Meeting Topic: "Using Arrays of Loudspeakers for Focusing or Diffusing Sound"
Speaker Name: Mike Seltzer, Jasha Droppo, Ivan Tashev, Microsoft Research; James (JJ) Johnston*, Microsoft Corporation (* now with Neural Audio)
Meeting Location: Microsoft Research in Redmond WA
The PNW Section visited Microsoft Research in Redmond WA for its February 27, 2008 meeting on "Using Arrays of Loudspeakers for Focusing or Diffusing Sound." 23 AES members and 41 nonmembers were counted.
Using a "tag team" format, each speaker covered a topic, then a lengthy Q&A was held, door prizes awarded, and finally free time with the research lab demo array hardware devices.
Mike Seltzer of the Microsoft Research Speech Technology group started with an overview of reasons for small speaker arrays. These arrays were oriented toward home and office use, instead of large reinforcement systems. They could be used to create a more uniform sound field, or let several listeners hear something different in the home or a retail setup. He noted an existing Yamaha product capable of some of these features; 40 small drivers/2 woofers in a modest, wide cabinet.
Next, Jasha Droppo, also of the Microsoft Research Speech Technology group, continued with "Focusing Sound principles and generic beamforming." He covered the basics physics of why arrays work and steering the the horizontal line array beam by controlling each driver's delay. However, since real drivers aren't ideal or identical, their answer was to develop some math to solve the problem, which was a segue to the next speaker.
Ivan Tashev, also of the Microsoft Research Speech Technology group, continued with "Robust beamformer design for loudspeaker arrays." Ivan showed formulas for beamforming and filter design to focus sound and be as silent as possible elsewhere. When solved for an array, it works fine until real-world drivers are taken into account. It is impractical to do complicated measurement/compensation for each driver in a home system, so a tweaked and maximized formula helps accomodate driver variations/tolerances.
James (JJ) Johnston, Audio Architect of the Windows Live division at Microsoft finished with "Generating diffuse sound with arrays." He noted that sound only needs to be decorrelated enough to make the ears think it is, fortunately. The main problem here is if we play a diffuse wave through a point source, it is now heavily correlated. Psychoacoustics can solve that if we create a pattern that is diffuse to the ear. In each critical band, scramble the onsets of the signal envelope; the onset is different for adjacent critical bands, and scrambling the frequency response helps.
45 minutes of Q&A were followed by the awarding of some door prizes.
Demo hardware was available the remainder of the evening. These showed:
-Focusing sound - Beam focused towards the array broadside
-Multiple beams - Two beams, one broadside, one under 45 degrees
-Diffuse sound - Stereo sound track, with both direct and diffuse radiation
-and the Yamaha 1100 speaker array for home surround sound
Report by Gary Louie, PNW Secretary
