Icon Parsec loudspeaker Measurements

Sidebar 3: Measurements

I use a mixture of nearfield, in-room, and quasi-anechoic FFT measurement techniques (using primarily DRA Labs' MLSSA system with a B&K (DPA) 4006 microphone, but also an Audio Control Industrial SA-3050A 1/3-octave spectrum analyzer with its calibrated microphone) to investigate objective factors that might explain the sound heard. The speaker's nearfield low-frequency responses and impedance phase and amplitude were measured using the magazine's Audio Precision System One.

The Parsec's plot of electrical impedance magnitude and phase is shown in fig.1: with a phase angle between –3° and +33° in the audio band and dropping below 7 ohms only in the midbass, it implies that the speaker will be an easy load for even inexpensive amplifiers to drive, especially with a measured sensitivity (with a 1/3-octave-wide 1kHz warble tone) of around 91dB/W/m (this somewhat higher than specification). The port tuning is indicated by the minimum at 22Hz and the tweeter "oil-can" resonance by the dimple at 25kHz. Note, however, the glitch just above 130Hz in both magnitude and phase plots; this indicates the presence of a serious cabinet resonance at this frequency.

Fig.1 Icon Parsec, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.)

To look at the cabinet behavior further, I placed the Parsecs facing each other, positioned as closely as possible, and drove them in antiphase with the MLSSA pseudo-white noise signal. In this way, nearly all the direct output of the speakers would be canceled, leaving the cabinet panels as the main radiating surfaces. I then placed the measuring microphone very close to each panel and calculated the impulse response of that panel with MLSSA (fig.2). The resonance at 130Hz or so turned out to be due to the Parsec's back panel which, as noted during the auditioning, was extremely "live."

Fig.2 Icon Parsec, impulse response of back panel just above the terminal panel with drive-unit output canceled (250ms time window).

Fig.3 shows the cumulative spectral-decay plot calculated from its impulse response (a 250ms, 1/5 of a second, time window is shown): both back and side panels could be felt to vibrate strongly between 60 and 90Hz, and a strong ridge can be seen centered on 65Hz. As this is also the frequency of the fundamental floor-to-ceiling mode in my listening room, it is possible that this has affected the measurement. Note instead the ridge at the cursor position, 133Hz, which indicates a strong panel resonance at this frequency. Feeding the speaker with a pure 133Hz sinewave tone revealed that the back panel did indeed vibrate very strongly!

Fig.3 Icon Parsec, cumulative spectral-decay plot calculated from microphone output placed adjacent to center of rear panel just above the terminal panel (MLS driving voltage to speaker, 7.55V; measurement bandwidth, 1kHz).

As tapping the back panel at different places produced different "notes," I repeated this measurement at a point about 6" below the top of the cabinet. The resultant cumulative spectral-decay plot is shown in fig.4. Again, it is possible that the two low-frequency resonances are room modes, but there is a series of strong resonances noticeable as ridges at 385, 280, 205, and 170Hz. (Close inspection of figs.3 and 4 reveals that all the resonances are present in both plots, but at different intensities at the different positions on the rear panel.) These were all frequencies at which this panel felt very lively with sinewave tones. The side panels were generally less live-sounding as judged by the knuckle-rap test; measurement proved this to be the case, though mild resonant modes at 135Hz, 205Hz, and 283Hz were found.

Fig.4 Icon Parsec, cumulative spectral-decay plot calculated from microphone output placed adjacent to center of rear panel 6" below the cabinet top (MLS driving voltage to speaker, 7.55V; measurement bandwidth, 1kHz).

I went into this detailed analysis of the Parsec's cabinet resonances because of the large radiating area of the panels and the persistent colorations I noted on male voice, timpani, and piano to which they undoubtedly contributed. Their less-noticeable effect on other kinds of music I can only put down to the fact that as these resonances appear to be of reasonably high Q and therefore will need to be excited by a signal pretty much having the same frequency, only relatively sonically pure music having a strong content between 100Hz and 300Hz will be colored to any significant effect. In addition, as it was the rear panel that was most resonant and as this faces away from the listener, the subjective effects will be reduced in amplitude. Nevertheless, the presence of the higher-frequency cabinet modes might well have contributed to the subjective hardness noted at high playback levels and both to the occasional feeling of confusion in the lower midrange and the unevenness on treble piano notes.

Turning to the time domain, fig.5 shows the Parsec's impulse response captured on the midrange axis at a distance of 48"; fig.6 shows the step response, which indicates that all three drive-units are connected in positive acoustic polarity and that the step of each smoothly blends into that of the next lower in frequency, which implies optimal crossover design.

Fig.5 Icon Parsec, impulse response on midrange axis at 48" (5ms time window, 30kHz bandwidth).

Fig.6 Icon Parsec, step response on midrange axis at 48" (5ms time window, 30kHz bandwidth).

The speaker was placed on a 36"-high table outside for these measurements, which gave a time window of more than 7ms before the first reflection of the sound, from the floor, reached the microphone. This means that the quasi-anechoic response derived from this impulse response will be accurate down to a frequency of approximately 150Hz. Because of noise from neighbors and traffic, eight separate measurements were averaged to produce the final curve shown. This response, actually averaged across a ±15° lateral window, is shown to the right of fig.7. Impressively smooth through the midrange and treble, it is broken by a degree of prominence in the low mids and the top octave of treble, with a corresponding "saddle" in the low treble. As this latter range corresponds to the brightness region, this slight depression will correlate with the rather mellow treble featured by the Parsec.

Fig.7 Icon Parsec, anechoic response on midrange axis at 48" averaged across 30° horizontal window and corrected for microphone response, with nearfield responses of woofer and port.

Fig.8 shows how the Parsec's anechoic response changes as the listening height changes, these again derived from impulse responses taken outdoors. The curve at the bottom/front was taken on the port axis. This is an unrealistically low axis for listeners other than Snow White's seven companions, but the trace reveals that Dave Fokos has appropriately arranged the drive-unit phasing so that the inevitable crossover cancellation notches are well away from the listening axis. The next curve was taken midway between the ports and the midrange unit, while the central curve is on the midrange axis, this some 36" off the floor, equating with my usual listening height. The next curve was taken on the tweeter axis, while the top/rear curve shows the response 15° above the cabinet top, representing the balance heard by a standing listener at the back of the listening room. This has an excess of midrange energy—remember that I found the speaker's balance on this axis to sound "squawky"—but apart from that, the Parsec should not be too finicky regarding listening height. Listening with your ears either level with the midrange axis or just above will give the smoothest balance, as noted in the auditioning.

Fig.8 Icon Parsec, vertical response family at 48", normalized to response on midrange axis, from back to front: differences in response 10–5° above axis, reference response, differences in response 5–10° below axis.

The lefthand side of fig.7 shows the responses of the woofer and ports, measured in the nearfield. The woofer can be seen to peak up a little in the midbass, with then a reasonably slow rollout, reaching –6dB at a low 39Hz. Although not shown, repeating the nearfield measurement with the MLSSA system set to a bandwidth of 1kHz reveals the woofer to have a very clean spectral decay, with minimal resonant behavior apparent until well above its passband. The response of the ports is somewhat enigmatic, as although their apparent tuning of 20Hz agrees with the plotted impedance in fig.1, it rolls off much less smoothly than usual, with blips noticeable at 60Hz and 80Hz, these probably not coincidentally the third and fourth harmonics of the port tuning. A blip can also be seen at 133Hz, the rear-panel resonance frequency. The port's output also rolls off more gradually than usual, being only 12dB down at 300Hz.

How the output of the ports, with their minimal radiating area, and that of the woofer integrate is hard to assess from the individual nearfield measurements, which is one reason why I carry out an in-room 1/3-octave measurement, spatially averaged across a 72"-wide window at the listening position to minimize the effects of room resonances. This is shown in fig.9 and was taken with the speakers at the original 5' distance from the side walls. As suspected from the auditioning and predicted by the nearfield curve in fig.7, there is a degree of mid- and upper-bass exaggeration in-room, perhaps due to the closeness of the floor boundary to the drive-unit. A generally sloping-down response from the upper bass to the midrange is also apparent.

Fig.9 Icon Parsec, spatially averaged, 1/3-octave response in JA's Santa Fe listening room.

Note, however, the Parsec's astonishingly smooth response throughout the upper midrange and treble. This is due both to the speaker's intrinsically flat response in this region and its wide, even dispersion, and will contribute to the Parsec's natural reproduction of treble instruments and voices. In the low bass, the ports reinforce the woofer output to give a –6dB point referenced to the level at 200Hz of 25Hz—excellent in-room bass extension if not flat to the port tuning frequency, this presumably due to the overdamped low-frequency alignment.

Finally, fig.10 shows the cumulative spectral-decay plot for the speaker on the midrange axis at 48", derived from the impulse response in fig.5 and featuring a frequency resolution of 90Hz or so. The tweeter resonance again can be seen at 25kHz; apart from that, the treble is pretty flat and clean, apart from a degree of liveliness noticeable around 3kHz and 5kHz, probably due to midrange-unit breakup modes. The cursor position shows that there is something peculiar happening around 1250Hz, there being a complicated pattern of a ridge broken by reflections at this frequency. This behavior might again correlate with my feeling of occasional hardness in the low treble at high levels and the propensity for some high-piano notes to jump forward.—John Atkinson

Fig.10 Icon Parsec, cumulative spectral-decay plot on midrange axis at 48" (0.15ms risetime).