What do mic sensitivity, noise floor and signal-to-noise ratio really mean?

How to read a microphone spec sheet: what dBV/Pa, dBA self-noise and SNR actually measure, why a 4 dBA mic will not hide your fridge, and which numbers matter for a voice in a normal room.


Two browser tabs, one purchase. The first spec sheet says "Equivalent Noise (A-Weighted): 4dBA" and "Sensitivity: -32 dB re 1 Volt/Pascal". The second, for a Shure SM7B, gives a sensitivity of -59 dBV/Pa and no noise figure at all. Read quickly, the first microphone looks quieter and stronger, so it must be the better pick for calls from a spare bedroom with a fridge humming behind the wall.

That reading is wrong in a useful way. On a call from that bedroom, the 4 dBA microphone will not give your colleagues less fridge. Depending on where you put it, it can easily give them more. Neither number lies; each answers a question the buyer did not ask.

Below, microphone sensitivity, noise floor and signal-to-noise ratio are read off real spec sheets, the arithmetic is done for an ordinary room, and the article ends with the few numbers that predict how your voice will come across.

If you are still choosing between microphone types and have no spec sheets open yet, start with What is a condenser microphone, and is it right for your room? instead. This article is for the reader who already has the numbers in front of them and wants to know which ones to believe.

Four real spec sheets side by side

Every figure below comes from the manufacturer's published specification. The SNR column is either the stated value or 94 minus the self-noise, which is how the industry defines it (more on that below).

MicrophoneType, outputSensitivitySelf-noiseSignal-to-noise ratio
RØDE NT1 5th Generationcondenser, XLR and USB-C-32 dB re 1 V/Pa (25 mV)4 dBA90 dB (94 minus 4)
Audio-Technica AT2020condenser, XLR-37 dB re 1 V/Pa (14.1 mV)20 dB SPL74 dB (stated)
Shure SM58dynamic, XLR-54.5 dBV/Pa (1.85 mV)not givennot given
Shure SM7Bdynamic, XLR-59 dBV/Pa (1.12 mV)not givennot given

Three things jump out. The condensers are roughly 18 to 27 dB "hotter" than the dynamics. Only the condensers list a noise figure. And the spread in noise between the two condensers (4 against 20) is huge on paper. The rest of the article explains why the first fact costs you money, the second is normal, and the third matters far less in a normal room than it seems.

Microphone sensitivity: how many millivolts per pascal

Sensitivity tells you how much voltage the microphone produces for a fixed amount of sound. The reference is 1 pascal of sound pressure, which by international convention equals 94 dB SPL. That is loud: the CDC's noise chart puts normal conversation around 60 dB and a whisper around 30. So the spec sheet describes a test tone roughly as loud as someone speaking right into the grille.

The figure comes in two forms that say the same thing:

  • Millivolts per pascal. The NT1 5th Generation puts out 25 mV for that 94 dB tone; the SM7B puts out 1.12 mV.
  • dBV/Pa (or "dB re 1 V/Pa"). The same values expressed as decibels relative to 1 volt. Since every microphone produces far less than a volt, the numbers are negative, and closer to zero means more output: -32 is hotter than -59.

The gap between those two is 27 dB, which is a voltage ratio of about 22 to 1. Treat it as a number about gain, not quality. A low-sensitivity microphone is not worse; it just needs a stronger preamp to reach the same level.

What low sensitivity costs in practice

Shure's own support article on the SM7 puts it in figures. Typical speech at 3 inches (about 7.5 cm) is 84 dB SPL, and at that level the SM7 puts out -69 dBV. Shure's advice is a preamp with at least 60 dB of gain, and it warns that many interfaces offer only 40 to 50 dB. Phantom-powered inline boosters such as the Cloudlifter, Triton FetHead or sE DM1 add 25 to 28 dB, which is why they exist. Shure's own answer came in 2023: the SM7dB, an SM7B with a built-in preamp that adds 18 or 28 dB. The plain SM7B is still on sale as of October 2026, so check which of the two a listing actually offers.

Now hold that against a real interface. Focusrite lists the Scarlett Solo 4th Gen's mic input at a 57 dB gain range. Plug the SM7B into it and you are 3 dB short of Shure's own minimum before you have said a word. The NT1, 27 dB hotter, reaches the same level at around 33 dB of gain and never gets near the top of the knob. If this situation sounds familiar, Do you need a Cloudlifter or FetHead for a dynamic mic? covers the fix, and What is gain staging, and how do you stop your mic from clipping? covers where to set the knob once you have enough range.

The myth: a sensitive mic "hears more of the room"

Search for "mic sensitivity" and two kinds of answer keep coming up: guides to the input slider in Windows or macOS, and warnings that a sensitive microphone picks up more background noise. The second claim mixes up two things.

A more sensitive capsule makes everything louder by the same amount: your voice and the fan behind you both rise by the same number of decibels. The ratio between them does not change. What does change that ratio is the pickup pattern and, above all, how far your mouth is from the capsule. A condenser on a desk often does capture more room, but that is because people tend to place it further from their mouth, not because of the sensitivity figure.

The slider in your system settings is a different thing again. On a Mac it is the input volume in Sound settings, and it adjusts gain after the microphone, not the capsule's sensitivity. How do you change microphone settings on a Mac? shows where it lives and what it does.

USB microphones: dBFS instead of volts

A USB microphone never hands you a voltage. Its preamp and converter are inside, so its sensitivity is stated in dBFS: how close to digital full scale a 94 dB tone lands. The Shure MV7+ is a good example because it has both outputs. Its XLR sensitivity is -55 dBV/Pa (1.78 mV). Its USB-C sensitivity is -33 dBFS/Pa at minimum gain, with up to 36 dB of gain available on top.

Read that as: a loud close-up voice at minimum gain sits about 33 dB below the clipping point, and the gain control gives you up to 36 dB to bring it up. You cannot compare -33 dBFS/Pa with -55 dBV/Pa; they are measured on different scales, and a USB figure that leaves out the gain setting tells you very little.

Noise floor: what the microphone hisses on its own

Self-noise, also called equivalent noise level, is the signal a microphone produces with no sound reaching it. Manufacturers express it as the sound level that would produce the same output, almost always A-weighted (dBA), a filter that roughly follows how human hearing treats quiet sounds. Neumann adds a warning worth keeping in mind: A-weighting is the method that produces the lowest figure, and some makers measure the electronics without the capsule, which flatters the number further.

Neumann's home-studio guide gives a useful scale for condensers:

Self-noise (dBA)Neumann's verdict
below 10extremely low noise
11 to 15still very good
16 to 19good enough for most purposes
20 to 23a pretty high figure for a studio microphone
24 and aboveunworthy of a studio microphone

By that scale the AT2020 (20 dB SPL) sits at the noisy edge of studio use and the NT1 5th Generation (4 dBA) sits in the top band.

Two caveats change how you read it. First, dynamic microphones rarely publish self-noise at all. Their noise comes mostly from the preamp they are plugged into, and Neumann puts a dynamic with an ultra-low-noise preamp at roughly 18 dBA. That is why the SM7B and SM58 rows in the table are blank: the hiss you hear from them is mostly the hiss of your interface's preamp turned up 60 dB. Interface makers publish this as EIN, equivalent input noise; Focusrite lists -127 dBu A-weighted for the Scarlett Solo.

Second, the "noise floor" you see in a recording is not the microphone's self-noise. It is the sum of everything: capsule, preamp, converter, and the room. Neumann is blunt on this: even a very quiet recording room contributes considerably more than 10 dBA of ambient noise. In most homes the room wins by a wide margin.

Static, crackle and hum are a third thing again. Those come from cables, ground loops, USB power and interference, not from self-noise, and no spec sheet figure predicts them. If you hear crackling rather than an even hiss, Why does your mic have static, and how do you get rid of it? is the page you need.

Signal-to-noise ratio: one subtraction

The signal-to-noise ratio on a microphone spec sheet is a fixed calculation, not a separate measurement. Neumann and Schoeps both define it the same way: the 94 dB reference tone minus the A-weighted self-noise.

SNR (dB) = 94 - self-noise (dBA)

RØDE NT1 5th Gen   94 - 4  = 90 dB
Audio-Technica AT2020   94 - 20 = 74 dB   (Audio-Technica states 74 dB)

Two consequences follow. A mic's SNR and its self-noise carry exactly the same information, so never count them as two separate advantages. And the 94 dB reference is far louder than you speak at a normal distance, so the SNR on the box is a best case. Your real ratio depends on how loud your voice is at the capsule and how loud the room is, and both of those are about you, not the microphone.

One trap: some older spec sheets use a microbar reference, which sits 20 dB lower at 74 dB SPL. A figure quoted against it will look 20 dB worse than the same microphone measured against 1 pascal. Check the reference line before comparing.

What is a good signal-to-noise ratio for a microphone?

Here is the arithmetic for an ordinary room, which is what the spec sheet never does for you.

Start from Shure's figure: speech at 3 inches is about 84 dB SPL. Sound drops roughly 6 dB every time the distance doubles, so your voice at the capsule is approximately:

Mouth to capsuleVoice level at the mic
7.5 cm (3 in)about 84 dB SPL
15 cm (6 in)about 78 dB SPL
30 cm (12 in)about 72 dB SPL
60 cm (24 in)about 66 dB SPL

Now take a home office whose background noise measures 35 dBA. That is an example value; measure your own with a sound meter app at the chair where you sit. Room noise does not get quieter when you lean in, because it arrives from every wall at once. Your voice does.

Noise sources add as powers, not as plain numbers, so the microphone's hiss only matters when it gets near the room's level:

Room 35 dBA + AT2020 (20 dB)     -> total 35.1 dBA
Room 35 dBA + NT1 5th Gen (4 dBA) -> total 35.0 dBA
Room 25 dBA + AT2020 (20 dB)     -> total 26.2 dBA

In the 35 dBA room the gap between a 74 dB and a 90 dB microphone is a tenth of a decibel, which nobody can hear. The general rule falls out of the same formula: once the room is 10 dB louder than the microphone's self-noise, the microphone adds less than half a decibel to the total. Only in the very quiet 25 dBA room does the AT2020's hiss start to count. Moving from 60 cm to 15 cm, on the other hand, raises your voice by 12 dB against the same room. That is 12 dB from distance against a tenth of a decibel from the spec sheet.

So the honest answer to "what is a good signal to noise ratio for a microphone":

  • For speech at close range in a normal home room: a stated SNR of about 74 dB or better (self-noise of 20 dBA or lower) means the microphone is not your bottleneck. Spend the next dollar on placement.
  • For quiet sources, a treated room, or a mic placed far away (voice-over in a booth, whispered narration, an acoustic guitar a metre out): self-noise starts to show, and 80 dB or better (14 dBA or lower) is worth paying for.
  • For dynamic microphones: ignore the missing figure and check the interface instead. Gain range and EIN tell you more than the mic sheet does.

There is also a case where a very low self-noise works against you. A sensitive, quiet condenser placed at arm's length in a lively room records the room in loving detail. The hiss you were worried about was never going to be the audible problem; the reflections and the fridge were.

Laptop, headset and webcam mics: no numbers at all

Most of the microphones people actually use on calls publish none of these figures. Apple describes the MacBook Pro's microphones as a "studio-quality three-mic array with high signal-to-noise ratio and directional beamforming" and gives no number. Many headsets and webcams list nothing useful either.

That is less of a loss than it sounds. In those devices the result depends on built-in processing (beamforming, in Apple's case) and on the fixed distance to your mouth: a headset boom a few centimetres away has the distance problem solved by design, while a laptop mic 50 to 70 cm away starts the arithmetic above from a weak position no matter what its capsule measures.

Where software fits, and where it does not

Once your microphone's self-noise sits well below the room, the remaining noise on a call is the room itself: the fan, the fridge, the keyboard, the street. No spec sheet figure fixes that. The options, in order of cost:

  1. Move the microphone closer. Halving the distance is worth about 6 dB, free.
  2. Use the call app's own noise suppression. Many call apps have a built-in filter; Which noise cancelling software for your mic actually works on calls? compares them.
  3. Process the voice before it reaches the app. On a Mac, TunedMic adds a virtual microphone that any app can select. During a one-minute Sound Check it takes more than 30 measurements of your voice, your microphone and your room, then cleans and shapes the sound on the Mac itself, with less than a tenth of a second of delay, and the voice is never uploaded.

In the 38-second video on the TunedMic home page you can watch that step happen, from the minute of reading aloud to a call in Zoom, Google Meet or Discord. No particular mic is on test there, so judge the result in your own room during the free trial.

The limits are worth stating plainly. TunedMic needs a Mac with Apple silicon (M1 or newer) and macOS 14 or later; there is no Windows or Intel version. It does not make a weak mic sound like a specific expensive one, and it does not change your voice into someone else's. It also cannot add detail that a Bluetooth link never sent, so a wired or USB microphone gives it more to work with.

A checklist with numbers, before you buy

1. Condenser: self-noise 20 dBA or lower (SNR 74 dB or higher) is enough for close speech.
2. Dynamic: check the interface, not the mic. SM7B-class (about -59 dBV/Pa) wants 60 dB of gain.
3. Interface gain range under 60 dB with a low-output dynamic: budget for an inline booster (+25 to 28 dB).
4. USB mic: sensitivity in dBFS only means something next to the gain setting it was measured at.
5. Measure your room in dBA at your chair. If it is 10 dB or more above the mic's self-noise, the mic is not your problem.
6. Plan the distance: every halving gains about 6 dB of voice over room.

If your room turns out to be the loud part of the equation, the next step costs nothing: try TunedMic on your own microphone with the free 7-day trial. It starts at first launch and needs no card or account. The license is $19 once on the pricing page, and since there are no refunds, the trial is where to decide.

Keep reading