The Harbeth's impulse response on the tweeter axis, measured with the DRA Labs MLSSA system and a B&K 4006 microphone calibrated to be flat on-axis, is shown in fig.7. The multiple-humped shape betrays the presence of the high-order crossover, while the step response (fig.8) reveals the tweeter to be connected with positive polarity (the sharp up-down spike at 3.5ms), and the midrange/woofer with the opposite polarity (the slower negative-going hump peaking at 3.9ms). The two are obviously not time-coherent. The cumulative spectral-decay, or waterfall, plot (fig.9) reveals an initially clean decay throughout the midrange and treble, disturbed by some general low-level hashiness and interference effects.

Fig.7 Harbeth LS3/5a, impulse response on tweeter axis at 45" (5ms time window, 30kHz bandwidth).

Fig.8 Harbeth LS3/5a, step response on tweeter axis at 45" (5ms time window, 30kHz bandwidth).

Fig.9 Harbeth LS3/5a, cumulative spectral-decay plot at 45" (0.15ms risetime).

Using a simple PVDF accelerometer (footnote 1), I investigated the speaker's structural resonances. Two modes dominated the vibrational behavior of the side panel (fig.10): a main mode at 355Hz that coincides with the small wrinkle in the impedance plot (fig.1), and a lesser one at 120Hz. The 355Hz mode could also be detected on the top and back panels, but to a lesser degree. Both modes were very audible on music while listening to the cabinet talk using a stethoscope. I suspect that part of the LS3/5a's reputation for upper-bass bloom is its cabinet's vibrational behavior, not just the rather undamped woofer alignment.—John Atkinson

Fig.10 Harbeth LS3/5a, cumulative spectral-decay plot of accelerometer output fastened to side of enclosure (MLS driving voltage to speaker, 7.55V; measurement bandwidth, 2kHz).

---

Footnote 1: See Stereophile, June 1992, p.205; and September 1992, p.162.