The Synclavier Story (Part 1)
As Fairlight fight for survival the mighty Synclavier undergoes a dramatic upgrade. Chris Meyer takes us on the first of three tours of the new state-of-the-art Synclavier.
Twelve years after its launch, the world's most sophisticated sampler/synthesiser/digital recording system finds itself in its fourth revision - adapted for dedicated post-production and Macintosh-based systems.
THE LAST THING anybody could ever blame New England Digital for is planned obsolescence. The original Synclavier had a multi-button front panel and continuous rotary controller just like the top system today, and all along the way NED have tried to make upgrades as smooth as possible for existing owners.
Although this has been an enormous boon to those faithful early customers, some of this dedication to evolution has arguably hampered the Synclavier. Add-ons often meant using the existing buttons and functions in non-intuitive ways, and the hardware has sometimes taken on an ungainly appearance. Sampling was particularly hard to add, because the original system was synthesis-based. Also, sampling was only monophonic for a couple of years before polyphonic sample playback became available. However, the new systems are comfortably modular and allow the Synclavier to be neatly split into multiple systems and easily expanded.
The main control processor, or the brain of the system, is the NED-designed "Able" 16-bit computer, better known to some as the "Model D". Based on a RISC (Reduced Instruction Set Computer) architecture, the latest incarnation of the NED CPU features 256 times the memory capacity of the previous Model C, and the ability to keep track of up to 396 voices and 128 channels of MIDI data before losing track of timing (the Model C could only handle 40 voices). It runs the control programs and manages the user interface. Conversations with the various pieces of user interface hardware - the Synclavier keyboard, Mac II graphics workstation, the new Controller/Editor/Locator hardware interface, the MIDI module, the guitar controller interface and so on - is performed over RS422 serial links. These links (called "DAWN," for Digital Audio Workstation Network) pass information such as motion control, system status, and file management from one part of the system to another.
"Input anti-aliasing filters are not the oversampling digital variety, but good old inductors and capacitors, which don't add nearly as much noise."
Using its current software version, the Model D actually draws almost all the graphics you see appearing on the Mac IIs high-res colour screen. As part of the "let's not alienate the old users" program, a graphic representation of the hardware button-and-rotary-controller panel is even recreated on the Mac, and can be accessed via trackball (this is the only screen currently being controlled by the Mac). Future plans include putting more of the user interface burden on the Mac II, but all the actual sound production will stay on the Synclavier side.
Using a Mac II allows Mac-based MIDI programs to run concurrently with the Synclavier under Multifinder. These programs can drive the Synclavier by running MIDI cables from the Mac II to the Synclavier itself. Unfortunately, applications running on the Mac II cannot take advantage of the Synclavier's built-in SMPTE and VITC synchronisation capabilities (see Time Exposure feature elsewhere in this issue). Although there are no current cross-links that I'm aware of, NED are hinting that Mac-based applications will be able to edit Synclavier sound files and exchange sequencer files eventually. The Mac II comes with all systems (3200, 9600, and PostPro) complete with 2Mb of RAM and a 20Mb hard disk. The 3200 comes with a 16" monitor; the 9600 and PostPro come with a 19" monitor.
The main sound processor is the Polyphonic Synthesizer, whose purpose in life is to play up to 32 channels of sampled audio out of RAM - synthesis requires a different card. Sampling voices can be added four at a time, synthesis voices (a combination of additive and FM) can be added eight at a time, and RAM can be added 4 or 16Mb at a time up to a maximum of 32Mb for the 3200 and 96Mb for the 9600. A 3200 can have up to 32 sampling voices, but cannot have any synthesis voices. The 9600 can have from 32 to 96 sampling voices and from zero to 32 synthesis voices with a combined total of 96 sampling and synthesis voices max. Multichannel outputs may be added eight or 16 at a time for the 3200 and 9600. The PostPro comes with eight audio outputs that can be expanded to 16 in groups of four.
RAM cards for the Synclavier used to be outrageously priced (over $10,000 in the States for a 4Mb card). Not surprisingly, bootleg memory cards started to appear. All was fine for the bootleggers until someone left a bootleg card in a machine sent to be serviced. This discovery was followed by a stern memo from NED that the next revision of software would self-destruct a system if it detected an alien board. No system ever did away with itself, but by the same token, official RAM boards have come down drastically in price since.
Mass storage is truly massive for all systems. The 3200 comes with an 80Mb hard disk; the 9600 comes with a 320Mb drive. Both can handle multiple 320Mb hard drives, a Kennedy tape drive, and a 2 gigabyte WORM (Write Once/Read Many) optical drive. The PostPro comes with an 80Mb drive for the "system", and a minimum 50 minutes of record time at 5OkHz/16 bits (we'll get more into that in, due course). Connection to the external storage is done over SCSI. Connection between the control processor and the two central audio processors in the PostPro system is also performed over SCSI. You can chain hard disks out to the limit of SCSI.
"The time compression algorithm seems to have been created more for dialogue editing or as an abstract exercise in data compression than for music."
An additional processing board for the Synclavier has recently become available - the DSP card. Using the Motorola 56001 DSP chip (as used by Digidesign, NeXT, WaveFrame...), it will be the driving force behind future sound modifications. The first application - time compression - is described below.
HOW DOES THE outside world speak to these units? For starters, the 3200 and 9600 come equipped with a 2 in/8 out MIDI card. This card may be added to the PostPro as an option. Also on the horizon is MIDInet - "an 8 in/8 out MIDI processor featuring simultaneous processing on all channels with any combination of available algorithms (filtering, channelisation, echo, transposition, merging, keyboard mapping and scaling) married to a 16 in/16 out patchbay". You can stack as many of these 16x16 patchbays as you like onto a system. Considering that they're touting the 3200 as the ultimate MIDI control centre, this will be a welcome addition.
The 9600 comes with the famous 76-key piano-response, poly-pressure keyboard first seen in the Sequential Prophet t8, along with the traditional button panel and rotary encoder carried over from the very first Synclavier. For those wondering what NED were going to do for a keyboard now that Sequential are no more and Pratt-Reed don't make synth keyboards anymore (well, I was wondering), NED inform me that they have recently been granted a patent on a new keyboard design that looks the same on the outside, but is totally different on the inside. Look (inside, I guess) for it this Autumn.
The Synclavier has synced to SMPTE timecode for some time now, and after some initial timing problems with their sequencer, they report that their SMPTE interface has recently won a Monitor Award for technical excellence from the ITS (International Teleproduction Society). A VITC SMPTE interface is also now available that will allow video editors to crawl frame by frame through a paused video with the Synclavier in tow.
Fine - but how do you get sound in and out of these beasts? NED pride themselves on having some of the highest sound quality in the world. You can sample in mono or stereo in 16-bit linear format at almost any rate 1-100kHz. Audio input enters all three systems through a custom Analogic ADC that runs at up to 200kHz - two channels are multiplexed through one ADC. Input anti-aliasing filters are not the latest oversampling digital variety, but actually good old passive inductors and capacitors, which don't add nearly as much noise as op amps, and rival the low distortion levels induced by some digital designs. The output DACs are 16-bit linear with a 12-bit amplitude envelope for pretty smooth dynamic shaping. There are no output filters of any kind for sound reconstruction or alteration. Apogee input filters (the current favourites in the digital multitrack world) will be offered as an option "in the imminent future".
"The main sound processor is the Polyphonic Synthesizer, whose purpose in life is to play up to 32 channels of sampled audio out of RAM."
Why bother with sampling rates higher than 48kHz? Well, you actually gain something by distributing sampling noise and distortion over a wider (and mostly inaudible) frequency range. Also, it takes some of the load off the anti-aliasing filters, and it means you can transpose downwards over an octave without having to worry about imaging and clock noise. Yes, the Synclavier is still a variable-clock system, as opposed to the newer fixed-clock systems like the WaveFrame.
Because voices running at variable clock rates cannot be mixed cleanly in digital, submixing is actually done in the analogue domain. You can buy "multichannel outputs" in increments of 8 or 16 voices for the 3200 or 9600. The PostPro comes with eight standard and can add outputs in increments of four. There is also a digital I/O card available that complies with SDIF (Sony 630), ProDigi (Mitsubishi), and AES/EBU standards. If the PostPro is locked to SMPTE timecode and house sync (a very stable clock that runs at twice the frame rate - common equipment in all video studios for aligning equipment) and the house sync speed changes, the speed of the digital transfer stays in step.
One curious aspect of the new Synclaviers is how stereo imaging is done on the 3200 as opposed to the 9600. The 9600 has an analogue panning circuit after the DAC - nothing special. What's different is that the 3200 actually uses two voices (one left, one right) playing identical information to create a stereo field. The relative loudness of the two voices determines the stereo position. The two voices are locked tightly enough to avoid phasing problems. This approach eats up voices twice as quickly, but some users have actually reported that the 3200 sounds better since the audio signal is going through one less analogue stage (the panner) than the 9600 on the way to the outputs.
So is digital mixing out of the question from NED? They're rather sly about the answer, but no - when asked at a recent AES show where their digital mixer was, they replied to the effect "we have no related product announcements at this time". A more recent reply was "New England Digital are developing technology to address all aspects of an integrated digital recording environment". We'll just have to wait and see.
"The 9600 Synclavier system comes with the famous 76-key piano-response, polypressure keyboard first seen on the Sequential Prophet t8."
ALL THE HIGH-END samplers I've worked with or seen demonstrated do not have sample editing features as sophisticated as the earliest versions of Sound Designer, Sound Filer, or Alchemy - let alone the most recent versions. Sound Designer itself came into being because the main people in Digidesign (then called Digidrums) looked at the Fairlight and were appalled by the relatively primitive sample editing features. Such is life.
The Synclavier's sample editing isn't actually all that bad. You have a wide range of view resolutions, a scrub wheel for locating sections of the sound, and all the usual reverse, cut, paste, extract, delete, exchange, combine, mono, crossfade, copy, fill, loop, scale, normalise, reverse, mix, invert commands you'd expect from a stereo sample editor. You can even trim DC offsets that may have existed in the original sample. All previous edited versions of the sound are stored in sample RAM as long as there's memory available.
Nonetheless, some things are still a little clumsy. For example, you have to type in values for some edit functions and you cannot audition a loop before committing to a crossfade or discard the remainder of the sample past the end point. The Synclavier has an autoloop-detect function, but the manual only claims 50% success. If you don't get the loop points just right on sight, you have to retrieve the last version of the sample and try again. This is an area where further Mac-based screen developments will be very welcome (let's hope the Synclavier can eventually cope directly with Sound Designer or Alchemy).
Sounds are stored in directories that more closely resemble the IBM school of file management rather than the friendly Macintosh interface. Soundfile names are also pretty short and semi-cryptic. On the other hand, search functions are available and whole sets of multisamples can be called up with one name. In other words, the filing system is antiquated but functional and thorough.
As mentioned, a 56001-based DSP card is now available, but the only currently implemented DSP function is time compression. And it's an odd time compression algorithm at that. It seems to have been created more for dialogue editing or as an abstract exercise in redundant data compression than for music. (To be fair, many people use the Synclavier for dialogue editing and the like). As opposed to uniformly compressing or expanding the sound. it looks for silences, such a pause between words, or repeated data, such a vowel sustained too long, or a sustained portion of a musical sample. to remove. How much it removes and how large the spaces are between extractions are partly under algorithm control - there are three algorithms that are mainly differentiated by their speed and precision - and partly under user control, such as the largest segment to remove, and a guess at the pitch of the sound. Since a DSP board was not installed in the unit I got to play with, I didn't try this out.
The next application being developed for the DSP card is sample rate conversion - already a feature of higher-end sample editing software and a useful function of E-mu's EIII. Hopefully, other applications such as digital filtering are to follow.
AS I AM typing this, I have beside me a 1981 ad declaring the Synclavier II to be the most powerful synthesiser ever made. Kind of quaint, when you think that today the Synclavier is primarily a sample-based system. But when you start to work with one for a while, you realise that the old synth framework proved to be a good home for samples too. I confess I'm as quick as anybody to call the Synclavier a dinosaur. But I'm also willing to admit that all sampler manufacturers could learn from NED when it comes to animating a sample upon playback.
Each key can play up to four samples (previously known as 'partial timbres"). Each sample can have a different velocity response curve and range. All partial timbres are started in perfect phase synchronisation, making things such as velocity crossfades really workable. The lack of phase locking on most semi-pro samplers results in horrible phase cancellations if the sounds are similar at all.
Each partial timbre can be affected by a wealth of modulations. For starters, there's a standard ADSR volume envelope, with delay and real exponential decays - very natural-sounding. Attacks can be up to I5 seconds long; decay and release can be up to 30 seconds long. There are six vibrato waveshapes, including random, adjustable from 0-50Hz, and each can be inverted. Vibrato depth can be up to two octaves deep and it can be quantised to act in steps as opposed to a smooth glide. Vibrato attack can be up to ten seconds long. Tremolo (amplitude modulation) is just as versatile, and tremolo can be synchronised for effects like Leslies or vibes. Tremolo, by the way, is also a perfect example of how the old button panel has become non-intuitive - the Stereo Wave, Depth and Rate buttons access tremolo. More Mac screens...
Portamento can be up to a minute long, and is smooth.
Partial timbres (individual samples) can be tuned against each other in 0.1Hz increments. The Synclavier creates chorusing by playing two samples for every partial and allowing detuning between the doubled samples in FM-type harmonic intervals (frequency multi-plication factors) from 0.000-10.000. Stereo positioning of each partial can be specified and modulated by keyboard position or a number of LFO waveshapes - including the ability to synchronise panning motions to keystrokes. And yes, you can set up the nominal volume levels for each partial, along with positional crossfades across selectable sections of the keyboard.
All this would be fun even if you couldn't modulate them in real time. But you can. Called "real time effects", you can patch velocity, pressure, either of two pedals, mod wheel, ribbon controller (9600), breath controller (9600), and keyboard position to modulate the envelope parameters, partial tuning and volume, all vibrato and panning parameters, portamento rate, chorus depth, envelope amount, and even the parameters of the Synclavier's built-in arpeggiator (even the Synclavier has a built-in arpegiator).
That covers the basics of the new systems. Next month we'll look into making some music with the Synclavier, and the following month we'll look into some of the more unique functions multitrack hard disk recording and FM/additive synthesis. May the Force...
Gear in this article:
Feature by Chris Meyer
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!