A serious evaluation of a loudspeaker system requires both objective and subjective tests. For the objective tests, we've used an electro-acoustic measurement system, the DAAS32 (Digital Audio Analysis System) with a dedicated acquisition card and Windows-based software. The test microphone was a 1/4" Neutrik 3382. This system can be used for many types of measurements, but we limited our test to frequency response using an overall measurement system that is obviously subject to the acoustical faults of the environment in which the test is conducted. So we then complement the overall rating with a test of the subwoofer alone (using the close-up method) to get a good idea of what it can do and also what the crossover range with the satellites is.
To plot the curve shown here, we used weighting and a third octave plot. This smoothes out frequency response accidents but makes things much clearer for non-experts. The raw curve produced by our measurement system is very finely defined (in 1/24 octave), and requires knowledge of electro-acoustics and a certain amount of practice to read it with any degree of accuracy. We do of course use it for our analysis work.
All the tests are conducted under conditions that are near typical for actual setups - with the satellites on a desktop and the subwoofer on the floor. The results are therefore representative of the results you would get in practice, with a properly installed system.
For the subjective listening tests, we used a PC with a Terratec Aureon FireWire sound card, a high-end (24 bits/192 kHz) multimedia card. We vary the music we listen to in order to evaluate sound quality and any defects in the main registers. The media are DVD Audio and CD Audio. We then perform tests with several games and video DVDs.
The tests are complemented by an evaluation of the maximum sound level possible without audible distortion. For this test, we've created a test signal to suit our purposes, using three pink noise tests encoded in stereo (right, left and center). This signal, compared to broadband pink noise, is heavily filtered in the sub-bass range (practically absent from most kinds of music) and with a fairly gentle low-pass filtering, starting at 500 Hz (-3 dB).
This third-octave analysis of our test signal shows exactly how the frequencies sent to the speakers are distributed.
The crest factor is 12 dB. We haven't tried to reach maximum levels, but rather to be realistic in terms of the musical reproduction the test is supposed to simulate - while using a constant signal for reliable measurement. Like all pink noise (what the signal is based on), this signal is random in nature.
This 3D spectrum analysis (sound level, frequency, and time over 5.5 s) gives an idea of what our signal is like.
Sound levels were measured at one meter from the speakers installed in a typical way (satellites on a desk and the subwoofer on the floor) using a sound level meter with a resolution of 0.1 dB SPL and configured with a standard "Long" response time to obtain a stable value representative of the mean level obtained. The resulting value is of course in dB SPL (Sound Pressure Level) without weighting, as it should be for sound ratings of broadband noise. But don't forget that this is a comparative test, so the value is not absolute. In other conditions, these systems would produce different sound level ratings.
The sound level obtained in dB SPL "L" (Linear) as seen on the screen of our meter