The Role of PWM in Signal Reconstruction: Applications and Techniques

Pulse Width Modulation (PWM) is one of those unsung heroes of modern electronics. You see it everywhere—from dimming an LED to regulating a DC motor’s speed—but its most elegant and often misunderstood application lies in signal reconstruction. And nowhere is that more vivid, more tactile, and more demanding than in the world of micro servo motors.

Think about it: a tiny servo, smaller than your thumb, can position its output shaft to within a degree or two, holding that position against a light load, all because of a single square wave. That square wave is PWM, and the process of turning that digital pulse train into an analog position is a textbook example of signal reconstruction.

In this blog, we’ll strip away the abstraction. We’ll look at how PWM actually reconstructs a control signal for a micro servo, what techniques make it work reliably, and what real-world challenges (and clever solutions) emerge when you push these tiny actuators to their limits.

---

What PWM Actually Does in the Context of Servo Control

Let’s start with the basics, but not the boring kind. PWM is not “just” a way to vary power. In servo control, PWM serves as a digital-to-analog converter that encodes a desired angular position.

A standard micro servo (like the ubiquitous SG90 or MG90S) expects a 50 Hz PWM signal—meaning a pulse every 20 milliseconds. The width of that pulse, typically ranging from 1 ms to 2 ms, directly maps to the servo’s angular position. A 1.0 ms pulse commands 0°, a 1.5 ms pulse commands 90°, and a 2.0 ms pulse commands 180°.

But here’s the key: the servo’s internal electronics don’t just “read” the pulse width. They reconstruct a continuous voltage that represents the desired position. This reconstructed voltage is compared to the actual shaft position via a potentiometer feedback loop. The difference drives the motor until the two match.

So when you send a PWM signal, you are not sending position data directly. You are sending a time-domain encoded signal that the servo’s control circuitry must reconstruct into a stable analog reference. That is signal reconstruction in its purest form.

---

The Anatomy of PWM Signal Reconstruction in a Micro Servo

The Pulse Width as an Analog Proxy

The micro servo’s internal circuit typically uses a monostable multivibrator or a dedicated servo control IC. When a rising edge arrives, the circuit starts a timer. The falling edge stops it. The resulting time interval is converted into a voltage by charging a capacitor through a constant current source.

This is a classic pulse-width-to-voltage conversion technique. The capacitor voltage rises linearly during the pulse, and the peak voltage is sampled and held until the next pulse. That held voltage becomes the reference for the servo’s position error amplifier.

The quality of this reconstruction depends on several factors: - Capacitor tolerance and leakage – affects voltage stability over the 20 ms period. - Current source precision – determines linearity of the pulse-to-voltage mapping. - Sampling jitter – introduces position noise if the pulse edges are not clean.

In cheap micro servos, these components are often mediocre. That’s why you sometimes see a servo jitter or drift slightly even when the PWM signal is perfect. The reconstruction is imperfect.

The Role of the Pulse Repetition Frequency

Standard micro servos expect a 50 Hz refresh rate. Why 50 Hz? Because the internal hold capacitor will droop over time due to leakage. If the refresh rate is too slow, the voltage decays too much between pulses, causing position drift. If it’s too fast (e.g., 200 Hz), the capacitor may not fully discharge, leading to a cumulative offset.

This is a classic sampling theorem problem in disguise. The PWM signal is a sampled representation of the desired analog position. The servo’s reconstruction filter (the RC time constant of the capacitor and its load) must be designed to smooth the pulses into a steady DC level without introducing excessive ripple. The 50 Hz rate is a compromise between ripple amplitude and response time.

---

Techniques for Improving PWM Signal Reconstruction in Servo Applications

1. Active Filtering vs. Passive RC Smoothing

Most cheap servos use a simple RC low-pass filter to reconstruct the DC level from the PWM train. But this approach has a fundamental trade-off: a larger time constant gives less ripple but slower response to position changes. A smaller time constant gives faster response but more ripple, which manifests as audible humming or visible jitter.

Advanced techniques use active filtering with operational amplifiers. An active integrator can average the PWM signal over multiple cycles, producing a much cleaner DC level with faster settling. Some high-end servos even use synchronous demodulation, where the PWM signal is multiplied by a reference square wave to extract the DC component with minimal noise.

For a micro servo, you rarely have the space or budget for an op-amp. But if you are designing your own servo driver, adding a second-order Sallen-Key low-pass filter can dramatically improve position stability without sacrificing responsiveness.

2. Dithering and Pulse Width Dithering

Sometimes the PWM resolution is not fine enough. A typical microcontroller generates PWM with 8-bit or 10-bit resolution. At 50 Hz, an 8-bit PWM gives you only 256 possible pulse widths across a 1 ms range. That’s about 0.7° per step—noticeable as coarse positioning.

Dithering solves this by alternating between two adjacent pulse widths over multiple cycles. The servo’s mechanical inertia and the RC filter average these pulses, effectively creating a fractional pulse width. For example, if you want a 1.53 ms pulse but can only generate 1.52 ms or 1.54 ms, you can alternate 3:1 ratio to achieve an effective 1.53 ms.

This technique is a form of sigma-delta modulation applied to servo control. It pushes quantization noise to higher frequencies where the servo’s mechanical system cannot respond, giving you sub-degree precision from a coarse PWM generator.

3. Feedforward Compensation for Nonlinearities

The relationship between pulse width and shaft angle is not perfectly linear in real servos. Dead zones near the endpoints, potentiometer nonlinearity, and gear backlash all distort the reconstruction.

A feedforward technique involves pre-distorting the PWM signal to compensate. You first characterize the servo’s actual response curve—measure the angle for each pulse width—then build a lookup table that maps the desired angle to a corrected pulse width. This is essentially predistortion of the reconstructed signal.

For micro servos used in camera gimbals or robotic grippers, this technique can reduce positional error from ±2° to ±0.3° without any hardware changes.

---

Real-World Signal Reconstruction Challenges with Micro Servos

The 50 Hz Limit and Why It Hurts

The 50 Hz refresh rate is baked into the servo’s design. But for applications requiring fast response—like a quadcopter’s control surfaces or a high-speed pick-and-place robot—50 Hz is painfully slow. The reconstruction filter inside the servo cannot track rapid changes, so the servo lags behind the commanded position.

One workaround is to increase the PWM frequency beyond 50 Hz, but this only works if the servo’s internal circuit can handle it. Some micro servos (especially digital ones) accept up to 200–300 Hz. At these higher rates, the hold capacitor has less time to droop, and the reconstruction filter’s cutoff frequency can be raised, giving faster response.

But there’s a catch: at higher PWM frequencies, the pulse width range may shrink because the servo’s timing circuit misinterprets very short or very long pulses. You must experimentally determine the valid pulse width window for your specific servo at the new frequency.

Electrical Noise and Ground Loops

PWM signals are square waves with fast edges—rich in high-frequency harmonics. In a noisy environment (like near a brushless motor driver), these harmonics can couple into the servo’s reconstruction circuit, causing false triggering or jitter.

A common fix is to add a Schmitt trigger on the PWM input line to clean up the edges. Additionally, a series resistor (100–470 Ω) at the servo’s signal pin, combined with a small capacitor (10–100 nF) to ground, forms a low-pass filter that attenuates high-frequency noise without distorting the pulse width too much.

For critical applications, differential signaling over twisted pairs can reject common-mode noise entirely. But that’s overkill for most micro servo setups.

Temperature Drift in the Reconstruction Circuit

The capacitor and current source inside the servo are temperature-sensitive. As the servo heats up during operation, the reconstructed voltage drifts, causing the shaft angle to change even if the PWM signal is constant.

This is a classic analog drift problem. Digital servos mitigate this by using a microcontroller to measure the pulse width directly (via a timer capture) and compare it to the potentiometer reading digitally. The reconstruction is done in software, not analog hardware. This makes digital servos far more stable over temperature.

If you are stuck with an analog micro servo, you can compensate by measuring the servo’s temperature with a thermistor and adjusting the PWM command accordingly. But that’s a complex hack for a simple component.

---

Advanced PWM Reconstruction Techniques for High-Performance Micro Servos

1. Dual-Edge PWM for Increased Resolution

Standard PWM uses only the falling edge to encode information. But by using both edges of the pulse, you can double the effective resolution. This is sometimes called “dual-edge modulation” or “pulse position modulation with width encoding.”

In practice, the servo’s internal timer measures both the time from the rising edge to the falling edge (pulse width) and the time from the falling edge to the next rising edge (gap width). The sum of these two is the period. By varying both, you can encode more information without changing the fundamental frequency.

This technique is rarely used in consumer servos but appears in industrial-grade micro actuators where sub-millimeter precision is required.

2. Predictive Reconstruction with Kalman Filtering

This is an exotic technique for robotic applications. Instead of treating each PWM pulse as an independent command, a predictive filter (like a Kalman filter) estimates the true desired position based on past pulses and a motion model.

If a pulse is corrupted by noise or missing entirely, the filter predicts the most likely position based on the previous trajectory. This reconstructed signal is then fed to the servo’s control loop. The result is smooth motion even with a noisy or intermittent PWM source.

Implementing this requires a microcontroller with enough processing power to run the filter in real time. But for high-end micro servos used in surgical robots or drone camera stabilization, it can make the difference between shaky and silky-smooth operation.

3. Adaptive Refresh Rate Based on Motion

Instead of sticking to a fixed 50 Hz, some advanced drivers adjust the PWM refresh rate dynamically. When the servo is idle, the refresh rate drops to 20 Hz to save power. When rapid motion is commanded, the rate ramps up to 200 Hz.

The reconstruction filter in the servo must be designed to handle this variable rate without introducing artifacts. This usually requires a phase-locked loop (PLL) inside the servo that locks onto the incoming pulse rate and adjusts its internal timing accordingly.

This technique is still experimental for micro servos, but it points toward a future where PWM-based signal reconstruction becomes adaptive and intelligent.

---

Practical Tips for Working with PWM and Micro Servos

Choosing the Right PWM Frequency

Start with 50 Hz. If your application requires faster updates, try 100 Hz first. Monitor the servo for overheating (higher frequency can cause the motor to run continuously, increasing current draw). Also, check if the servo holds position when the pulse width is constant—if it drifts, the internal reconstruction circuit may not be stable at that frequency.

Cleaning Up the PWM Signal

Use a dedicated PWM pin on your microcontroller, not a software-generated one. Software PWM often has jitter due to interrupt latency. If you must use software, implement a timer-based PWM with minimal overhead.

Add a 1 kΩ resistor in series with the signal line and a 10 nF capacitor to ground at the servo connector. This simple RC filter removes most high-frequency noise without affecting pulse width accuracy.

Handling Multiple Servos

When driving several micro servos, do not share the same PWM signal line. Each servo should have its own pin. If you use a servo driver board (like the PCA9685), note that it generates PWM at a fixed frequency (typically 50–1600 Hz). Make sure the frequency matches your servos’ specifications.

Also, power each servo from a separate regulator if possible. The inrush current when multiple servos start moving simultaneously can cause voltage dips that corrupt the PWM signal reconstruction.

Testing Reconstruction Quality

To evaluate how well a servo reconstructs the PWM signal, send a slowly varying pulse width (e.g., a sine wave from 1 ms to 2 ms over 10 seconds) and measure the actual shaft angle with a potentiometer or encoder. Plot the commanded angle vs. actual angle. Look for: - Hysteresis – different paths when increasing vs. decreasing angle. - Nonlinearity – deviations from a straight line. - Noise – high-frequency jitter in the measured angle.

These metrics tell you how good the reconstruction really is.

---

The Future of PWM in Micro Servo Signal Reconstruction

We are moving toward digital servos that communicate via serial protocols (I2C, SPI, or even CAN bus). These eliminate PWM entirely. But PWM remains dominant for cost-sensitive applications because it is simple, requires only one wire per servo, and works with almost any microcontroller.

Emerging techniques like PWM with embedded data (where the gap between pulses carries additional information) could extend the life of PWM in servo control. Imagine a single wire that carries both position command and feedback data—that would be true bidirectional signal reconstruction.

For now, though, PWM remains the workhorse. Understanding how it reconstructs a continuous position from a train of pulses is essential for anyone who wants to get the most out of a micro servo. It is not just a square wave. It is a carefully crafted time-domain message that, when decoded properly, translates into precise mechanical motion.

And that, in a nutshell, is the quiet brilliance of PWM in signal reconstruction.