|
|||||||
|
|
|
|||||
|
|
|||||||
Look inside and you'll discover a mostly empty PC board. There's a connector for the wall-wart power transformer (if you've got a rack module), or a transformer and some big capacitors and an IC voltage regulator (if you've got a keyboard). You'll notice a monster VLSI chip in the center of the PC board, some RAM, a whole buncha ROM chips, and a processor to scan the keyboard and handle MIDI input and output. Then there's the DAC chip with some resistors and capacitors hanging off it...and that's about it. The average modern synthesizer contains fewer electronic components than your average high-end microwave oven. And almost as few as a universal infrared remote control.
'Twas not always thus. Open up an older digital synth, and you'll be shocked by how *many* chips there are. Pop open a Synergy II from 1982 and your eyes will bug as you survey several hundred ICs. Back in the early days of digital synthesis, all synths were built from what was known as "jellybean TTL." That is, you'd hook together some high-speed ALUs (arithmetic and logic units) and then perform all the other synthesis functions with hundreds upon hundreds of AND and NAND and OR and XOR and NOR gates... scores of hex buffers and counter chips &c. Gary Morrison built a synth this way from scratch, an accomplishment akin to building the Notre Dame cathedral out of Lego blocks. It cost ungodly amounts of design hours and skullsweat, but this approach to synth design had certain advantages that modern all-in-one-chip synth designs do *not* enjoy.
At the heart of the old digital synths used to be an antique 8-bit microprocessor, usually a 6809 or a 6501 or a Z-80. This microprocessor ran herd on the rest of the hundreds of chips and essentially told the entire synth what to do. The CPU would not only handle I/O (pre-MIDI, there were several competing early digital communications protocols for hooking one digital synth to another of the same brand, or to a digital drum machine, etc.) but it would also look several hundred times per second at every knob and button on the synth's front panel, AND the very same dinky CPU would also handle *all* of the synthesis functions. To a modern engineer this seems incredibly weird: the same dinky little 8-bit CPU not only sent information out and processed information coming in over a digital line from/to other synths, but the very same chip also calculated vibrato for wavetables and incremented phase counters to get frequency ramps from glissandi wayyyyyyyyy down at the waveform level.
Today, this no longer happens. One CPU chip handles "talking" to the outside world through MIDI or a computer connection & scanning the keyboard and knobs and buttons on the front panel of the synthesizer: but another entirely different and generally much higher-powered chip handles all the synthesis functions. So in today's synths, the same chip NEVER calculates vibrato or finds an increment to a wavetable AND handles MIDI, floppy disk I/O, front panel knobs, etc.
This has profound consequences for tuning. In the old synths because one CPU handled every function in the synthesizer, you could add new features by telling the engineers to rewrite the ROMS and, for example, add an 8096-part-per- octave tuning table. Provided you had enough free RAM (and in 1982, eight kilobytes of RAM was no laughing matter--that cost real money!) and enough space in the 16K ROM to burn new instructions, the engineers would indeed literally change the entire design of the synthesizer simply by rewriting the instructions of the ROMs.
But today's synthesizers segregate the functions of synthesis and MIDI. This means that the feature- set of the synthesizer is no longer under the engineers' control once the silicon chip at the heart of the synth has been designed and debugged.
If someone at Ensoniq says to the engineers, "We want you to add a 65536-part-per-octave tuning table to the new ASR-10," the engineers' response will be, "No can do. It's physically impossible without resdesigning the chip."
This is a particularly poignant example of how much we've lost as well as gained by moving to modern VLSI digital electronics. Synths offer many more features than ever before, they're less noisy, more reliable and contain many more sounds "burned in" at the factory...yet at the same time today's synths are far less flexible than those of the 1977-1985 period. All modern synths tend to have envelopes with the same number of segments, they all tend to use the same phase and ring-mod and digital filtering effects, and they all tend to have the same tuning resolutions because those limitations are built into the chips from which modern synthesizers are made.
Sadly, this applies even to synthesizers like Peavey's DSP-based line. You'd think that because Peavey's synths are built out of programmable DSP chips, the engineers could easily change the entire design of the synth: alas, not so. 80% of the cost of a modern synthesizer design lies not in the cost of building the parts, but in writing the software that does all of the hundreds of millions of additions and subtractions and multiplications and divisions per second that add up to a sound coming out when you press the key. Remember that every one of those additions and subtractions and multiplications and divisions must be PERFECTLY glitchless: one dropped bit, one bad sample, one mismatched level, and you've got a click that sounds like hell and sends your synth designs back to the drawing board to rewrite the operating code.
Believe it or not, one dropped sample is audible. It sounds awful. And that requirement of absolute perfection in handling the audio stream is one of the main reasons for the high cost of modern synths. Thus, even synths like Peavey's which use programmable DSPs wind up being boxed into a corner. Once a few hundred million man-hours have been invested in writing a glitchless and perfectly smooth-sounding set of routines for handling vibrato and portamento and velocity sensitivity, no engineer wants to entirely rewrite all that DSP code just to add a feature like a tuning table.
"If it ain't broke, don't fix it."
And when you realize that the most common DSP in audio use, the 5600, boasts only 20 bits after round- off, you quickly understand that even rewriting the entire operating system for the Peavey line of synths still wouldn't give you more than 12 bits of tuning resolution--that is, 4096 parts per octave. (Remember the rule of thumb: always count on throwing out at LEAST four bits of precision to rounding error.)This is a far cry from the proposed tuning resolutions put forward by Erv Wilson or the people on this forum: generally in the neighborhood of 10,000 or 20,000 parts per octave.
Alas, there's no prospect of getting this kind of tuning resolution wth any modern synthesizer. The economics and the technology simply prevent it.
--mclaren
|
|
|
|
|