Pulse width modulation carries information in the duration of each pulse rather than its height. Demodulation is the reverse step: recovering the original message from a train of pulses whose widths vary. This page explains how that is done, the block diagram usually asked for in exams, and the waveforms at each stage.
What is PWM demodulation?
In a PWM signal every pulse starts at fixed intervals but ends at a point that depends on the amplitude of the message signal at that instant. A large instantaneous amplitude produces a wide pulse; a small one produces a narrow pulse. The amplitude and the frequency of the pulses stay constant throughout.
PWM demodulation therefore has one job: convert varying pulse width back into varying voltage.
Block diagram of a PWM demodulator
The standard demodulator has four stages in sequence:
- Reference pulse generator — produces a clock synchronised with the leading edges of the received pulses.
- Ramp generator — starts a linear ramp at the beginning of each pulse, so the voltage reached at the end of a pulse is proportional to how long that pulse lasted.
- Sample and hold circuit — samples the ramp voltage at the trailing edge and holds it until the next sample arrives, producing a staircase approximation of the message.
- Low pass filter — smooths the staircase into the original continuous message signal and removes the high-frequency switching components.
A simpler arrangement, used where accuracy is less critical, feeds the PWM signal directly into a low pass filter. Because the average value of a pulse train is proportional to its duty cycle, the filter output follows the message. This is the method used in most microcontroller circuits.
Waveforms at each stage
- Received PWM — constant amplitude, constant pulse rate, varying pulse width.
- Ramp output — a sawtooth rising during each pulse, reaching a higher peak for wider pulses.
- Sample and hold output — a staircase whose step heights trace the shape of the message.
- Filter output — the recovered message signal, a smooth reproduction of the original.
Why the low pass filter is essential
The sampled staircase contains the message plus harmonics of the pulse repetition frequency. The filter’s cut-off is set above the highest message frequency but well below the pulse rate, so the message passes and the switching components are attenuated. Without it the output would be a stepped waveform rather than a faithful copy of the input.
Advantages and limitations of PWM
Advantages
- Noise affects amplitude, and PWM carries no information in amplitude, so it is far more noise-resistant than PAM.
- Simple to generate and demodulate with common components.
- Well suited to power control — motor speed, LED dimming, switching supplies.
Limitations
- Pulse width varies, so the transmitted power varies from pulse to pulse.
- Requires more bandwidth than pulse amplitude modulation.
- Synchronisation between transmitter and receiver must be maintained.
PWM compared with PAM and PPM
- PAM — the message varies pulse amplitude. Simplest to produce, but noise corrupts amplitude directly, so performance is poorest.
- PWM — the message varies pulse width. Good noise immunity, variable transmitted power.
- PPM — the message varies pulse position. Best noise immunity and constant power, but the receiver needs accurate synchronisation.
Frequently asked questions
What is the full form of PWM?
Pulse Width Modulation.
Which circuit actually recovers the message?
The low pass filter. Everything before it converts width into voltage; the filter turns that stepped voltage into a smooth signal.
Can a PWM signal be demodulated with only a low pass filter?
Yes. Since the average value of the pulse train is proportional to its duty cycle, filtering alone recovers the message. It is less accurate than the ramp and sample-and-hold method but adequate for most control applications.
Why does PWM resist noise better than PAM?
Noise mainly alters amplitude. PWM carries no information in amplitude, so amplitude disturbance can be removed by clipping without losing the message.

1 comment
Anonymous
Sir i have two questions rising in my mind:
1) If the clipper being used here gets faulty and it starts to clip voltage levels within the rising ramp function/s. Can you suggest any modification or any alteration in the above mentioned block diagram that could make you get rid of the fault.
2) let suppose if synchronous pulse generator will generate a pulse waveform such that the pulse will end at the ending of each PWM pulse. How this slight variation in the operation of the system would hamper or shall not hamper the final output of the system
Comments are closed.