The Celtic Macintosh (Part 2)
AN ALBUM PRODUCED ENTIRELY ON A PERSONAL COMPUTER
Part 2. Paul D. Lehrman takes up where he left off last month in his detailed description of how he recorded and digitally mastered an entire album using an Apple personal computer, the Total Music multitrack software, and a bunch of MIDI synths. Read all about it!
Paul D. Lehrman concludes his report about how he recorded an album of music entirely by himself using a handful of MIDI synths, a 512K Apple Macintosh computer and the Southworth Total Music' multitrack sequencing software.
To recap briefly on last month's article, the author had reached the point where the music had been constructed using the software and the song orchestrations more or less finalised. Next came the task of cleaning up the tracks and preparing for the digital mastering...
It was at this stage that I had to deal seriously with the idiosyncracies and limitations of the various machines I was using to create The Celtic Macintosh. The Kurzweil, although its MIDI implementation is extensive, had some very odd features. For example, when the instrument was in 'Multi' mode (receiving polyphonic information on more than one MIDI channel), only one channel will respond to controllers (pedals, modulation wheel, etc) or to program changes. Therefore, most of the sounds I wanted to get from the Kurzweil had to be preassigned to existing MIDI channels, and if I wanted to use any other instruments, they would have to be assigned to the Kurzweil's 'basic' channel, which I set up to be MIDI channel 6. Because there were 10 channels available to assign to the Kurzweil (16 MIDI channels total, minus one for the DX7, one for the TR-707 drums, and four for the Casio CZ-101), this was not a serious drawback.
As Figure 1 shows, the instruments I was using the most were pre-assigned to separate MIDI channels, and channel 6 ("active channel") was reserved for 'specials'.
However, if MIDI channel 1 were pre-assigned to Grand Piano, I found I could not use the sustain pedal on it.
When a Piano track was recorded on channel 6, and then moved to 1, the pedal information was ignored, so on those occasions the track had to be edited so that the notes I wanted to sustain were made longer. I didn't use pitch-bend on the Kurzweil at all - if a pitch-bend command is sent to the instrument's 'basic' channel, it bends all of the channels at the same time. (I understand that the new MIDI software for the instrument, implemented after I finished the album, takes care of many of these quirks.)
On the plus side, the Kurzweil handles MIDI patch changes in a way that I believe is unique [Editor's note: the Ensoniq ESQ-1 now has this valuable feature.]: if you send it a patch change while a note is sustaining (either because the key is held down or the sustain pedal is on), that note will continue to sound, without changing, until the key or pedal is released, and the patch change will only affect the next note played. The note-ons and patch changes can be sent extremely quickly to the Kurzweil without ill effect, so I was able to use one MIDI channel for several different, virtually simultaneous voices.
For example, I wanted a timpani roll to end on a downbeat with a chime note. The timpani roll was started at the beginning of beat 4, and a patch change for the chimes was inserted on the 95th clock tick of the beat (the Total Music software breaks down beats into 96 parts). Then the chime note was inserted on the downbeat (clock 0) of the next measure, followed by a patch change back to the timpani on clock 1, and the final timpani note on clock 2. Any delay between the notes was inaudible.
Pitch-bend was also a problem on the Casio CZ-101. The Casio can receive homophonic (single-note - sometimes called, confusingly, 'monophonic') information on four different (adjacent) MIDI channels, with independent patch changes on each channel, but it will only respond to pitch-bend on one channel, which may not necessarily be the 'basic' MIDI channel (which in Casio's case, is always the lowest number of the four). Therefore, I had to be careful assigning voices in which I wanted to use pitch-bend, and also make sure the Casio was always set up correctly.
Another interesting Casio trait (which is shared by only a few other synthesizers) is that if portamento is applied to a note, it will only be heard if there is no interruption between the note and the following one. In other words, if two notes are played legato while portamento is 'on', there will be a 'glide' between them, and the second note will have no separate attack; but if there is any separation at all, there will be no glide and there will be a fresh attack on the second note. This meant that in those sections in which portamento was used on the Casio, note durations had to be very carefully edited.
The DX7's most annoying quirk is the way it handles patch changes. If a patch change is received while a note is sounding, the note does not finish, or even politely shut off (as it does in the CZ-101). Instead, the instrument immediately executes the final decay stage of the new patch, which most often results in an ugly, metallic glitch. Even if the note is no longer being played by the computer, if the patch change occurs any time before the synthesizer envelope is completely finished (and many sounds I was using had a long 'ring' to them), the glitch will occur.
As I was doing the final editing, I discovered a more subtle, but equally annoying characteristic: under certain circumstances, the DX7 will not respond to a patch change right away, and will actually delay the onset of a note following a patch change by a few milliseconds. Therefore, patch changes had to be placed extremely carefully, and sometimes notes had to be cut off (or pedals turned off) earlier than I might have liked in order to make room for the patch changes.
The Lexicon PCM 70 digital effects unit I used on The Celtic Macintosh album responds to MIDI program changes quickly, but at a cost. If a patch change is sent to the unit, the input is momentarily cleared and the contents of the delay registers are unceremoniously dumped. This doesn't cause a glitch, but it does introduce a 'hole' into the sound which can be disconcerting. Since I was using fairly straightforward reverb programs, this problem could have been avoided by using the Lexicon's 'Dynamic MIDI' features, and instead of actually changing the program, I could have used a keyboard controller to adjust the decay time and/or the wet/dry mix, which results in completely smooth transitions.
I chose not to do so for several reasons. One was the Total Music program's limited controller-editing capabilities, which meant that if I made a mistake it would be difficult to correct. Another was that the Lexicon PCM 70 only has a mono input, and if the wet/dry mix were under MIDI control, the stereo image would collapse towards the centre as the reverb was diminished. As it worked out, it was perfectly acceptable to limit MIDI program changes to the pauses between the reverb selections, and there were no audible effects. Because I was addressing the PCM 70 only with program change information, and no notes or controllers, I could use one MIDI channel to control both the Roland TR-707 drum machine (which, of course, ignores patch changes) and the PCM 70 reverb.
Irish tunes, especially fast ones, are rarely played one at a time. Instead they are performed in groups of two or more tunes, one right after the other, sometimes with abrupt tempo and key changes. The word processor-like editing functions of the Total Music software for the Macintosh provided an ideal environment for experimentation in constructing such suites. From the catalogue of recorded tunes I had assembled on the computer, I could audition various combinations and contexts, not committing anything to disk until I was satisfied with it, and even then keeping alternate versions on disk in case I changed my mind later.
Because I could not only move sections of music around, but also duplicate them and put them in different places, I was able to do 'rondo form' constructions ('ABACA' etc) with ease. With each repetition of the 'A' section, I could alter it slightly as to orchestration, tempo, key, or whatever, so that the result sounded as if it was a continuous performance, rather than built up from pre-existing sections.
These experiments were often done by using the software's 'sequence chaining' function. Each section would be loaded in as a separate sequence, and then a sequence 'call' would be inserted at the end of the section, commanding the next section to start. The calls could be placed at any point (and sequences could be overlapped if necessary), and any sequence could call any other, so the order and timing of the sections could be adjusted with great flexibility and accuracy - which was advantageous, because I found that since I had not used much rhythm-correction or step-time recording, lining up the entrances of tunes exactly on bar lines didn't always work. Once I was happy with the construction of a suite, I usually merged the various sequences into one, to simplify loading the computer for the final mix.
Under some circumstances, if two tunes were concatenated together, there would sometimes be an apparent tempo change (even though I had programmed none), due to the characteristics of my playing style, or the nature of the music itself. A single slight tempo change, inserted at just the right point, cleared up the problem. Patch changes were also a problem, especially with the DX7, and sometimes note durations and placement of the patch changes would have to be carefully adjusted at the transition points.
Sony F1-type PCM convertors, especially when used with consumer-grade half-inch video recorders, allow no editing. Just hitting 'Pause' on the VCR between selections puts an audible click on the tape which is impossible to remove. Therefore, if I was to avoid the expense of transferring the tape to the professional 1610 digital format, I knew I would have to record The Celtic Macintosh in just two passes, one for each side of the album.
I assembled the sides in much the same way as I assembled the individual selections, by chaining them together with sequence calls, and adjusting the pauses between them by ear. (I also had to take into consideration the time necessary to execute smooth-sounding program changes between tunes on the Lexicon PCM 70.)
This arrangement also let me try out the track selections on each side in different orders, without going through the old-fashioned hassle of continually rewinding and splicing (and possibly damaging) a conventional master tape.
My first priority was to vary the tempos, so that there weren't two fast or two slow tunes in a row; and the second priority was song keys - I discovered on my first assembly of Side 1 that there were three cuts in a row all in the key of D-major. Rather than transpose the middle tune (which would have been a simple operation, but it might have subtly changed the sound, forcing me to redesign synth patches or worse), I simply exchanged it with a tune I had originally planned to put on Side 2.
I also made sure that the transitions between cuts was dynamically smooth - that a loud tune following a softer one didn't jump out and scare the listener. This turned out to be of concern at only one point, and I compensated for it by thinning out the orchestration of the first note of the louder tune.
Since each side of the tape was over 20 minutes long, I knew I would have to keep mixing down to a minimum, because if I were to make a serious mistake any time during the pass, I would have to start over again. Fortunately, I had left all of the TOA D-4/D-4E's mixer levels pretty much alone throughout all of the recording process, which meant that the balances were more a function of how I played the individual tracks than how the mixer was set up. When it became apparent that some level adjustment was necessary, I would change the velocity of the track in question with the software (although if it was a long track, or one with many notes, this could be very time-consuming), or design a new patch for the track in question, identical to the one I used to record it, but with a softer or louder overall level.
Using these methods, I managed to get to the point where there were fewer than a dozen moves on each side of the tape that had to be performed in the mix. One distortion problem made itself apparent at this point: the Kurzweil keyboard, especially when it's playing multiple channels, can put out a whopping 20 volts peak-to-peak, and this was just too much for the input stage of the D-4 mixer, even with the input trim control all the way down.
Therefore, a couple of the mixing adjustments were done from the Kurzweil's volume slider. I kept these to a minimum, however, because the slider has an unfortunate side-effect: it not only lowers the level, it also reduces the instrument's dynamic range.
The input level on the Sony PCM-701 I used for the digital master was adjusted by my finding the loudest spot on the album, and setting the level so that it peaked at -3dB. After that, mastering the tape was simply a matter of loading the video cassette into the Panasonic VCR I had hooked up to the Sony PCM-701, loading the various sequences into the Macintosh computer, and hitting 'Record' on the Panasonic.
I deliberately took my time doing the mixes, and took long breaks between passes to keep my ears fresh. The first side of the album took four passes, and was completed in an afternoon, while the second side, which took three tries before I was satisfied, was done the next morning.
During the mastering of the second side, I decided to try something a little risky. On a lengthy violin solo on the final cut, I removed all of the modulation wheel data, and instead 'played' the mod wheel while the master tape was running, being careful not to make any gross mistakes so that I wouldn't have to start over. The risk paid off: the result was a more spontaneous, live-sounding track, which ended up (according to most of the folks who have heard the tape) being the most exciting part of the whole album. The philosophical implications of this action I will leave for others to debate...
The final step was to record reference tones on the master tape, for which a single-operator patch on the DX7 served just fine.
The digital master was taken to a local duplication facility, equipped with various analogue and digital mastering machines, and about four dozen good quality Aiwa cassette decks for real-time duplication. I told the engineers where the programme peaks were, and explained that to keep the noise floor as low as possible, I wanted the cassettes recorded at as high a level as their equipment and tape stock would tolerate.
The first run didn't come out well, as the overall record levels were far too low. The engineers explained that they heard some of my more 'buzzy' synth patches on the tape, and figured it was tape distortion, so they backed off the level until (they thought) it went away. After a few well-chosen words from both sides, we all sat down and went over their procedures. We decided that a little bit of limiting would not hurt the sound, and would improve the noise level, so a dbx limiter was patched in, and set to 2:1 limiting at +6dB. It kicked in briefly at two points on the first side and once on the second. We did a partial run of 30 tapes, which would be enough for the St. Patrick's Day show, which by this time was only 24 hours away!
The next week I rented the Sony PCM-701 once more, and remixed the first side, with an eye towards keeping the levels more consistent, and thereby (I hoped) making the tape duplicator's job a little easier. I took the tape back to them, and they did several more runs without a hitch.
How a cassette tape is packaged for the customer is not usually a subject for a technical article, but it should be mentioned that the Apple Macintosh computer, besides creating the music, was used to create all of the tape labels and cassette inlay cards as well (not to mention the diagrams accompanying this article). A Macintosh program called 'PageMaker' allows you to mix text and graphics and place them with great precision on a page. I obtained the dimensions for some peel-off blank cassette labels that the duplicators had available, and used PageMaker to design and replicate a page's worth - twelve, to be precise. This 'document' was printed out on an Apple LaserWriter at a local computer store, and was then placed in a Xerox photocopier which could print directly onto the blank labels.
The inlay cards were likewise done with 'PageMaker' and the LaserWriter, and were reproduced on green card stock, in keeping with the spirit of the Irish music. The results were very professional-looking, and few people would ever guess that they were not done by a typesetting house.
Besides the immediate need for the music, there was one major reason for doing The Celtic Macintosh - to prove it could be done.
I have always had delusions about being my own orchestra, so that I could be sure all the parts of my music were played just the way I heard them in my head. This obsession led me, when I was a teenager, to experiment with overdubbing myself on two tape recorders feeding each other, and later to spend all of my spare cash on multitrack studio time, and then, as the technology began to develop, to get involved with computerised multitracking instruments like the alphaSyntauri. Only now are the hardware and software available to allow me to do solo projects whose sound I can be truly happy with, without blowing my bank account.
From this point, the technology can only improve and get cheaper. By the time you read this, there will be even less expensive FM synthesizers and samplers on the market, as well as cheap, sophisticated MIDI-controlled signal processors, and even MIDI-controlled mixers. The software is getting faster and more comprehensive, and editing systems that can automatically and rapidly perform sophisticated composing and arranging tasks will soon be commonplace. MIDI delay problems are being overcome by systems that use multiple MIDI data streams, and synchronisation between MIDI and the rest of the world is becoming more accurate and easier to use.
In some ways, the growth of music technology has had the effect of putting more barriers between the musician and his music. Multitrack tape recorders, mixing consoles that need six or eight hands, automation systems that take hours to programme, and digital processors that require a physics degree to make sense of them, all mean that there are even more people who have to be involved in the realisation of what might have started out as one person's musical ideas. While this can mean that there is more creative input into a project, it can just as easily mean that there are more ways to distance the artist from the finished product.
With computers, however, we can return to a world where a musician can realise his or her own vision, unencumbered by budgets, schedules, cranky A&R departments, finicky recording equipment, under-enthusiastic players, or overworked engineers. I think Bach would have loved it, and I know Varese certainly would!
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 | Part 2 (Viewing)
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!