While analog enthusiasts tout the benefits of that approach, Texas Instruments is bragging this week about the digital engine that powers Korg’s RADIAS synth, providing both analog emulation and digital processing. Unlike synths, Digital Signal Processing (DSP) chips don’t get misspelled English names. Instead, the brain of the RADIAS is the Texas Instruments TMS320VC5502 DSP. (Catchy.) This chip is what allows the Korg to process 24 voices and throw in extras like a vocoder. Check out all the techie details in the press release there, but suffice to say this is what digital is all about.

I’m not about to get into a digital versus analog debate here: they do different things. You can’t plug control voltage into the RADIAS for a modular analog system, just like you can’t play 24 voices on a Moog or other analog piece, and they do sound different. Instead, it’s more important to note that in the age of faster CPUs, dedicated DSP chips are still surviving, because of their specialized advantages and low cost. Even as Korg touts the benefits of Intel CPUs in their high-end OASYS, DSP chips make up a significant part of keyboard hardware you use. For a really up-close-and-personal look at the innards of a digital synth, you can read the manual and tech specs for that TI chip. Even with a passing knowledge of this stuff, you can see how specialized and lightweight DSP chips are when compared to the “sledgehammer, process any computations you throw at me” approach of our computers.

  • Thanks for daring to post about the dark and mysterious world of DSP!

    The frustrating thing about incorporating DSPs into a hardware design is that you still need a separate microcontroller to handle mundane tasks like polling the front panel, driving LEDs and displays, keyboard scanning, and MIDI/USB processing.

    By the end of the design process, you find yourself having built a fairly capable little computer. As an example, the prototype synth sitting on my design bench actually has several times the memory of a vintage Commodore 64 and a 40x faster clock speed – and that’s without factoring in the sound generation circuitry.

    To keep R&D costs down, most designers like to build new machines around chips from existing MCU and DSP chip families, which lets them port bits and pieces of existing software to shiny new hardware. And — in an effort to get costs even lower — a chip company called Atmel actually produces single-chip DSP/microcontroller systems optimized for synthesis and audio effects (the ATSAM3103 / ATSAM3108 series, if anyone's interested in Googling them). I hope that companies like TI, Motorola, and Analog Devices take a similar "all in one" approach — it will help to fuel development of the next generation of hardware synthesis.

  • I guess the question is, would TI, Motorola, or Analog Devices have enough impetus to do that; the DSP chips are getting sold for very different applications that wouldn't require that particular microcontroller.

    Thanks for the added insight, James!

  • The combination of a microcontroller and dsp core in a single package makes sense for a variety of common non-music applications, and companies are beginning to release "all in one" DSP systems (Microchip's dsPIC lineup is a good low cost example). So it's only a matter of time… 🙂