The Listening Test

The Recording Session

At the recording session, we placed all eight microphones into a small pickup zone that was on a plane approximately 5” in diameter and 12-15” from the singers. The mics were recorded simultaneously to separate tracks of a multitrack recorder running at 30 IPS. In addition, a separate music bed track was recorded for some of the selections. Each singer sang one a cappella selection and one accompanied selection, with the music bed remaining unmixed. We also recorded excerpts of pink noise and two different commercial recordings, played back through a Genelec 1031 speaker (that we later used for the listening session) 12” from the microphones.

After the singers finished, we transferred all the recordings to hard disk (using Pro Tools), I calibrated the tracks for equal levels (using the NLA software in the TEF 20 system), and set up playback so we could easily listen to any four microphones with or without a mixed music bed.

The Listening Session

The listening session took place a few days after the recording session. All of the singers were present, as was the session engineer Mark Torre, three Berklee faculty (Peter Alhadeff, Carl Beatty and Wayne Wadhams) and Alex Case. We listened over the single Genelec 1031 set up on the median plane through 10 ABCD trials, comparing all microphones . Each singer performed live as well, so that the panel could use the original source as a reference for comparison with the recordings. We did the same with the commercial recordings as well, playing them back through the Genelec in comparison to their recorded-through-the-mics versions. The listeners were asked to rate each microphone on a scale from 0 to 10, where 0 represented unrecognizable reproduction and 10 represented reproduction indistinguishable from reality. We also asked each listener to make comments on each microphone. Microphones were identified by randomly assigned letters. No listener knew the identities of any of the microphones.