Before I got involved with modern modular synthesizers, I worked in the video and film industry. Before that, I designed synths, samplers, software, and digital audio recorders for the likes of Sequential (Studio 440, among others), Digidesign (Q-Sheet), and Roland (DM-80). But if there’s one project people seem to know me for, it’s my involvement with the Sequential Prophet VS, and for creating what is now known as Vector Synthesis. Here is the story of their birth. Earlier versions are floating around the internet; this one is more complete, and with (hopefully) fewer typos.
It all started some time in 1985. We were still working on the Prophet 2000 sampler, and as resident “historian” (in other words, I had the biggest archive of synth magazines and manuals), another engineer asked me to explain how various instruments performed waveform crossfades.
I had finished discussing the Fairlight, and had moved on to the PPG – explaining its wavetables, and the ability for it to scan a group of waves first in one direction and then back again. While I was scrawling this back and forth motion in my notebook to demonstrate, suddenly a little twinge went off in the back of my head, and my hand drew the next line arcing down the page…and the concept of crossfading between waves in two dimensions, not just one, was uncovered.
Trying to imagine what it would sound like to wander around a space with different waves in each corner was a little beyond most people’s grasp at the time. So, I decided to patch it up at home on my rag-tag modular synthesizer.
Prototyping the Patch
For some reason, I was originally convinced that all four waveforms had to be at exactly the same frequency, with their only differences being their waveshape (and therefore, timbre). I plunged into my already heavily-modified Oberheim TVS 1-A (Two Voice Synthesizer – a pair of old SEM modulars, a sequencer, and a three octave keyboard in a road case) and found a way to sync all four oscillators. I then dialed up four different waveforms by playing with the waveshapes, sync tuning, and filters. Since the TVS has only two filters, to create more variations I used my hand-built linear-based PAiA modular for the other two, employing a Korg MS-02 exponential-to-linear control voltage converter to make sure they all tracked the Oberheim’s keyboard together.
I then patched the four resulting waves into VCAs contained in a custom cabinet built by Gentle Electric (pictured at right), using the various control voltage mixers built into it. To help keep the correct correspondence between VCAs (i.e as one goes up, it’s opposite goes down), I employed the inverters and bias offsets in a Dennis Electronics Control Voltage Processors to perform the crossfades. The timbre mix was animated by envelopes from a Sequential Model 700 programmer and an LFO patched in from the Oberheim. This was all mixed together and fed into a final VCA, which was also controlled from the Oberheim. In the end, one voice took up almost my entire collection of analog synths, spread across modules from six different manufacturers.
(I’ve read comments elsewhere to the effect of “oh, someone probably patched this up years ago on a Buchla or something and just didn’t make a big deal out of it” – maybe; but it would have required a pretty large system. The trick is not just in mixing together four oscillators, but carefully controlling their blend so the sum always added up to 100% regardless of the modulation etc. Today, crossfading oscillators such as the Synthesis Technology “Morphing Terrarium” perform this crossfading, sometimes in 3 dimensions…but they miss out on the fundamental feature that the user gets to pick the four waveforms the oscillator fades between, rather than the sequence of waves being pre-determined when the module was designed.)
It took about an hour to learn the patch. Trying to mentally map a pair of ADSRs into imaginary two-dimensional space while my hands probed around a maze of wires…lets just say it took some thinking. Yet even on this limited system, some pretty interesting timbres started to emerge – struck attacks that faded into shimmering flutes; clarinets that opened up into raw sawtooths, etc. I took this tape into work and played it for a few engineers, and after ribbing me about my bad playing, they got excited.
At this time no back-of-the-building engineers were part of the product planning meetings at Sequential. But as luck would have it founder and pioneer Dave Smith was talking about starting work on Sequential’s first digital synth, and trying to find a technology that would allow the waveshape to be changed in realtime. Just like people say, timing is everything.
As interest started to build, I started lobbying Sequential’s marketing department to consider this new idea. I played them a tape of The Fixx (where the keyboardist plays a Prophet 5 and a PPG Wave), and boldly declared then that we could design one instrument that could make all of the sounds. (The whole issue of what exactly the VS would end up sounding like finally reached a head when the director of marketing demanded – well into the project – “Can it make the sound of a DX7 slap bass? If it can’t do a good slap bass, I want out of the project right now.” We just laughed at him and said “We don’t know what it’s going to sound like!!!” Thankfully, the VS didn’t let us down.)
The team of primary engineers became myself, Josh Jeffe and Tony Dean (the latter two eventually moved on to E-mu Systems). Tony ended up being the main hardware engineer on the project, and Josh the main software engineer and project manager. We spent a lot of time trying to refine what I was calling “The Diamond Patch” (so-called for its brilliant, shimmering sounds…and other reasons I’ll explain later), and figuring out how to design the hardware. Keep in mind that none of us had any background in DSP or psychoacoustics; we had to make it all up from scratch. As it turns out, some of our ignorance paid off in taking paths that textbooks would have told us to avoid – for better and for worse.
Without any background, Tony and I theorized how to do a wavetable synth, and came up with two strategies – either have a waveform of short, fixed length and vary how fast you played it back, or start with a very long version of a waveform and skip samples in it to alter how long it took you to read it out once (and therefore alter your final pitch). Unable to decide, we acquired a PPG Wave 2.3 and a Korg DW6000 to see how they pulled it off. We felt vindicated to learn that the PPG did it exactly the first way we theorized, and the Korg the second. (In the process, Josh became one of the few people in the US who knew how a PPG worked internally. The company didn’t release schematics, advising would-be techs “Don’t loose (sic) time – return to factory immediately!” if they had a problem). We liked the pure balls of the PPG better (plus the scheme that used shorter waveforms meant we could stuff more waves into the instrument), so we chose the transposition method.
This choice is the one place our ignorance paid off the most. Transposing a waveform way down in pitch causes strange upper harmonics – images of the original harmonic pattern – to appear if not “properly” filtered. These images are what gives the PPG (and VS) its brightness as well as vicious bass. I was raving for months (much to the bafflement of Josh and Tony) in particular about the sound a PPG makes where only the lowest and highest harmonics seem to be present. One day, as I walked into the lab while Josh was testing the VS hardware, I happily exclaimed “That’s it! That’s it!! That’s the sound I mean! What’s the waveform?” It turned out it was only a sine wave, but it was tuned so low that the image of its one fundamental harmonic was audible as a very high, airy harmonic. I’ve heard many say they feel the Korg Wavestation – which uses a form of sample rate conversion to do the pitch shifting, rather than varying the sample rate – lacks the bite of the “imperfect” Prophet VS.
(For an interesting side excursion, read this forum thread to hear how John Bowen decided to design the wavetable section of his Solaris synth based on our experience developing the VS.)
Detuning and Mixing
Working on a sampler (the Prophet 2000) at the same time we were developing the VS really tuned our ears to the difference between natural sounds and synthesized ones. Most synths of the time went through some interesting gyrations with their envelopes, but then stopped any harmonic development when they hit their sustain stages. Also, real sounds have a lot of harmonic phasing and beating going on, which are too complex to recreate with just a simple LFO.
The first major change to the Diamond Patch was to allow the four waves to be separately tunable. Detuning provides a type of motion and beating that no amount of timbre variation seems to be able to replace. Josh tried all sorts of algorithms to simulate detuning in oscillators that were actually synched together, but got all sorts of weird artifacts for his troubles. In the end, it was best that we ended up with four independent oscillators per voice. (Speaking of sync, we considered that, but Tony didn’t feel he could guarantee that it would work reliably in the custom chips. Oh well…)
Figuring out how to mix so many oscillators together and present them in analog to the rest of the voice chain ended up being one of the weak links in the VS’s hardware design. I wanted to keep all 4 oscillators per voice in tune, to save on mixing issues: add the four waveforms digitally, shift two bits off to divide by 4, and feed down the normal analog chain. The new detuneable plan required the oscillators to run at different rates, so this simple digital add-and-divide scheme would not work.
The first idea for the VS oscillator mixing was a VCA per oscillator per voice, but that was too expensive. (Remember, you don’t need just the chips, but support circuitry per chip as well – all of which gets multiplied 8 or more times by the time you reach the retail list price.) Next, we explored TI DSP chips or more 68000 CPUs to mix digitally, but that didn’t work out either expense-wise: I remember we ran the clock cycle numbers, and felt it would take more than one DSP chip to do all the oscillators. Remember that chips were a LOT more expensive back then than they are now (slower, too); a Mac with a 8 MHz 68000 was a $2500 computer then (in mid-80s $$$), compared to now. It was hard to get TMS chips down to the low $20s each. And they needed all the support RAM, ROM, and chips. Plus no one in house had programmed a TMS chip before.
As I recall, Tony then thought he could fake VCAs with a bunch of sample and holds, varying the reference voltages to the holds, to get the mixing action. We talked Doug Curtis (yes, of Curtis filter fame) into making the S/H chips, while Tony did the digital oscillator clock chip. Alas, as I remember it, the oscillator chip came back with a bug, plus – without assigning blame – the analog S/H chips did not exactly sound pristine when married to our digital hardware design. Remember: our design required 32 oscillators, with a high enough clock master frequency to integer-divide down into several octaves of equal temperament tones at audio rate. Darn high for analog back then.
The tuning issues with those divide tables is another discussion: Josh went for dithered (averaged) division cycles rather than bad intonation on the high notes, resulting in all those sidebands you hear when playing the VS in higher registers. Tony had all sorts of problems with crosstalk etc. There was a time when we thought we would need another silicon run to make it work (devastating, given Sequential’s weakened financial position at the time), but they found a kludge to sort of make it work, in exchange for more noise and distortion (and parts cost) than we intended. As it turns out, those chips have been one of the most common points of failure inside the VS.
Refining the Voice
Back to happier subjects: Those wonderful envelopes. The original inspiration was the looping envelopes in the Buchla 400. I had spent a little time with that particular beast (Sequential briefly considered licensing it from Don), and enjoyed the extra motion I could get from its independent, ten-stage, loopable envelopes. Each envelope became a sort of complex LFO, operating at a frequency that was different than all the other envelopes and LFOs. This came the closest to imitating the almost-random motion we were hearing in real sounds.
Josh and I argued over the number of steps in the envelopes for some time: I wanted more steps than the normal ADSR, but he didn’t want it to be too complex for the musician to program. He finally came up with the idea of a five-point envelope, where the adjustable point “0” that the attack phase started from could be used as an instant-on attack without wasting a whole stage with the envelope rate set to zero. The envelope could then loop through these early stages, which acted as an additional complex LFO. I wish more envelopes copied our design today, including the back-and-forth looping that gives a smooth result – most if not all “looping” envelopes today merely re-trigger themselves when they finish, which is far less interesting that transversing the envelope shape back and forth. (For the record, most of the stages of the Prophet VS’s envelopes were linear, to make them easier to calculate; just the release was exponential.)
Scott Peterson (my immediate boss, and one of the original Sequential employees) came up with the idea of the joystick to control the waveform mix. This proved to be a lot easier and more intuitive to program than the discrete X and Y axis envelopes I used in the original Diamond Patch. It also fit well with the idea to include user waves in RAM – the user could then easily dial up a mix with the joystick, and press a button, and save their own custom waveshape. We felt this feature would really make the VS a timeless instrument, since the user could keep renewing and personalizing the instrument as they went on. Alas, this has never really been exploited until the advent of the Wave-O-Rama portion of the Interval Music’s long-since-discontinued VS WaveWrangler patch program.
By the way, the influence of the original name “Diamond Patch” has ended up being a curse for all implementations of vector synthesis to this day. It would be much easier to use and internally program if the four oscillator positions were the corners of a square. But I didn’t like the name “square”; I thought it was too boring and uninspirational. An unintended result of my new name was that Josh felt the name “diamond” demanded that the user interface arrange the oscillators and the corresponding joystick like a diamond – which everyone has followed ever since.
So, how did it end up being called the Vector Synthesis instead? Well, in truth, no one but me was really thrilled with the name “Diamond” or the term “Diamond Patch.” We also didn’t want to give it just another number – although its project code was 2400, and the “Prophet 2400” was indeed considered. Josh’s creativity again saved the day. He was sick one week with a bad cold, and was taking a lot of Sudafed Plus one day to try and get over it. While in this sick+medicated state, Josh somehow got the idea to call the process of defining the waveform mix “vector synthesis”. We called Dave Smith in Hawaii (where he was preparing for another Iron Man tournament), and surprisingly, the incredibly sick Josh had little difficulty convincing the incredibly fit Dave to go along with it.
The original waves for the VS were created three ways: extracting single cycles from sampled sounds, using a custom additive synthesis program, and using a program Josh slapped together called “Hacker” where you could draw the waveshape. These were fed straight from the computer through the filter and VCA of a Pro-One to figure out what the result might sound like in a patch. The lead engineers plus Sequential patch master John Bowen contributed to the creation of the waveforms; there may have been others as well.
Contrary to some rumors, no PPG waveforms appear inside the VS. We had access to them from their ROMs, but in the end our consciences got the better of us. We did attempt to recreate the harmonic structure of some of the waveforms from the Korg DW6000, but only by looking at the harmonic drawings on the front panel and trying to imitate them in our additive synthesis program – we did not resample them or copy the ROM.
The very first patch I remember coming from the VS was created by Josh. It was pretty simple – four disparate waveforms in each corner, with one LFO panning between A and C and the other panning at a different rate between B and D. The resulting extreme washes between radical harmonics and simple sine waves ended up being an example of the signature sound of the Prophet VS.
The Random Patch feature came from another Dave/Josh exchange. In the lab, we had one prototype VS wake up brain dead. In the process, it scrambled its patch memory, and was producing some of the most bizarre sounds as a result. Dave – an aficionado of bizarreness – therefore wanted a patch randomizer built in, but Josh was against it. Finally Dave promised Josh a six-pack of his favorite beer if he would put it in, and in one weekend Josh slapped it together. (Ah, the scientific process…) Some of the factory patches in the 90s are the result of this random patch generator.
What did we leave out? We briefly considered adding sampled attack transients in front of the waves, but ruled it out for lack of RAM. (I, for one, was sick in my stomach when the Roland D-50 came out with sampled attacks). Sync, velocity to attack rate, and another stage in the envelope are on my short list of additional features that didn’t make the cut.
One of the most despised features of the VS is its aftertouch sensors. Remember that the VS was Sequential’s first keyboard with mono aftertouch (the t8 had poly aftertouch – expensive), and we didn’t quite get the hardware design right. First off, the VS’s were always sitting on large flat benches when we were working on them, and quite often sitting on a flat table in the demo room. Unfortunately, The VS’s flat steel case wasn’t stiff enough – we didn’t know to rib it for strength; it would bow if supported by anything other than its feet. The problem with the case bowing causing aftertouch to be on all the time didn’t show up until the VS appeared in users’ hands. If you know what the problem is, you can slide the arms of a keyboard stand out very near the feet, or even add a board underneath to give it a flat surface. (I “fixed” the Stevie Wonder’s VS during a trip to LA by telling his tech about the arms.)
Second, the first generation of the sensors themselves were prone to failure. Again, the mechanical hardware is out of my sphere of knowledge, but I remember that the problem had something to do with a combination of too small a pressure point from the keyboard and too high a voltage through the sensor – it burned them out. Replacing the sensors, with I believe a mod that used a lower voltage, helped. There’s a trimmer inside to tweak that as well. Still, I’ve got one broken key on my VS from pushing down too hard to get the aftertouch I want (I would even wrap my thumb underneath the metalwork and squeeze to get more force!). Sequential didn’t get it all that much better on the Studio 440; sensors wear out in those, too.
Ironically, neither Josh, Tony, nor myself worked on the follow-ups from Yamaha (SY-22; TG-33) or Korg (Wavestation). Still, it is a point of pride that Vector Synthesis enjoyed a few years in the sun, and that Prophet VSs are one of the few instruments that didn’t really had a big dip in value in between being “new” and being “classic.” I’m proud to have been a part of it.
originally printed in the VS WaveWrangler User Guide, (c)1991 Interval Music Systems; updated 7/99, 6/01, 10/13, 5/15, and 1/17 by the original author