that 1 arc-minute is acceptable for the operator; 5–6 arc-minutes may not be.

3.4 Mitigation Options for Chain

If chain must be used, the following reduce ripple and backlash:

1. Use the largest practical sprockets (40T or more) to reduce chordal action and lower meshing frequency.
2. Use high-quality roller chain with matched pins, factory-preloaded.
3. Add an idler sprocket on a spring-loaded tensioner to maintain constant wrap angle and absorb slack.
4. Run chain with light oil film lubrication; dry chain wears rapidly and increases impact energy.
5. Mount the encoder on the load side of the coupling rather than the motor rear, eliminating the motor-bearing vibration path.

4. Timing Belt Drive Engineering Analysis

Synchronous (toothed) belts are the de-facto standard for CNC servo coupling on knee mills and small rotary tables. They combine low backlash, low noise, no lubrication, and predictable resonance.

4.1 Belt Profile Selection

For a 4-1/2" center distance and 1" width envelope, a 21" 5mm HTD or 21" XL belt typically fits. Goodyear Eagle NRG is explicitly marketed for compact drives and offers a stiffer elastomer than commodity HTD, which reduces the spring rate of the belt span and pushes the first torsional mode higher.

4.2 Belt Resonance

The first torsional resonance of a belt span can be estimated from:

fn = (n · c) / (2 · L)

where c is the belt wave speed (typically 1500–2500 m/s for steel-cord HTD), L is the span length, and n is the mode number. For a 21" (533 mm) HTD belt with c = 1800 m/s:

f1 = 1 × 1800 / (2 × 0.533) ≈ 1690 Hz

This is well above the velocity-loop bandwidth and typically causes no servo problem. By contrast, a chain span with the same center distance has a much lower effective stiffness, and the first mode often falls in the 200–600 Hz range – squarely inside the velocity-loop bandwidth. This is the dominant reason chains are not preferred on high-bandwidth CNC axes.

4.3 Backlash in Belt Drives

Modern HTD and RPP belts have near-zero geometric backlash because the teeth fill the sprocket grooves. The remaining sources of lost motion are:

• Belt elastic stretch – linear in transmitted torque, generally negligible at 1:1 ratio.
• Sprocket runout – dependent on manufacturing quality, typically <0.05 mm for 21T steel pulleys.
• Belt tooth deflection – the elastomer compresses by ~0.1 mm under rated torque, contributing a small repeatable error that is mostly absorbed by the controller's friction model.

Total angular backlash for an HTD belt at 4-1/2" center distance is typically under 0.5 arc-minutes at the driven shaft – on the same order as the existing worm backlash.

5. Anti-Backlash Gear Options

If envelope permits, an anti-backlash split gear set is the stiffest, lowest-backlash option. The configuration: two identical gears mounted on the same shaft, one fixed, the other rotationally preloaded by a torsion spring. The spring force spreads the driven gear's tooth flanks against both sides of the mating pinion, eliminating backlash.

5.1 Spring Rate and Slip

The preload torque must be large enough to overcome the transmitted torque without slipping, but small enough to avoid excess friction and heat. The rule of thumb is:

Tpreload ≥ 0.5 × Tpeak transmitted

For a typical Siemens 1FK7060 with rated torque ~6 Nm and peak ~18 Nm, preload is 9 Nm on the split gear – achievable with a small clock spring in a machined pocket on the hub.

5.2 Geometry Constraints

A 1:1 gear set with 32 diametral pitch, 20-tooth pinion, 60-tooth gear has a center distance of 2.5" (63.5 mm), which fits inside the 4-1/2" envelope. Face width on a precision gear set is typically 0.25–0.50" – within the 1" width limit if mounted in a single plane. The constraint is the gear “envelope” – the axial space required for the gearset plus bearings. The source case ruled this out on space.

6. Encoder Feedback Considerations

Modern Siemens servos in the 1FK7 / 1FT6 / 1FT7 family use:

• Incremental encoders: typically 2048–8192 ppr with sin/cos (1 Vpp) interpolation to 20+ bits in the drive.
• Absolute encoders: EnDat 2.1 / 2.2 protocol, 25-bit resolution (~33 million counts/rev) for 1FT7 motors.
• Resolvers: legacy 2-pole, 4096 Hz excitation, low-noise but lower resolution (~14 bit).

The encoder is mechanically coupled to the motor shaft either directly, through a hollow-shaft mount, or via the rear shaft extension. Vibration transmitted through the bearings from a chain or gear coupling can couple into the encoder scale and create periodic position errors.

6.1 Quantifying Encoder Noise

A typical high-resolution encoder has a noise floor of ~5–15 arc-seconds RMS in a benign mechanical environment. Under chain-impact excitation at the chordal frequency, this can rise to 30–60 arc-seconds, depending on chain quality and motor bearing stiffness. For a trunnion axis with arc-minute tolerable error, this is acceptable; for a spindle or a finishing pass on a diameter, it would not be.

6.2 Mitigation: Encoder Relocation

If the coupling choice is unavoidable (for example, the user already owns a Siemens motor with rear encoder), the encoder can be relocated to the driven shaft using:

• Belt-mounted encoder on the jackshaft, reading a small timing pulley and an EnDat or TTL disk.
• Direct-mount encoder on the worm shaft with a flexible coupling, eliminating the jackshaft from the feedback loop.
• Linear encoder on the table face for full-closed loop (rare on rotary axes).

7. Resonance and Servo Stability

A coupling introduces a 2-mass torsional system: motor inertia, coupling stiffness, load inertia. The natural frequency of this system is:

fn = (1 / 2π) × sqrt(K × (Jm + JL) / (Jm × JL))

where K is the torsional stiffness of the coupling in Nm/rad, and J are inertias. The stability rule of thumb for a PI velocity loop with bandwidth f_BW:

fn ≥ 3 × fBW

For a Siemens SINAMICS S120 drive with default 1FK7060 motor, the velocity-loop bandwidth is ~200 Hz. The coupling resonance must exceed 600 Hz.

7.1 Effective Stiffness Comparison

These values are first-order estimates and depend heavily on the actual load inertia reflected through the 4:1 jackshaft and worm. If the worm is highly loaded, the load inertia at the motor shaft can be 10–20× the motor inertia, pushing all coupling resonances down. In the source case, the motor was selected as “way bigger than needed,” so the inertia ratio is favorable.

8. Sizing the Coupling

The transmission torque on the rotary table is set by:

Ttable = (Fcut × rworkpiece) / (ηworm × ijack)

For a typical 30" P&W rotary table holding a 20" diameter workpiece at 500 lb cutting force and a 0.4 worm efficiency, 4:1 jackshaft:

Ttable = (500 × 10) / (0.4 × 4) = 3125 lb·in = 353 Nm

After the 4:1 reduction, the jackshaft sees 88 Nm. After another 1:1 (or whatever ratio the coupling runs at), the motor shaft sees the same 88 Nm steady state, with acceleration transients up to ~150 Nm.

For belt selection, the design torque should be at least 1.5× peak, or 225 Nm on the coupling. An HTD 5mm 1" wide belt with 21T steel sprockets is rated for ~3.0 kW continuous at 3000 rpm – well above the 1.5–2.0 kW peak expected here. A #25 double-strand chain with 21T sprockets is rated for ~1.5 kW continuous – marginal at peak.

9. Installation Considerations on a Pratt & Whitney Rotary Table

The 30" P&W rotary table has specific mechanical features that constrain the coupling design:

1. Cast housing access: the worm housing is a sealed casting. Drilling the housing for an external mount requires careful deburring and sealing against chip ingress.
2. Jackshaft bearings: the original jackshaft is supported by sleeve bearings. When converted to servo, the bearing play and runout must be checked; if excessive, replace with tapered roller bearings.
3. Worm & wheel: typically bronze worm wheel meshing with a hardened steel worm. The wheel is shrunk-fit into the table base. Disassembly for inspection is non-trivial.
4. Encoder mount: the tach housing on the original motor may need modification. In the source case, the tach housing required ~0.030" of material removal to clear the new encoder body. A 3mm sprocket pilot was reamed to 1/8" (3.175 mm) to match the machine's other encoders.
5. Shaft length: the source initially feared the output shaft would need shortening, but a 21" belt with same-size pulleys provided enough clearance, requiring only ~0.25" of grinding on internal webbing.
6. Motor bracket: a cast or fabricated bracket must be added to mount the servo. The bracket must be stiff enough to resist motor reaction torque without flexing into the coupling resonance.

10. Recommended Coupling Decision Matrix

For the source case (4-1/2" CD, 1" width, 1 arc-minute acceptable worm backlash, 5th-axis trunnion), the user's final decision – a 21" 1" wide belt, same-size pulleys, with minor internal grinding – is the textbook-correct outcome. The chain option was abandoned primarily on harmonics and encoder noise concerns, both of which are quantitatively justified above.

11. Commissioning and Verification Procedure

1. Mechanical pre-check: with the coupling installed but the drive de-energized, rotate the motor by hand and feel for binding, rough spots, or backlash. Backlash should be < 0.5° at the motor shaft, or ~0.125° at the jackshaft.
2. Belt/chain tension: use a belt tension gauge (e.g., Gates Sonic 507C) to set static tension to 1.5× the recommended static value for 1:1 ratio. For chain, deflect 1/64" per inch of span under finger pressure.
3. Encoder alignment: power the drive in commissioning mode and verify the encoder count increments match the commanded motion direction. Reverse the motor connector if inverted.
4. Auto-tune: run the Siemens SINAMICS STARTER or TIA Portal auto-tune routine. The default 1FK motor database should be selected. The drive will measure coupling resonance and set the velocity-loop filter.
5. Following-error test: command a 0.1 rev move at 100 rpm and capture the following error on the trace. The steady-state following error should be < 1 encoder count (1 Vpp analog) or < 0.001°.
6. Roundness test: mount a test indicator on the table face and command a full rotation at 50 rpm. Total indicated runout should be within the original P&W spec (typically 0.001–0.003" on a 30" table).
7. Surface test cut: take a light climb cut on a circular workpiece with a 0.020" depth of cut and 5 ipm feed. Inspect the surface for periodic patterns at the chordal or mesh frequency – a chain drive will show this on a fine finishing cut; a belt will not.
8. Thermal check: run the axis at 50% rated torque for 30 minutes. Bearing and coupling temperatures should stabilize below 60°C.
9. Long-term monitoring: log the following error and motor current over 8 hours of typical use. A chain drive will show increasing impulse amplitudes as the chain stretches; a belt will not.

12. Field-Proven Notes and Edge Cases

• Servo with rear shaft and integrated encoder: when the encoder is mounted on the rear of the motor (common on 1FT6/1FT7), the encoder-to-coupling vibration path is short. Both chain and belt are workable, but chain will show measurably higher encoder noise.
• Resolver-equipped motors: resolvers are inherently more tolerant of vibration than optical encoders. A 1FK motor with resolver is more forgiving of a chain drive than the equivalent 1FT7 with EnDat.
• Goodyear Eagle NRG: marketed for compact drives with no flanges required on the pulleys. The reduced hub diameter saves axial space – a significant advantage in the 1" width envelope. Goodyear Engineered Products publishes part numbers and load ratings by belt width and length.
• Worm backlash accumulation: the 1 arc-minute reported in the source is the table backlash under no load. Under cutting load, deflection in the worm/wheel mesh can add 50–100% of the no-load value. A coupling with 5–6 arc-minutes of backlash (chain option) on top of the loaded worm can easily exceed 8–10 arc-minutes – visible as part taper on a diameter.
• Motor sizing: the source motor was “way bigger than needed” for the application. This is a feature, not a bug – the larger motor inertia reduces the load-to-motor inertia ratio, which raises the coupling resonance and gives the velocity loop more headroom.
• Lubricant choice on chain: if chain is mandatory, use a light synthetic chain oil (e.g., Klüber Syntheso LM 220) rather than grease. Grease attracts abrasive dust and accelerates sprocket wear in this environment.
• Service access: the user noted that disassembly is required to inspect the coupling. A belt that can be inspected visually (cracks, glazing) is preferable to a chain that requires tension measurement. A chain that runs dry for 6 months will fail without warning.

13. Summary Decision Path

For a CNC rotary table retrofit with a 4–5" center distance, 1" width envelope, and a worm-driven table with ~1 arc-minute of existing backlash, the decision tree is:

1. If the coupling resonance (estimated by the formula in section 7) is above 600 Hz, all three options are viable.
2. If envelope allows a gear set with anti-backlash preload, choose gears for maximum stiffness and minimum backlash.
3. If envelope forces a flexible coupling, choose HTD 5mm or Goodyear Eagle NRG belt over chain unless the environment prohibits elastomer (chemicals, extreme heat).
4. If chain is mandatory, use the largest practical sprockets, add a spring-loaded idler tensioner, mount the encoder on the load side, and accept ~6 arc-minutes of total backlash.
5. Verify with a roundness test cut after commissioning. Re-evaluate the decision if periodic surface marks appear at the chordal frequency.

The source case – a 21" 1" wide belt with same-size pulleys and minor internal grinding – is the recommended outcome for this envelope. The chain option was rejected on quantitative grounds (resonance, encoder noise, backlash), and the gear option was rejected on envelope.