An automated screwdriving spindle tightens a fastener to a target torque of 8.0 Nm ±10% (7.2–8.8 Nm). The spindle's servo motor draws a current proportional to the torque it produces — but only approximately. Friction in the gear reducer, viscosity variation in the lubricant, and the motor's own cogging torque all consume current that is measured by the drive but is not transmitted to the fastener. A current-based torque estimate might read 8.0 Nm while the actual torque at the fastener head is 7.1 Nm — a 12% error that pushes the process Cpk below 1.33. A direct torque transducer mounted on the output shaft reads 7.1 Nm and commands the motor to keep turning until the fastener sees 8.0 Nm. This guide explains when current-based estimation is good enough and when you need a torque transducer.
How Current-Based Torque Estimation Works
A permanent-magnet synchronous motor (PMSM) or brushless DC motor produces torque proportional to the q-axis (torque-producing) current: T = kt × Iq, where kt is the motor's torque constant (Nm/A). The servo drive measures the phase currents, transforms them into the d-q reference frame via the Clarke-Park transform, and multiplies Iq by kt to estimate motor shaft torque. This is the torque value displayed on the drive's screen and reported to the screwdriving controller.
The torque estimate is accurate at the motor shaft — typically within ±3–5% of actual motor torque under steady-state conditions. The problem is that between the motor shaft and the fastener head is a gear reducer (planetary, cycloidal, or harmonic), an output shaft, a bit holder, and a driver bit — each of which introduces friction that consumes some of the motor torque before it reaches the fastener. The gear reducer alone has an efficiency of 70–90% depending on the reduction ratio, gear type, and lubricant condition. A 10:1 planetary reducer running at 80% efficiency means that 20% of the motor torque is lost to gear friction — and critically, that friction varies with temperature, speed, and wear. The same motor torque reading of 8.0 Nm might correspond to 7.5 Nm at the fastener when the gearbox is cold (thick grease) and 7.9 Nm when it is warm (thinned grease) — a 5% variation that the current-based estimate cannot distinguish because it is not measuring output torque at all.
How Direct Torque Measurement Works
Direct torque measurement places a strain-gauge-based torque transducer in the drivetrain — typically between the gear reducer output and the bit holder — where it measures the actual torque transmitted to the fastener. The transducer outputs a millivolt signal (typically 1–2 mV/V at full scale) that is amplified, digitized, and sent to the screwdriving controller. The controller closes a torque loop around this signal: when the transducer reads the target torque, the controller stops the motor regardless of what the motor current was doing.
The accuracy of a strain-gauge torque transducer is typically ±0.25–0.5% of full scale, with temperature compensation (the strain gauge bridge is inherently temperature-sensitive, but the transducer includes compensation resistors or digital temperature correction). A 0–20 Nm transducer with 0.5% accuracy reads within ±0.1 Nm — an order of magnitude better than current-based estimation and, critically, immune to gearbox friction variation, lubricant aging, and motor parameter drift.
When Current-Based Estimation Is Sufficient
Current-based torque control is acceptable when the torque tolerance band is wide enough to absorb the estimation uncertainty. The uncertainty stack-up for current-based estimation includes:
- Motor torque constant variation: ±2–3% (kt tolerance + temperature drift of permanent magnets)
- Gearbox efficiency variation: ±3–8% (temperature, speed, wear, manufacturing tolerance)
- Bit/socket friction: ±1–3% (bit wear, socket fit, fastener head condition)
The RSS (root-sum-square) of these uncertainties is approximately ±5–10% of the estimated torque. If the fastener's torque specification is 8.0 Nm ±15% (6.8–9.2 Nm, a 2.4 Nm window), and the current-based estimate has ±10% uncertainty (±0.8 Nm), a Cpk of 1.33 is achievable — the process window is wide enough relative to the measurement uncertainty. If the specification is 8.0 Nm ±8% (7.36–8.64 Nm, a 1.28 Nm window), current-based estimation cannot achieve Cpk ≥ 1.33 because the measurement uncertainty alone consumes over half the tolerance window.
The Process Capability Threshold
The decision rule: if your torque tolerance exceeds ±12% of nominal, current-based estimation is generally sufficient and the cost savings (no torque transducer, no transducer amplifier, simpler mechanical integration) are substantial — typically $1,000–3,000 per spindle saved. If your torque tolerance is below ±10%, direct torque measurement is justified because current-based estimation cannot deliver the required process capability.
The WEBER C5S digital screwdriving control unit supports both current-based torque estimation for cost-sensitive applications and direct torque transducer input for precision fastening — with 15 programmable fastening programs and CANopen interface for line-level integration. For precision fastening applications spanning multiple spindle types and torque ranges, the wider servo drive and motor catalog includes drives with integrated torque control functions for both indirect and direct measurement architectures.
Current-based torque control measures what the motor produces. Direct torque measurement measures what the fastener receives. The difference — gear friction, bit wear, lubricant viscosity — can be 10% or more of the measured value. If your torque tolerance is tighter than ±10%, that difference is too large to ignore.
