The Synclavier Story (Part 3)
PART 3: DIRECT-TO-DISK AND SYNTHESIS
In the final part of our look at the latest models of the state-of-the-art Synclavier, Scott Wilkinson homes in on the Direct-to-Disk recording system.
In the final instalment of the Synclavier Story, we investigate the depths of Direct-to-Disk recording and synthesis in the new generation Synclaviers.
AS YOU MAY recall from the first part of this series, the Synclavier actually started life as a synthesiser. As a matter of fact, I remember working with the first prototype of the Synclavier that was developed at Dartmouth College (Hanover, New Hampshire, USA) back in the mid-'70s. At that time, it was a 16-voice additive synthesiser controlled by a Data General minicomputer. The system also had two enormous hard disks that held 1.5Mb of data each.
While the computing power and hard disk capacity have improved greatly in the intervening years, the basic synthesis capabilities have not. Of course, the Synclavier is not primarily a synthesiser any more, and the timbre frames and resynthesis capabilities (explained below) were not present in those early days. Now the system can accommodate up to 96 voices of synthesised sound, but the basic means of creating a sound have remained essentially the same.
Synthesis on the Synclavier, which is available only on the 9600 system, is basically an additive process. You can construct waveforms by specifying the relative strength or volume of up to 24 sine wave harmonics. These harmonics are fixed in the harmonic series, which means that their frequencies are whole number multiples of the fundamental frequency. For example, the frequency of the first overtone above the fundamental is exactly twice the fundamental frequency. The next overtone is three times the fundamental frequency, and so on. This is fine for sounds with timbres that fall into this pattern, but there are plenty of sounds for which this is not true, such as bells and drums. Of course, you can always use sampling to reproduce these sounds.
Synthesised waveforms can be assigned to the partial timbres that we discussed in the first instalment of this series (see MT, June '89). As you may recall, up to four partial timbres can be assigned to each user-defined keyboard zone, and these zones can be as small as a single key (but watch out, each partial timbre uses up one synthesiser voice). These waveforms can also be combined with samples in different partial timbre locations and mixed, balanced, detuned and so on.
Various modifiers can be applied to each partial, including a volume envelope and several real-time effects such as vibrato, panning and chorus. This allows you to combine different groups of harmonics at different pitch levels to achieve those non-harmonic sounds.
The synthesis process is controlled from the FM Timbre Page (more about FM in a moment). This page has several modes for displaying the sound data for each partial timbre. The Graphic Display shows you a bar graph that indicates the level of each harmonic and the shape of the volume envelope controlling each partial timbre. As you change a parameter on the keyboard, these displays change in real time. The Numeric Display includes the same information in the form of numbers with the addition of partial tuning, volume, and the real-time effects. The Partial Timbre Display shows more detailed information about each partial timbre individually.
The FM (Frequency Modulation) capabilities of the Synclavier are fairly limited. A single sine wave modulator can be applied to each partial timbre. It has its own volume envelope, which is displayed as a dotted line superimposed on the graph of the partial timbre envelope. The frequency of the modulator can be specified only in a ratio with the frequency of the partial timbre. For example, if the ratio is set to 1.000, the frequencies of the modulator and the partial timbre will be the same no matter which key is played.
As you'll know if you've looked into FM programming, you can achieve non-harmonic timbres like those of bells by setting the modulator frequency ratio to something other than a whole number. This is also possible with the Synclavier. With four independent partial timbres per key available, this just about makes up for the limitation of exclusively harmonic overtones. However, the lack of pitch envelopes (or even volume envelopes for each overtone) seems quite limiting in the Synclavier after working with a system like Digidesign's SoftSynth.
RESYNTHESIS IS ONE area in which the Synclavier excels. Resynthesis is the process whereby a sampled sound is analysed and recreated using synthesis of one form or another. The Synclavier achieves this by dividing a sample into many small time segments called "timbre frames". It then analyses the harmonic content of each timbre frame and automatically reconstructs the waveform additively, using as many as 128 harmonics. Upon playback, the Synclavier plays the timbre frames back, each crossfading into the next.
The more timbre frames into which you divide a sample, the more accurate the reproduction. Within the current memory limitations of the system (although it's hard to imagine, there is a limit to the memory in a Synclavier), a sound can be divided into as many as 300 timbre frames. Typically, however, most sounds can be resynthesised quite accurately with only 20 or 30 timbre frames. For example, I heard a speaking voice resynthesised with remarkable accuracy using 25 or so timbre frames.
At the moment, you must manually mark the timbre frames into which you wish to divide the sound. This is a personal process unique to each resynthesist, but generally, the beginning and end of clearly repetitive waveforms within the sound are the best candidates for marking. Once you have done this, the system automatically calculates all the parameter values necessary to reproduce the sound as accurately as possible.
"Synthesis on the Synclavier is additive you construct waveforms by specifying the relative strength or volume of up to 24 sine wave harmonics."
After that, you can edit the sound in a number of ways. The available parameters include crossfade, or "splice" time, delay, shape (linear or logarithmic), peak level (volume of the frame), pitch offset (up or down as many as 48 semitones from the previous frame), and harmonic coefficients (volume levels of all harmonics). You can also copy one timbre frame to other frames within a sound and modify them in any way. Each frame can be looped for evaluation as long as you hold down a key. Timbre frames typically follow each other, but they can be made to occur at the same time. This can be used to achieve non-harmonic tones by applying different pitch offsets to simultaneous frames.
An entire resynthesised sound can be assigned to a partial timbre and treated in the same manner as sampled and additive/FM waveforms. With this you can do some amazing things. For example, the non-harmonic possibilities are greatly enhanced. You could also create the sound of an oboe that ends up as a trumpet, or a plucked marimba.
"What about crossfading samples?" you might ask. Of course, this is also possible, but timbre frame resynthesis provides more opportunity for creativity. You can perform infinitesimal modifications on individual harmonics and frames, and each frame can be at a different pitch (this provides the only "pitch envelope" in the system). You can also apply a randomiser to the pitch of each frame and adjust the pitch range in which it will be active. This adds the human touch of imperfection to resynthesised sounds. An additional advantage over sampling is that the sound does not distort or exhibit aliasing anywhere on the keyboard.
I MUST SAY that I was quite impressed with the direct-to-disk system (or D-to-D as New England Digital call it). As you would expect, the D-to-D software is standard on the Synclavier Post Pro model, but can also be used with the 3200 and 9600 systems. The standard configuration includes eight tracks that can each record up to 25 minutes of sound at a sampling rate of 50kHz. This rate can actually be set to any value from 1kHz to 100kHz, which will also determine the maximum recording time per track.
Four 320Mb hard disk drives are required to support this configuration, each recording two tracks of digital audio. You can use fewer hard disks to record fewer tracks if your budget is tight. Standard configurations also include streaming tape drives for backing up that all-important data. Because each tape cartridge backs up about 18 minutes per track from each hard disk, you need eight tape cartridges to completely back up an eight-track system. A parallel interface between the hard disks and tape drives means that backing up takes less than real time.
By installing a software and hardware upgrade called MaxTrax, the system can record 12.5 minutes on each of 16 tracks at 50kHz. You could also add four more hard disks and record 25 minutes on 16 tracks. Other expansion options include additional hard disk drives that can boost the maximum recording time to 100 minutes on each of eight tracks at 50kHz. Keep in mind that each track is played by a Synclavier voice, so you must have as many voices installed as you have tracks.
The D-to-D system supports all the standard digital I/O (input/output) formats including PRODIGI, AES/EBU and SDIF (Sony Digital Interface Format) at 44.1kHz, 48kHz and 96kHz. If you intend to use these formats to transfer the digital information directly to or from the Synclavier, you must set your sample rate accordingly. It can also run concurrently with the 200-track sequencer described in last month's instalment.
THE D-TO-D SYSTEM is organised into three software pages: the Project Directory, Track Display, and Audio Event Editor. Typically, you start a new project from the Project Directory. This is where you can specify the start and end times for all tracks, set the sampling rate, and view various status parameters and recording time for each track in a project. You can manage up to 50 projects from this page, depending on their size. In a standard configuration, 50 30-second commercial spots are equivalent to 25 total minutes. The Project Directory also provides the means of retrieving and backing up projects to tape cartridge.
The Track Display simulates a normal eight-track tape deck, providing linear recording. There are several Track parameters including Title, Status, and Record mode. With these modes, you can monitor the playback from the disk or the signal coming into the system. You can also record several versions of a cue and play them in any order. This is used for sound effects, flying in vocals and so on.
The majority of the D-to-D system can be controlled from the Audio Event Editor. In fact, you can do everything of which the system is capable except tape backup. Unlike the rest of the current Synclavier software, this page is completely mouse-driven.
When you first enter this page, most of the screen is blank. But not for long. At the top of the page are several "buttons". When you click on one of these buttons, a "panel" appears on the screen that provides a specific set of controls. For example, the Project Manager panel provides many of the same controls that are found on the other two software pages. In fact, you can display several panels at once, all of which are active simultaneously. There are also four user-definable buttons that can each be used to call up a series of panels that you use most often.
"Resynthesis: the process whereby a sound is analysed and recreated - the Synclavier achieves this by dividing a sample into small time segments called 'timbre frames'."
THERE ARE THREE basic steps in producing a finished project: recording, editing, and assembling. Each panel addresses one of these steps. For example, the Sync panel provides time controls for the D-to-D system. With it, you can specify the time reference in minutes and seconds, measures and beats, feet and frames, or SMPTE time. You can also set the SMPTE format and offset as well as compute event times, generate SMPTE, and display the offset.
The Record Control panel allows you to record cues directly. There are three record modes that allow you to record on the next available free disk space, punch in to a selected cue without affecting adjacent cues on the disk, or take full manual control of a recording. This panel also allows you to start and stop recording on a disk track at a specified times, compress or expand the time of a cue without affecting its pitch, and transfer a Synclavier sample directly to a disk track.
A list of the cues in a project can be displayed in the Cue Directory panel. Not only that, each cue can be heard by simply clicking on it. This is particularly useful for recording and auditioning several versions of a voice-over. For example, suppose that you're given a radio commercial for a car company that is to play around the country. While most of the spot will be identical for all areas, you must insert a different tag indicating the specific car dealer that local listeners should see to buy that new car. By recording the tags and listing them in the Cue Directory panel, you can play the spot over and over, each time triggering a different tag.
Another application of the Audio Event Editor panels is automatic dialogue replacement (ADR). As you may already know, much of the dialogue you hear in a film or TV show isn't recorded during the filming of the scene. It's dubbed in afterwards in order to have greater control over the acoustics and ambience of the dialogue in the soundtrack. As you watch the film or video, you find the moment at which you wish to begin recording. You can then trigger the system to record the dialogue onto a disk track at a specific start time. After recording several versions, you can trigger each one to play at the appropriate moment and then select the best one.
ONE OF THE primary advantages of disk-based editing is that it is non-destructive. Unlike tape, the raw material is never destroyed. If you don't like what you've done, you can always go back and start again. Also, the quality of the recorded sound doesn't degrade as you manipulate it, and it's much faster than tape editing.
In the D-to-D system, most of the editing is done from the Cue Editor panel in the Audio Event Editor. This panel includes a standard Macintosh horizontal scroll bar that allows you to "jog" through the contents of a disk track. The recorded material can also be displayed as a waveform. You select edit points by positioning the scroll box and clicking on the Edit button. This places a scissors icon in the edit window at the selected point. These edit points can be dragged to new positions with the mouse as well. After selecting a section of the material, you can perform all the usual edits including (but not limited to) cut, copy, paste, delete, fill, and slide (which allows you to move parts of the cue around). All this is accomplished without losing the synchronisation or length of the cue.
The Cue Editor panel also allows you to make microscopic edits, such as removing lip smacks or shortening pauses that are too long. You can even use it to transfer a portion of a disk track to the Synclavier sample memory. This is extremely useful for correcting the pitch of an out-of-tune solo or vocal line. The corrected portion can then be transferred back into the disk track.
ONCE THE CUES have been recorded and edited, they must be assembled into a finished spot. This is done in the Event List panel, which gives you access to the Synclavier sequencer tracks. The cues can be mixed with sequenced parts and are triggered to play at a specified SMPTE time directly from the disk.
Cues can be placed in a sequencer track any number of times in one of several ways. You can manually select a cue from the Cue Directory in real time as you watch the picture. You can also trigger a cue automatically by specifying its start time, end time, or any marked time within the cue. For example, if you need to place the sound of a plane that appears on the screen after its engine sound is to be heard, you can trigger it to play at the correct moment by marking and referring to the time at which it appears on the screen, rather than guessing at its position. Cues that are already synchronised can be placed automatically with their sync time as well.
Once a cue is placed in a sequencer track, you can slide its position by specifying a new SMPTE time or you can globally move groups of events. Once you're happy with the spot, you save it to a separate system disk with other sequences and system software. That way, you don't have to take up space on the D-to-D disks.
NED HAVE DEVELOPED an alternative hardware controller for the D-to-D system called the DESC (you might expect a cute acronym, but NED inform us that there is none - it's just a distinctive way of spelling the name of the furniture at which you work). This device provides a user-interface that's already familiar to audio professionals. Included on the DESC are "transport" and other dedicated buttons, as well as a jogger wheel. This provides complete control over the Post Pro system, including track assignments, muting, soloing, and full editing. Actions on the DESC are reflected on the Mac screen in real time.
AS WE ARRIVE at the end of our tour, it seems clear that the Synclavier lives in the stratosphere of computer music systems. However, as technology continues to increase in power and decrease in cost, other systems will begin to challenge the Synclavier in one way or another.
Of course, it will be a while before these other systems are likely to provide the same level of integration between so many varied functions. And the quality of anything produced on the Synclavier is undeniably excellent. With improvements to the user interface and more reliance on the Macintosh itself, the Synclavier will maintain its place in the heavens as the ultimate musical tool for some time to come.
The author wishes to extend his thanks to Sean Callery and Ted Pine of New England Digital for their help and patience during the research for this series of articles.
Gear in this article:
Feature by Scott Wilkinson
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!