Perrin Meyer, Meyer Sound Labs, on DSPs
Acoustic consultant and audio designer Sam Berkow launches a new column exploring audio technology through interviews with product engineers. Here he talks to Perrin Meyer, who works in advanced R&D at Meyer Sound.
An audio consultant asks leading AV engineers what goes into their product designs.
Quick bio: Perrin Meyer works in advanced R&D at Meyer Sound. Son of company founders John and Helen Meyer, he joined the company in 2000 to head up the development of MAPP Online Pro and continues to define the technology architecture for future product development. He has published numerous papers on topics ranging from computer protocols and monitoring for audio systems to green mobility. Meyer holds a Masters in Scientific Computing from the Courant Institute of Mathematical Sciences and a Bachelors in Science and Applied Physics from Columbia University.
Sam: You work in digital signal processing, one of the fastest changing areas of audio. What changes have you seen and what changes can we expect in DSP designs?
Meyer: When I started college 20 years ago, DSP was considered magic because people assumed if a processor was digital it had to deliver better sound quality than analog. This is actually not always the case because errors in code can lead to poor-sounding digital gear–worse than analog in some cases. Artificial boundaries were created by old-school electrical engineers (analog) and new-school computer science grad students (DSP), but the lines aren't actually black and white. For example, a discrete analog/digital converter chip has analog electrical signals (voltages that correspond to pressure) as well as voltages that correspond to ones and zeros ("digital"). But is it an analog or digital device? I think we should promote "Audio Processing Algorithms," not DSP, because DSP is just one component of audio processing.
In terms of DSP evolution, these are exciting times. CPUs continue to get faster and [single instruction, multiple data] vector extensions are useful for audio DSP. Graphics chips are morphing into general purpose DSP engines and [field programmable gate arrays] are getting fast enough to be used as audio DSPs. These advances will make debates like "48-bit fixed-point is better than 32-bit floating-point" obsolete because the audio DSPs of the future will use combinations of fixed-point, vector, floating-point, and other formats, mixed together.
Sam: You headed up the design team for Galileo, Meyer Sound's DSP-based loudspeaker controller. With the broad adoption of powered loudspeakers, do you expect to see system controllers like Galileo replaced by distributed DSP systems with user-controlled DSP in each loudspeaker?
Meyer: From a user perspective, it doesn't really matter where the processing takes place, and/or if those algorithms use analog or DSP processing. All that matters is that the sound quality is excellent and reliable. So as we come out with new products, we continue to use our systems engineering approach and put the processing components where they make the most sense, which will probably vary depending on the product and application.
Sam: As with most things audio, there is a distinct lack of standardization in input and output formats, levels and data transmission systems, loads and outputs. Does this make the design process more difficult or more fun?
Meyer: Lack of standardization hurts the audio industry as a whole because it makes it harder to ensure good and consistent sound. This lack of standardization drives our research to expand to cover more of the audio chain in order to guarantee the best sound quality to our customers. This was the initial motivation behind the drive towards self-powered loudspeakers.
Sam: How closely do you work with end-users when designing new products?
Meyer: We work very closely with the people who use our products in the field, and we listen very closely to their feedback because it's invaluable. A good example is the recent Compass software upgrade to Galileo, which was developed as a direct result of customer feedback. Pro audio technology for its own sake is not very interesting to me; I want to create audio technology that enables the people using our products to enhance the audience's enjoyment of music. Because after all, people don't really care about the technology, they just want to hear the performers they came to see.
Sam: It seems that DSP boxes are particularly hard for users to compare using spec sheets. Which specs, if any, do you feel give the most useful information for an engineer looking to make an informed choice?
Meyer: This is a hard problem to tackle because human hearing has more dynamic range that our eyes, and also works over a higher bandwidth, making it difficult to come up with specifications that capture this remarkable "instrument." The obvious specifications include total harmonic distortion and frequency response, and those should be as low and as flat as possible. Measurement systems like SIM [Meyer Sound's audio analyzer] are useful at creating even frequency responses in rooms, but other aspects of sound are hard to quantify. Blind listening is still a good final measurement.
Sam Berkow is the founding partner of SIA Acoustics, an acoustical design firm with offices in New York and Hollywood, Calif.