You are here

DAVE ROSSUM: Inside Creative Labs

Interview | Manufacturer By Paul White
Published February 1998

DAVE ROSSUM: Inside Creative Labs

Creative Labs' Chief Scientist, Dave Rossum, looks into his crystal ball to predict the future of soundcards. Paul White listens in.

Dave Rossum's interest in synthesis began when he was a graduate student in biology and happened to help the music department unpack their new Moog synth. From then on he was hooked, and the following summer, he bankrolled a group of six students from UC Santa Cruz and Caltech to build a synthesizer. At the end of the summer of '71, his high school friend Scott Wedge joined him and the seeds of Emu were sown...

"Almost exactly 25 years ago, Scott and I got our business license to start selling the modular analogue electronic music synthesizer we'd been working on for the previous year. For the next 20 years I was Emu's primary technologist [Chief Wizard — Ed], and designed the Emu modular system, the Audity computer‑controlled polyphonic analogue synth, the Emulator I, the Drumulator, SP12 and SP1200. I also led the design team for the Emulator II and Emax samplers. I designed the G‑chip and H‑chip, which continue (as G‑chip 2 and H‑chip 1.6) to be the primary DSP chips in many of Emu's products.

We firmly believe that even gamers care deeply about sound, and multiple outputs are not wasted on them.

"In 1993, when Emu merged with Creative Labs, I was appointed director of the Joint Emu/Creative Technology Centre. I continued to occasionally contribute to new Emu products, particularly in the design of chips, but was primarily focused on building a team of researchers and engineers to develop advanced technology for Emu and Creative.

"In 1996, I was promoted to the position of Creative's Chief Scientist [still Chief Wizard! — Ed], where I'm responsible for directing Creative's technological strategy. That's a fancy way of saying I pretty much do what I want. And I continue to primarily focus on audio."

Card Games

DAVE ROSSUM: Inside Creative Labs

To date, Creative Labs soundcards have offered exceptional value, but seem in some ways to be hampered by the need to maintain backwards compatibility (and lowest prices) for the games market. Is this a real problem in designing a card that will also benefit serious project studio musicians?

"Backwards compatibility is less a technological burden than a marketplace burden. There's no technological problem with designing a card which can work superbly for the musician, yet carry all the features necessary to keep the gamer happy. But the volume sales associated with such a card would be much lower than the majority of Creative's products, because to give the discerning musician the performance he craves costs lots more than keeping a gamer happy.

"Fortunately, the cost of silicon keeps plummeting. So we've designed the EMU10K1 chip, which does contain incredible features for the musician, yet can still support the needs of the gamer, and is low‑cost enough to survive in both markets. The first card we're producing — the SoundBlaster LIVE! — will prove to be a real bargain for musicians. It may not have every feature they want, but I'm really pleased with what we can provide, and the fidelity is as good as the best of Emu."

Do Creative Labs intend to enter the high‑end, multiple‑output soundcard market, or is a broad market base of games users essential for all new products?

"Creative will continue to address products to a broad base, which largely includes game users. But we firmly believe that even gamers care deeply about sound, and multiple outputs are not wasted on them. With the advent of PC‑DVD using Dolby Digital (AC‑3) and MPEG‑2, we see that broad base of users as a great opportunity to sell multi‑output cards as well as decent speakers. Serious project studio owners need multiple audio outputs to enable them to mix tracks externally — unless they're using something like Cubase VST, which needs a very powerful computer.

Is this something that you can provide, either as standard or as a hardware add‑on?

"Yes. SB LIVE! can supply up to eight channels of 48kHz audio."

Nobody ever has enough MIDI ports. Is it possible to do a daughterboard‑style add‑on to your cards to provide multi‑port support without the user having to buy third‑party hardware or install additional drivers?

"In contrast to output channels, MIDI ports aren't of much utility to non‑musicians. Of course, MIDI is dirt cheap, but frankly I doubt this one will get solved by our products in the immediate future. I'd add, however, that Creative are very interested in IEEE‑1394 (Firewire), and that this is already a consumer interface, as well as being eyed by many music companies as a digital audio buss. So perhaps the MIDI problem will go away, and musicians will be using 1394 in the not‑too‑distant future."

Filters Or Modelling?

Emu have made a number of genuine advances in electronic instrument design, not the least of which is Morpheus‑style filters. Can these be integrated into a soundcard without making the result too complex for the typical user?

"One of the biggest problems with the Morpheus (H‑chip) filters is guaranteeing stability under any combination of user parameters and excitation waveform. They sound wonderful, but inherent in that analogue sound is a certain tendency toward oscillatory behaviour. In Morpheus, we had to test each filter thoroughly, and many people don't know it but there's actually software which shuts down a voice if it detects its filter going into oscillation. These factors tend to make the software for this sort of thing pretty complex; I expect more maturity in the MI market is required before we'll put it in a soundcard. Also, in contrast to the G‑chip technology, the H‑chip technology is still somewhat expensive. But there are many of us who think putting this sort of technology in a soundcard will bring benefits not only to musicians, but also for 3D audio modelling and sound effects."

Do you see more applications for physical modelling on soundcards, or can you achieve similar results through the appropriate use of your complex filters?

"My take on physical modelling is that it's an ideal application for host‑based signal processing. Since there is no 'generic' physical model in the way that a generic wavetable voice can make any sound you want, the software for each model is different, and continuous refinements are desirable. So I'm not a big fan of physical modelling on the soundcard itself, though I think we'll continue to support physical models in software synthesis. The complex filters can, indeed, produce very similar results in many ways with a general model, subject to the issues of stability I already mentioned. And a teaming of the two techniques is very powerful. In fact, looking into my crystal ball, I'd say a lot of the future of synthesis will involve new ways of combining the techniques that we are experimenting with today."

Is it practical to build a soundcard that functions as a multitimbral digital sampler at the same time as generating synth sounds and audio?

"Sure. That's why Creative acquired Emu!"

Backwards compatibility is less a technological burden than a marketplace burden.

A typical soundcard comes with several separate pieces of software and a whole bunch of manuals. Can this be simplified?

"Of course it can be simplified. The question is, does the customer really want us to simplify it? The issue is the tremendous cost in time to market to really integrate a technology into the system. By having a bunch of little pieces of software, we can change them independently, and offer both great value (easy to negotiate when you can change content quickly) and the latest technology. Were we to integrate everything, it would be older, more stale technology. I can argue that such a vast majority of our customers don't really get in and make use of these hot programs and features because they are difficult to use. But to some extent, it's the sales guys and magazine reviewers who do get into the product that determine the sales, and these folks are into the hottest technology.

"My philosophy at Emu was always to offer a nicely integrated, highly useful product. This was successful largely because the people who use our gear really care about getting something done; if they can make more music in less time, they'll pick our product and pay for it. But in the more consumer‑oriented world of Creative, we're selling more dreams and less productivity. So the product target is necessarily different. But still, I think if you look at how Creative's products have evolved since Emu was acquired, you'll see my touch in them."

Soundcard installation causes a lot of problems for some people (yours less so than most, I might add). Is it practical to include diagnostic software that can spot incorrect system settings or conflicts?

"That's a rather interesting area of research. We're working on it, but most of the time it's invisible because once the software has figured out the problem, it adjusts the functions accordingly and things work. You only notice when it doesn't work. The PC is a brutal playing‑field in this regard, because there are so many motherboards and chipsets and add‑in cards which were never designed with a strict specification and testing. That's why the Apple Macintosh is so good compared to the PC in this respect. But we continue to improve in doing this sort of thing. And to directly answer the question, yes, there is more we can do."

PCI Architecture

What benefits and challenges arise when you move to PCI‑card architecture?

"The biggest challenge in PCI is legacy compatibility. Since you don't have a DMA chip, and can't get specifically numbered interrupts, you need some way to fool the software into thinking these exist. The best way is to actually get hardware to do it, and the schemes for doing this — PC/PCI and Serial IRQ (SIRQ) — unfortunately require something called 'sideband signals'. With sideband signals you either have to mount the function on the motherboard, or run some separate wires up to the card in addition to the PCI connector. There are other schemes for the DMA function, including one called Distributed DMA (DDMA), which don't require sideband signals and work almost all the time. But since there is really no 100% reliable alternative to SIRQ, you really need sideband signals for 100% legacy compatibility.

"The PCI buss has a much wider bandwidth than the ISA buss, but this is really of no benefit if you're just playing 'legacy' audio as wave files, or using FM synthesis. But with wavetable synthesis and multi‑channel audio, you want to be able to transfer many channels of audio at once, and PCI enables this. So PCI really opens up a lot of new capabilities."

Foreseeing The Future

What impending technological advances do you think will have the most impact on the future of computer‑based recording and music production?

"Wow — that's the question, isn't it? It may sound clichéd, but I'd actually say I think the Internet may play a bigger role than anything else. And not because it will change the way music is made, but because it will change the way music is sold! Today, you need to sign with a major publisher to get into the CD players of the world via the record stores. But with a cheap PC and a CD‑R, you can put up a web site which allows people to sample your music and order a CD which you can make one‑off. If we get the bandwidth up, even the CD‑R may become unnecessary. This is sort of the equivalent of the invention of the printing press for musical performances. I think it will spawn a new diversity of music, perhaps finally overturning the 'superstar' mentality of recent years. And as a biologist I'm very much in favour of diversity!"

Quality Improvements

Last year we checked out the AWE64 Gold, which had pretty good audio quality, but the synth sounds seemed quite noisy and grainy. What was the technical reason for this, and can it be eliminated in future low‑cost products?

"Hmmm... The hardware wavetable synth in the '64 Gold is the same EMU8000 wavetable engine, based on Emu technology, that's been in the whole AWE family. I would be surprised at any problem with audio quality from that synth itself, though of course the quality depends on the samples, and I think there was a lot of room for improvement in the original General MIDI ROM for the AWE. It could also be that you were listening to the software synthesizer voices, and those are largely running sub‑sampled at 22kHz, which would degrade the perceived quality. But in answer to the second half of the question, the EMU10K1 has even better math specs that the Emulator IV, so I expect you'll find the quality superb. And I expect to see it run down the cost curve quite quickly."

The noise performance of the audio side of soundcards seems to be heavily influenced by the environment inside the computer. Can anything be done, in the way of improved grounding or screening, to squeeze better signal‑to‑noise ratios from on‑card converters?

"The best thing, of course, is to use S/PDIF or AES/EBU or, eventually, 1394 [see main text] to get the audio out of the harsh environment of the computer, then use external converters, perhaps even ones inside your speakers or amp. We continue to work on getting better in‑box converters, and still have some tricks to use, but you'll always have the variations in power supply and grounding from the system manufacturer, to say nothing of the results of inductive peripherals like disk drives. One thing that I've found makes a world of difference is ensuring that the grounding screws on the card brackets are tight! Use a lock‑washer, even. A little resistance in that chassis ground path can hurt the audio a lot."