MAX98357A (Adafruit / Clones): Persistent Crackling & Noise Fix

[Eyenseo]

This took way too long to figure out, so I’m leaving this here to save others from going through the same ordeal.

I'm building a speaker using an ESP32-S3 and a MAX98357A I2S breakout. It should be a straightforward project - great software libraries exist, and the soldering is minimal. However, no matter what I tried, the audio sounded terrible, with constant crackling and distortion.

What I tried (without success):

• Swapping speakers (different impedance, wattage, and designs)
• Switching power supply between 3.3V and 5V
• Replacing the MAX98357A module with a new component from a different supplier
• Trying a different ESP32-S3 board and swapping GPIO pins
• Testing 10+ different audio formats, bit rates, and sample rates
• Testing 5 different audio clips (music, bells, chimes, pure tones)
• Switching to different ESP32 audio libraries
• Checking signals on an oscilloscope (BCLK, LRCK, DIN looked fine)
• Adding two capacitors (0.1µF and 470µF)
• Adding resistors to I2S wires
• AI ...
• ... maybe something else

The Solution

Somewhere online (sorry for not quoting/linking, but I lost the post), I came across a suggestion to remove the four output ferrite beads. I initially ignored it - I shouldn't have.

On the Adafruit-style MAX98357A breakout board:

• Desolder or remove the 4 ferrite beads located near the speaker output terminal (marked by the red line in the picture).
• Put solder across the inner two pads to bridge the connection each, directly to the output terminals (marked by the green lines).

After bypassing those ferrite beads, the sound became crystal clear.

Hope this saves someone else from wasting hours debugging perfectly good code and wiring!

[schreibfaul1]

Thank you for your contribution; I’ve linked to it in the readme.md file.

[yellobyte]

Pls allow me to throw in my two cents on this matter:

1) According to spec the MAX98357A can be connected directly to a speaker load without any filtering. This will work.

2) However, the chip is a Class D amp with chopping frequency of ~0.3MHz and if the lenght of the cables going to a load exceed ~12inches you run the risk of producing and emitting (!!) harmonics of even higher frequency. To prevent that Analog Devices suggest the use of an EMI filter consisting of a ferrite bead and a cap for EMI suppression. They recommend the bead to have a low DC resistance, high current rating and 100-600R impedance @600MHz. Disappointingly they do not give a special maker/model reference for engineers to use. Not even their Evaluation Board specs & schematics give any concrete hint.

3) Quality FBs combining a small form factor, very low DC resistance, high current rating and acceptable impedance at that stated frequency are not cheap and contribute noticeably to the board price. Those Ali clones might feature components on their boards that are not up to the task. This could contribute to your problem.

4) Typical 4...8Ω speakers exhibit series inductances in the 10μH to 100μH range. If your particular speaker (incl. cabling) doesn't fall into this range this might be a contributing fact as well.

In contrast, both the Adafruit Amplifier breakout board and the Yellobyte board YB-ESP32-S3-AMP feature the MAX98357A chip with EMI filter (FB & Cap) at each output. I extensively tested both boards and haven't encountered any such problem with crackling & noise. As for the Yellobyte board: If any such issues were to arise, I would certainly like to know about them, as I am constantly striving to optimize my boards.

Regards
yellobyte