Floor Loudspeaker Reviews

Sidebar 2: Measurements

Fig.1 shows the MG2.6's electrical impedance magnitude (solid line) and phase (dashed line). With a value averaging 5 ohms in the bass and 3.5 ohms in the treble, with a rise in between due to the crossover, the speaker is a reasonably demanding load in terms of current draw, though the phase curve indicates it to be resistive—ie, not significantly departing from a 0° phase angle—over much of the audio band. The MG2.6s shouldn't give good tube amps any problems as long as they're driven from the 4 ohm output taps. The very slight wrinkles at 50 and 60Hz presumably indicate the Magneplanar panel's fundamental "drumskin" resonances. Its A-weighted sensitivity measured around 86dB/3.83V/m, which means that a 100W amplifier would provide more than enough drive to raise satisfyingly loud levels in normal-size rooms.

Given the quite different sounds produced by the Maggie with the amplifiers with which I used it, I looked at the frequency response of my source and amplification components measured at the loudspeaker terminals (fig.2). The response with the Audio Research Classic 60 amplifier (solid line) can be seen to droop slightly in the treble and further emphasize the entire midrange region as a result, though the lack of HF air when compared with the Mark Levinson No.23.5 (dashed line) was more audible than the 0.5dB difference above 10kHz shown in the graph would indicate.

Turning to the MG2.6/R's frequency response, the top curve to the right of fig.3 shows the quasi-anechoic response averaged across a 30° horizontal window midway up the ribbon at a 44" distance. (The B&K measuring microphone's response has been subtracted mathematically from this curve to give the true speaker response. The sharp notch at 20kHz is an interference effect and should be ignored.)

Drawing inferences from such measurements of large panel speakers is fraught with difficulty, particularly at the relatively close distance mandated by an FFT measurement technique when performed in a room. The fourth edition of Martin Colloms's High Performance Loudspeakers, pp.147–151, examines this problem using the 1986 work of Stanley Lipshitz and John Vanderkooy, and indicates that the close microphone distance will result in a measured mismatch between the speaker's on-axis output above and below 1kHz, the treble region apparently being shelved down. Remember, however, that I did find the 2.6/R to sound somewhat warm and thickened in the midrange. There is probably still a residual mismatch, therefore, in the speaker's mid and high treble compared with its low treble and upper midrange (though it's possible that this will diminish as the size of the listening room increases).

The lower curve shows the low-pass rollout of the Magneplanar panel, with its level approximately matched to that of the complete speaker. Coupled with the fact that the electrical drive signal to the ribbon is down 6dB at 750Hz, this indicates that the actual crossover frequency appears to be around 700–800Hz. Though this is lower than the 1kHz specified, it seems sensible in that most instrumental and vocal musical fundamentals will be handled by the panel, most of the harmonics associated with those fundamentals by the ribbon.

To the left of fig.3 is the low-frequency response of the Magneplanar drive-unit measured with the microphone capsule almost touching the grille cloth in the dead center of the panel. Again, interpretation of such a measurement with such a large diaphragm is dangerous, but it would seem to indicate that the fundamental bass tuning at 60Hz is of reasonably high Q, with a rapid rolloff below that frequency. While all but the lowest notes of the double bass will be satisfactorily reproduced by the MG2.6, this curve implies that it will never be a speaker for low-bass freaks, again something that was observed during the auditioning.

Remember that I found the exact degree of toe-in to be critical with this speaker when it came to getting the optimum balance between the midrange and treble. I therefore investigated the manner in which the speaker's response changes as the listener moves to either side of the ribbon axis. Fig.4 shows the differences in response that occur as the listener moves from 15° off-axis on the ribbon side of the speaker to 7.5°, 15°, then 30° off-axis on the bass panel side. (To produce this graph, the response on the ribbon axis has been subtracted from all the other curves so that just the differences that would occur if the ribbon-axis response were perfectly flat are revealed; the latter therefore appears as a straight line.) Ignoring the spikes and dips due to innocuous interference effects, it can be seen that moving to the panel side of the speaker—placing the speakers with the ribbons to the speakers' outside edges and carefully adjusting the toe-in—alleviates the midrange prominence by introducing a degree of suckout at the crossover frequency. (With side-by-side drivers, such crossover lobing appears in the horizontal rather than the vertical plane, which is usual with conventional dynamic speakers.)

Fig.5 shows a family of curves analyzing what happens as the listener moves from a position with his or her ears level with the base of the ribbon—an unrealistically low listening height of 30"—to one level with the top of the ribbon—this representing a tall person standing. Again, just the differences are shown: they imply that the speaker will sound too shrill for a standing listener and too dull for someone sitting too low. The best balance would appear to be the second curve from the front, this representing a listening height of 40", with again a lack of energy in the upper midrange compensating for what would otherwise be too "thickened" a tonal balance.

One can draw inferences about how a speaker will sound in a room from the anechoic response and response differences in figs.3 through 5, but there's no substitute for actually measuring the perceived in-room balance. This, taken in my listening room, spatially averaged to minimize the effect of low-frequency room resonances, and with the microphone response compensated for, is shown in fig.6. Admirably smooth and flat from the upper midrange upward, which doubtless correlates with the speakers' seamless presentation of instrumental and vocal colors, an excess of energy in my room can still be seen in the mid-to-lower midrange; this corresponds to the warm tonal balance noted. The in-room bass appears well-balanced and drops off rapidly below 38Hz or so, as implied by the specification.

Turning to the time domain, fig.7 shows the speaker's step response, which supports my suspicion that the ribbon is inverting. Processing the time data to show how the speaker's response changes as the impulse decays gives the "waterfall" plot in fig.8. This appears somewhat hashy in the treble, but I'm not sure whether this is due to the presence of HF resonances in the ribbon or to the physical nature of the driver (footnote 1). The Magneplanar panel's time-domain behavior in the midrange also appears from this graph to be complicated.

I did note some treble hardening of the sound at very high levels with music program, as well as a somewhat "dirty" sound to pure tones in the low treble, so I looked at the speaker's distortion performance in this frequency region. The MG2.6/R appears to be a very linear loudspeaker at sound pressure levels below 86dB at 1m, with the main distortion component, the third harmonic, being below 0.3%, or –50dB, over most of the audio band. Raising the spl of a tone at the bottom of the ribbon's passband (at 1.6kHz, where the woofer drive is 27dB down) to a very loud measured level of 96dB at 1m (footnote 2), however, just under the point where the 2.5A tweeter fuse would blow, gave around 1.4% of third-harmonic distortion with around 0.5% of sub-harmonic content present, centered on 1kHz.

This behavior seemed confined to a span of an octave or so in the low treble, as can be seen from figs.9 through 11, which show the distortion spectra of 800Hz, 1.6kHz, and 5kHz tones, respectively, all taken at an indicated 91dB at 1m. The MG2.6/R reproduces the 800Hz (fig.9) and 5kHz (fig.11) tones at this high spl with just 0.3% (–50dB) of distortion in total. By comparison, the distortion of the 1.6kHz tone in fig.10 is three times as high, at 1.0% THD (–40dB), with the third harmonic at 0.6% being joined by appreciable amounts of second and fourth harmonic, as well as some sub-harmonic content. While the 800Hz and 5kHz tones did sound relatively pure—though the third harmonic of the latter at 15kHz is still within my hearing range, its level was below audibility—the third harmonic of the 1.6kHz tone at 4.8kHz, an octave-and-a-fifth above the fundamental, was surprisingly audible (perhaps not so surprising, considering the fact that the lower harmonics of a 1.6kHz sinewave fall in the region where the ear is most sensitive). The presence of subharmonics lent the tone an unmistakably buzzy quality.

Remember that these distortion measurements were made at a very high level, not too far below that at which the speaker's tweeter fuse blows. However, they reveal that Magnepan had to compromise ultimate dynamic range to produce a relatively affordable Magneplanar loudspeaker that incorporated a ribbon drive-unit. Magnepan's necessary decision to make the MG2.6/R a two-way design means that the tweeter has to be taken to a sufficiently low frequency to meet the panel woofer, but it then runs out of excursion capability relatively early. (I'm reminded of the similar dynamic-range problems possessed by the Avalon Eclipse and Celestion 3000, both two-way speakers that again cross their high-frequency drivers over below 1kHz.) If you like the sound of the MG2.6/R but find that it doesn't quite go loud enough for your tastes, you should turn to the three-way MG3.3/R at $2850/pair, which features a dedicated Magneplanar driver covering the midrange from 200Hz to 1kHz.—John Atkinson

---

Footnote 1: It has been postulated that though large, nonrigid-diaphragm drive-units such as electrostatic, ribbon, and Magneplanar designs are driven uniformly, their actual behavior can be chaotic (in mathematical terms). In effect, each little part of the diaphragm "shivers" about the panel's average position as the latter moves back and forth under the influence of the electromagnetic or electrostatic drive signal. The corrugation that Magnepan and other manufacturers of ribbon units apply to the diaphragm acts to keep it moving as a whole. The ribbon's motion is still complex, however. Shining a flashlight upon its rear surface with music playing reveals a degree of local twisting behavior, this excited by high levels of lower-midrange energy and bass-rich transients but constrained by the ribbon's alternate side clamping.

Footnote 2: My comparative MLSSA measurements of the same loudspeaker sample at sea level and at Santa Fe's 7000' altitude indicate that a drop of sensitivity of approximately 3dB is to be expected as a result of the reduced air loading here in the Desert Southwest. 3dB should therefore be added to my indicated SPL figures to obtain the equivalent distortion performance at sea level. In other words, if I started to be bothered by some treble hardness at 93dB, you should be able to push a pair of 2.6/Rs to an SPL of 96 or 97dB before you experience the same amount of hardness.