The Wirecutter recently asked me to conduct an extensive test of powered, 2.0-channel speaker systems designed for desktop/computer audio. In the test, I and a panel of listeners compared several eight models; there were three more I excluded because I thought they had zero chance of being picked as the best, second-best or even fourth-best. And on The Wirecutter, once you're past fourth-best, you're out of the running.
With so many systems in my house, I couldn't resist putting them up on my measurement stand and seeing how they perform in lab tests.
I measured the frequency response of each system, which gives you a good indicator of how well-engineered a system is. Ideally, the frequency response of the blue trace in each chart (which corresponds to a measurement from 0 degrees, directly in front of the speaker), would be flat or close to it. And ideally, the green trace in each chart (which shows the average of responses at 0, ±10, ±20 and ±30 degrees horizontally) would be slightly downturned at the right side of the chart, as the test frequency approaches 20 kilohertz, which is the generally accepted theoretical limit of human hearing.
I did these measurements using quasi-anechoic technique, with the speaker atop a 2-meter-high stand and the MIC-01 measurement microphone at 1 meter, using the gating function on my Clio 10 FW audio analyzer to eliminate the acoustical effects of surrounding objects. I adjusted the microphone height, within reason, to try to get the optimum response from each speaker. Bass response was measured using ground plane technique, with the microphone on the ground 1 meter in front of the speaker, then splicing the result to the quasi-anechoic curves somewhere between 160 and 180 Hz. Quasi-anechoic results were smoothed to 1/12th octave, ground plane results to 1/6th octave. Results were normalized to 0 dB at 1 kHz.
Incidentally, when I calculate the plus/minus dB numbers, I discard everything below 200 Hz because the scaling of the bass response to the quasi-anechoic response relies to some degree on guesswork. I calculate the bass response limit by taking the peak output below 200 Hz and subtracting -6 dB.