Electric Motors: The Complete Engineering Guide to Torque Constants, Stepper Sizing, BLDC Commutation, and Driver Selection

A dense engineering guide to electric motors for machine builders: torque constant and back-EMF math, stepper microstepping physics, BLDC commutation and FOC, thermal and magnet limits, and fully worked motor sizing for 3D printers, CNC spindles, and drones.

·

Electric Motors: The Complete Engineering Guide to Torque Constants, Stepper Sizing, BLDC Commutation, and Driver Selection

A ₹450 hybrid stepper will hold a 1.5 kg CoreXY gantry dead still against a 0.45 N·m disturbance — and then miss every third step at 300 mm/s on the wrong driver supply voltage, silently producing a part that is 0.2 mm too short with no error message anywhere. A 500 W brushless spindle will happily chew aluminium at 180 W of true cutting power, then stall instantly in mild steel when the operator doubles the depth of cut, because nobody ever computed the specific cutting energy. Electric motors are the most deceptive component in a build: they look interchangeable (it's all copper, iron, and magnets), yet every motor problem a maker will ever face — layer shifts, burned drivers, hot frames, spindles that bog, drones that sag — reduces to three equations that fit on a Post-it note:

The global electric motor market runs on the order of USD 180 billion a year (IMARC/Grand View estimates for 2024), and roughly 45% of global electricity consumption passes through electric motors. A modern EV carries one to four PMSM traction machines; a Voron 2.4 carries five steppers and two BLDC fans; the Gcode that drives both is identical physics. This guide covers the unified motor model, the four motor families (brushed DC, BLDC, stepper, servo), commutation and field-oriented control, loss and thermal limits, three fully worked sizing examples, driver electronics, an India-anchored pricing table, and the failure taxonomy that follows from getting any of it wrong.

1. The Universal Motor Model: Three Equations, Every Motor

Every electromechanical machine is a Lorentz-force device. A conductor of length L carrying current I perpendicular to a flux density B experiences a force:

A motor is just N conductors arranged on a rotor of radius r, bathed in an air-gap field produced by permanent magnets or field windings. Summing the tangential forces gives torque proportional to current — the torque constant k_t:

The same conductors sweeping through the same field generate a back-EMF proportional to speed — the voltage constant k_e:

where \omega is angular speed in rad/s. The two constants are the same number in SI units (k_t in N·m/A equals k_e in V·s/rad), and the proof is one line of power balance: electrical power converted is EI = k_e\omega I, mechanical power produced is \tau\omega = k_t I \omega. Equating them gives k_t = k_e. Anyone who has ever seen "9.549" in a motor datasheet has seen this identity in disguise — see Section 3.

Kirchhoff around the armature gives the terminal equation (steady state, ignoring the L\,di/dt term):

Rearrange and eliminate I to get the torque–speed line — the single most useful curve in motor selection:

with the two intercepts \tau_{stall} = k_t V/R (stall torque — all supply voltage dropped across resistance, zero useful output) and \omega_{nl} = V/k_e (no-load speed). Three consequences fall straight out of this line, and each one is a real-world failure waiting to happen:

  1. Maximum mechanical power occurs at half of no-load speed, P_{max} = V^2/(4R). A motor loaded for maximum power is running at 50% efficiency and dissipating as much heat in copper as it delivers to the shaft. "The motor is rated 500 W, so 500 W of cutting is fine" is only true at one speed.
  2. Ideal efficiency is \eta = \omega/\omega_{nl}. Real motors deviate below this line due to no-load current I_0 (friction, windage, iron loss): \eta = (I - I_0)(V - IR)/(VI) approximately. The practical takeaway: motors are happiest lightly loaded near their top speed — which is exactly why a 500 W spindle at 180 W of cutting runs cool for hours.
  3. Stall is a heater, not a failure mode. At stall the entire VI input becomes I^2R in the windings. A motor that can run continuously at 3 A can burn out in tens of seconds at 30 A stall — no load on the shaft, all of it heat.

2. Brushed DC Motors: The Commutator's Elegant Brutality

The brushed PMDC motor solves the direction-switching problem mechanically: a commutator (segmented copper cylinder) and carbon brushes reverse the armature current every half-turn, so torque always points the same way. It is the cheapest self-commutating machine ever built — a 12 V, 300 RPM geared PMDC with a plastic gearbox costs ₹150–400 in India — and it remains the right answer for wipers, window actuators, peristaltic pumps, and anything that runs off a battery and doesn't care about efficiency.

The engineering trade-offs are all in the contact. Brushes wear (1,000–5,000 h typical for small PMDC at rated load), produce conductive carbon dust, arc under commutation transients, and set an upper speed limit (commutator arcing and brush float). Speed control is trivial: PWM the terminal voltage at 1–20 kHz and the torque-speed line shifts down with duty cycle, V_{eff} = D \cdot V_{supply}. Stall current is V/R with R typically 1–10 Ω on small motors — this is why cheap gearmotor robots kill battery packs and why every brushed-drive design needs a current limit in the controller.

When to choose brushed DC: budget below ₹500, gear reduction ≥ 50:1 so you don't care about motor efficiency, intermittent duty, and replaceable brushes acceptable. When not to: anything continuous, anything above ~100 W where brush dust and heat matter, anything where 1,000-hour replacement cycles are unacceptable. The BLDC exists precisely to delete the commutator.

3. Brushless DC: Deleting the Commutator, Adding a Controller

A BLDC motor is a PMDC motor turned inside out (or outside in — outrunners put the rotor shell outside the stator for higher torque per gram). The three phase windings are energised in six sequences per electrical revolution — six-step trapezoidal commutation — with rotor position sensed by three Hall sensors (spaced 120° electrical) or, sensorless, by watching the back-EMF zero-crossing on the floating phase.

The maker-facing spec for BLDC is the Kv rating (RPM per volt, no-load). It is just the reciprocal of the voltage constant in different units:

Worked examples from real datasheets: a 2212 920 KV quadcopter motor has k_t = 9.549/920 = 0.0104 N·m/A, so 12 A peak gives 0.125 N·m — matched to a 1045 propeller at ~800 g thrust. A 2306 2400 KV racing motor has k_t = 0.00398 N·m/A: at 30 A burst it produces only 0.12 N·m, but at 2400 × 16.8 V ≈ 40,000 RPM the power is 500 W because P = \tau\omega. Low Kv = torque; high Kv = speed at the same power. Choosing the wrong Kv and compensating with the throttle is how people burn motors: prop load scales roughly with \omega^3, so a 5-inch prop that draws 15 A at 12,000 RPM demands ~44 A at 17,000 RPM on a higher-Kv motor.

Trapezoidal commutation has a fundamental flaw: current jumps phase to phase, producing a ~13% torque ripple at six times the electrical frequency — audible as the characteristic BLDC whine. The fix is field-oriented control (FOC): sample the phase currents, transform to a rotating frame with the Clarke and Park transforms,

then regulate i_q (torque-producing current) and i_d (magnetising, held at zero) with PI loops, and synthesise smooth sinusoidal phase voltages via space-vector PWM. Torque ripple drops below 5%, efficiency peaks at 85–92%, and the motor becomes silent — FOC is why every serious EV runs sinusoidal control and why open-source controllers (VESC, ODrive, SimpleFOC) are worth their ₹15k–40k price tags when a ₹500 ESC's six-step whine and mid-throttle cogging won't do.

Sensorless start: without Hall sensors the back-EMF is unmeasurable at standstill, so sensorless BLDC must start open-loop — a fixed voltage ramp at increasing frequency — before closed-loop commutation can lock on. This is why sensorless hobby motors stutter on startup under load, and why anything that must start under full torque (traction, e-bike hub motors, spindles) either carries sensors or runs an HF-injection sensorless scheme.

4. Hybrid Stepper Motors: The Open-Loop Positioning Workhorse

The hybrid stepper — the NEMA 17 bolted to every 3D printer axis — is a 50-pole-pair PMSM with a rotor designed to sit still. A toothed rotor (50 teeth) aligns with toothed stator poles; alternating pole excitation advances the rotor one tooth pitch per four electrical steps, so 50 \times 4 = 200 full steps per revolution = 1.8°/step. The two phases carry sine/cosine currents I_A = I\cos\phi, I_B = I\sin\phi, and the static torque as a function of electrical angle is approximately:

Microstepping interpolates between full steps by setting the phase-current vector to intermediate angles. The per-step resolution on a linear axis is:

For a 20-tooth GT2 pulley (z\cdot p = 40 mm/rev) at 1/16 microstepping: s = 40/(200 \times 16) = 12.5 µm. At 1/256 (TMC interpolation): 0.78 µm. But resolution is not accuracy. Two hard truths:

  1. Microstep holding torque collapses. The holding torque at a microstep position is \tau_h\cos(\phi) — at a 1/2 step that's 0.707\tau_h. The incremental torque available to move from one microstep to the next is \tau_{inc} = \tau_h\sin(\pi/2m): 9.8% of holding at 1/16, 0.6% at 1/256. Below about 1/16 microstepping you are buying smoothness (less noise, less resonance excitation), not accuracy — the motor cannot reliably hold positions 0.78 µm apart against real friction.
  2. True positioning accuracy of a hybrid stepper is ±5% of a full step (±0.09° on a 1.8° motor), dominated by detent torque (0.01–0.03 N·m on a NEMA 17) and pole non-uniformity — not by your microstep count. If you need guaranteed absolute position, you need a closed loop (encoder + servo, or a closed-loop stepper kit), full stop.

The torque–speed reality check. Stepper torque collapses with speed for two reasons: winding inductance limits current rise (di/dt = V/L) and back-EMF eats supply headroom. The canonical GeckoDrive rule sets the maximum sensible drive voltage:

A typical 42BYGH NEMA 17 (2.3 Ω, 3.2 mH, 0.45 N·m holding) computes to a 57 V ceiling; a 24 V supply is comfortably inside it, and 12 V is the classic mistake — at 450 RPM the motor delivers roughly half the torque it would at 24 V, and "random" skipped steps begin. Datasheets publish the pull-out curve (torque vs speed, the boundary beyond which the rotor loses synchronism): check your operating point against it at maximum feed, not at standstill. Mid-band resonance around 1–2 rev/s is a real phenomenon (the rotor's inertia against the magnetic spring of the detents); 0.9° steppers (400 steps), spreadCycle/stealthChop choppers, and mechanical dampers all mitigate it.

The driver landscape (current ratings are peak chopper limits, ₹ prices are Indian street prices for genuine modules):

Driver · Microstepping · Current · Standout features · ₹

A4988 · 1/16 · ~2 A · Cheap, loud, adequate · 80–150

DRV8825 · 1/32 · ~2.2 A · Higher microstepping, mixed decay tuning · 120–200

TMC2209 · 1/256 (interp.) · ~2 A RMS · stealthChop2 silence, StallGuard sensorless homing, UART · 250–450

TMC5160 · 1/256 · external FETs · SPI, StallGuard2, spreadCycle, high current · 800–2,000

Setting the current limit: the chopper trips when the sense voltage hits V_{ref}, so I_{trip} = V_{ref}/(\text{gain} \times R_{sense}) — gain 8 for A4988, 5 for DRV8825, with R_{sense} typically 0.05–0.1 Ω. Run steppers at 0.6–0.8× rated current: holding torque scales linearly with current, but winding heating scales with I^2 — the last 20% of torque costs 56% more heat and a motor too hot to touch (Section 6).

5. AC Servos: The Closed-Loop Apex

The AC servo is a PMSM (or, in older machines, a permanent-magnet synchronous reluctance hybrid) with an encoder on the shaft — 17-bit absolute (131,072 counts/rev) is now standard even on budget units — and a drive running three cascaded control loops: current (5–20 kHz bandwidth), velocity (0.5–2 kHz), position (100 Hz–1 kHz). Ratings come in pairs: a typical 400 W servo delivers 1.27 N·m continuous at 3,000 RPM and 3.8 N·m peak (3×, for a few seconds). The continuous rating is thermal, the peak rating is magnetic saturation plus thermal time constants — the same physics as Section 6, quantified.

The rule that decides whether a servo will actually perform: inertia matching.

Exceed the ratio and the velocity loop oscillates: the motor has the torque authority to accelerate itself 20× faster than it can accelerate its load, so every gain setting that's stable unloaded rings violently loaded. Between motor and load there is often a gearbox — and J_{load} is referred through a reducer at 1/n^2, which is why a 100:1 harmonic drive lets a 0.02 N·m motor position a robot arm. Sizing torque for a trajectory:

Servos command ₹18k–35k for a 400 W kit in India (Section 9) — the price of guaranteed position, 20× speed range, and 3× peak overload. Closed-loop stepper kits (₹2,500–5,000) split the difference: stepper simplicity with encoder correction, losing steps detected and re-acquired instead of silently ruining the part.

6. Losses, Heating, and the Limits That Actually Matter

Every motor rating is a thermal rating in disguise. The loss budget has three components:

The Steinmetz exponents are the useful part: hysteresis loss grows ~linearly with frequency and as B^{1.6}, eddy-current loss grows with f^2 — which is why iron is laminated (eddy loss in a solid rotor scales catastrophically) and why the same motor running faster on field-weakening gets proportionally hotter even at the same output power. Copper loss hides a temperature trap: copper's resistance rises 0.393%/K, so R(T) = R_{20}[1 + 0.00393(T-20)] — at 120 °C the resistance is 39% higher than cold, and the extra I^2R heat drives the temperature higher still. A motor at its thermal limit can thermally run away on a hot day with no change in load.

Temperature rise follows a simple ohmic analogy: T_{motor} = T_{ambient} + P_{loss} \cdot R_{\theta}, with R_{\theta} the frame-to-ambient thermal resistance (a 60 mm BLDC is ~1.2 K/W: 50 W of loss = 60 K rise). Two temperature ceilings define every motor:

Insulation class (IEC 60085): A = 105 °C, E = 120 °C, B = 130 °C, F = 155 °C, H = 180 °C. The class letter is the winding hot-spot temperature the varnish survives for 20,000 h — the "10 °C rule" halves insulation life per 10 K of over-temperature.

Magnet grade (NdFeB max operating temperature): N = 80 °C, M = 100 °C, H = 120 °C, SH = 150 °C, UH = 180 °C, EH = 200 °C. Exceed it and demagnetisation is irreversible: the motor comes back cooler but permanently weaker, with higher no-load speed and lower torque. Cheap drone motors and cheap steppers are N-grade — the failure mode that converts a ₹600 motor into a ₹600 paperweight after one long hot run. (Worth knowing: China processes ~90% of the world's NdFeB rare-earth magnets, which is the real reason high-grade SH/UH motors carry a price premium and why induction machines remain the default for cost-driven applications.)

7. Sizing: Three Worked Examples

7.1 CoreXY 3D printer X-axis — NEMA 17 stepper

Moving mass m = 1.5 kg (toolhead, carriage, bearings), target acceleration a = 10 m/s², 20T GT2 pulley (pitch radius r = z\cdot p/2\pi = 6.37 mm):

Add 30% for belt friction and bearing drag: 0.124 N·m at the operating speed. Now check speed: 300 mm/s gives \omega = v/r = 47 rad/s = 450 RPM. A 42BYGH 0.45 N·m motor with k_e \approx 0.35 V·s/rad generates 16.5 V of back-EMF at this speed — fine on 24 V (headroom to keep driving current), marginal on 12 V. From the pull-out curve the motor retains roughly 0.25–0.35 N·m at 450 RPM on 24 V: a 2–2.5× margin over the 0.124 N·m demand. Current draw: \tau/k_t \approx 0.35 A RMS at low speed — comfortably inside the 1.5 A rating, which is why the motors survive years of printing. The whole selection lives on one curve; the "my printer skips steps at high speed" failure is this table computed with 12 V and 1/16 torque rolloff instead.

7.2 CNC spindle — BLDC/VFD with specific cutting energy

Spindle sizing goes through material removal rate and the specific cutting energy u_c (energy per unit volume removed):

u_c depends strongly on material — this is why depth of cut "feels" non-linear:

Material · u_c (J/mm³, approx.)

Aluminium 6061 · 0.7–1.1

Mild steel · 2–3

Stainless 304 · 3–4

Ti-6Al-4V · 4–5

Worked: 6 mm 2-flute carbide end mill in 6061 at 18,000 RPM, f_z = 0.05 mm/tooth → v_f = 1,800 mm/min = 30 mm/s. With a_p = 2 mm, a_e = 3 mm: MRR = 180 mm³/s, P_c = 0.8 \times 180 = 144 W, and at 80% efficiency P_{elec} = 180 W. A 500 W spindle has 2.8× headroom — correct for tool wear and transients. The same tool in mild steel at a conservative 1 × 1.5 × 10 mm³/s MRR demands 3 \times 15 = 45 W at the cut — but wait: mild steel at 6 mm cutter width is usually run at 4–8,000 RPM with lower feeds, so MRR craters and power rises; the ₹6,000 hobby spindle stalls exactly here. Torque sanity check at 18,000 RPM (\omega = 1,885 rad/s): \tau = P_c/\omega = 0.076 N·m — minuscule, which is why high-speed spindles are tiny; but at 500 RPM the same 0.076 N·m limit explains why you can't swing a 16 mm drill in a spindle built for 3 mm end mills. VFDs compensate with low-speed torque boost up to their current limit.

7.3 Drone motor — Kv arithmetic in action

A 2306 2400 KV motor on 4S (16.8 V): k_t = 9.549/2400 = 0.00398 N·m/A. At 30 A burst, \tau = 0.12 N·m and P = 16.8 \times 30 = 504 W — spinning a 5-inch tri-blade to ~1.2 kg of thrust. Notice the pattern from the worked examples: torque comes from current and geometry, power comes from speed, and the same 0.12 N·m at 40,000 RPM is 500 W while at 1,000 RPM it would be 12.6 W. Power ratings without a speed are meaningless — always ask "at what RPM".

8. The Driver Is Half the Motor: Power Electronics Essentials

Every DC-side motor drive is an H-bridge: four switches (MOSFETs) steering current through the winding in either direction. The failure modes of motor electronics are physics, not luck:

Open-source controller reality check for makers: a 30 A BLHeli ESC costs ₹400–800 and flies a quad fine; a VESC 6 (₹25k+, FOC, regenerative braking, telemetry) or an ODrive/SimpleFOC board (₹10k–30k) buys sinusoidal commutation, torque control, and the ability to treat the motor as a programmable actuator rather than a throttle. For most CNC/drone/robot builds, the ₹500 ESC is the right call — until you need silence, torque control, or regeneration, at which point the open-source FOC ecosystem is the cheapest path.

9. Selection and the Indian Price Landscape

flowchart TD
    A[What does the motion need?] --> B{Positioning, < 3 N·m, budget?}
    B -->|Yes| C[Hybrid stepper + microstepping driver]
    C --> C1[Lost steps unacceptable? Closed-loop stepper]
    B -->|No| D{Continuous rotation?}
    D -->|High speed, low torque| E[BLDC + ESC or FOC]
    D -->|High torque, low speed| F[Brushed DC + gearbox]
    D -->|Mains power, constant speed| G[Induction motor + VFD]
    A --> H{Guaranteed accuracy + high dynamics?}
    H -->|Yes| I[AC servo + encoder drive]

Component · Spec · Indicative ₹ (2026)

NEMA 17 hybrid stepper · 42 mm, 0.45 N·m, 1.5 A · 350–700

NEMA 23 hybrid stepper · 57 mm, 1.9–2.8 N·m · 900–1,800

Stepper drivers · A4988 / DRV8825 / TMC2209 / TMC5160 · 80 / 150 / 280 / 900

Closed-loop stepper kit · 42 mm + encoder · 2,500–5,000

Drone BLDC · 2212 920 KV / 2306 2400 KV · 350–600

ESC · 30 A BLHeli / 4-in-1 60 A · 400–800 / 3,500–6,000

Brushless spindle + VFD · 500 W / 1.5 kW (80 mm) · 6,000–9,000 / 12,000–18,000

AC servo kit · 400 W, 17-bit encoder · 18,000–35,000

BLDC ceiling fan · 28–35 W · 2,500–5,000

Sourcing notes: Indian robotics distributors (Robu.in, Robocraze, ElectronicsComp, Thingbits, 3idea) stock the entire stepper/BLDC/driver stack with datasheets; spindles and servo kits are cheapest from machine-tool supply channels in Rajkot, Ahmedabad, and Coimbatore, with genuine Delta/Leadshine/motor-clone pricing roughly 30–60% below branded Western equivalents. One efficiency aside worth a ₹ calculation: a BLDC ceiling fan draws ~35 W against ~75 W for an induction fan — a 40 W saving over 8 h/day at ₹7/kWh is ₹818/year, paying back the ₹1,500–3,000 premium in 2–4 years. The same arithmetic is why every EV, e-rickshaw, and drone in the country runs PMSM/BLDC rather than induction.

10. Failure Taxonomy: What Actually Goes Wrong

  1. Layer shifts / skipped steps (3D printers): operating point above the pull-out curve — too much speed on too little drive voltage, or current set too low. Fix with V_{drive} \le 32\sqrt{L} headroom, not with more microstepping.
  2. Demagnetised motor (drones, hot steppers): N-grade magnets run above 80 °C once too often. Symptom: higher no-load RPM, less torque, runs hotter than before. Permanent; replace or spec SH/UH next time.
  3. Burned driver: stalled motor drawing V/R with no current limit, or a powered stepper unplugged mid-run (inductive kick). Always set the chopper current limit; never hot-plug steppers.
  4. Bearing failure: overtensioned belts and radial loads — GT2 belts at 15–20 N of preload eat the small 625-series bearings in NEMA 17s. Support the pulley shaft with an outboard bearing; keep belt tension at the low end of functional.
  5. Resonance stalls (steppers, 1–2 rev/s): the rotor-mass/spring-mode band. 0.9° motors, spreadCycle, or a mechanical damper.
  6. Spindle stall in harder materials: u_c jumps 3–4× from aluminium to steel (Section 7.2). Reduce a_p/a_e or change the tooling — the spindle isn't "weak", the MRR math was wrong.
  7. EMI and erratic behavior: stepper chopping at 20–50 kHz is a broadband noise source. Twist the motor leads, shield the encoder cables, star-ground the frame, and keep signal cables out of the motor-cable bundle.
  8. Overheated "rated" motor on a hot day: copper resistance +39% at 120 °C vs 20 °C; the thermal headroom the datasheet promised on a 25 °C test bench doesn't exist at 45 °C ambient (an Indian reality). Derate ~15–20% above 40 °C ambient.

11. The Ten-Point Motor Selection Checklist

  1. Write the three equations first — V = IR + k_e\omega, \tau = k_t I, P = \tau\omega — and mark your operating point on the torque–speed plane before buying anything.
  2. Convert Kv to k_t (k_t = 9.549/Kv) and check torque at your working current, not the burst rating.
  3. Steppers: set the drive voltage inside 32\sqrt{L} — 24 V for typical NEMA 17s; never 12 V if you want speed.
  4. Set the chopper current to 0.6–0.8× rated and verify by touch after 10 minutes of motion — warm is fine, untouchable is wrong.
  5. Plot the pull-out curve at maximum feed: your operating point must sit under it with ≥1.5× margin.
  6. Respect resolution vs accuracy: microstepping past 1/16 buys smoothness only; true stepper accuracy is ±0.09°.
  7. Compute J_{load}/J_{motor} for servos — under 10:1 general, under 5:1 for high dynamics.
  8. Check the two temperature ceilings: insulation class and magnet grade — and derate for a 45 °C Indian summer.
  9. Spindles: compute P_c = u_c \cdot MRR in the worst material you'll cut, then take 2–3× headroom.
  10. Never hot-plug a powered stepper; always current-limit everything.

Conclusion

The entire discipline of motor selection collapses into a surprisingly small kit: the Lorentz force, the k_t = k_e identity, one straight line on the torque–speed plane, two temperature ceilings, and the habit of computing load demand before reading the marketing wattage. The difference between a printer that runs 3,000 hours without a skipped step and one that eats ₹600 motors is not expensive hardware — it is a 24 V supply instead of 12 V, a chopper current set to 0.7× rated instead of 1.1×, a pull-out curve checked at maximum feed, and a spindle sized from u_c \cdot MRR instead of horsepower folklore. Get the three equations right and every motor in the shop — from the ₹350 stepper in the next 3D printer to the servo kit in the next robot arm — becomes a solved problem rather than a mystery that fails at the worst possible moment, halfway through the part.

More FabFlow blog posts