Servo Sizing for Machine Builders: Torque Margin, Inertia Ratio and the Motor That Fits the Job
A custom machine shipped with a servo that was generously oversized, and it spent its life apologizing for it: more inertia on the axis, more cost on the BOM, and a loop that never needed the torque it carried. The engineer who sized it picked by habit, not by load, and the machine paid the habit back every day. Servo sizing is the small, teachable discipline that sits between “it moves” and “it moves right,” and it is where the number of machines that feel sluggish or nervous is decided. This article walks the sizing path for machine builders: the moving load, the inertia ratio, the torque profile, the speed check, and the choice that makes the motor fit the axis instead of overwhelming it.
Start With the Moving Load: Speed Curve First, Forces Second
Sizing begins with the mechanical load described as a motion profile, because the motor’s task is defined by what the axis must do, not by the load it is attached to. The profile is the speed-versus-time law: the acceleration ramp, the traverse speed, the deceleration, the dwell, and the cycle time. From that profile come the two numbers that dominate the size: the peak speed the axis must reach and the acceleration it must survive. A machine that needs to move a gantry in half a second across a meter is a different machine, and a different motor, than one that carries the same gantry over seconds.
The forces enter as the load torque against which the motor must work. Cutting force, friction in the guides, gravity on a vertical axis, and any process resistance add to the torque requirement, and the honest load torque is the sum of the worst-case contributions that occur at the same moment. The acceleration torque is added to that: the inertia of the whole system, times the angular or linear acceleration, converted through the transmission. The peak torque of the servo must cover the sum at the worst instant, and the continuous torque must survive the average over the cycle, because a motor is sized by both its maximum and its ability to shed heat.
Inertia Ratio: The Number the Machine Feels as Vibration or Sluggishness
The inertia ratio, the load inertia reflected to the motor shaft divided by the motor’s own rotor inertia, is the single most informative number in servo sizing. A ratio near one gives a loop that is stable and responsive, where the motor does not have to fight the mass it moves; a ratio of five or ten begins to feel sluggish and demands a faster loop with more gain, and a ratio far beyond what the servo can manage produces the nervous, oscillating axis that machine builders spend weeks debugging. The inertia ratio is why two motors with similar torque spec out differently on the same load.
The ratio is adjusted at the transmission, not the motor choice alone. A gearbox or a belt drive with a reduction divided by the radius squared brings the reflected inertia closer to the rotor, making a heavy load tractable for a modest motor; the same reduction that solves the inertia also changes the speed and torque at the load, so the transmission is tuned in the same calculation. The designer’s job is to pick the reduction that lands the reflected inertia in the servo’s comfortable band, and to verify it numerically rather than assuming the gearbox is only about speed.
The Torque Profile: Average, Peak and the Cutting Truth
The cycle’s torque is not a single number, it is a history, and the motor is sized by two averages folded into it. The root-mean-square torque over the whole cycle must stay under the servo’s continuous rating, because the heat the motor sheds is set by the RMS value, not the peak. A cycle with brief heavy cutting and long dwells runs cool despite a high peak; the same peak repeated without pause overheats the motor. The RMS calculation, folded over the cycle, is the honest measure of the motor’s thermal contract.
The peak torque, meanwhile, sets the acceleration and the ability to hold against load spikes, and it is the number that trips up designs whose loads were approximated. A servo rated for continuous torque may still shear its coupling or trip its drive at a peak it was never asked to carry; sizing must check both the average and the maximum, and the margin between them is the design’s breathing room. Torque margin, typically fifteen to thirty percent above the calculated worst case, is the cushion that keeps a motor from living at its limit on the shop floor’s unpredictability.
Speed and the Transmission Check: What the Motor Shaft Actually Sees
The speed check verifies that the motor, through the transmission, can reach the axis’s required traverse speed while still having margin for the dynamic behavior the loop needs. A motor whose rated speed is just barely met by the axis’s fast move has no headroom, and its velocity loop runs at the top of its range, hunting and complaining. The practice is to select a motor and reduction whose combination places the nominal operating speed comfortably within the servo’s envelope, with a few hundred RPM of margin above the fast move’s peak.
The transmission check also catches the matched errors: a reduction that solves the inertia ratio but pushes the motor speed over its rating, or a belt drive whose compliance adds a resonance into the very loop the inertia ratio was meant to stabilize. The transmission and the motor are sized together, three times in one calculation, inertia ratio, speed and torque, and checking them together is what separates a coordinated drive from a motor bolted to a box. The motor that “just fits” on paper and shakes on the machine is the motor whose transmission was never checked.
Regeneration, Braking and the Energy That Has to Go Somewhere
Every decel that slows a driven load turns kinetic energy into electrical energy in the motor, and that energy has to go somewhere. On a modest axis it is dissipated in the drive’s internal resistor; on a heavy vertical axis or a large inertia that stops often, the regenerated energy exceeds what the drive can absorb and a dynamic braking resistor, external and sized, is required. Sizing the braking path is part of the thermal contract of the drive, and a machine that faults on decel or burns its internal resistor is a machine that skipped this page.
The mechanical brake has its own sizing story for vertical axes and safety-hold requirements. A brake that must hold a gravity load against an unexpected power loss is sized for holding torque, not for stopping, and its engagement time and wearing surface are checked against the axis’s worst-case load. The combination of regenerative braking for normal deceleration and mechanical braking for hold-and-emergency is the mature design, and each plays the role the physics demands rather than the role habit assigns.
Drive, Loop Gains and the Motor That Becomes an Axis
The servo motor is half the system; the drive and its loop are the other half, and the same motor with a poorly tuned loop is a nervous axis. The loop gains, proportionally raised until the axis tracks the command and stops short of oscillation, are applied in the classic steps: velocity loop first, position loop second, feed-forward to cancel lag. A motor sized honestly gives the loop a comfortable inertia ratio to work with, and the tuning then lands quickly; an oversized motor with a small load is exactly as hard to tune as an undersized one, because the loop must fight the extra inertia.
The drive’s current and voltage limits are the motor’s own envelope. A motor that demands more current than its drive can deliver never reaches its torque rating; a drive starved of bus voltage cannot accelerate the axis to the profile’s speed. The drive selected with the motor, rated for the motor’s peak current and continuous current, and fed with the motion profile’s worst-case demand, is the contract that makes the sizing real. Servo sizing is complete when the motor, the drive and the transmission together survive the cycle’s torque history and speed envelope, on paper and then on the machine.
The Practical Sizing Checklist for a Machine Builder
The discipline condenses to a working list. Draw the motion profile and read off peak speed, acceleration and cycle time. Sum the load torque from cutting, friction, gravity and process at the worst instant. Compute the reflected inertia through the transmission and pick a reduction that lands the ratio near one. Check peak and RMS torque against the continuous and peak ratings, with margin. Verify the motor’s speed envelope with headroom. Size the braking path for the energy of every decel and the safety of every vertical hold. And tune the loop in the classic order, with measurements confirming the response. Each step is a number on a page, and the page is what the machine carries instead of a habit.
The machine whose servo “felt right” by guesswork is replaced by the machine sized by the profile, the inertia and the torque history, and the difference shows up in the parts, not in the spec sheet. A servo that fits its job runs quiet, holds its position, survives its thermal cycle and asks for nothing during service. The motor that was oversized apologized with every extra gram of inertia; the motor that fits simply works, and the machine builder who sizes the servo instead of choosing it buys the machine’s whole life of good behavior with an hour of calculation.
Linear Motors and Direct Drive: When the Transmission Disappears
Not every axis needs a rotary motor and a transmission, and the direct options change the sizing equation rather than removing it. A linear motor drives the load directly, with no leadscrew or belt, so the inertia ratio problem vanishes into a simpler force-and-mass balance: the accelerating force is the moving mass times the acceleration, and the motor’s continuous and peak force must cover the profile without the lever arm of a transmission to adjust it. The linear motor’s practical cost is the moving mass of its own magnet track and the cooling it needs; the sizing discipline is the same torque-and-speed logic translated into force and velocity.
Direct-drive rotary systems, a torque motor on the rotary axis with no gearbox, similarly trade the inertia ratio for stiffness and simplicity. The motor’s rotor inertia now participates directly in the ratio against the load, and the motor is selected so its own inertia plus the table inertia stays in the servo’s comfortable band. Where the transmission was once the tuning lever, the direct-drive design tunes with the motor’s size and the loop itself. The sizing checklist reads the same, with the mechanical reduction replaced by a careful look at the weight the direct motor adds to the axis it must accelerate.
Multi-Axis Interplay: The Motor That Serves the System, Not Itself
A single axis’s sizing is not complete until it is checked in the company of the others. A gantry’s two drives must match in their dynamic response so the structure does not rack; a rotary axis’s torque demand changes with the posture of the arm or table it carries, so the sizing is checked across the full range of motion, not at a convenient midpoint. The servo that was perfectly sized alone and misbehaves in the assembled machine is usually a sizing error that only appeared at the system level.
The system check also catches the resonance problem that individual sizing cannot see. A structural natural frequency that sits near the loop crossover makes the axis oscillate at exactly the gain the single-axis tuning wanted; matching the drives’ dynamics, adding a filter at the resonance, or stiffening the structure are the system-level corrections. The disciplined builder verifies the assembled axis with a frequency-response measurement, confirming that the sized torque, the tuned gains and the real structure live in a stable arrangement. The servo is sized as a member of the machine, and the machine is tuned as a system of servos.
Feedback and Resolution: What the Motor Actually Knows
The servo’s position truth comes from its feedback, and the sizing story includes the resolution of that feedback against the axis requirement. An encoder count that is too coarse for the positioning spec leaves the motor stepping between its own quantization; a resolver or encoder with more resolution than the loop can use adds cost without behavior. The rule is to match the feedback resolution to the finest position step the axis must hold, with margin, and to verify that the loop can actually command and settle that step without hunting.
The feedback also tells the sizing story about velocity and acceleration. A motor whose velocity feedback is noisy limits the loop’s gain and makes the axis sluggish or jittery at low speed; the feedback chosen with the motor is the difference between a smooth slow traverse and a walking one. The combination of motor, drive and feedback is the complete servo package, and each element of the package is sized against the same profile, the same cycle and the same accuracy that the machine promises.
Verification: The Commissioning Dance That Confirms the Sizing
Every sizing calculation earns its truth at commissioning, when the motor is asked to live its profile for real. The commissioning checklist mirrors the sizing page: measure the actual torque history and compare it to the calculated RMS and peak; sweep the axis through its full speed range and watch for resonance; check the motor and drive temperatures over repeated cycles, looking for the thermal margin the calculation promised; and tune the loop in the classic order with the actual load attached. A sizing that survives commissioning, with margin measured rather than assumed, is the sizing the machine can be trusted with.
The margin measured on the floor is the final signature. A motor whose commissioning temperature runs well below its rating, whose peak torque demand clears the limit with room, and whose loop settles cleanly across the whole envelope, is a motor sized for its life, not for its first hour. The hour of calculation at the drawing board is repaid in the commissioning week that finds no surprises, and in the years of production where the driven axis simply does what it was asked. That is the honest payoff of servo sizing: the unremarkable smoothness of an axis that was engineered to behave.
