PORTLAND, Ore.—Analog Devices Inc. (ADI) aims to bring high-fidelity audio to industrial, business and consumer applications with the world's first micro-electro-mechanical system chip to package a MEMS microphone with an industry standard 24-bit data integrated interchip sound—I2S—interface.
"For industrial applications and for non-handset consumer devices, the key feature on the our new iMEMS mic is the I2S interface," said Kieran Harney, ADI's MEMS product line manager "For new designs, engineers can literally plug it straight into their signal processor, DSP or micro controller."
Handset manufactures have standardized on a pulse density modulated (PDM) encoding which is built into their main ASIC on their cell phones, allowing two microphones to be time multiplexed on a single data line using a single clock. However, non-handset consumer device designs are abandoning the PDM encoding in favor of I2C, which provides similar functionality, but uses pulse code modulation which is supported by standard microcontrollers and DSPs. Applications adopting I2S include surveillance cameras, IP Phones, video conferencing, Bluetooth headsets, game controllers, video cameras and other non-handset consumer applications.
The ADMP441 iMEMS microphone has an extended frequency response from 100 Hz to 15 kHz, a 61 dBA signal-to-noise ratio and 80 dBFS of power-supply-rejection along with ADI's patented dust filter, entire-package back-volume and high-quality audio signal processing. The one millimeter thick package measures 4.72 -by-3.76 millimeter in a reflow solder compatible surface mount package.
Analog Devices' 441 iMEMS digital microphone has a tiny silicon diaphram contained on a separate die measuring just one square millimeter.
Micro-electro-mechanical system (MEMS) pioneers Analog Devices just introduced their first solid-state microphone two years ago, but already it is ploughing new furrows with the world's first MEMS mic to use a state-of-the-art digital interface, rather than the legacy interface inherited from non-MEMS mics. Look for MEMS microphones to displace electrets in all but the most cost sensitive applications over the next few years.
I'm curious why non-handset product designers prefer I2S over a PDM (sigma-delta) interface. The serial PDM bit stream coming from the microphone's ADC needs to be decimated somewhere in the chain, but I don't see the advantage of putting the decimation filters in the MEMS device rather than in the DSP or ASIC -- where digital logic gates or MIPS are very cheap.
I agree with Frank. PDM provides a more efficient implementation of the microphone making smaller, lower power, and less expensive. Also in some systems you can process PDM with lower latency without ever doing decimation.
The reason is that I2S is a ubiquitous interface for digital audio, permitting the use of a wide range of existing audio ICs without needing to write DSP code for the decimation step or adding an ASIC where one would not otherwise be required. In some cases, it is possible to make do without a DSP entirely by routing I2S directly between ICs thus reducing cost and time to market significantly.
When using I2S for microphone output, it is true that you have more opportunity to interface with other audio devices. But you also burden the mic device with a more complex I2S format structure (is the output Left or Right justified, is it MSB first or LSB first, is it a slave or master device,etc.?) Unless all modes of I2S are supported, you will run into devices you cannot interface to even though the interface is supposedly a flavor of I2S.
To Frank's curiosity, there are very few (if any) off-the-shelf DSPs/uPs that have dedicated PDM audio input hardware. Sure, one can write FPGA code, burn MIPs, or throw a bunch of gates into an ASIC to handle the decimation filtering but now you are treading into make or buy scenarios with all the associated tradeoffs (as janekm points out) that include power consumption, number of parts, price, size of implementation, etc.
ADI has made it very easy to add a digital microphone to any audio processing system that supports I2S input (and almost all digital audio circuits do.) In applications that need audio in, the I2S mic dramatically reduces the number of parts to implement this function. The mic has a glue-less interface to uP/DSPs that support I2S. These uP/DSPs support the various flavors of I2S (as kdboyce points out, several exist.) The ADMP441 doesn't support every I2S flavor, it supports slave I2S format (a popular flavor) so the mic is not burdened by having it handle all of the I2S flavors. Sure, there might be some devices out there that can't take this format in, but my money says these situations are very few and far between.
The I2S mic also provides the freedom to put the audio input and output wherever it is needed without the worry of analog signal degradation and potentially eliminating a separate audio codec. Digital I2S Class D amps are available allowing one to place the amplifier near the speaker itself, reducing EMI, and preserving signal integrity up to the speaker since all audio is delivered digitally. A digital I2S mic, complements this architectural approach, reducing parts, preserving audio quality, and utilizing a common I/O interface.
The I2S microphone addition completes a value portfolio of ADI MEMs microphones. If you want analog, ADI has an analog out MEMs mic. If you want PDM, ADI has that. If you want I2S, ADI has that...the customer gets choices and that is usually what customers want. Analog, PDM, I2S...ADI has a mic for that.
Thanks for the reply, and it's good to know that ADI also has MEMS mics with PDM outputs. Although I2S is an old, ubiquitous standard audio interface, I would not be surprised if we start seeing more PDM support on audio CODECs.
Blog Doing Math in FPGAs Tom Burke 3 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...