The Celtic Macintosh (Part 1)
AN ALBUM PRODUCED ENTIRELY ON A PERSONAL COMPUTER
Part 1. American, Paul D. Lehrman, documents how he recently recorded an album of synthesized Irish music entirely on an Apple Macintosh computer using the Southworth Music Systems 'Total Music' software.
In the music industry, as in most other fields, the grandiose promises of the personal computer revolution have yet to be entirely fulfilled. To the majority of musicians and producers, computers have so far been merely aids to production, to be used in conjunction with conventional recording tools and techniques, like multitrack tape (analogue or digital) and discrete hardware for mixing and processing. There are still many steps and many people involved in a composer's realisation of his (or her) musical ideas; the traditional way of recording music - ie. hiring a spacious studio and populating it with a group of musicians whose job it is to make sense out of music charts inscribed by the arranger, who in turn has done his best to meet the needs of the composer - is still very much with us.
But the next step, completely automated music production by an artist working alone, is ready to be taken. With the right software and in the right hands, computers can replace almost everything standing between the artist and the music.
This report is about how this author took that next step. It chronicles how I recorded an album of music entirely by myself, at home, using equipment that almost any serious musician can afford to own (or hire), working on my own schedule - which meant that a session started when I was good and ready, which was sometimes not until three in the morning!
Everything from the individual performances of tracks to the final digital mastering was in the hands of one person. There were no other musicians; there were no instruments other than electronic 'synthesizers' (a weak term for the state of the art sound-generating hardware now available) controlled by a computer, which also handled most of the balancing and processing chores as well; there were no outside engineers or producers; and there was no intermediary tape stage, as the mix was done 'live' to a 2-track digital master.
The album was very different from what most people associate with 'machine' music - the requirements of the project were that it sound basically like a live band; an acoustic/electric Irish band, to be precise. It was done quickly - from concept to completed master took only two weeks.
Most important, the project was done on a budget of practically nothing. After an initial equipment investment of about $20,000 (which, I hasten to add, was not really my money anyway and, as shall be seen, could easily have been cut by at least half), the cost of producing the finished tape was less than $100: half a dozen computer disks, two high-grade video cassettes, and a few days' rental on a PCM convertor.
Before a project like this could become feasible, two significant developments had to occur: the availability of truly convincing-sounding electronic musical instruments, and the availability of sophisticated MIDI control programs (commonly called 'sequencers', a term that should soon be consigned to history), that allow a composer or arranger to produce complete, fully orchestrated and mixed compositions, with nary a tape recorder in sight.
The instruments used on The Celtic Macintosh were a Yamaha DX7 synth, a Casio CZ-101 synth, a Roland TR-707 drum machine, and a Kurzweil K250 sampling keyboard (the model with the original, somewhat limited, MIDI implementation). The K250 was a loan unit from Kurzweil, which I was fortunate enough to have for several months while I was working on a software development project for the company. (Although the project was, by mutual decision, never completed, Kurzweil Music Systems were graciously willing to let me hold onto the instrument for a while afterwards.)
Mixing was handled by a TOA D-4 keyboard mixer with their D-4E expander, allowing ten inputs, stereo effects, and stereo outputs. The only signal processing in the chain was provided by a Lexicon PCM 70, also a loan unit.
Monitoring was done on a variety of speakers, including TOA 280MEs (very bright and unforgiving), Auratone 5Cs (for near-field monitoring and that 'cheap cassette player' sound), and a pair of Rectilinear X1s that I've had since my college days, and therefore know very well - which for me is the most important criterion for choosing a monitor speaker. The final master was produced on a Panasonic home VCR, with the digital coding handled by a modified Sony PCM-701ES rented from a local classical music production company.
Controlling everything was a standard Apple Macintosh 512K computer, equipped with an external floppy disk drive (but no hard disk) and Total Music software, one of several MIDI sequencing programs available for the Macintosh. The early version of the program I had at the time (Revision 0.99) suffered from some software 'bugs', but since I was on the original development team, I knew where the bugs were, and how to avoid them, and so I managed to keep it running relatively smoothly.
It must be noted that were I financially responsible for the equipment I used, the Kurzweil and the Lexicon would not have been my choice. They are both marvellous devices, but a little out of my (or any struggling musician's) price range.
Either one of them could have been replaced with cheaper units, and although the sound quality would probably not have suffered unduly, the project would likely have taken considerably longer.
The Celtic Macintosh, as the album ended up being called, was the result of a request by a friend. Sharon Kennedy is a Boston-area storyteller who specialises in Irish stories and songs for both children and adult audiences. For a St. Patrick's Day show, however, Sharon wanted some taped music - to be played before and after the show and during the interval - that would put the audience in the mood of being on the Emerald Isle. The music had to sound 'acoustic', and could contain few elements unfamiliar to the audience. (Making my task more difficult was the fact that two excellent live musicians, a fiddler and an Irish harpist, were also on the programme, and they didn't approve of the idea of computerised background music one bit - no pun intended.)
About three weeks before the performance, I started arranging a few tunes. Some, like the well-known jig 'The Irish Washerwoman', I did off the top of my head. Others - airs, hornpipes, reels, and laments - I gleaned from published collections by Robin Williamson and Miles Krassen. Still others I pulled off records, especially those of the modern Celtic bands, Silly Wizard and Relativity. At first I did simple arrangements, and after about four days I had four tunes more or less completed. Sharon was so delighted with what she heard, and I was so pleased that the work was going so fast, that I decided to keep working right up until the show, to see if I could finish an album's worth of material by then. Soon, the music began to take on a life of its own, and on top of arrangements that sounded like traditional dance bands, I found myself applying Vaughan Williams-style symphonic orchestrations, Phil Spector-ish sound-washed backbeats, Springsteen-ian bass lines, Fairport Convention-ish electric sizzle, and Windham Hill-like ambiences.
'The tune', a single melody line, often without a fixed harmonisation, is the starting point of just about all Celtic music. I started building most of the compositions on the album by choosing a lead instrument and playing the tune through once or twice. Because the tunes were being both arranged and 'recorded' simultaneously, there was no necessity to start with a rhythm section, and in fact this was never done. Instead, a metronome generated by the Total Music software through the Macintosh's internal audio circuitry, and amplified through the TOA mixer, was used as a click-track most of the way through the production, often even after drums (in those songs in which they were used) were laid down.
Since the instrumentation could easily be changed at any time after a track was recorded on the Macintosh, I was free to use just about any sound I wanted for the initial track, which was usually a DX7 violin, a Kurzweil guitar, or a Casio pennywhistle or flute. There was also no need to use step-time entry of notes, as the rhythm could always be corrected after the fact, and I wanted to maintain as much of a 'human' feel on the tracks as possible. (As it happened, step-time note entry was used only once on the entire album, to create a countermelody for one of the faster jigs. Due to discrepancies in MIDI implementations, the software could not read step-time information from the Kurzweil, which was normally used as the master keyboard, and so the Yamaha DX7 keyboard was pressed into service for the task.)
After the initial tune was laid down, obvious mistakes were cleaned up immediately, and then an accompaniment part was added, using bass of some kind and a rhythm instrument like piano, electric piano, accordian, or guitar. These would be recorded at a slow tempo, to keep rhythmic errors at a minimum. I would then play with the orchestration of the lead line, and spend some time looking for (or designing) an instrument sound that would fit.
Irish music, particularly dance music, relies largely on repetition, and therefore much use was made of the Total Music software's ability to generate automatic repeats. The software could also transpose sections at will, either in-place or moving the transposed sections elsewhere. Continuously repeating sections, with the pitch raised by a half-step at each iteration, is an effective means of creating musical tension and movement. If more tension was needed, small accelerandos could be programmed in from time to time.
As all this was going on, I could change flutes to horns (or double them at the octave), guitars to harpsichords, basses to tubas, etc. Once the initial musical elements were 'recorded' (stored in the computer's memory), the compositional process moved quickly, as I attempted to modify and assemble those elements in a pleasing and coherent way. Occasionally I found that after changing an instrument, the playing style didn't match the new sound - chords played on a piano have a very different articulation from chords played on a guitar, for example - and I would have to re-record the segment in question, thinking 'guitar' to myself as I did so. Most alternative versions were saved to disk, so if I later wanted to mix and match between the various versions, I would have all of them at my disposal.
Some of the orchestrations were deliberately planned, while others were the result of experimentation. For example, there was a line playing on the Yamaha with a 'singing foghorn' patch, which I intended to double on the Kurzweil with a string section. By mistake, I called up a 'vibraphone' patch on the latter instrument, and the result was so pleasingly unearthly that I let it stay that way.
At no time was the Roland drum machine's internal sequencer used. A computer, with it's visual feedback, allows rhythmic patterns to be constructed much faster, and with a far greater degree of control over both timing and dynamics. The program would be set up to loop a four-bar segment, starting at the drum entrance, and a basic drum pattern was built, using the Kurzweil keyboard to 'play' the individual drum voices on the TR-707. Then the pattern was rhythm-corrected, and extended out to the desired length using the Macintosh software's copy function. Drum fills were recorded on a separate sequence (the software can play up to eight sequences simultaneously), so that they could be rhythm-corrected and edited separately without affecting the basic track. For example, the basic track might be corrected to eighth notes, but I would want to rhythm-correct a drum fill to triplet-16ths. Drums in the more 'symphonic' sections of the music were recorded as individual notes, instead of repeating patterns. Many of these orchestral sections used drum sounds from the Kurzweil instead of, or in addition to, the Roland.
Because of the 'acoustic' nature of the Irish music, the Kurzweil's factory presets - grand piano, strings, horns, woodwinds, etc - were entirely appropriate, and so I did no editing of them (which would have taken lots of extra time, anyway). I tried to resist the temptation of including the Kurzweil's beautiful but already-cliched 'Chorus' patch, but I ended up sneaking it in on the last note of the album! I already had a good library of original patches for the Casio CZ-101 synth, including clarinets, woodwind choirs, basses, Minimoog-type lead synth sounds, and nice, thick growls, and few changes were necessary. I also made extensive use of the flute and whistle factory presets on the Casio.
Many DX7 patches I had on hand were useful, but a few needed tweaking. For the aforementioned 'Singing Foghorn' sound, I modified an acoustic bass patch by adding infinite sustain and detuning the DX7 operators drastically so that they would 'beat'. For a Mark Isham-style Oberheim-y vocal sound, I used a Yamaha stock 'Voice' patch and sped up the attack, at the same time smoothing out the pitch envelope and (again) detuning the operators. DX electric pianos were modified to give more bite, and an interesting 'Chiff' voice designed by Gary Rottger was made less strident so that it blended better with the Kurzweil strings and horns. Several stock organ patches were made velocity-sensitive.
An accordian was fashioned from the DX7 factory 'Harmonica BC' patch by slowing down the attack and moving the expression control from the Breath Controller (which I did not have) to keyboard aftertouch. To make a solo violin, the vibrato on Bo Tomlyn's fine 'Agitato' patch was made more coherent by un-detuning the operators, and made slightly brighter and given a heavier velocity-based attack.
Because none of the available Macintosh-based patch editing programs allow you to listen to the results of an edit within the context of a multitrack sequence, which was essential for hearing how the sounds were blending, I did most of the DX7 programming directly on the synthesizer. Extensive use was made, however, of the voice library facilities of the OpCode Systems 'DX/TX Patch Librarian' program, for patch storage and retrieval.
After the songs were constructed, and the orchestrations more or less finalised, came the task of cleaning up the tracks. There were often subtle mistakes in some parts that needed to be removed, and this was the most time-consuming part of the project. Certainly if I had been more careful fixing tracks immediately after I had played them, this stage could have been made significantly simpler, but that would have interfered with the flow of the composition process, and so I didn't mind taking the extra time at this later juncture.
Some mistakes were deliberately left in. For example, a chord on an electric piano may have been played raggedly, but rather than time-correct it so that all the notes played simultaneously (or at least as simultaneously as the DX7 would allow, which isn't very!), I left it ragged, which served to give the track a much more 'live' feel. Drum fills were sometimes slightly off too, and I tended to leave them that way, as it helped create rhythmic tension.
On one hornpipe, I switched the bass line from acoustic bass, to horn, to electric bass as the song progressed. The Kurzweil horn, however, was originally a relatively high-pitched sample, so its attack in the bass register was quite slow, which made it sound late in the context of the ensemble. The solution was simple: I moved the horn track back in time a small fraction of a beat. The Total Music software allows some editing functions to be 'Undone' with a single keystroke, so I was free to experiment with placement of the track until it sounded just right.
To keep things simple, the tracks were initially recorded without aftertouch or modulation wheel information. This type of information is very data-hungry, and indiscriminate use of it can lead to trouble, especially since the version of the software I was using (Revision 0.99) did not have its controller editing facility fully implemented. Doubling a MIDI track containing controller information onto another channel, doubles the amount of information that the MIDI data stream as a whole has to deal with, and this increases exponentially the possibility of MIDI delays being introduced when the sequence is played. Also, if a segment of a track has controller information on it (eg. modulation wheel set to full) and you want to move it into a different section where no modulation is wanted, it can have very disconcerting results.
Therefore, controller effects were added afterwards, as overdubs on the appropriate MIDI channels (with one exception, which will be explained later). Aftertouch could be added to a track by playing a single note on the DX7, far removed in pitch from the other notes on the track, and holding that note through the entire recording, varying the aftertouch as required. After the recording was done, the note itself could be removed with a single motion of the Macintosh mouse, leaving the aftertouch data intact.
At one point, I found myself on the horns of a serious dilemma. I had taken one tune off a Silly Wizard record, and when I listened to the record again, I discovered that I had got much of it wrong: a melody line went up when it should have gone down, and I had three incorrect chords. It was a long tune, with a very complex arrangement involving many repetitions, and I had almost finished it when I discovered the error.
In a normal recording situation, I probably would have had to record everything (except the drums) over again - there were far too many mistakes to try to correct them by punching in on each track. But this was not a conventional situation, and I realised that the corrections should not be that hard to achieve.
In the best of all possible computer worlds, there would be a program that would recognise all of the mistakes and would perform a logical operation on them that would correct them (eg. find all of the F#'s that appear in the context of a B-minor chord and turn them into G's), but unfortunately such a program is not yet available for MIDI applications on the Macintosh (although there are a few programmers who are working on it). Instead, I had to find all of the offending notes and chords by listening and then alter them by hand. Since I knew where they all were, they were easy to find, and since the computer allows very quick alteration of pitches without affecting duration or velocity, it was simple to make the required changes. Therefore, the complete overhaul of the harmonic and melodic structure of a six-minute song took less than an hour and a half.
[In the conclusion of this report next month, Paul D. Lehrman relates how he overcame the limitations of some of the MIDI instruments he used on the project, how the album was mixed and digitally mastered, and how an Apple Macintosh can be employed to add a professional touch to cassette packaging.]
The Celtic Macintosh is available from (Contact Details)
This article was originally commissioned by America's Recording Engineer/Producer magazine, and a version of it will be appearing in their October edition.
Part 1 (Viewing) | Part 2
Gear in this article:
Feature by Paul D. Lehrman
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!