Stand Loudspeaker Reviews

Editor: Art Dudley's rave review of the Harbeth Monitor 40.1 in the October issue, and its swift placement in Class A of your "Recommended Components" section, is most gratifying. Three cheers to AD and John Atkinson for their efforts. Critical acclaim is wonderful, but what is even more important to us is the unanimously positive feedback from customers, both consumer and professional, affirming our belief that we have achieved our design objective: to take all the strengths of the previous generation of the BBC-style three-way monitors and make them even more domestically friendly.

Stereophile should be applauded for their continuing efforts to balance subjective reviews with technical measurements, even though we have differences of approach and interpretation. If I, as a speaker designer, were sitting in the editor's shoes, it would be with considerable trepidation that I would contemplate making technical measurements of the loudspeakers that cross my desk. I would surely have sleepless nights, wondering if I'd overlooked something through my inexperience or human error. That's the thing with measuring speakers—just when you think you really understand them, they surprise you.

Nevertheless, we speaker people are well served by the audio test-equipment/software industry, and acoustic-test equipment packages have been available to even impoverished DIY speaker enthusiasts for years; even the most basic equipment can give surprisingly useful results. That leaves three significant areas of uncertainty when measuring loudspeakers: the appropriateness of the measuring technique and its variables (often concealed), the environment in which the speaker is measured, and, crucially, the interpretation of the results.

As I mentioned in my previous comments (October 2008, p.203), I believe that the most useful and consumer-friendly way to measure a loudspeaker is under conditions where the room is completely removed from the measurement. This implies a nonreverberant environment (an anechoic chamber or, more honestly, a freefield room), or placing the speaker on a high pole on a warm and windless day, outside in the open. Assuming similar temperatures and equivalent path length from the loudspeaker to the nearest reflective surface, in my experience the anechoic chamber and the back yard should give and can give comparable results. In fact, even the BBC's anechoic chamber has noticeable reflections from the metal arm supporting the microphone boom, and from the metal-grid floor (which, for safety reasons, cannot be completely removed). There are also other small, irritating reflections—from the thermostat housings, light bulbs, and cables—that are absent in backyard measurements. These would not have been noticeable in the 1950s using paper-trace pen charts, but modern computer systems clearly reveal all imperfections onscreen.

So there are reflections in real-world anechoic chambers that limit their usefulness. At lower frequencies, the wedges are neither long enough nor absorbent enough to prevent the low-frequency sounds from bouncing off the chamber walls and disturbing the measured response. Though imperfect, anechoic chambers around the world of similar dimensions and with similarly absorptive linings can be expected to provide a comparable frequency-response measurement of a loudspeaker under test. They are, then, an international yardstick by which loudspeakers can be evaluated, with little or no argument or need for interpretation.

We, the industry, have to admire John Atkinson's energy and willingness to even attempt the measurement of loudspeakers without the convenience (and confidence?) of an anechoic chamber, or an industrial hoist capable of lifting heavy speakers at least 3m skyward. As I understand it, Stereophile's measurements are made on or near the ground, with the microphone brought very close to the speaker drive-units. If the speaker (like the M40.1) has only a single pair of terminals, the cabinet is not opened to disconnect the unwanted drive-units—all drive-units are fully operational during all measurements. I could not design a speaker without being able to isolate the contribution of the individual drivers; and, conversely, I could not accurately deduce each driver's (or port's) unique contribution under measurement conditions in which all drivers are being driven, due to cross-pollution at the microphone at the low frequencies/long wavelengths.

The nearfield measurement proposed by Keele was a development of Small's earlier work, when he proposed drilling a hole in the cabinet to measure the driver+vent composite pressure inside at low frequencies. I have some limited experience of comparing the Keele nearfield measurements against those made in the farfield in an anechoic chamber. In my testing, a sealed box (no port) with the Keele-method microphone placed within 2–3mm of the cone, plotted and overlaid on the anechoic response, is often revealing of the limitations of the chamber. The real challenge is using this technique on a loudspeaker that has multiple bass units, and/or that has one or more ports, and scaling their contributions. In the anechoic chamber, when the microphone is some distance from the loudspeaker, it will collect the soundwaves from all sources, regardless of whether the low frequencies are from the bass unit(s) or port(s). The microphone, just like the ear, is unable to distinguish the true source of the sound, so the frequency response is a composite of the entire energy output at low frequencies.

Keele's approach assumes that a microphone placed extremely close to the drive-unit or port will drown the reverberations so that, at an extreme, a kitchen or bathroom could be used as the test room. His objective was to produce a "poor man's" anechoic response without the need for a nonreverberant space. But I don't think that he was attempting to predict how a speaker would measure (or sound) in a domestic listening room; that's a wholly different matter. But if I understand it correctly, Stereophile's particular measurements strategy from Keele's (quasi-anechoic) technique is offered as a prediction of how the speaker will behave in the reader's room—not the chamber. From a quasi-anechoic response prediction to an in-room prediction is a significant intellectual leap for a rather basic measurement technique. Of course, there may conceivably be a speaker and room combination that measures in-room as it would anechoically (or quasi-anechoically), but that's probably rare. But what listening room would the curves actually represent? A New York apartment or a Miami beach house? It's impossible to predict or define a "standard" listening room—were it that simple, it would have been done at the start of the stereo era; instead, the money was spent on designing and building horrendously expensive anechoic chambers to remove the room from the measurements.

During the design, the basic driver parameters are determined anechoically to confirm a smooth bass extension and response. It's then crucial to measure how the speaker behaves in a real listening room. Armed with this knowledge, the crossover can be designed to distribute energy across the spectrum, on and off axis. It is conventional to present this composite speaker/room frequency response divided into narrow blocks of similar frequencies—the so-called (third-) octave response. This gives a far greater insight into the speaker/room response because it usefully reveals trends rather than the individual lumps and bumps every room imparts to even the flattest-measuring speaker. This traditional, tied-and-trusted method, which is how concert halls and studios are evaluated and treated, presents visual information in a similar way to how the ear actually interprets sound. That's been my finding. It also permits measurements to be taken and meaningfully averaged throughout the entire room. Measured in-room, using 1/3- or 1/6-octave analysis, the M40.1 is extremely flat, and this is indeed how it sounds, as Art Dudley clearly appreciates. I believe that these narrowband octave measurements would be a very useful additional tool for the reader to better understand energy-spectrum trends that may be significantly independent of the room. They do not need the leap of faith that turns a quasi-anechoic response into a real listener's room response!

We're all aiming at giving the consumer not only the best products, but disseminating knowledge that empowers him to make the best choices for his room, budget, and taste. My plea is to make maximum use of all the wonderful software tools and measurement techniques available, but to take the greatest care that they are used only within their known, proven confidence limitations—or to openly discuss the limitations of technique and the extreme difficulty of interpreting graphs and numbers. That said, it's the sound that ultimately matters; so much more intellectually satisfying when it's counterbalanced by exceptionally smooth technical measurements.

Thank you for allowing the space for me to register my two cents' worth.—Alan Shaw, Harbeth UK

It is not generally my policy to respond to manufacturers in this space; I have had my say in the review to which they are responding. However, Alan has now twice referred to possible cross-contamination with nearfield measurements, once in his comment on the original review findings and now in response to my Follow-Up in our November issue, in which I examined the M40.1's measured response in Art Dudley's and my rooms (p.125). I believe that the difference in loudness at the microphone position between the output of the drive-unit being measured and the possibly interfering radiator is sufficiently large that the measurement is not significantly affected in the frequency region of interest. That this is the case can be seen from looking, for example, at my nearfield measurements of the Thiel CS3.7's drive-units in this issue (fig.6). The difference in level between the outputs of the woofer (blue trace) and the passive radiator (red trace) at the reflex tuning frequency is 21dB, implying that any cross-contamination from the latter's output into the former's measurements is at most 9% and probably much less.

Ultimately, the worth of a loudspeaker cannot be judged by any one measurement but by the totality of its measured performance and its sound in the listener's room. And on that score, the Harbeth M40.1 is indeed a winning design.—John Atkinson