Experimental Active Noise Cancellation Prototype

The beginning

When I decided that I wanted to start building engineering projects to develop my skills and learn engineering by actually doing it, I started looking around me for problems I could work on.

I wanted to pick a real problem, understand it as much as I could, and see if I could build something that would help with it. I knew that while doing that I would probably have to learn things I did not know yet, and that was part of why I wanted to build projects in the first place.

Why noise?

For more than a year and a half, there was a construction site near my home. From around 9 a.m. to 9 p.m., I could hear drilling almost every day.

So I started asking why. Why does something like drilling or deep construction take this long even with all the machines and engineering we have today?

I started searching about construction methods and different technical solutions. A lot of the ideas I found were expensive, needed a lot of power, had safety problems, or were simply much bigger than something I could realistically work on.

At that point I left the construction side alone and started searching for ways to reduce the noise itself.

That was how I found active noise cancellation.

What is ANC?

Active noise cancellation, or ANC, uses a microphone to pick up sound, a processor to work with that signal, and a speaker to produce another sound that reduces the original sound through interference.

I had already studied waves and destructive interference in physics, so this immediately connected with something I already knew from class. The idea itself looked simple on paper. You have one wave, create another wave against it, and reduce the sound.

Then I started learning what happens when you try to do that in real life.

My first idea

My first idea was honestly unrealistic.

I was imagining a system with many microphones and many speakers that could reduce noise by 80 percent or more, work over a large area, and even deal with difficult changing noise like drilling.

That was what I had in my head before I actually started working on it.

When I started researching and building, I realised that what I was imagining would need years of research and development. There were problems with delay, timing, the position of the microphones and speakers, changing sound, the size of the area, and a lot of things I did not understand yet.

I realised I had made the idea much bigger than what I could actually do.

So I decided to make the simplest version I could.

One processor. One microphone. One output speaker.

I also stopped expecting it to be a product ready for someone to use. I decided to treat it as an experiment and only ask one question: can this small system produce a measurable reduction in sound?

Enlarge ↗

Close

The experiment was designed to compare sound level with ANC OFF and ANC ON while keeping the other experimental conditions fixed.

Enlarge ↗

Close

Enlarge ↗

Close

The system uses an INMP441 microphone to capture the incoming sound and send the digital audio signal to the ESP32 through I²S.

The ESP32 processes the microphone signal and generates the anti-noise control signal. The MAX98357A receives the digital I²S output from the ESP32 and converts it into an amplified audio signal for the speaker.

The speaker then produces the anti-noise wave, which combines with the original sound and reduces it through destructive interference.

Signal path

Sound→INMP441→ESP32→MAX98357A→Speaker→Anti-noise

Code

The code continuously reads the microphone input and focuses on the target noise frequency used in the experiment.

The ESP32 generates a sinusoidal signal at the same frequency. The code adjusts its phase and amplitude to find a lower sound level at the microphone position.

The selected signal is then continuously sent to the speaker as the anti-noise output.

Enlarge ↗

Close

• Six positions were measured from x = 0 cm to x = 50 cm.
• The measured reduction ranged from 1.1 dB to 4.5 dB.
• Maximum measured reduction: 4.5 dB at x = 0 cm.
• Mean measured reduction: 2.8 dB.
• Estimated uncertainty in each sound-level reading: ±0.5 dB.
• Estimated uncertainty in the calculated reduction ΔL: ±1.0 dB.
• The measured reduction became smaller as the measurement position moved away from the intended cancellation region.

To do this experiment, I applied what I learned in the laboratory physics classes I took in high school and the theoretical physics I learned, especially the waves and interference chapters. I filled the information gaps between what I learned in class, the theory, and the electronics by searching for what I was missing and using AI models to help me understand it. More about how I use AI →

I also want to give my appreciation to my physics teacher, Mr. Loay Khalil, who taught me physics.