Chrysalid Requiem by Toby Twining
MIDI Sequences and Recording work by Mark Johnson

Summary of My Roles

• Created MIDI sequences for guide tracks by transcribing the manuscript scores literally, and converting the microtonal accidentals to pitch bends for each note. Got Toby’s suggestions for tempos and rubatos and created tempo tracks as I would conduct the piece live
• Recorded practice materials with each individual singer’s parts highlighted
• Created full scores for each movement by importing the MIDI sequences into Finale^®. These were printed without accidentals, which Toby filled in by hand.
• With Toby, co-directed rehearsals with the singers, helping them to understand the complex rhythms and unfamiliar intervals in the piece
• For live performances, I ran the computer, adjusted headphone levels, started the sequences, gave cues, and conducted the ensemble, while singing the Bass 1 part
• For the recording sessions, I set up & coordinated the synth guides in the studio, cued and conducted the singers as necessary, and acted as primary session producer (when I wasn’t recording bass and tenor parts)
• In post-production, I edited and pre-mixed all tracks on my Macintosh G4 Cube using Digital Performer 2.7, and sent the resulting Performer documents to Mark Rubel in Champaign for final mixing with Toby

Early Sequencing Work

The first recording of a portion of Chrysalid Requiem was a demo of movement 3, the “Gradual & Tract,” recorded in Baltimore in October, 1998 by Toby, Martha Sullivan, and myself. For this project Toby had made synthesizer guide tracks by multitracking his Yamaha DX7, which he scale-tuned to accommodate the pitches needed for each section of the music. He played each of the 6 parts individually in real time against a click track and some spoken counting. He sent us practice CDs with our parts highlighted in the mix, and the same tracks also served as guides for the recording session. There were some slight rhythmic inaccuracies, and an occasional mistuned pitch, but the tracks were as good as we could hope for on a tight budget, and certainly more precise than Martha’s & my abilities to sing the music at that point.

After we made that recording I wondered how difficult it would be to create MIDI guide tracks for the Requiem myself. I had fairly extensive experience with MIDI transcription beginning in 1994, when I provided the music of Gustav Holst’s “The Planets” for a light show at the Whitney Museum’s “Performance on 42nd” series. I got a Roland JV-880 (which I still use), and used EZVision 1.0 to sequence reductions of the music of all 7 movements, complete. I then performed as a one-man band, playing the sequences on the synth and adding some texture at a grand piano.

In the meantime I had experimented with making a MIDI rendition of a simpler piece of Toby’s, “Hotel Destiné.” This piece progresses slowly through a series of modulations that make it end up in a key about 65 cents lower than where it started. My method for tuning this sequence was to send scale-tuning Sys-Ex messages to the synth every so often, as new tones were required. (Keeping track of the current set of pitches needed at every point in the piece proved to be too much of a headache, and I never finished the project.)

Sequencing the Requiem

I knew that execution of the complex polyrhythms and very unfamiliar microtonal intervals in Chrysalid Requiem would benefit from precise computerized guide tracks. I began by designing a new synth patch for vocal tuning that consisted of a harp attack and a filtered sawtooth wave to sustain. I made it play brighter at higher key velocities, and sustain just enough so singers could continue to tune to it for long notes. It also makes any inaccuracy in just intonation quite clear to the ear, because the rich harmonic spectrum of the sawtooth wave makes strong beats.

I needed a different method for tuning from what I’d used before, because some of the harmonic modulations in the Requiem cover a great distance very quickly, and often there are just too many different pitches at once to accommodate with scale-tuning. I found that I could use MIDI pitch bends for the microtonal accidentals instead, and then I just had to decide what pitch bend range the new synth patch should have, because many of the accidentals Toby had written bent the pitch much more than the standard 1 semitone up & down. I decided that a major third up & down (4 semitones) would be enough to accommodate most of the music. (At the extreme, there are some notes in the “Dies Iræ” that are flattened about an octave and a half by accidentals. For this I needed to change the pitch bend range temporarily via Sys-Ex.)

The calculations involved in converting accidentals to cents to pitch bend values of ±8192 were repetitive, so I set up various spreadsheets to do these in bulk. Before starting to sequence a new movement, I took inventory of the accidentals used, to be sure I wouldn’t have to redo the pitch bend work with a different range. And I never found a way to automate entering these calculated values into the sequencer, beyond copying & pasting pitch bend events. I had to type most of them in by hand, and proof the results aurally. Vision’s list view was absolutely essential for this work.

Performances

All complete performances of the Requiem so far have been done with the singers wearing headphones to hear the synth guide tracks. For the first performance (at the Concertgebouw Kleine Zaal, Amsterdam, 10 November 1999), we fed each singer a separate headphone mix playing only his or her part. This made the tuning of large chords very good, but the ensemble was poor, because there were no aural cues. The only guides the singers had were the sound of the others singing, and my conducting. It was also a nightmare for setup, because the JV-880 has only 4 discrete outputs. So we used three identical synths, and I had to add MIDI pan information all over the sequences to make sure everyone was being fed the right parts.

After that Toby & I agreed that we should try constructing accompaniments (reductions) for each movement, with pitches and clicks for cues, so that everyone could listen to the same headphone mix. This was much more effective for ensemble, but slightly weakened the strict just intonation of the chords, because there was much more information in each singer’s ear. We dream of a day when singers can become familiar enough with these intervals to sing them accurately without guides, but I don’t expect that to happen in our lifetime.

The Recording

Mark Rubel, who has worked with Toby since the mid-1980s, agreed to come to New York for our sessions at Dumbo Studios in Manhattan. We recorded to tape on DA-88 and DA-38, and later transferred the audio to hard disk for editing and mixing in Digital Performer 2.7. One convenient aspect of the tape was that all 12 movements for up-to-8 tracks fit on one tape, and could be backed up easily every night.

We started out on the first day recording several singers at once on one simple movement, but this proved to be too time consuming for production, because proofing each recorded track for accuracy wasted too much of everyone else’s time. So after the first day, we recorded all of the remaining movements one singer at a time. This was at least as tedious as my initial sequencing work, but it was the most efficient way to be sure everything got recorded right.

I wanted to make every tool available to guide the singers in the sessions, so we had the newer accompaniment tracks, the old individual parts, various clicks, other singers’ tracks (sometimes live but mostly previously recorded), and my conducting through the window for cues and fermatas. We recorded the accompaniments to tape in 2-track stereo, with the computer (my old Macintosh Quadra 630) synched to SMPTE from the DA-88, and we added to the headphone mix the singer’s vocal line and various click tracks live from the synth during each take. The balance of all these plus the right amount of the singer’s own mike in the cans was very critical; if one element was out of balance, it became hard for the singer to stay in tune, and often they couldn’t tell us what was wrong. Even the singers who were more experienced with multitrack recording found this an especially challenging project.

During the sessions we found a few errors in the sequences, which I was able to correct on the spot. Also, when a singer needed a special rhythmic guide for a difficult passage, I was able to create a custom click track in a few minutes and feed it into the headphone mix.

The piece includes a wide variety of textures for the 12 singers, including 6-part and 4-part writing (2 and 3 on a part, respectively), and chord-pyramiding. This created some particularly tough mixing challenges, because the singers could not hear each other to blend their sounds. It was mostly up to me to regulate each singer’s timbres and dynamics for each take.

In the control room, we usually monitored the singer, the individual synth part (rather louder than in the headphone mix) and clicks, to put a microscope on precision. As session producer, I had to judge whether the single line we just heard, with human inaccuracies, would blend with the whole ensemble in our imagination. There was very little time to listen to rough mixes until everything was recorded. Because we ran out of time in New York, Toby had to record most of his parts at Pogo Studio in Illinois, after September 11th, and without me present.

Editing & Mixing

Mark Rubel transferred the audio to computer in Illinois and assembled the tracks into Digital Performer files, then mailed me CD-Rs containing the audio and DP files. Several of these files had many more than 12 tracks in them, including multiple takes and separated passages where singers were roadmapped to different parts in the score. I edited these down to 12 clean tracks, one for each part in the score, regardless of who sang which passage. After some practice I was able to correct ragged entrances and cutoffs without the use of crossfades, by matching the singer’s waveform.

I found that precise editing entailed a certain amount of rough-mixing, so I could hear the results clearly. After the first couple of simpler movements, it was clear that the more rough-mixing I did, the easier it was for Mark & Toby to concentrate on refinements, so in some cases I did as much mixing work as they did.

I sent back the edited DP files via e-mail, and that worked fine, because we both had the same audio files on disk. I couldn’t be with them in Illinois for the final mixing, so they mailed me mix CDs and I had to offer my comments via e-mail as well.

We did not use any digital pitch shifting to correct intonation, mainly because we didn’t have time. (But it’s also nice to be able to say we didn’t use it.) So the tuning you hear on the CD is exactly as we sang it.

The final mastering was done at Sony Studios in NYC by the wonderful Charles Harbutt. A bit of overall EQ was applied uniformly across the whole piece, and normalization work was very painstaking. It took a total of 12 hours in 2 sessions to complete the mastering.