MEMS Mic Datasheet: SNR, AOP, PSRR Explained

Concept render

SNR tells you the noise floor, AOP tells you the loudest sound before heavy distortion, sensitivity sets where speech lands in digital code, and PSRR tells you how much supply ripple leaks into the audio. For a pendant or clip recorder, compute equivalent input noise from SNR first, then check the others against your own supply and housing.

A MEMS microphone datasheet lists many numbers. For speech on a worn device, four of them do most of the work: sensitivity, SNR, acoustic overload point (AOP) and power supply rejection (PSR, often called PSRR). This guide reads those four from two public Infineon digital (PDM) datasheets, then shows the arithmetic that turns them into a shortlist check. It contains no Duobay price and no measured result. All figures below are bare-part datasheet values, read on October 31, 2026. We did not read other vendors' datasheets for this article.

This guide assumes a recorder with a visible indicator while the mic is on and a consent step before capture. These specs matter for capture that the wearer can see and has agreed to.

What do the numbers on the datasheet say?

Each row below is quoted from the vendor tables. "Not in our ledger" means we did not verify that row, so we leave it out instead of guessing.

What do SNR, AOP, sensitivity and PSR mean in practice?

• SNR is the gap between a 94 dB SPL reference tone and the mic's own noise. A bigger number means a quieter mic. It depends on clock mode, weighting and bandwidth, so read those conditions with it.
• AOP is the sound level at which distortion reaches a stated value. For the IM69D120 that value is 10% THD. It is a ceiling.
• Sensitivity in a digital mic is a level in dBFS for a 94 dB SPL tone at 1 kHz. It decides where speech lands in the digital code, and where digital full scale sits in SPL.
• PSR is how much ripple on the supply shows up in the output, expressed in dBFS for a stated ripple. A lower (more negative) number is better.

How do you compute a usable comparison?

You need only the datasheet rows and one assumption you supply. The formulas:

EIN (dB SPL)        = 94 − SNR (dB)
full scale (dB SPL) = 94 − sensitivity (dBFS)
level (dBFS)        = sensitivity (dBFS) + (SPL − 94)
PSR margin (dB)     = level (dBFS) − PSR (dBFS)         for a stated ripple
dynamic range (dB)  = AOP (dB SPL) − EIN (dB SPL)

Units: SNR and margins are in dB, sensitivity, PSR and level are in dBFS, EIN, AOP and SPL are in dB SPL. The worked results are in the table below. The 60 dB SPL speech level and the 100 mVpp ripple in the last two rows are assumptions for you to replace.

The ripple margin only compares two dBFS numbers. Whether 20 dB of margin is enough depends on your product, which is why the board test below matters.

Shortlist check: speech margin over EIN at distance

Sound pressure from a point source falls by about 6 dB for each doubling of distance. That is an inverse-distance rule for free field, and a body-worn device sits in a messier field, so treat the table as a first screen. Assume 60 dB SPL at 0.3 m (your assumption; replace it). Margin is speech level minus EIN. The last column uses the IM69D120 at the 768 kHz clock: EIN = 94 − 64 = 30 dB(A) SPL.

The margin shrinks with distance and with the quieter clock mode. Whether any of these margins gives speech you can understand is a question this guide cannot answer. The datasheet figures are for a bare part under the vendor's test conditions, which this guide does not restate. Only a recording made on your own device answers it.

What are the common datasheet traps?

These follow from the numbers above. They are engineering reasoning about how the figures behave. None describes an incident at any company.

• Reading AOP as clean headroom. The IM69D120 AOP of 120 dB SPL is defined at THD = 10%. The same table lists 0.5% THD at 94 dB SPL and 1.0% at 118 dB SPL. Plan your loud-event headroom from the distortion level you can accept.
• Comparing SNR numbers taken under different conditions. The IM69D120 lists 69 dB(A) at a 3.072 MHz clock and 64 dB at 768 kHz, and the second row states a 20 Hz to 8 kHz bandwidth. The IM66D130M row is measured over 20 Hz to 20 kHz. Check clock, weighting and bandwidth on each row before you rank two parts.
• Choosing the lowest-sensitivity part without checking code resolution. A sensitivity of −37 dBFS puts digital full scale at 131 dB SPL, so the part is unlikely to clip digitally first. Speech then lands at a lower code level (−71 dBFS against −60 dBFS in our 60 dB SPL example), so each dB of speech uses less of the code range. This follows from the arithmetic above. Whether it matters depends on your audio chain, which you must test.
• Treating two PSR numbers as the same test. The IM69D120 lists one PSR for a 100 mVpp sine sweep and another for a 217 Hz square wave. They test different disturbances, and one is A-weighted. Match the test to the ripple on your own rail.
• Trusting bare-part numbers inside a housing. These figures come from the vendor tables for a bare part. Our ledger does not state the test conditions beyond the table rows, and we have no data for a housing, port, mesh or clothing. Anything that changes the acoustic path changes the result, so measure the assembled device.
• Mixing up the part number in the document. Confirm the ordering code on the cover page and with the vendor before you commit.

What should you measure on your own board?

The datasheets do not tell you how the part behaves in your housing, on your supply, on a body. We have not measured these for you, and no result appears here. Set the pass line yourself for each test, and record the basis for it.

The pass line is set by you per product requirement, and you record the basis for it. Fix it before you run the test.

What should you think through before you choose?

• What distance from mouth to mic does the product need to cover, and who wrote that distance down?
• Which clock mode will we ship, and have we used the SNR figure for that mode instead of the best case?
• What loud events must the recording survive, and what distortion level is acceptable?
• What disturbances does our supply rail see during radio bursts, charging or display activity, and which PSR test is closest to it?
• Does the vendor publish the number we need at our conditions? If not, which test on our board will produce it?
• Is the recorder’s status light visible while the mic is on, and does the consent flow work before audio is captured?
• Who signs off the pass line for each test, and where is the basis for it written?
• Which assumptions in our calculation are measured on our own board, and which are still guesses?

If you are choosing a microphone for a pendant or clip recorder, tell us about your device. You can also read how our engineering gates handle part selection and bench tests.

Sources

All pages accessed October 31, 2026. Datasheets are revised by the vendor, so re-read the current version before you decide.

• Infineon, IM69D120 datasheet: https://www.infineon.com/assets/row/public/documents/24/49/infineon-im69d120-datasheet-en.pdf (accessed 2026-10-31)
• Infineon, IM66D130M datasheet: https://www.infineon.com/assets/row/public/documents/24/49/infineon-im66d130m-datasheet-en.pdf (accessed 2026-10-31)