How loud, and what stops it
The NS4168's datasheet gives 1.2 W into 4 Ω at 3.6 V; on the TK100's 3.3 V that works out to about 1 W, and the kit's 3 W speaker takes it easily. When a sound asks for more than the supply allows, the chip turns its own gain down rather than clip. Running it from 5 V would be louder, and it breaks the logic levels.
About a watt on 3.3 V
The NS4168's datasheet gives its output into a 4 Ω speaker as 1.2 W at 3.6 V and 2.5 W at 5 V, both at 10 % distortion. It gives no figure for 3.3 V. Power goes with the square of the supply, so 1.2 × (3.3 / 3.6)² comes to about 1 W, worked out, not measured.
The kit's speaker is rated 3 W, so on this board it is never the limit.
Why not 5 V
More supply would be louder, but the chip's idea of a logic HIGH moves with its supply: 0.7 × VDD. At 3.3 V that is 2.31 V, which a 3.3 V GPIO clears easily. At 5 V it is 3.5 V, above anything an ESP32-S3 pin can give, so BCLK, LRCLK and DIN would be read unreliably. The datasheet's own tables also disagree about where 5 V sits against the limits. Power the TK100 from 3V3, as printed on every pin.
Too loud gets turned down, not cut
An amplifier cannot swing its output past its supply. Asked to, most cut the tops of the wave flat, which is heard as a harsh crackle. The NS4168 watches for that and turns its own gain down instead, within about 10 ms, by at most 10 dB, and lets it back up over about 1.1 s. The datasheet calls this NCN anti-clipping. Loud passages come out a little quieter instead of distorted.
The supply when it pulls hard
A Class-D amplifier draws current in bursts that follow the sound. The board carries small ceramic capacitors on its supply; the datasheet's typical supply filter is 100 µF with 1 µF near the chip. If loud passages crackle, a 100 µF capacitor across 3V3 and GND at the board is the first thing to try.
When it does not work
The supply is probably dipping when the amplifier pulls hard. The datasheet's typical supply filter is 100 µF with 1 µF close to the chip, and the board carries only small ceramics. A 100 µF capacitor across 3V3 and GND at the board usually helps; so do short, thick supply wires.
Not with an ESP32 driving it. At 5 V the chip wants 3.5 V before it reads a HIGH, and an ESP32 pin gives 3.3 V, so the clocks and the data fall short. Power it from 3V3, as the board is printed.
That is the anti-clipping at work: the chip heard its output clip and turned its gain down, by up to 10 dB, then lets it back up over about a second. Lower the level in the sketch and it stops.
Yes, while CTRL is HIGH: the datasheet gives 13 mA at 5 V with no sound, and no figure at 3.3 V. Driven LOW, CTRL shuts the chip down to about 1 µA, which is the way to save a battery between sounds.
Edit this page — content/books/i2s-amplifier/how-loud.mdx
Questions about this product
See what other owners have asked, and read their solutions.
I2S Amplifier
Loading discussions…
Discuss this article
Ask about this page. The answer stays here, on the page it belongs to, for whoever hits the same wall next.