One by One: the Cantus Recording Project Capturing Cantus part 2
I made test recordings at 44.1kHz. The sound from both pairs of mikes had superb clarity, but it was perhaps a bit too immediate, particularly from the omnis. Luck favors the unprepared: Erick found a large oriental rug in another anteroom, and we put that on the floor between the singers and the omnis, which tamed the treble balance.
The soundstage from the cardioids was evenly balanced and well-defined. However, the soundstage from the omnis favored the tenors a little, even though the measured distances of the two mikes from the center of the choir were identical. I therefore lowered the left-hand mike by 2" and moved the right-hand mike in toward the center by another 2", the combination of which evened out the amplitude balance. I could adjust for the difference in arrival times in post-production.
And so we continued through four days of recording. Balance problems were solved by the singers moving closer or farther away from the mikes. For their solo on "Mpwaga lindu, lindu nye-nye," the fifth of the Songs from Matengo Folktales, for example, the two basses moved close to the left-hand B&K. The opposite was the case for the marimba featured on Matengo Folktales and "Let Your Voice Be Heard," the instrument ending up at the back of the stage so that it didn't overpower the voices.
Once thing I could not fix was the presence of low-frequency noise. The custodian had turned off all the heating and air-conditioning pumps in the concert hall, but it turned out that the sound was coming from a building across the road, to which we could not get access. We had to live with it. Fig.1 shows a high-resolution spectral analysis of the background noise as picked up by the omnis. The highest-level components are actually at very low frequencies, 25Hz and 17Hz, which will mitigate their audibility, but even though the 120Hz component lies at -74dB, which might imply it should be low enough not to be heard, the absence of any low-frequency masking in the music made the noise disturbingly obvious. This was something that would have to wait for post-production, unfortunately.
Fig.1 Spectral analysis of background noise on raw omni pickup, 20-bit A/D conversion (6dB/vertical div.).
The Surround Option: Soon after Cantus' Princeton recital, I found myself back in New Jersey, visiting AT&T's Jim Johnston at the company's Shannon Labs in Florham Park. "JJ," as he is known to denizens of the Internet newsgroups, was treating me to some of the surround recordings he had made with his seven-mike, single-point array. (See "As We See It," February 2001)
Impressed by the solidity of the reproduced soundfield and the absence of the sharply defined sweet spot typical of so much surround reproduction, I invited him to bring his rig to Northfield the following March, to capture Cantus in surround. This JJ did, setting his mike array (see photo) slightly to one side of and slightly lower than my Neumann mikes, and storing the seven channels of 16-bit data—up, down, front left, front center, front right, rear left, rear right—on an 8-channel digital ADAT recorder.
Visitors to JJ's demo room at Home Entertainment 2001 last May will have heard the superb result of the 7-to-5 mixdown. You were literally placed in the Carleton College concert hall, around row 20. I swear that you could even hear the presence of the reflecting cloud above the stage! Perhaps we can negotiate with AT&T to put at least one song recorded with JJ's setup on a future Stereophile surround DVD-Audio or CD. Watch this space.
We'll Fix It in the Mix: Back home, the editing proved straightforward. Cantus had been sufficiently well-prepared that several songs could be recorded in complete takes. The other, longer ones were recorded in short sections, in order to make sure that the pitch remained true from beginning to end. Assembling the master was just a matter of choosing the best takes of each section.
But then there was the sound. When I had aligned all four tracks on the Sonic Solutions digital audio workstation to be exactly coincident in time for a soundsource in the precise center of the stage, and mixed them all the same level, the sound was vividly real. But it also made the sound of the bright, clear, well-focused hall sound too real. Yes, you were in that hall, listening to those singers, with superb fidelity—but it was glaringly obvious that you were the only audience member present!
I had realized at the sessions that some equalization was going to be necessary, in order to reproduce the singers' correct tonal quality. The fairly close omnis were picking up too much high-frequency information, while the distant cardioids were bass-shy. And I had to do something about the acoustic hum from the next building. But this didn't solve the in-your-face nature of the overall presentation.
No way did I want to use artifical reverberation. The solution, as it had been with my Beethoven piano sonata project, was to experiment with the time alignment of the microphone pairs. I ended up moving the cardioids back the equivalent of around 3' from the theoretically correct position. I also backed off the cardioids by 3dB compared with the omnis. The changes gave a more comfortable feeling for the listener, resulting in both a better-defined picture of the singers in an arc on the stage and a clearer picture of the hall's reverberant dieaway. I sent off a CD-R with this unequalized mix and the first round of edits to Erick Lichte in Minnesota and awaited his reaction.
Fortunately, it was positive! Erick and tenor Michael Hanawalt arranged to fly out to Brooklyn so we could spend an August weekend doing the second round of edits and finalizing the mix.
First on the agenda was to tidy up some of the edits, replacing some sections with others that proved to work better, and deciding on how much silence to leave between each cut. (I had recorded a couple of minutes of room sound at the sessions to use between songs, so we wouldn't have to fade to black.) Erick and Michael also wanted less of the cardioids in the mix. However, dropping the Neumanns too far resulted in the singers at far left and far right starting to fold over, back toward the center. We settled on keeping the cardioids at -6dB with respect to the level of the omnis.
More problematic was the dynamic range. In "Rainbow 'Round My Shoulder," the final chorus features one of the singers banging a railroad spike in time with the music. These transients alone used up 6dB of the recording's dynamic range. Coupled to that, the songs featuring the marimba used up another 6dB of dynamic range, meaning that the level of the other songs had to be dropped by 12dB. As a result, the average level of the CD was going to be very low, meaning that it would play too quietly even on high-resolution systems, let alone boomboxes.