The Effect of Motor Torque and Speed on System Stability

Micro servo motors have become the unsung heroes of modern mechatronics. From the nimble fingers of robotic arms to the precise gimbals of camera drones, these tiny powerhouses define the boundary between smooth operation and catastrophic failure. Yet, one question haunts every engineer who has ever tuned a PID loop: How do torque and speed interact to destabilize an otherwise perfect system?

This article is not a textbook derivation. It is a practical, hands-on exploration of the torque-speed-stability triangle, grounded in the real-world behavior of micro servo motors. We will strip away the theory and look at what actually happens when a micro servo is pushed to its limits—and how you can predict, measure, and mitigate instability before it ruins your next prototype.

The Micro Servo Motor: A Miniature Dynamo with Attitude

Before we dive into stability, we need to understand the beast we are dealing with. A typical micro servo motor (e.g., SG90, MG90S, or a high-end digital servo like the Hitec HS-35HD) is a DC motor coupled with a gear train, a potentiometer for position feedback, and a control board that interprets PWM signals.

Torque Characteristics of a Micro Servo

Micro servos are gear-reduced, which means they trade speed for torque. A standard SG90 offers about 1.8 kg·cm of stall torque at 4.8V. That is not much—roughly the force needed to lift a small apple—but inside a 9-gram package, it is astonishing.

• Stall Torque (T_stall): The maximum torque the motor can produce when the rotor is locked. This is the point where current is highest, heat builds fastest, and stability vanishes.
• Rated Torque (T_rated): The continuous torque the motor can deliver without overheating. For micro servos, this is often 50–70% of stall torque.
• Dynamic Torque (T_dynamic): The torque available while the motor is rotating. This varies with speed and load.

The critical insight: Torque is not constant. It drops as speed increases, following a roughly linear inverse relationship—until you hit the motor’s mechanical limits or the controller’s bandwidth.

Speed Characteristics: The Other Side of the Coin

Speed in a micro servo is usually expressed as the time to rotate 60 degrees under no load. For a typical analog servo, that might be 0.12 seconds at 6V. For a digital servo, it could be 0.08 seconds or faster.

• No-Load Speed (ωnoload): The maximum angular velocity when the output shaft has zero external load. This is the theoretical ceiling.
• Loaded Speed (ω_loaded): The actual speed under a given torque load. This is always lower than no-load speed.
• Speed Regulation: How much the speed drops as load increases. Poor regulation means the motor slows dramatically under load, which can trigger instability in closed-loop systems.

The Stability Problem: When Torque and Speed Collide

System stability, in the context of a micro servo, refers to the ability of the control loop to maintain a commanded position without oscillating, overshooting, or hunting. The torque-speed relationship is the primary culprit behind instability.

The Torque-Speed Curve: A Double-Edged Sword

Every DC motor has a characteristic torque-speed curve. For a micro servo, this curve is steep. A small increase in load torque results in a large drop in speed. This is because micro servos have high internal resistance and low rotor inertia.

Consider a typical micro servo operating at 5V:

• At zero load, it spins at ~200 RPM (after gearing).
• At 50% of stall torque, speed drops to ~100 RPM.
• At 90% of stall torque, speed is barely 20 RPM.

This nonlinear relationship is a recipe for instability. When the servo is commanded to move to a position, the control system (usually a PID controller) calculates an error and applies a voltage to the motor. If the load is high, the motor cannot accelerate fast enough. The error persists, the integral term winds up, and the motor overshoots. Then it corrects, undershoots, and oscillates.

The Role of Back-EMF in Stability

Back electromotive force (back-EMF) is the voltage generated by the motor’s rotation, which opposes the applied voltage. For a micro servo, back-EMF is significant because the motor spins fast relative to its size.

• At high speed, back-EMF reduces the effective voltage across the motor windings, limiting current and torque.
• At low speed (under heavy load), back-EMF is low, allowing high current and high torque.

This creates a natural damping effect—but only if the control system can handle it. If the controller’s gain is too high, the back-EMF can cause phase lag, turning the motor into an oscillator.

Gear Train Effects: The Hidden Instability Amplifier

Micro servos use plastic or metal gears to multiply torque. But gears introduce backlash, friction, and compliance.

• Backlash: The slop between gear teeth. When the motor reverses direction, the gears must take up this slack before torque is transmitted. This creates a dead zone in the control loop, leading to limit cycles.
• Friction: Static friction (stiction) requires a higher torque to start moving than to keep moving. This can cause stick-slip oscillations, especially at low speeds.
• Compliance: The gears and output shaft twist under load. This torsional spring effect adds a resonant frequency to the system. If the control loop’s bandwidth approaches this resonance, instability is guaranteed.

Measuring Instability: What to Look For

You cannot fix what you cannot measure. In micro servo systems, instability manifests in several distinct ways.

Jitter: The High-Frequency Shake

Jitter is a rapid, small-amplitude oscillation around the setpoint. It is often caused by:

• Excessive proportional gain (Kp): The controller overreacts to tiny errors.
• Noise in the feedback signal: Potentiometer wiper noise or PWM jitter from the microcontroller.
• Insufficient deadband: The servo tries to correct for noise that is below the mechanical resolution.

Jitter is especially problematic in camera gimbals or laser positioning systems, where even 0.1 degrees of movement ruins the output.

Overshoot and Ringing: The Low-Frequency Wobble

When a servo overshoots its target and then oscillates back and forth before settling, you are seeing the classic symptom of insufficient derivative gain (Kd) or excessive integral gain (Ki).

• Overshoot > 20%: The system is underdamped. The torque is too high relative to the inertia and damping.
• Ringing frequency: This is related to the natural frequency of the motor-load system. For a micro servo with a small load (like a lightweight camera), the ringing frequency might be 5–10 Hz. For a heavier load (like a robotic arm link), it could drop to 1–2 Hz.

Hunting: The Slow, Persistent Drift

Hunting is a slow, periodic oscillation where the servo never quite settles. It is usually caused by:

• Integral windup: The integral term accumulates error over time, causing the output to saturate.
• Load variations: The load changes during motion (e.g., lifting a weight against gravity), and the controller cannot adapt fast enough.
• Friction asymmetry: Different friction levels in different directions cause the servo to overshoot one way and undershoot the other.

Hunting is the most insidious instability because it can look like normal operation to an untrained eye. The servo appears to be “searching” for the position.

Practical Case Study: A Micro Servo in a Pan-Tilt Mechanism

Let us ground this discussion in a real example. You are building a pan-tilt mechanism for a small camera. You use two MG90S micro servos (metal gears, ~2.5 kg·cm torque). The camera weighs 50 grams. The load on the tilt servo is asymmetric: lifting the camera requires more torque than lowering it.

Step 1: The Naive Tune

You set Kp = 1.0, Ki = 0.1, Kd = 0.0. You command a 90-degree pan movement.

Result: The servo overshoots by 30 degrees, oscillates three times, and settles after 1.5 seconds. The tilt servo, lifting the camera, overshoots by 45 degrees and never fully settles—it hunts with a 2-degree amplitude at 0.5 Hz.

Step 2: Diagnosing the Problem

• Pan axis: The overshoot indicates low damping. The load (camera) has inertia, but the torque from the servo is high relative to the inertia. The solution is to increase Kd.
• Tilt axis: The hunting indicates integral windup. When lifting, the integral term builds up because the error persists. When lowering, gravity helps, so the integral term is too high, causing overshoot in the opposite direction.

Step 3: The Torque-Speed Fix

Instead of just tuning the PID, you analyze the torque-speed curve. At 5V, the MG90S delivers ~1.8 kg·cm at stall. The camera’s static torque requirement is 0.3 kg·cm (lever arm × weight). That is only 17% of stall torque—plenty of headroom.

But the dynamic torque during acceleration is higher. To move the camera 90 degrees in 0.3 seconds, you need an angular acceleration of ~1000 deg/s². The required torque is:

Tdynamic = I × α + Tgravity

Where I is the moment of inertia of the camera (roughly 0.0001 kg·m² for a small camera). This gives a dynamic torque of ~0.1 kg·cm, plus the 0.3 kg·cm gravity load—total 0.4 kg·cm. Still well within the servo’s capability.

So why the instability? Because the speed is the issue. The servo’s no-load speed is 0.10 sec/60°, or 600 deg/s. Under load, the speed drops to ~400 deg/s. The controller expects the servo to move at the commanded speed, but the actual speed is lower. The error grows, the integral term winds up, and you get hunting.

Step 4: The Solution

You implement two changes:

1. Reduce the maximum speed command to 300 deg/s. This matches the loaded speed capability, preventing integral windup.
2. Add a feed-forward term that accounts for gravity on the tilt axis. The controller pre-computes the torque needed to hold the camera against gravity and adds it to the output.

Result: The pan axis settles in 0.4 seconds with 5% overshoot. The tilt axis settles in 0.6 seconds with no hunting. Stability is restored.

Advanced Topics: When Torque and Speed Push the Limits

Torque Saturation and Stability

Every micro servo has a maximum current limit. When the commanded torque exceeds this limit, the motor saturates. In a PID loop, saturation causes the integral term to wind up because the error cannot be reduced. This is called integrator windup.

• Effect: The servo overshoots massively when the load is removed (e.g., when the camera is released from a tilted position).
• Solution: Implement anti-windup logic (e.g., clamping the integral term or using a conditional integration method).

Speed-Dependent Damping

The damping coefficient of a DC motor is proportional to the back-EMF constant. At high speeds, damping is high. At low speeds, damping is low. This means:

• At low speeds (high torque), the system is underdamped. Oscillations are more likely.
• At high speeds (low torque), the system is overdamped. The response is sluggish.

For a micro servo that operates across a wide speed range (e.g., 0–600 deg/s), the damping varies by a factor of 10 or more. A fixed-gain PID cannot handle this. You need gain scheduling—different PID gains for different speed ranges.

The Microcontroller’s Role

The stability of a micro servo is not just about the motor. The microcontroller that generates the PWM signal has its own limitations.

• PWM resolution: A standard 8-bit PWM gives 256 steps. For a 180-degree servo, that is 0.7 degrees per step. This quantization can cause limit cycles if the deadband is smaller than the quantization step.
• Update rate: Most micro servos accept PWM updates at 50 Hz (20 ms period). This is the control loop’s sampling rate. According to the Nyquist criterion, the control loop cannot respond to disturbances faster than 25 Hz. If the mechanical resonance is above 25 Hz, the controller cannot damp it.
• Timer jitter: Inconsistent PWM pulse widths due to interrupt latency can introduce noise into the control loop, causing jitter.

Practical Guidelines for Stable Micro Servo Systems

Based on the torque-speed-stability analysis, here are actionable guidelines for your next project.

1. Match the Servo to the Load

• Torque margin: Choose a servo with at least 2x the maximum static torque requirement. For dynamic loads (fast acceleration), use 3x or 4x.
• Speed margin: The no-load speed should be 1.5x to 2x the maximum commanded speed. This prevents the servo from operating in the steep part of the torque-speed curve.

2. Tune the PID with Torque-Speed in Mind

• Start with Kp: Set Kp low enough that the servo does not oscillate at no-load. Then increase until you see the first sign of overshoot, then back off 20%.
• Add Kd: Use derivative gain to dampen the overshoot. The optimal Kd is proportional to the load inertia. For micro servos, Kd is usually small (0.1–0.5).
• Limit Ki: Integral gain should be just enough to eliminate steady-state error. Too much Ki causes hunting. For micro servos, Ki is often 0.01–0.1.

3. Implement Anti-Windup

Always include integral clamping or a conditional integration method. A simple approach: stop integrating when the output is saturated.

4. Use Digital Servos for Critical Applications

Digital servos have a higher update rate (300 Hz vs. 50 Hz) and allow finer PWM resolution. They also have faster response times and better torque linearity. For systems where stability is paramount (e.g., drone gimbals, surgical robots), digital servos are worth the extra cost.

5. Measure the Mechanical Resonance

Use an accelerometer or a high-speed camera to measure the natural frequency of the servo-load system. If the resonance is within the control loop’s bandwidth, you need to either:

• Increase the stiffness of the mount.
• Add a notch filter in the control loop to reject the resonant frequency.
• Reduce the control loop gain at that frequency.

The Future: Smart Servos and Torque-Speed Optimization

The next generation of micro servo motors is incorporating on-board microcontrollers that handle the torque-speed-stability trade-off internally. These smart servos (e.g., Dynamixel, T-Motor) allow you to set target positions and velocities, and the internal controller manages the torque profile.

• Torque limiting: You can set a maximum torque, preventing the servo from stalling or overheating.
• Speed profiling: The servo accelerates and decelerates smoothly, avoiding the torque spikes that cause instability.
• Real-time feedback: You can read back actual torque, speed, and position, allowing external controllers to adapt.

These features do not eliminate the need to understand torque-speed effects. They simply move the problem to a higher level. You still need to choose the right servo for the load, and you still need to tune the higher-level control loop.

Final Thoughts

The effect of motor torque and speed on system stability is not a theoretical curiosity—it is the daily reality of anyone who works with micro servo motors. The small size of these motors amplifies every nonlinearity: gear backlash, friction, back-EMF, and current limits. The torque-speed curve is the Rosetta Stone that deciphers why a seemingly perfect servo oscillates, hunts, or jitters.

By understanding the interplay between torque and speed, you move from a reactive debugger to a proactive designer. You can predict instability before it happens, choose the right servo for the job, and tune your control loop with confidence.

The next time your micro servo starts shaking, do not just turn down the gain. Look at the torque-speed curve. Measure the load. Calculate the dynamic torque. And remember: in the world of micro servos, stability is not a gift—it is an engineering choice.