Digital Microphone (DMIC)

Overview

The audio DMIC interface provides access to digital microphones.

Digital microphones typically output a PDM (Pulse Density Modulation) bit stream instead of analog audio. PDM uses a 1-bit, high-rate signal in which audio amplitude is represented by the density of 1s and 0s over time.

Because applications usually consume PCM (Pulse Code Modulation) samples rather than raw PDM data, a DMIC controller converts the microphone stream into PCM audio. This conversion includes filtering and decimation: filtering removes the high-frequency noise shaped into the PDM stream, and decimation reduces the bit-stream rate to the desired PCM sample rate.

From a Zephyr point of view, a DMIC device is an audio capture peripheral. The application configures the microphone and controller, starts capture, and reads PCM buffers from the driver.

Key concepts

The DMIC API separates configuration into three pieces that mirror the capture path:

  1. How the microphone is driven on the PDM side,

  2. How incoming PDM channels are arranged, and

  3. How converted PCM samples are delivered to the application.

These pieces are combined in dmic_cfg and passed to dmic_configure().

PDM I/O configuration (dmic_cfg.io)

This describes the electrical and timing requirements of the microphone interface, such as the supported PDM clock frequency range, duty cycle, and any controller-specific signal polarity settings.

Channel configuration (dmic_cfg.channel)

This tells the driver which physical PDM controller and left or right microphone lane should appear as each logical audio channel in the PCM output. It also declares how many channels and streams the application wants to use.

PCM stream configuration (dmic_cfg.streams)

This defines the PCM data delivered by the driver, including sample rate, sample width, block size, and the k_mem_slab used to allocate receive buffers for each enabled stream.

Typical application flow

Typical use of the DMIC API is:

  1. Get the DMIC device, usually from devicetree.

  2. Fill a dmic_cfg structure with I/O, channel, and stream settings. See the Buffering section below for more details on how to define the memory buffer the DMIC uses to store received PCM data.

  3. Call dmic_configure(), passing the configuration structure.

  4. Start capture with dmic_trigger() using DMIC_TRIGGER_START.

  5. Fetch PCM data with dmic_read().

  6. Stop, pause, or reset capture with additional trigger commands as needed.

Buffering

Received PCM data is returned through buffers owned by the driver. Applications provide the backing memory through a k_mem_slab referenced by each configured PCM stream.

A common pattern is to declare this slab statically:

Declaring a memory slab statically for PCM receive buffers
 K_MEM_SLAB_DEFINE_STATIC(mem_slab,
                          SAMPLES_PER_BUFFER * sizeof(int16_t),
                          BUFFER_COUNT,
                          sizeof(void *));

In this example, each slab block stores one PCM buffer and BUFFER_COUNT sets how many buffers can be queued internally before the application reads them using dmic_read().

Shell commands

When CONFIG_AUDIO_DMIC_SHELL is enabled, a set of dmic commands is available. They allow capturing audio from a DMIC device interactively without having to write a dedicated application.

Each subcommand takes the DMIC device name as its first argument, optionally followed by the audio capture parameters (sample rate, number of channels, and PCM sample width).

The following subcommands are available:

dmic read <device> [<count> [<rate_hz> [<channels> [<pcm_width>]]]]

Capture count audio blocks and print the peak level per channel. The default count is 5; each block is 50 ms of audio.

dmic vu <device> [<rate_hz> [<channels> [<pcm_width>]]]

Display a live, color-coded level meter with peak-hold and clip indication, one bar per channel. Runs until any key is pressed.

dmic dump <device> [<seconds> [<rate_hz> [<channels> [<pcm_width>]]]]

Capture seconds of audio and print it as base64-encoded raw PCM, together with the host commands needed to decode and play it back. The default duration is 2 seconds.

The built-in help (e.g. dmic read --help) lists the arguments and their defaults.

Configuration Options

Related configuration options:

API Reference

Digital Microphone Interface

Related code samples