Questions arise from time to time about how to interpret the measurements shown with each speaker review. Here’s an explanation of each of the curves in our charts from top to bottom, starting with that showing frequency responses.
Pink Noise Sweep: Here at AIG we measure speakers in a room, using some absorbent materials to minimize reflections. I’ll say more about this below, but several kinds of test signal are employed. The one that by its very nature provides the best indication of a speaker’s overall frequency balance, and therefore its general tonal character is this measurement, which we have developed by combining two measurement technologies.
Pink noise is a broadband impulse noise signal with very even distribution across the whole audio band. Thus when it is used with a 1/3d-octave realtime analyzer, a picture can be seen of a speaker’s frequency response. Our standard microphone distance for this is 1 metre, which minimizes room effects.
Normally with our LMS measuring system, a frequency sweep is generated through the speaker to measure its response. In this case we replace that with pink noise to provide a more averaged response that is particularly useful in judging the bass performance, its smoothness and extension, in a given speaker. It also gives a good indication of how the speaker will sound in the midrange at the listening position, the pink noise tending to approximate what speaker designers call a Sound Power Response.
Summed Axial Response: Sound Power is measured by summing or averaging the response of a speaker on many vertical and horizontal axes, usually a 60o horizontal radius and a 30o vertical one. We also do this sort of measurement by adding together the 0, 15, 30, and 60o curves seen below, each of these in our tests derived from 3 vertically averaged measurements to minimize inaccuracies caused by driver spacing or driver lobing or other interaction effects.
The SAR is overlaid on the PNS, and the closer they correlate, the smoother the dispersion and more even the radiation of the speaker over a wide listening axis. This is especially important in the midrange where most music happens. Typically, if there are a lot of variations in midrange response in the axial curves, the SAR will look less like the PNS curve. What we are looking for in a good loudspeaker is consistency as well as smoothness in the way the speaker couples to the room on all axes.
Quasi-Anechoic Curve: The nuts-and-bolts, warts-and-all curve, made of three vertically averaged measurements directly on the speaker’s centre axis using the gated mode of LMS, which by tracking and filtering the test signal at each frequency hears only the impulse as it is first generated by the loudspeaker, effectively removing any room reflections; this reveals the naked truth about a speaker’s driver complement. It shows what you’d hear if you sat close to the speaker with it aimed directly at you. Crossover problems, spikes and dips, and other anomalies of response can be clearly seen here. Sometimes these things are very audible, but don’t show up as well in the other measurements.
Axial Curves: Different types of speakers radiate their energy into the room in a variety of ways. Direct unipolar radiators become less directional with decreasing frequency, while dipoles create lobes to surprisingly low frequencies, allowing bass tuning with position by utilizing the side cancellations, while bipolar speakers have a more uniform radiation all around them. Omnidirectional speakers are coming back into vogue because they do approximate the radiation characteristics of many musical instruments, but can be very room placement sensitive.
Our axial measurements explore these radiation aspects to show how a speaker will interact with a room and its side and rear surfaces. Some speaker designs, whatever their radiation patterns, have quite uneven frequency response to their sides, this a result of driver interaction, that is, acoustic reinforcements and cancellations, often due to phase shift above and below crossover points. Such effects can occur with first order designs where the overlap covers several octaves in the middle frequencies.
The measurements at 0 and 15o should ideally look as much alike as possible, smooth through the midrange (600-4000 Hz), and extended and flat in the treble (4000-20,000 Hz). Those at 30 and 60o are generally characterized by some upper frequency rolloff; if they weren’t, the speaker could sound bright and etched from energy reflected off side walls. If there’s excessive midrange energy on these axes, especially if it’s greater than at 0 and 15o, the speaker will interact more with the room and might sound quite coloured. Getting back to the SAR, if the sum of these axial curves significantly varies from the PNS, then we have a speaker which may require significant room treatment to sound good. And, referring back to the quasi-anechoic curve, the less smooth it is on axis, the more likely are anomaliesas we move to the sides. A speaker is a system, with complex interactions that the best designers control by both electrical and acoustic means.
Impedance: In a perfect world with a perfect amplifier, it won’t matter how much the electromechanical device resists the force driving it. In general, speakers with low impedance that does not vary with frequency will draw more current and play louder, provided the amplifier can provide that current; conversely, speakers with widely varying impedance with frequency will not demand current delivery, but will find their frequency response starting to look like their impedance curve with amplifiers that also have a high output impedance. This is why it is so important for speakers used with single-ended designs to have smooth impedance curves with all values above 8 ohms. Speakers with impedances that go below 4 ohms, especially in the bass frequencies where the most current is required, may be unsuited to tube amplifiers and require ample solid-state power.
Electrical Phase: This curve is derived mathematically from the impedance curve in our LMS program to show the phase angle of the loudspeaker over its frequency range. In general, extremes, especially through crossover or at high frequencies, can make a speaker a more difficult load for an amplifier. A very linear electrical phase usually goes with a minimum of acoustic driver interaction and overall phase coherence in listening. If the drivers are interacting electrically, which is what the phase curve shows, then they may be out of phase acoustically, too. An exception to this rule are 1st order crossover designs where the drivers are kept as much in phase as possible electrically, but because of the necessary acoustic overlaps through crossover will have some cancellations and reinforcements. For more on this complex question, see the article, Loudspeaker Phase & Musical Timing by speaker engineer Roy Johnson in our Winter 97 issue.
Listening Tests: I prefer to audition speakers in our big 32′ x 12′ listening room so they can be readily compared to our Energy Veritas v1.8s using a simple switching box. Standard cables are Kimber Select KS-3035 (v1.8) and van den Hul The Revolution (SUT), and the amplifier is a Bryston 3B ST. I use a CD-R on which I have assembled from our own and other recordings, selections which show the various areas of speaker performance discussed. This is, perhaps, the secret of reviewing so many speakers, in that it focuses our listening, and makes the evaluation process more efficient and consistent.