Why DSP, and Why Ours is Special

For some reason, probably due to early generation, not so great sounding conversion and electronics and poor deployments back in the 80s and 90s, DSP got a bad name in the studio monitor world. To this day, the very phrase can elicit a grim reaction from many a seasoned veteran in the recording world. However, live sound engineers have been running almost exclusively DSP controlled PA systems for years in prestige venues, and some of the most advanced technical innovation in loudspeaker design in the last thirty years has been made possible by advances in computing and DSP power. The power and precision of crossover filter design, limiter protection, proper delay compensation and offset that even simple fixed point DSPs offer a loudspeaker far exceed what’s possible with analog tools alone. And ours ain’t no simple fixed point DSP…

To start with, let’s consider the benefits DSP offers us in crossover design and implementation. Traditional passive loudspeakers are quite limited in their crossover slopes and shapes – much beyond a 4th order network becomes extremely expensive to implement passively, the inductors and capacitors in the network waste a significant amount of the amplifier power thrown at them as heat, and the combination of multiple drivers and a crossover on a network presents a complex impedance load to the amplifier driving it, making it much more difficult for the amplifier to power. Active analog crossovers can offer greater flexibility, but can be noisy, imprecise and still necessarily exhibit phase shift, as all analog filters do. DSP on the other hand allows for a huge range of filter shapes, including complex asymmetrical crossovers and linear-phase filters. Even in situations where a simple, symmetrical crossover is called for, modern, high precision digital filters are often more accurate and lower distortion than even the best analog counterparts. In our application, the first thing our powerful DSPs allow us to deploy is ultra-steep, 8th order crossovers that reduce distortion and improve efficiency in our drivers by halving the amount of out-of-band information they have to reproduce.

DSP also allows for much more precise delay correction than physical baffle offset or analog networks – at 192khz, where our DSP runs natively, timing differentials of as little as 5 microseconds can be adjusted and corrected for. Complex EQ curves to correct magnitude irregularities inherent to a given driver, or even from batch to batch can also be deployed with ease in a DSP, whereas attempting to implement a similar degree of magnitude correction with analog filters would be costly, noisy and necessarily introduce significant phase shift. A DSP based system can also introduce protection limiters to prevent over-excursion or clipping in a loudspeaker system without the need for additional componentry in the signal path.

The elephant in the room, despite all these benefits, is of course the perception that the additional conversion loop necessary to get a signal in and out of the digital domain causes irrevocable harm to fidelity of the playback, however with modern, high quality conversion and clocking design, this is simply no longer true. The AKM Velvet Sound based conversion and ultra high precision reclocking circuit Danville Signal designed for us in this application in fact introduces far less distortion and deviation than even the highest quality analog crossovers would impart. Our DSP circuitry actually enables us to use a shorter, simpler, cleaner signal path than would be required just to deploy our crossovers in the analog domain.

All these benefits alone more than justify the use of DSP based loudspeaker system, but our deployment goes far beyond just this. We do not use a simple fixed point DSP. Every loudspeaker we reproduce instead gets its own floating point SHARC DSP, which allows us the computing headroom to employ a far more advanced approach. After all the initial crossover, protection and limiting coding, each loudspeaker gets its own individual measurements taken, and using a proprietary process written in house from scratch, a unique correction is generated for the loudspeaker that flattens not just the magnitude response, but the phase response as well. This correction method is so accurate and so powerful that each resulting loudspeaker exhibits linear phase response within +/- 15 degrees and linear magnitude within +/- 1db across its rated range, or it doesn’t leave our QC line. Which, perhaps more than any other aspect of our design, is what creates the uncanny sense of space, imaging and phantom center that defines our monitors. It also means that any two serial numbers can be combined as a stereo pair, or deployed in a surround setup, without any degradation in matching.