Given its Canadian pedigree, it was fitting that the first journalists to tour the Samsung Audio Lab are from Canada: yours truly, plus four other writers based in Vancouver, Edmonton, Toronto and Montreal. Over a two-day period in late May, we toured the lab, participated in subjective tests, and heard informative presentations by the Audio Lab's senior engineers.
Located in an industrial park 50km northwest of downtown Los Angeles, the Audio Lab houses two anechoic chambers and two custom-built listening rooms, plus labs for transducer and system design. Three of its research staff are PhDs; four more have Masters degrees. Five are active musicians.
The basic methodology underlying the Audio Lab's research is based on Toole's work at the NRC, and later refined at Harman. Objective measurements are conducted in the anechoic chambers, where four-foot sound-absorbing wedges suppress acoustic reflections that would otherwise mar the results. And as the NRC did, Samsung conducts controlled subjective tests, where trained listeners evaluate and compare speakers in carefully designed listening rooms.
As Toole and his colleagues established, first at the NRC and later at Harman, flat frequency response is essential for a speaker to sound good. But frequency response has to be measured - and interpreted - correctly.
When we're listening to a speaker, direct sound is only part of what we're hearing, explained Principal DSP Engineer Pascal Brunet in his presentation on loudspeaker measurements. Our ear-brain systems integrate direct sound with the first reflections from adjacent walls, floor and ceiling into a single event. Secondary reflections from multiple boundaries, occurring 30msec or longer after the direct sound, generate the sound power of the room/speaker combination. As Dr. Brunet explained, direct sound accounts for 12% of the typical in-room response of a speaker. First reflections account for 44%, as does the sound power. So to understand how a speaker will behave in a reflective environment, you have to measure its response at different angles.
In Samsung's main chamber, the speaker under test is placed on a pedestal in the centre. Around the perimeter of the chamber are test microphones mounted on two metal tubes: one vertical and one horizontal. The test microphone on the horizontal tube can shifted laterally under computer control, allowing measurements to be taken at different angles. The vertical tube has microphones located at different angles relative to the speaker. Once the speaker is in place and the door to the chamber is closed, multiple measurements are taken from microphones at 70 different locations. The resulting data is used to generate three different frequency response curves.
The first curve is derived from measurements straight on-axis, plus three points to the left and right, and one additional point above and below, all within ±30°. These nine measurements are blended into a spatially averaged response curve through the primary listening window. As Brunet notes, this is far more useful than a single on-axis measurement.
In a single measurement, response irregularities can be the result of acoustic interference between waves bouncing off different elements of the speaker. Or they can arise from resonances in the speaker itself. "Acoustic interference is position-dependent," Devantier notes. "Resonances aren't." In a spatially averaged curve of multiple measurements in the main listening window, irregularities from acoustic interference are averaged out, leaving irregularities arising from resonances. That lets engineers zero in on the most important characteristics of a new design. "Resonances are about 10 times more audible than acoustic interference," Devantier says. "You've got to fix them."
The second curve is derived from measurements taken further off-axis, to ±60° horizontally and vertically. This provides a representation of early reflections. As Devantier explains, early reflections from adjacent room boundaries enable a stereo pair to create an image that extends beyond the speaker plane.
The third curve is a weighted average of 70 measurements, taken all around the speaker - in front, behind, above and below - through 360°. This indicates the response of the total sound power radiating into the room.
If the three curves are reasonably flat and similar to each other, the speaker will sound spacious, and will have a wide listening window. The directivity characteristics can be shown on a curve, derived from the difference between direct sound and sound power (the Sound Power Directivity Index) or between first reflections and sound power (First Reflections Directivity Index, a measurement developed by Devantier at Harman). As Devantier explains, "From a single plot, you can learn about the speaker's frequency balance, timbre and spatial characteristics."