PAM8302A Amplifier Module: Both Speaker Terminals Are Driven

A small class-D amplifier can look like a two-wire audio source while behaving very differently from a ground-referenced headphone output. Build a modest mono module, calculate its input network and low-level power estimate, and learn how to measure a differential speaker signal without shorting an output to ground.

This is an unbuilt design candidate, with no acoustic, electrical, distortion or thermal measurement claimed. Its software is an offline calculation tool, not firmware or a transistor-switching simulation. Begin with a regulated 5 V source, a 0.25 A current limit, a documented 8 Ω speaker and low input level. The chip’s advertised wattage is not a verified rating for this printed circuit board (PCB).

What class D and BTL mean here

The Diodes Incorporated PAM8302AADCR is an audio-power integrated circuit (IC), combining a switching amplifier and its control functions. The selected order code uses the SO-8 small-outline package with eight leads. PAM8302AASCR instead selects MSOP-8, a different physical package. Do not choose a footprint from the shared “PAM8302A” name and assume every suffix fits.

Class D describes a switching amplifier arrangement. Its power outputs change rapidly between electrical states; their switching pattern carries the audio information. The letter does not mean that a digital audio data bus enters the chip. This part accepts an analog input signal, a continuously varying voltage rather than a sequence of audio sample words.

Bridge-tied load (BTL) means the speaker connects between two actively driven outputs. Its useful differential signal is the voltage difference between those outputs. Neither speaker terminal is the circuit’s zero-volt reference. A connector marked “minus” identifies a differential terminal; it does not automatically identify ground. This distinction matters before connecting a headphone jack, oscilloscope, audio interface or another amplifier.

The complete low-level module

Bias is the steady operating level around which a changing signal moves. The input capacitors let the amplifier establish its own internal bias without forcing the source to share that steady voltage. The capacitors selected here are not polarized. A different polarized part requires checking which side remains more positive in all intended conditions, not copying a generic plus mark from another board.

SD is the shutdown control. The board defaults to muted operation through R3 and uses a manual shunt at J4 for the first experiment. A shunt is a removable conductive jumper, not the current-sense resistor used in other projects. This board does not export a 5 V enable signal to a processor input. Ground, abbreviated GND, remains shared between the input source and the power supply, while the speaker pair remains separate from it.

Gain and the input high-pass response

Gain is output amplitude divided by input amplitude within a chosen operating condition. The data sheet describes nominal internal feedback resistance of 150 kΩ and minimum internal input resistance of 10 kΩ. With 47 kΩ added externally to each input branch, the nominal voltage gain is Av = 150 kΩ / (10 kΩ + 47 kΩ) = 2.6316. Expressed in decibels (dB), a logarithmic ratio, 20 log10(Av) is approximately 8.404 dB. These internal values are not precision calibration resistors.

Each 100 nF coupling capacitor and its nominal 57 kΩ input resistance forms a high-pass filter: very slow changes are reduced more than faster changes. The simple corner calculation is fc = 1 / (2πRC) ≈ 27.92 Hz, where frequency in hertz is cycles per second. At that corner, the ideal single-branch amplitude ratio is 1/√2, approximately 0.707. Source impedance, component tolerances and the real chip’s behavior still matter.

The input network is deliberately symmetrical even though the external source is single ended. “Single ended” means its signal is referenced to GND rather than being a second driven differential pair. IN− reaches that reference through its own resistor and capacitor. Do not bypass this branch merely because the source has one signal wire: the two inputs need a compatible bias arrangement.

For a 1 kHz sine wave at 0.1 V root mean square (RMS), the ideal high-pass factor is about 0.9996 and the estimated audio output is about 0.263 V RMS. RMS is an averaging measure related to heating for a specified waveform. Across an ideal 8 Ω resistance, V²/R gives approximately 8.65 mW. That calculation describes a low-level audio estimate, not a measured speaker power or sound-pressure level.

Why two output readings require subtraction

A differential signal is obtained by subtracting one terminal voltage from the other. A common-mode component is the part that appears on both terminals in the same direction. In a simplified reconstructed-audio example, two terminal averages centred at 2.5 V can move oppositely by 0.5 V. Their difference then swings by 1 V about zero. The real pins carry switching waveforms, so the illustration below intentionally does not draw them as measured smooth sine waves.

An ordinary oscilloscope ground clip can be connected to protective earth through the instrument. Clipping it onto SPK_MINUS may short a driven output through that earth path. A mathematical channel-subtraction button does not repair an unsafe connection already made. Use a properly rated differential probe, or a suitable two-channel method in which both ground clips connect only to circuit GND and the channel probes separately observe the two outputs. Check the instrument’s voltage, common-mode, grounding and bandwidth limits first.

A handheld meter placed across the two outputs may also give a misleading number if its bandwidth or input arrangement responds to the switching carrier. “True RMS” on a meter does not promise separation of audio from switching content. The manufacturer describes a filtered measurement setup for performance characterization. Use an appropriate audio-band measurement method before claiming output power, frequency response or distortion.

Complete arithmetic tool

This program computes the nominal gain, the ideal coupling response, a resistive audio-power estimate and an ideal rail-swing warning. It also demonstrates subtraction of two matched sample arrays. It does not model switching edges, a loudspeaker’s complex impedance, clipping, electromagnetic emissions or the chip’s protection behavior. Its ideal sine limit is an upper-bound thought experiment, not a guarantee of clean output below that number.

"""Ideal small-signal arithmetic for the PAM8302A bench module.
No PWM, speaker, acoustic, noise, efficiency or measured distortion simulation.
"""
import math

def audio_budget(input_rms=0.1, frequency_hz=1000, rin_ohm=47000,
                 coupling_nf=100, load_ohm=8, supply_v=5):
    values=[float(v) for v in (input_rms,frequency_hz,rin_ohm,coupling_nf,load_ohm,supply_v)]
    if not all(math.isfinite(v) for v in values):
        raise ValueError('All inputs must be finite')
    input_rms,frequency_hz,rin_ohm,coupling_nf,load_ohm,supply_v=values
    if input_rms<0 or frequency_hz<=0 or rin_ohm<0 or coupling_nf<=0:
        raise ValueError('Invalid signal or input network')
    if load_ohm<4 or not 4.75<=supply_v<=5.25:
        raise ValueError('Outside the module bench conditions')
    resistance=10000+rin_ohm  # nominal internal10k, plus external resistance
    gain=150000/resistance  # nominal internal feedback150k
    cutoff=1/(2*math.pi*resistance*coupling_nf*1e-9)
    fraction=frequency_hz/math.hypot(frequency_hz,cutoff)
    output=input_rms*gain*fraction
    ideal_limit=supply_v/math.sqrt(2)  # ideal rail-to-rail differential sine only
    return dict(nominal_gain=gain,nominal_gain_db=20*math.log10(gain),
                highpass_hz=cutoff,highpass_fraction=fraction,
                ideal_audio_rms_v=output,ideal_audio_power_w=output**2/load_ohm,
                ideal_rail_sine_limit_v=ideal_limit,
                exceeds_ideal_rail_limit=output>ideal_limit)


def differential_samples(plus, minus):
    if len(plus)!=len(minus) or not plus:
        raise ValueError('Equal nonempty sample arrays required')
    result=[]
    for a,b in zip(plus,minus):
        a,b=float(a),float(b)
        if not math.isfinite(a) or not math.isfinite(b):
            raise ValueError('Finite samples required')
        result.append(a-b)
    return result


if __name__=='__main__':
    import json
    print(json.dumps(audio_budget(),indent=2))
    print('Subtract channel samples:',differential_samples([2,3,2],[2,1,2]))
    print('This subtraction does not make an unsafe earth-clip connection safe.')

Run python3 audio_budget.py, then python3 test_audio_budget.py. The actual function tests include the corner frequency, limiting cases, square-law power relation, load scaling, invalid values and common-mode cancellation. Host execution establishes those mathematical/software behaviors only. No firmware compilation, target execution, physical waveform or measured distortion is claimed.

A controlled listening and measurement experiment

1. With every connection unpowered, inspect the exact SO-8 package, pin-one orientation, no-connect pin, supply polarity and speaker-pair routing. Check that neither speaker pin has been connected to the ground header or a common headphone socket.
2. Remove the enable shunt and leave the speaker disconnected. Connect a regulated 5 V bench source while it is OFF, set the 0.25 A input-current limit, then enable power. Verify VDD at U1 and the default LOW shutdown state.
3. Power OFF before connecting a documented 8 Ω speaker suitable for the intended low power. Keep speaker leads short and together. The initial experiment is one mono channel, not a bridged combination of two boards.
4. Connect a ground-referenced audio generator through J2. Verify its actual signal amplitude and DC offset independently. Begin with the calculated 0.1 V RMS, 1 kHz example, then enable the amplifier using J4.
5. Record supply voltage/current, source level and the observation method. Increase input only in small steps while remaining within the speaker and module test conditions. A louder tone is not evidence that a low-distortion wattage claim has been met.
6. If making waveform measurements, first verify the safe differential connection described above. Compare audio-band differential amplitude with the ideal estimate and record the instrument/filter arrangement. Stop if protection cycling, abnormal heating or unexpected current appears.

The IC’s recommended supply range is 2.0–5.5 V; 6 V is an absolute maximum stress limit rather than an operating target. The selected prototype narrows the bench input to 4.75–5.25 V. Its data-sheet power figures refer to specified supply, load, frequency and total harmonic distortion plus noise conditions. Total harmonic distortion plus noise describes unwanted harmonics and noise relative to the signal; a 10% test condition is not the same claim as 1%.

Keep the supply bypass near pins 6 and 7, route the two speaker outputs as a short pair and keep their high-current loop away from the input components. The SO-8 package’s published thermal resistance does not automatically apply to this custom copper layout. This design therefore makes no 2.5 W continuous module promise. Actual electrical, thermal and electromagnetic compatibility qualification remain work for the physical prototype.

Challenge

Two scope channels each show a large voltage above ground, yet their properly measured audio-band difference is small. Must the speaker be receiving the sum of those two ground-referenced readings?

Downloads and primary reading

Complete source and experiment archive · Blank observations CSV. CSV means comma-separated values, a plain-text table format. The published ATtiny85 passive-piezo article is a different acoustic-load experiment; a piezo branch must not be assumed equivalent to an 8 Ω speaker.

• Diodes Incorporated PAM8302A data sheet, DS41333 Rev. 6-2: exact orderable packages, application circuit, gain equation, limits and suggested footprints.
• Diodes product documentation and change notices: verify applicability to the exact orderable part.
• TDK selected supply capacitor: nominal characteristics and dimensions.
• Distributor-hosted original Diodes DS41333 Rev. 6-2: the actual PDF used for visual pinout, ordering and SO-8 pad-layout verification.