Adding And Subtracting

When we glance over the specifications of a new audio component, how do we know if it will suit our sound system needs. Not only must the unit provide the audio quality and reliability we need, it also needs to fit into a chain of audio components that make up the sound system. While it may be possible to create complex systems from very few physical components, thanks to DSP, even these systems must consider the accumulation of unwanted changes to the sound that inevitably occur between the transducers.

While the electrical signal passes from the output of the microphone capsule through to the loudspeaker drivers, the side effects of adding gain, equalization and dynamic control will increase the noise and distortion, in addition to reducing the bandwidth. While a good sound system may render these effects inaudible, it will only happen if the components in the signal chain are carefully selected and adjusted. This includes the system gain structure, impedances, filters (built into the device by design and those you can adjust), and the cables that connect them all together.

Most current sound systems jump back and forth between the analogue and digital domain to amplify and process the audio. Over the past two decades A/D and D/A conversion technology has improved to the point that many of the original concerns have been put to rest. However, some aspects of digitizing the audio more than once can still produce unwanted changes to the sound. In high performance systems (located in quiet venues), the increase in noise floor from multiple stages of dither (low-level distortion masking) may be more obvious. This may be where the latest 24-bit converters offer the greatest benefit, as the digital noise lies below the pre-existing analogue noise of the sound sources. If you can choose 24-bit converters for all of your digital processing devices, then this form of noise may not be an issue. However, 24-bit by name is not necessarily 24-bit by nature. The noise is not simply a product of the converters; there are other potential noise sources related to both the DSP code and the configuration of the processing.

Unlike a chain of analogue components, digital gear will typically have exactly the same bandwidth in each unit. Adding identical bandwidths together doesn't give you that same bandwidth at the end of the chain. Instead, two identical bandpass filters will increase the roll-off by 3 dB. Yet another reason to digitize the signal only once. The bandwidth of two processors with a 48 kHz sampling rate is a pretty good match, and going to 44.1 kHz won't improve things. Sure you might try a 96 kHz sampling rate but wait until you try to send that signal any distance and find out the conductor count (and cost) has doubled. However, there are not many options currently available for routing digital audio over long distances and then back into sound reinforcement devices. For many, the need to have flexibility in their choice of processing, and limitations of budget, force the signal path to pass through more than one A/D/A conversion. As much as we may want a fully digital signal path for quality's sake, the guy putting in the conduit would even be happier because we can finally share a length of pipe with all those other data signals.

Bad gain structure can occur within a multi-function DSP processor, just like in the analogue domain. To achieve the best performance, the signal should still clip each stage of (internal) processing every so slightly above where the input A/D converter is maxed-out. At least within DSP units you won't have to resort to hand-built resistor networks to optimize the gain structure. However, many of these one-box sound system processors do not offer much help with measuring or tweaking the gain in and out of each stage. On the other hand, systems like Peavey's MediaMatrix include very comprehensive test-point and monitoring capabilities. In all cases, the problems related to impedance, cable induced noise and poor connections disappear once your signal is inside!

Whether it's the latest digital gear or a trusted analogue box, the issues of noise addition and bandwidth subtraction remain. The trick is getting a combination of processors configured for each leg of the signal path to achieve the most suitable signal quality for the job. Don't worry about accumulating a 2 dB drop at 20 kHz, if the loudspeakers are 100 ft away from the listeners, you probably have other problems getting that gig to sound right. However, you don't want the guy sitting in the expensive seat listening to the "clock whine" of bad gain structure just because there's way too much makeup gain in the compression driver's EQ. While combining multiple stages of processing might subtract more from the audio than we expect, exercising good practice and good judgment will always add up.

United Entertainment Media Inc.
460 Park Avenue South, 9th Floor
New York, NY, 10016
Ph. 212-378-0400
FAX 212-378-2160

page construction