Setting up a desk noise test: what the work involves

The setup is where a noise test succeeds or fails. Nothing about the arithmetic afterwards can rescue a microphone in the wrong place. This is the physical work, in order.

What to take away

Position matters more than the instrument. A meter in the wrong place is precisely wrong.
Hold the recording position fixed between the before and after readings, to the centimeter.
Record for a set period and take an average, never a peak.
A phone microphone is a real instrument with real limits, and the limits are in the hardware.

Where the microphone goes

At the listener's head height, in the listener's normal position.
Not on the desk, which couples it to the surface you are testing.
Not in your hand, which adds movement and reflections.
Away from walls and corners by at least half a meter, because both raise low-frequency readings.
On something soft if it must rest on a surface.

Mark the position with tape. You will need exactly the same spot in a week.

How long to record

Sixty seconds of continuous typing is the practical minimum, and two or three minutes is better. Short samples are dominated by whatever happened in them. Take at least two runs on different days, because a single run can be ruined by traffic, an appliance or a neighbor you did not notice at the time.

What a phone is actually doing

Understanding the chain helps you respect its limits. Research proposing smartphones as environmental noise sensing units works through the sound capture and processing stages that sit between the microphone and the number on screen, from the built-in microphone through the device's audio path to the app's calculation. Each stage is a place where two phones differ. That is why comparing your reading to somebody else's phone is meaningless, and why comparing your phone to itself in the same spot is not.

How big a difference has to be to matter

Not every change worth making shows up as a big number. The decibel scale is logarithmic and built around perception: as a physics reference on the decibel scale from Georgia State University puts it, the factor of ten in the definition exists partly because roughly 1 decibel is the just noticeable difference in sound intensity for a normal ear, and the rule of thumb for perceived loudness needs a much larger change in intensity. Two consequences follow. A one decibel improvement is not worth spending money on. A three decibel improvement is real but modest, and a listener may not describe it as much quieter.

The setup checklist

Close the room to its normal state and let it settle for a minute.
Mark the microphone position with tape.
Note the time, the weather, and everything running.
Record silence for thirty seconds to capture the background.
Record the typing for at least sixty seconds.
Repeat the whole sequence after the change, with nothing else altered.

Step four is the one people skip and the one that explains most confusing results. If the background moved, the comparison is invalid.

After the setup

Once the routine exists it takes ten minutes and can be repeated for years. The full sequence around it is in what happens at each stage, and the wider question of what to change first is covered in the planning guide.

Common questions

Should the microphone point at the keyboard? Point it the way the listener faces. Most phone microphones are broadly omnidirectional at these levels, so orientation matters less than position.

Does a phone case affect the reading? It can, if it covers the microphone port. Use the same case for both readings, or none.

Do I need an external microphone? Only if the result has to convince someone else. For before and after in one room, the built-in microphone is adequate.

Can I test while working normally? Yes, and it is a good second test. Just keep a scripted run as well, since normal work is not repeatable.