source file: mills2.txt Date: Wed, 20 Dec 1995 13:02:56 -0800 From: "John H. Chalmers" From: mclaren Subject: The collapse of innovation in post-1988 synthesizer technology --- In a previous post, Your Humble E-Mail Correspondent mentioned the general excellence of Ensoniq's synths. They sound about as good as anything else out there, Ensoniq synths are rock-solid reliable, and their sampler operating systems are particularly intuitive and easy to use. Anyone who battled the hellish TX16-W Yamaha sampler operating system or the botched E-Mu sampler OS's from the late 80s is well qualified to appreciate the excellence of the EPS/ EPS-16/ASR-10's operating system. However, there's still plenty of room for improvement in the Ensoniq line. Someone posted a query about that--how can anyone say Ensoniq isn't up to date technologically? Here's how: [1] The ASR-10 needs more RAM. A *LOT* more RAM. Right now, the Kurzweill 2500 series can take up to 128 megs. This is a reasonable minimum amount: 256 megs would be more like it. But 128 megs is a start. The ASR-10, by contrast, is stuck with 16 megs. This is around 90 seconds of sampling time at 44.1 khz stereo. Completely unacceptable. Far too small an amount of RAM to be useful. Part of the problem is the kbd controller chip Ensoniq uses to address the RAM, part of the problem is the burden of maintaining backwards compatability with the EPS/EPS-16 &c. Backward compatability must go. The ASR-10's successor needs more RAM. A *LOT* more. With EDO, the people at Ensoniq need to start thinking in terms of *gigabytes* of RAM. (As always, readers will call this "insane" today, "a little over the top," in 6 months, and "very sensible, but somewhat conservative" in a year.) [2] A rule of thumb is that a decent saxophone or clarinet needs 25-40 multisamples. The ASR-10 allows 8 layers per instrument, 8 instruments total. This is utterly inadequate. Backward compatability must go. Dump the 8 layer limit. At least 127 A-B crossfades should be allowed per multisampled instrumnet, at least 16 MIDI channels/instruments at a time. [3] David Doty made a point about accessing various layers during performance. Clearly Ensoniq needs to upgrade the ASR-10's successor to allow MIDI access to each of the 127 layers within an instrument. Since these would often be used for alternate tunings, it's a particular priority. [4] Ensoniq may want to think about implementing some new technology. Ever since innovation ground to a screeching halt in the synthesizer industry, somewhere around 1988, industry pundits have wondered why sales of digital keyboard instruments have dropped steadily. There's no mystery. The Korg M-1 provided the original and ever since then all the keyboard manufacturers have concentrated on cranking out endless xerox copies of that instrument, all using exactly the same antique technology: sample playback. With the exception of the Yamaha VL-1M/VL-7 and the E-Mu Morpheus, all current synthesizers are nothing but sample playback machines that spit back digital recordings burned into ROM when you press a key. Now, there's nothing wrong with sample playback. It's a fine technology. But after a while, you get tired of hearing nothing but sample playback. Even today's samplers use exactly the same technology-- with the only wrinkle being that *you* get to choose what digital recording is regurgitated when you press a key, instead of *the synth company** choosing the sound. All today's synths are basically nothing but digital Mellotrons. Where's the synthesis? Does anyone remember the origin of the term "synthesizer"? These instruments are supposed to *generate new sounds.* Instead, we get yet another canned B-3 sample burned into ROM. And no matter how mich reverb, phasing, ring modulation, flanging or delay you slather on top of a sound burned into ROM, it all sounds pretty much the same. Around 1988, synth companies stopped making synthesizers. Instead, they all followed the cattle stampede toward the easy dollar. The net result is that there is today hardly any reason to buy one synth from one manufacturer rather than another. Ensoniq's tuning tables make a difference. But if they *really* want to increase sales, how about building some actual synthesizers for a change? Even Yamah has dropped the ball. Today, if you want to buy an FM synth you're out of luck. You have to buy one used, or pay for a Kyma. The brutal reality is there are *dozens* of synthesis methods: digital additive, subtractive, frequency modulation, amplitude modulation, Chebyshev distortion, Miller Puckette's formant synthesis, Lansky's LPC synthesis, phase vocoder analysis/ resynthesis, Daubechies wavelets, Walsh function analysis/resynthesis, Dashow's exponentiation synthesis, Hiller & Ruiz's physical modelling synthesis, waveguide synthesis, many others. Yet aside from Yamaha's VL-7/VL-1M physical modelling synth, not a single manufacturer has implemented *any* of these synthesis techniques in a currently available commercial synthesizer. Amazing. Shocking. Yet true. If Ensoniq wants to jump-start synth sales, they might think about implementing some of these synthesis techniques. Now that Korg's OASYS synthesizer has been discontinued--yet another case of classic vaporware--and the Gibson/G-WIZ labs' FAR Fourier resynthesizer cannot be purchased by anyone, anywhere, for any reason, at any time, any way, shape or form...well, now that these vaporware instruments have bitten the dust, what else is there on the horizon? Zero. Zilch. Zip. Squat. Diddly. Nada. These synths have joined the parade of vaporware formed by the Prism synthesizer (remember that one?), the additive synth built from the Amiga's Amy sound chip (256 additive partials--it gobbled the Amiga's entire CPU and memory so the company dumped it and licensed the rights to a start-up which was promptly sued out of existence), and the marvellous Technos 16pi...a synthesizer which, if it had ever existed, would have been superlative. Well, chances are this is all blue-sky fabulation. Chances are Ensoniq won't bother to actually build synthesizers. Too much work. And the MR rack tends to bolster that viewpoint. Yet another wannabe sample-playback box, yet another digital Mellotron. It's ironic that Ensoniq is giving up the opportunity to crush the Japanese synth companies. What with their little Yen crisis and teetering financial system, this is an ideal chance for American synth companies to steal back the initiative that was lost when the Japanese licensed FM technology and ground the American analog synth manufacturers into the dirt back circa 1983. In any case, these remarks should be understood inthe context of making Ensoniq's excellent products better. Rather than angering the engineers and management at Ensoniq, perhaps these words will irk them into improving already fine synths. N.B.: Even though the Kurzweil 2500 series offers gobs of RAM, the sampler does *NOT* have a full-keyboard tuning table. Thus my next sampler will be an ASR-10. Also, Dave Rossum at E-Mu needs to take a look at Ensoniq's multiple tuning tables and realize the *immense* importance of more than one tuning table. JI compositions which change key centers demand multiple tuning tables, as does work with (say) a Wilson 70-note hebdomekontany in only 128 MIDI notes. Allen Strange has mentioned that he gets only 3 octaves of Partch's 43-tone just scale in MIDI's 128 notes; using the same timbre on 3 MIDI channels tuned 3 octaves apart would increase his range to 9 octaves. And, as usual, Ensoniq is the *only* current manufacurer to support multiple tuning tables. Thus, for many microtonal applications, Ensoniq synths remain the only real choice. --mclaren Received: from eartha.mills.edu [144.91.3.20] by vbv40.ezh.nl with SMTP-OpenVMS via TCP/IP; Wed, 20 Dec 1995 22:57 +0100 Received: from by eartha.mills.edu via SMTP (940816.SGI.8.6.9/930416.SGI) for id NAA26137; Wed, 20 Dec 1995 13:56:54 -0800 Date: Wed, 20 Dec 1995 13:56:54 -0800 Message-Id: Errors-To: madole@ella.mills.edu Reply-To: tuning@eartha.mills.edu Originator: tuning@eartha.mills.edu Sender: tuning@eartha.mills.edu