Mozart Meets C-Lab Notator
Proving that hi-tech tools are not the sole preserve of pop musicians, Richard Gonski, Music Director of the Electric Symphony Orchestra, tells how he and Francis Monkman took Mozart's Piano Concerto In C Minor, K.491, and sequenced the whole work using C-Lab's Notator.
My first real contact with an orchestral score came about some 15 years ago when I was given Corelli's 'Christmas Concerto' for six-part string ensemble as my first conducting assignment. Filled with excitement and anticipation, I sat down at my rather dilapidated upright piano and set about learning the music. However, my enthusiasm was soon dampened - in no way could I get six voices under two hands, sustained notes disappeared, the polyphony was all but lost, and worst of all - it sounded like a piano and not a string orchestra. As I progressed to more complex scores the limitations of the piano became even more obvious. In the end, like many a budding conductor before me, I resorted to a combination of the piano, records, and my own imagination.
What I wanted then was a machine that would allow me to have complete control over a piece of music prior to that first rehearsal with the orchestra. I needed to be able to listen to a woodwind or string passage without the distraction of the rest of the orchestral sound, to experiment with different tempi, dynamics, articulation and balances - in short, I was dreaming of a sequencer.
I have therefore followed the development of music software and hardware with an eagle-eye, waiting for a product that was both affordable and would fulfill my requirements. A few months ago, the opportunity to sequence an orchestral score effectively finally presented itself. An Atari 1040ST computer, C-Lab's Notator software, and a forthcoming tour with the Electric Symphony Orchestra (ESO) provided the means and the framework. The synthesizers, expanders and mixer were provided by Yamaha, who have sponsored the ESO for the past two years - two TX802s, a TX81Z, the TX16W sampler, a DMP7 mixer, and a DX5.
The first step was entering the notes into the computer. A pattern in Notator contains all the recorded tracks between two musical points - these are then chained together to form an arrangement. It made sense therefore to input the concerto in chunks, with each pattern beginning at an important structural or orchestrational point.
The Mozart 'Piano Concerto In C Minor' is scored for flute, two oboes, two clarinets, two bassoons, two horns, two trumpets, timpani, strings (1st and 2nd violins, violas, cellos and basses) and keyboard. Each of the string, flute, timpani and keyboard parts were assigned a track and a MIDI channel; the remaining pairs of woodwind and brass were assigned to a single track, but with each instrument having its own MIDI channel. This allowed us to input both parts in one pass, edit them and then split them using Notator's displayed voice separation option, a new feature of the Version 2.0 software. This is achieved by drawing a line with the mouse between the notes, and although a little tedious, it does work. An option that would automatically split the voices on a high/low note basis without having to set a split point would have been very useful.
As the parts were entered one by one, we had the choice of editing each part separately or copying rhythmically identical segments from one track to another, and then transposing them to the correct pitch. I ended up using a combination of both methods. For example, if the 2nd violins mirrored the 1st violins between bars 20 and 30, I played in the part to bar 20, let the sequencer run on in record mode until bar 30, and then continued playing until the end of the pattern. A segment copy of bars 20 to 30 from the 1st violins to the 2nd's completed the operation. Notator 2.0's copying options are excellent, except for its mishandling of global bar numbers when copying from pattern to pattern.
Quantisation came next. We opted for placing almost everything directly on the beat. Classical musicians in general are not reknowned for their ability to play on the beat or to feel the small sub-divisions within the beat. I am sure that the reliance of classical music on notation is the main culprit. As a teacher, I have seen time and time again that the biggest stumbling block for children is translating the rhythmic values on the page - they eventually get it, but more often than not it remains an intellectual process, whereas a rock musician relies from the start on instinct and feeling.
Listening to the finished version of the Mozart, we were immediately struck by the excitement generated by a performance that was in strict tempo and always together. It became apparent that one could choose a slightly slower overall tempo for the movements because there was no slowing down in the fast bits. It was easy to insert ritardandos and accelerandos where needed, with Pseudoevents, although fermatas are more of a problem. Notator offers you the choice of inserting absolute or relative tempo changes - we used the relative option so that the overall tempo could subsequently be altered without destroying the flow of the music.
"What I wanted then was a machine that would allow me to have complete control over a piece of music prior to that first rehearsal with the orchestra - in short, I was dreaming of a sequencer"
One welcome feature of Notator is its ability to set different quantisation values within a track. In a passage of 16th notes followed by triplets, the computer makes the choice between the two, and quantises the values automatically. Although one can also quantise quintuplets or septuplets, they are not displayed as such - in the Mozart score they do not appear, but they do crop up frequently in music from the late 19th century onwards.
One of Notator's big advantages over its main competitor, Pro24, is its ability to display the whole score at once, allowing you to jump from track to track in the edit page. To my mind, this is an essential feature in any classical application, and its omission ruled out the use of the Steinberg program for our purposes. In any musical score there will be sections where a number of parts provide the harmonic underpinning or where a melodic fragment is doubled by other instruments. One instinctively wants to move vertically when editing note lengths and velocity on these sections, not jump back and forth between the main and edit pages. Unfortunately, Notator does not go quite far enough. While it is easy enough to group and then alter event parameters on a single track, it is not possible to edit different tracks simultaneously, unless it is a deletion or transposition. Shame.
In fact, editing of note lengths and velocities turned out to be not only the most time consuming task but the most musically illuminating as well. It also highlighted some of the hidden problems of combining electronic sounds, the limitations of MIDI, and the discrepancies that arise between notation symbols and the numerical length values in the event editor.
Note lengths first: Notator 2.0 has a resolution of 48 internal clock ticks per 16th note (96 in Version 2.1), and to begin with all the values were set to multiples of 48 minus 1. The subtraction of one tick should prevent MIDI clicks, but in reality, three or four ticks had to be subtracted in many cases. This, of course, had a detrimental effect on the phrasing, with some legato passages ending up not quite as smooth as one would want. We also had to go through each part checking for overlapping notes. There is a definite bug in the software - upon reducing the length of an overlong note, the lengths of subsequent notes change as well. Furthermore, a great deal of experimentation had to be undertaken to find the right length values for the different types of articulation ie. staccato, detache, etc.
One improvement to the software would have helped enormously - the addition of intelligent articulation markings to the notation library - ie. putting a dot on top of a note would automatically shorten it by a specified amount, say one half. At present (in Version 2.0), a staccato crotchet has to be transcribed as a semiquaver or quaver. This principle could then be extended to slurs, dynamics, trills, appoggiaturas, fermatas, repeats, etc. Another helpful feature would be a user-definable rhythmic or dynamic template. All music is based to some extent on repetition, and it would save literally hours of work if one could get the computer to search for a specified phrase and alter it automatically, just as in word processing software. The Transform function in Version 2.0 of Notator is a definite step in this direction, but is not quite flexible enough. What is needed is a language, allowing the user to create his own subroutines which could then be accessed at will. As things stand, I would recommend working on as much detail as possible in each voice before continuing with the next section - then if you need to copy a phrase to some point further along, all the MIDI parameters will be copied with it. I found it easiest to work in as high a display resolution as possible when editing note lengths - in 16th note resolution, a 32nd note is 0 24, whereas in 32nd note resolution it is 1 0. This saves you messing around with the slider and a considerable amount of time.
The setting of velocity levels posed some serious problems as well. When Mozart composed his piano concerto, he had specific instruments in mind. He knew what a violin sounded like, and he knew that all violins would respond more or less uniformly. With the hi-tech equipment at our disposal, we chose the sound patches that seemed most like orchestral instruments. We soon discovered, however, that the string sounds we had chosen on the TX802 expanders were not all that responsive to subtle velocity changes, and so quiet passages were set to a velocity value of 1 and loud passages to velocity 127. The sampled strings on the other hand were much more responsive to velocity, forcing us to set different levels on the sound modules themselves. After passing the disk on to a friend who owns a Roland, it transpired that his string sounds hardly responded to velocity 1 at all, and so we had no choice but to compromise and limit the dynamic range, especially at the quiet end. This also affected crescendos and decrescendos, as a velocity range of 20 (say) may be more than plenty for one machine and hardly noticeable on another. Eventually, this combination of velocity settings in the software, and volume settings on the expanders, sampler and mixer did the trick.
"One of Notator's big advantages over its main competitor, Pro24, is its ability to display the whole score at once, allowing you to jump from track to track in the edit page. To my mind, this is an essential feature in any classical application..."
Realising dynamic markings such as fp is not easy. A fortepiano in Mozart's time signified a short, loud attack followed by a quiet sustain. Due to the limitations of MIDI, there are only two ways of achieving a true fp - inserting a Control 7 (Volume) value in the event editor - say, a semiquaver after the Note-On - or placing a loud, short note on top of the quiet one. Both are unsatisfactory - with the first method one is sending a global volume setting to different machines with different responses (see above), and with the second, at least two oscillators per voice are needed. Even then, it doesn't always work if both oscillators have been busy until the fp event.
In the third movement of the concerto, there is a time signature change at the beginning of the coda from 4/4 to 6/8. Notator had no difficulty in coping with this, and it was only after I had completed entering the music that we realised that the click track (essential for performance with live musicians) was unusable. There it was clicking six times per bar, when the feel of the music is a rather fast two. We had no choice but to redo the whole section in 2/4. A similar problem has arisen in Beethoven's 8th symphony, which I am working on at present. The last movement is in a fast 4/4, but the feel is in 2. One should be able to vary the click settings to integer divisions of the time signature.
Even the most agile and flexible sequencer in the world will not generate a musical result unless the chosen sound patches are adequate to the task of interpreting the MIDI data. Undoubtedly, our use of a combination of sampled sounds and different synth expanders helped to achieve a convincing end result. Far more significant, however, was the realisation that it is all very well to have a convincing string or wind sound in a long, legato passage, but unless it continues to sound like a string or wind instrument in a fast passage full of repeated notes, you are back to square one. A balance has to be found between the shortest possible attack time (allowing the sound to 'speak' before the arrival of the next note) and retaining the characteristic of the instrument in question. Unless this balance is achieved, the short repeated notes will sound quieter than they should, and long ones will subsequently sound accented.
Solving this problem is a start, but one is then faced with the folly of treating velocity and volume as one and the same thing, which they clearly are not. Increasing the velocity value in a sequencer will alter the timbre [tone] of many electronic sounds - in most cases they become much more aggressive and abrasive. The absence of a Volume byte in the MIDI spec is a serious omission in my opinion. Notator's Realtime MIDI Generator is one way out - it is an excellent feature, with 16 faders assignable to MIDI Volume and other control parameters. The only problem with using it is that it inevitably clutters up the stream of MIDI data, uses up a lot of memory, and will cause glitches if used constantly and without care. (MIDI devices don't like receiving more than one controller message at a time.) Combining different digital sounds can also generate unwanted distortions, imbalances in different registers, and generally unpredictable happenings, despite their suitability when played on their own. We spent a lot of time trying out different combinations, including a trip down to Yamaha's R&D Centre in London for some extra help. This is a subject in its infancy and deserves to be investigated much more thoroughly.
Our intensive and time consuming work brought many rewards, some of them quite unexpected - new insights into the problems of orchestration was one of the most important, and they came through constant listening to small groups of instruments in both the editing and final mixing stages. Another was the necessity to define exactly what was wrong with an unsatisfactory passage. Since completing this project, I have discovered that my ability to pinpoint a problem in an orchestral situation has increased significantly and has saved a lot of time and unnecessary discussion in rehearsal.
In the end though, it was the interface with live musicians that brought the most satisfaction. It was not only the enthusiastic feedback from members of the Electric Symphony Orchestra that convinced us that we were on to a good thing, but also the potential that has been created for soloists who can now learn a concerto with a full orchestral accompaniment while retaining full control of all the parameters, for conductors who are studying the score, and for anyone with an Atari ST, C-Lab Notator, and a synthesizer who has wanted to stand on the podium and do their thing. My work for the next few years is cut out for me - you can expect to see Mozart and Beethoven on a computer screen near you any day now!
Richard Gonski, Electric Symphony Orchestra, (Contact Details).