Two microphones, one hollow sound

Something is wrong with the sound and you cannot describe it. It is not noisy. It is not quiet. It is not echoey, exactly. It just sounds thin, a little hollow, slightly cheap, as though the voice has been played through something rather than recorded.

Hollow sound has other causes too, the room, a flipped polarity, a microphone in the wrong spot, but if yours arrived the day you added a second microphone, this is the one to check first, and it has nothing to do with your voice or the money you spent. The mute test below settles it in under a minute.

Two microphones sitting at different distances from the same source hear the same sound at slightly different moments. Where the two versions disagree, they cancel each other. What survives is thinner than either recording would have been on its own.

Timing between microphones

Picture a recording mixed with a slightly delayed copy of itself. Some frequencies reinforce while others cancel, changing the tone.

Sound travels at a fixed speed. A microphone half a metre further from your mouth than another one receives everything a fraction later, every time, for the whole recording. The moment those two recordings are combined into a single track, which is what happens as soon as you play them together, some parts of the sound line up and add, and some arrive out of step and subtract.

Nothing is broken. This is arithmetic, and it happens whether or not anybody in the room notices.

Why does it sound thin rather than doubled?

Because the cancellation is uneven. Different parts of a voice are affected differently by the same delay, so some survive and some flatten, and what you hear is whatever was left over.

The effect has a name worth knowing, comb filtering, because an even delay strips out a regularly spaced row of frequencies and leaves the ones between them standing. It is the term almost everything else useful about this problem is filed under.

Which parts go first depends on the spacing, and at the distances this usually happens they tend to be the ones carrying weight and body. What remains is the part of a voice that sounds like a phone call: intelligible, present, and somehow not a person in a room. That is why this reads as cheap rather than as faulty. Faulty would be easier, because you would have known to fix it.

How can you hear it for yourself?

The test takes under a minute and needs no software you do not already have.

Play the microphones together, then listen to each alone. If the combined sound is noticeably thinner, investigate timing, polarity and the balance between the tracks. The mute test locates the interaction; it does not identify every possible cause.

Do this before you touch a single volume control. A recording with this fault in it cannot be balanced sensibly, because every adjustment is being applied to a sound that is already fighting itself, and you will spend an hour chasing a problem that moves each time you get near it.

Where does the second microphone go?

The three-to-one rule is a spacing guideline for reducing unwanted pickup when microphones cover separate sources. It is not a universal prescription for two microphones deliberately recording one source. For those, test their position, timing and combined sound.

If the second microphone is intended to capture the room, place it for the room sound you want and check it in the mix. If it is a backup, you may be better choosing one track than summing both.

Move deliberately and listen again. More distance changes both the delay and the amount of room sound; it is not automatically safer.

What if it is already recorded?

If the microphones were recorded separately, you may be able to improve the combination.

For a redundant recording of one source, adjusting timing may improve the combined sound. Check by ear as well as by eye. Multiple sources, reflections and intentional stereo spacing can make a single alignment inappropriate.

Alignment is the first thing this studio does to a redundant capture like this one, two microphones on one source meant to be summed, ahead of noise removal and ahead of any balancing at all. The reason is not tidiness. Every later decision is taken by listening, and until the two recordings agree with each other, the thing being listened to is not the recording. A deliberately spaced or stereo pair is the exception, because there the small timing difference is the stereo image itself, and aligning it away would flatten the recording rather than mend it.

Changes for the next session

Decide what the second microphone is for: another source, the room or a backup. Position and test it for that job, then listen to the combination before the session.

Then record a short test and run the mute check before the real take rather than after it. Thirty seconds of talking, listened to properly, will tell you more than any specification you can read about either microphone.

The reason this fault survives in so many recordings is that it never sounds like a fault. It sounds like the ceiling of what a home setup can do, which is a comfortable thing to believe and, in this one case, entirely wrong. If your audio has sounded thin since the day you added the second microphone, that date is the evidence. The mute test is the rest of it.

Sources and evidence

What here is established acoustics and what is our own method.

What is established. A thin, hollow sound from combining two microphones is comb filtering, the interference between a signal and a slightly delayed copy of it, documented by DPA Microphones and analysed for multiple sources in the peer-reviewed work of Clifford and Reiss (2011). The frequencies it hollows out are set by the delay between the microphones, which is why the effect changes with their spacing rather than being one fixed signature.

Our method. Keeping to the spacing that avoids it, and checking the summed signal rather than each microphone on its own, is the practice here.

References.

• DPA Microphones. The basics about comb filtering and how to avoid it. dpamicrophones.com. Multiple delayed versions of a signal can create comb filtering. Manufacturer education; diagnosis still requires listening and measurement.
• Clifford and Reiss. Calculating time delays of multiple active sources in live sound. eecs.qmul.ac.uk. Time offsets and multi-source interference. Specific technical context.