User's Manual

UMXXX
PDM modulated output. The sound inlet (port) can be at the top or at the bottom of the
package. Top-port microphones are easier to assemble (sound port can be easily protected
during assembly), but SNR and frequency response are not optimal; bottom-port
microphone are more difficult to use (soldering process must be very careful to avoid dust
or soldering paste enter the sound port, a precisely aligned hole must be carved in the PCB,
PCB thickness does matter), but SNR and frequency response are optimal.
The MP34DT05-A is an omnidirectional top-port digital microphone: the acoustic overload
point (AOP) is at 122.5 dBSPL and corresponds to 0dBFS (100% of digital Full Scale); the
reference audio at 1kHz and 94dBSPL corresponds to -26dBFS +/-3dB (5% of FS); the A-
weighted SNR is 64dB. A-weighting gives less weight to frequencies below 1kHz and above
10kHz, as the sensitivity of the human ear is reduced at those frequencies; on the opposite,
slightly more weight is given to frequencies around 3kHz where human ear has its peak
sensitivity. Roughly speaking, the equivalent input noise floor will be at 94-64 = 30 dBSPL,
which corresponds to -26-64 = -90 dBFS (0.003% of FS). The total harmonic distortion plus
noise THD+N at 94 dBSPL is 0.2%, at 110 dBSPL is 0.7%, at 120 dBSPL is 6%, best in
class.enabling the highest audio fidelity.
The frequency response is practically flat from 100 Hz to 5kHz, up to +0.5dB at 20kHz. This
is especially important, even if mic sensitivity variations are estimated and compensated in
real time, to maximize the performance of beamforming applications, when the output of at
least two microphones is combined to achieve some directionality and reject noise coming
from specific directions (see AcousticBF library in X-CUBE-MEMSMIC1 software expansion
on st.com).
PDM to PCM conversion
The output of the digital MEMS microphone is a bit stream at a rate equal to the clock given
to the microphone (1.2 to 3.25 MHz, 1.6 MHz is used on the BlueNRG-2 sensor node). The
frequency of ones is proportional to the sound pressure level; this is known as Pulse-
Density-Modulation (PDM). However, audio is usually represented by words with at least
16 bits, at a rate equal or greater than twice the max audio frequency, typically 16, 24, 48
kHz; this is known as Pulse-Coded-Modulation (PCM).
The PDM-to-PCM conversion can be done by low-pass filtering; in its simplest form, this is
equivalent to averaging, or counting the number of ones in a time interval, equivalent to the
sampling time of the PCM word. The conversion can be done in software (see PDM2PCM
library in X-CUBE-MEMSMIC1 software expansion on st.com, AN3998 PDM audio software
decoding) or in hardware (see AN4957 DFSDM filters on STM32 microcontrollers). In the
BlueNRG-2 sensor node the conversion is done in hardware by a dedicated block in the
ADC peripheral.