Time Dilation, Part 1
In an October 2004 "Industry Update" (p.29), Paul Messenger reported on a paper presented recently to the Institute of Acoustics in the UK on the subject of loudspeaker bass response (footnote 1). The paper begins with the observation that the subjective bass performance of different speaker designs can vary considerably, and concludes that phase behavior (group delay) and time-domain behavior (ringing) are the factors responsible for these disparities.
A friend in the UK audio industry who was also inspired to read the paper by PM's piece commented to me that this work actually tells us nothing we didn't already know—an assessment with which I concur, particularly as it offers no suggestion as to how phase and time-domain performance can be stirred together with the amplitude response into an overall figure of merit for speaker bass performance. But as a reminder of an aspect of loudspeaker behavior that remains poorly defined, the paper struck me as timely—not least because for some months I have been pondering another aspect of the problem.
For many of us on the front line of equipment reviewing, difficulties of interpretation are secondary to the problem of being able to measure loudspeakers accurately at lower frequencies. Although today's equipment reviewer can, without having to spend a small fortune on lab facilities, make most of the standard audio measurements to an accuracy similar to what manufacturers can achieve, this is one notable exception.
In fact, most reviewers who measure loudspeakers—John Atkinson and I fall into this category—routinely fail to make measurements that address lower-frequency behavior with anything like enough resolution. Note that I say lower-frequency, not low-frequency behavior—the problem typically begins to manifest itself below 1kHz. Actually, the principal problem area, as we shall see, is lower-midrange frequencies rather than bass frequencies per se, for which tricks exist to get us out of the mire.
To see why this is so, let's begin with a quick overview of the methods by which loudspeakers can be measured—or, rather, of the environments in which this can be done, which is the key to this story.
When loudspeaker measurements are referred to as freefield, it means that they are unaffected by the presence of a containing space. A true freefield measuring environment is difficult to achieve because it implies that all reflective surfaces are sufficiently removed from the speaker to have negligible effect on the measurement. Practically the best that can be done is to raise the speaker a long way off the ground, typically on a telescopic pole, in a location well away from buildings or other sources of reflection other than the ground itself (and, of course, from any sources of extraneous noise). This method was quite often used in the past, but its impracticality is obvious. Even assuming that you have a suitable wilderness available and can afford the telescopic pole, you remain at the mercy of the elements, particularly wind and rain.
It was to achieve almost freefield conditions in a more practicable way that the anechoic chamber was invented. Here the aim is not to remove all reflective surfaces to a safe distance, but to cover them with sufficient sound-absorbent material that they, in effect, disappear acoustically. Actually, size still matters: large anechoic chambers are better than smaller ones in that they permit accurate results to be achieved down to lower frequencies, although even the largest are rarely accurate below about 70Hz. Corrections are usually established to account for this, allowing reliable results to be had down to the lowest audible frequencies.
It goes without saying that building and equipping an anechoic chamber represents a substantial investment—one well beyond the individual reviewer or even magazine. This difficulty was sometimes sidestepped in years past by audio magazines hiring an anechoic chamber (a costly exercise in itself), but the bottom fell out of that market with the introduction of what are sometimes regarded, wrongly, as a panacea: time-windowed measurement systems, of which DRA Labs' MLSSA is unquestionably the best known.
Time-windowed measurement systems were a breakthrough because they made it possible for the first time to test loudspeakers in a normal reverberant environment. To achieve this, first they replaced the traditional swept sinewave with a broad-spectrum test signal. Initially, impulses were used (for example, by KEF, when it pioneered this methodology in the 1970s), although today maximum-length sequence (MLS) pseudo-random noise is routinely preferred for its higher signal energy. Second, they analyzed only that portion of the measured impulse response captured before the arrival of the first reflection from the nearest room boundary.
Eliminating reflections from the measurement in this way allows anechoic testing to be conducted in a normal room. Little wonder, then, that time-windowed measurement was greeted as a great democratizing technology, one that allowed almost anyone (actually, the expenditure involved is still not trivial) to perform speaker measurements without requiring access to an anechoic chamber. But this revolution (even though it did not occur in the 1960s) wasn't everything it was sometimes cracked up to be. By limiting the time window of a measurement, you also limit its frequency resolution. The relationship is a simple one: if you apply a time window of 6 milliseconds (0.006s—not unusual for measurements conducted in a typical domestic room), then the frequency resolution of the measurement will be limited to 1/0.006 = 167Hz.
Footnote 1: P.R. Newell, K.R. Holland, P. Mapp, "The Perception of the Reception of a Deception," paper presented at the Institute of Acoustics.