Harbeth M40.1 loudspeaker JA November 2008
Art Dudley reviewed the Harbeth M40.1 loudspeaker in October (p.153). A large, stand-mounted design in the British BBC tradition, the M40.1 costs $12,995/pair, plus $595/pair for its substantial, 14"-tall stands.
Art was extremely positive about the M40.1's performance. "Harbeth has the rare ability to bundle tonal neutrality with excellent performance in virtually every other regard that matters," he said, and concluded that, "Here, finally, is a loudspeaker that achieves flat frequency response without sucking all the life out of recorded music." However, he had problems getting the most neutral output from this pair of physically large speakers in his room (a point that was driven home when I tried the Harbeths in my own, much larger living room): "There were challenges in getting such a wide cabinet far enough from the sidewalls to avoid minor to moderate bass cancellations while still enjoying a sufficiently wide soundfield," he wrote. "Even the best-case placement, with the fronts of the speakers just over 6' from the front wall, didn't give perfectly smooth response throughout the bottom octaves. . . . The smoothness of the M40.1's bass response, though acceptably good in a small room, improved as the room boundaries became more distant. . . . Still, the M40.1's bottom-end response was solid to well below Harbeth's claimsonly 2dB down at 31.5Hz in my small room."
When I measured the big Harbeth's frequency response (See October, p.157, fig.3), it did appear that the speaker's bass response was elevated somewhat. But, as I went into in some detail in my 1999 article on measuring loudspeaker frequency response, not having access to a large, expensive anechoic chamber, I assess a loudspeaker's low-frequency response by measuring the drive-unit and port outputs in what is called the "nearfield"; ie, with the microphone capsule almost touching the radiator. First described by veteran speaker engineer Don Keele in 1973 (footnote 1), this technique gives a response that accurately reflects a loudspeaker's low-frequency output as assessed in the farfield, with the important caveat that, mathematically, it assumes that the loudspeaker's front baffle continues to infinity in all directions, which boosts the measured low-frequency response by up to 6dB. I have pondered over the years whether to apply a correction for this boost, but that correction will not be appropriate for any but very large rooms or anechoic chambers. In the end, I leave the curve uncorrected for this 2pi boost (which I did mention in the review text).
In his "Manufacturer's Comment" (October, p.203), Harbeth's Alan Shaw took exception to this measurement technique, feeling the 2pi midbass boost misleading. He had sent me a true anechoic frequency-response measurement for the Harbeth M40.1, this taken in the BBC's anechoic chamber, which is large enough to allow a reflection-free time window of 300 milliseconds for the measurement. His and my measurements were broadly similar in the midrange and treble, allowing for differences in mike distance and measuring axis, etc., and the shapes of the two curves were broadly consistent at low frequencies. However, compared to his anechoic curve, my nearfield curve was elevated by +5dB or so with respect to the reference level at 1kHz. "What is the real frequency response?" Alan asked in his comment. "Easy answer: Trust the anechoic measurements above all others!"
But no one listens in an anechoic chamber. They listen in rooms of varying sizes. If you think about it, a measurement taken in an anechoic chamber and a nearfield, 2pi measurement are at opposite ends of a continuum of measurement conditions. The first assumes the room boundaries are, in effect, infinitely far away; the second that one of the room boundaries is infinitely close. A speaker that measures flat in the low bass, such as the Harbeth M40.1, will sound as if it has a bass boost in a real room, the amount of that boost depending on the room size.
To investigate this matter, I took the spatially averaged response at the listening position of the Harbeth M40.1s in my Brooklyn room. I then drove the speakers back to Art in upstate New York, and took their spatially averaged response at the listening position in his room.
The results are shown in fig.1 (footnote 2). The fact that my room (blue trace) is larger than Art's (red) allows me to move the speakers usefully farther away from the sidewalls. This flattens out the peaks and dips throughout the midrange decade (200Hz2kHz), though the shape of the curve is similar. In addition, the Shindo tube amplifier I used at Art's has a much higher source impedance than the solid-state Musical Fidelity I used in my room, which increases the height of the peaks around 1kHz and 4.5kHz, which coincide with an increased impedance magnitude in those regions (see fig.1, October 2008, p.155).
Fig.1 Harbeth M40.1, spatially averaged in-room responses in JA's room (blue trace) and AD's room (red), taken under identical circumstances.
The M40.1's spatially averaged response rolls off a little more in the top octave in Art's room than it does in mine. I think this is due to three factors: 1) The Shindo's source impedance will probably increase above 10kHz, which will reduce the top-octave level a little (again, see the impedance graph); 2) my room may well not be as absorptive as his in the top two octaves; and 3) probably most significant, in my room I aimed the Harbeths straight at my listening chair, whereas Art arranges the toe-in so that he listens somewhat off-axis to the tweeters. Because of the speaker's very limited dispersion above 3.5kHz (see fig.4, October, p.157), this will reduce the level in this region with this sort of spatially averaged response. I certainly felt the Harbeth sounded more mellow, more polite in Art's room than it had in mine.
At the other end of the spectrum, Art's smaller room boosts the midbass more than mine does, but then, his lighter wall constructionmy room is solid-walled for the first 4' above the floordoesn't support the bottom octave quite as much. But the speaker still has a little more upper and midbass in my room than I have found to sound natural. While I thought the Harbeths scored highly in vocal reproduction and imaging in Art's room, with smooth, clean highs, the lumpy, excessive bass did eventually irritate me.
As is almost always the case with a large speaker that measures anechoically flat down to a low frequency, the Harbeth measures and sounds a touch midbass-heavy in a smallish room. Therefore, the larger the room, the better, especially as the smoothness of the speaker's midrange requires it to be well away from room boundaries. As Art found, the M40.1's treble balance can be fine-tuned by experimenting with toe-in. However, with the listener sitting straight on the tweeter axis and using solid-state amps, the treble does become fatiguing, I felt. While solid-state amps will better control the Harbeth's excessive in-room bass, a tube amp will probably work better with the highs.
Overall, the Harbeths did some things very well in Art's roommy new choral recording of Cantus sounded beautifully unforced, for example, with a very natural midrange quality. But on balance, I would rather live with the Audio Note AN/Es in Art's room, if not in mine.John Atkinson
Footnote 1: D.B. Keele Jr., "Low-Frequency Loudspeaker Assessment by Nearfield Sound-Pressure Measurement," Journal of the Audio Engineering Society, Vol.22, p.154 (March 1974); reprinted in R.E. Cooke, ed., Loudspeakers Vol.I, pp.330338 (Audio Engineering Society, 1978). See also C.J. Struck and S.F. Temme, "Simulated Free Field Measurements," Journal of the Audio Engineering Society, Vol.42, pp.467482 (June 1994).
Footnote 2: The in-room response of a reader's pair of M40.1s, measured with a TacT system can be found here. The same step down above 1kHz as in my and AD's rooms is evident. The top two octaves are similar to the speaker's response in my room, though the bass behavior appears to be closer to how the speakers behaved in Art's room.