Matching PDM Microphone Clock and PCM Decimation Rate

A PDM microphone stream is not a sequence of ordinary audio samples. It must pass through an appropriate filtering and decimation process before the resulting PCM samples have a defined audio rate. A wrong clock assumption can produce a recording that sounds plausible but has incorrect duration, pitch, or bandwidth.

Write the clock chain explicitly

Record the oscillator source, peripheral divider, actual microphone clock, capture arrangement, and decimation factor. The STMicroelectronics AN5027 application note explains that the PDM clock rate is the target audio sample rate multiplied by the decimation factor.

For a hypothetical 3.072 MHz PDM stream with a total decimation factor of 64, the resulting PCM rate is 48 kHz. That calculation is valid only if the captured stream and filter configuration actually use those values. A file header declaring 48 kHz does not prove the samples were generated at that rate.

Check capture organization

Determine which microphone data is valid on each clock edge and how the peripheral packs the received bits. Some arrangements combine two microphones on opposite edges; others use separate data lines. Follow the microphone and controller documentation for the exact topology.

Inspect a short raw capture and confirm bit ordering before applying the filter. Reversed bit order or interleaved channels can create noise that resembles an acoustic or power integrity problem. Preserve a raw sample block for independent analysis. Label its channel count and the packing convention used by the DMA transfer.

Treat decimation as filtering

Use the supported decimation filter and its required parameters. Simply keeping every Nth bit does not produce the intended audio signal. Review passband, stopband, gain, output word format, and any additional filtering stages.

Account for filter startup and delay when comparing channels or synchronizing audio to another sensor. A valid output buffer count can still include startup samples that should not be interpreted as steady sound. Verify how the software handles reset and sample rate changes.

Validate time and amplitude

Record a known tone and compare its measured frequency with the actual source. Also compare the number of generated PCM samples with an independently measured interval. These checks expose sample rate errors that a short listening test may miss.

Then examine clipping, background noise, and channel mapping using the intended gain and numeric format. Save clock measurements, filter settings, raw capture organization, and output sample rate together. A complete definition of the digital chain makes later microphone or firmware substitutions reviewable without relying on an apparently successful audio playback.