How to Compare Two Microphone Test Recordings

Why A/B-ing microphones by ear misleads you

Two mics, same script, same room. You listen back and one obviously sounds better.

The problem is that almost everything else is confounded. Whichever recording is a couple of decibels louder will sound fuller and more present, regardless of the microphone — this is well documented and it fools experienced engineers. Add a slightly different distance from the source, a different preamp gain, or a take where you happened to speak more clearly, and your comparison is measuring four things at once.

Looking at the frequency content directly removes the guesswork, and it answers a question your ears cannot: where the two mics differ, not just which you preferred today.

Set the test up so the result means something

Preparation matters more here than the analysis.

Match the levels first. Normalise both takes to the same level before comparing. Skip this and you are largely measuring gain. It is the single most important step.

Record the same source material. The same sentence, ideally back to back so your voice and the room have not changed. Better still, record both mics simultaneously if you have the inputs — that removes performance variation entirely.

Keep the distance identical. Proximity effect means moving a few centimetres closer changes the low end more than swapping microphones does.

Include some silence. Five seconds of room tone at the start of each take. This is what lets you compare noise floors, and it is the part people forget.

What to compare

Upload both takes to DiffALL’s audio comparison. You get a similarity score, a per-second chart, and mel spectrograms of both files with a difference view.

For this job the spectrograms are the main output and the score is secondary — you are not asking “are these the same”, you know they are not. You are asking how they differ.

A bright band at the bottom of the difference map — the mics differ in low-end response. Common between a large-diaphragm condenser and a dynamic, and the usual cause of one sounding “warmer”.

A bright band in the upper range — a presence or air difference. This is what makes one mic sound “crisper” and, pushed too far, “harsher” or “sibilant”.

Difference concentrated in the mid range — where voices live. Often the most consequential difference of the three and the hardest to hear as a difference rather than as a preference.

An evenly raised floor across the whole map during the silent section — one mic has a higher self-noise, or the preamp needed more gain. On quiet sources this matters more than the tonal difference.

The silence test

Compare only the room-tone sections and you get a direct read on noise. A mic that needs a lot of gain — most passive dynamics and ribbons — brings up preamp hiss along with the signal. In a quiet room with a loud source that is irrelevant. For quiet dialogue, whispered narration, or a noisy room, it is the deciding factor.

This is the comparison that most often reverses a decision made by listening to the loud parts.

Comparing a mic against a reference

If you already have a mic you trust, use it as your control. Record every candidate against it and compare each one to the same reference. You end up with directly comparable results rather than a chain of pairwise impressions that do not add up.

It also tells you something useful in its own right: how far a cheaper option actually sits from your known-good one, in specific frequency ranges rather than in general.

Comparing positions, not just microphones

The same method answers a question that matters more than the mic choice for most home setups: where to put it. Record the same line at three distances, or on- and off-axis, and compare. Position routinely makes a larger difference than the microphone does, and it is free.

What the score is good for

The overall similarity number is not the point here, but it does have one use: tracking consistency over time. Record the same test line at the start of every session and compare against your first one. A drift means something changed — mic position, room treatment, a cable going bad, or a gain setting nudged. Catching that on day two beats discovering it in the edit.

Try it

Compare two audio files → — normalise both takes to the same level first, then upload. The spectrogram difference view is where the answer is.