Harbeth P3ESR loudspeaker Follow-Up, October 2010

John Atkinson returned to the Harbeth P3ESR in October 2010 (Vol.33 No.10):

When I reviewed the latest iteration of Harbeth's P3 minimonitor, the P3ESR ($2195/pair), for the August issue, I found that this pocket-size speaker punched above its weight. Other than its necessarily limited low frequencies and lack of power handling, the P3ESR was not out of place in an expensive high-end system, particularly for those with a small room. However, I had a problem with the samples sent for review (serial nos. 0472L and R): Toward the end of the review period, I was giving the speaker a workout with some high-level rock when the soundstage lurched to the right and the balance became bright and brassy. At continuous levels below about 7V RMS (equivalent to 8W into the Harbeth's 6 ohm impedance), the responses of both speakers were the same. But at higher sustained levels, serial no.0472R developed a severe peak in its low-treble region. This peak disappeared when I backed off the level.

This didn't affect my review findings, but it turned out that this pair of speakers had done quite a lot of traveling around retailers and other reviewers before coming to me. It's possible, therefore, that a crossover component had been overstressed during that time and was now prone to intermittent failure.

Fig.1 Harbeth P3ESR, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.) Serial nos. 0672L and 0672R superimposed.

Walter Swanbon, of Harbeth's US distributor, Fidelis, sent me a second pair of P3ESRs (serial nos. 0672L and R). Before listening to these, I measured them using DRA Labs' MLSSA system and a calibrated DPA 4006 microphone. First, the impedances of the two speakers, measured with an Audio Precision System One, are shown in fig.1. The two speakers match very closely, with a single peak at 70Hz indicating the tuning of the sealed enclosure. The impedances are basically identical to those of the original samples, with a value that drops below 8 ohms only in the lower midrange. This speaker will be easy for the partnering amplifier to drive.

Fig.2 Harbeth P3ESR, anechoic response on tweeter axis at 50" without grille, averaged across 30° horizontal window and corrected for microphone response, with nearfield woofer response plotted below 300Hz: serial nos. 0672L (blue), 0672R (red), 0472L (green, offset by –3dB).

Turning to the frequency domain, the new samples (fig.2, blue and red traces) are superbly well matched, the two traces overlapping at almost all frequencies. The response of 0472L, taken under identical circumstances, is shown as the green trace in fig.2, reduced by 3dB for clarity. Above 1kHz, its response is almost identical to that of the newer samples: All are superbly flat. Below 1kHz, however, there is a little more energy in the middle of the midrange, and the peak in the upper bass is about 1dB lower in level. It's possible that this difference is due to the older samples having had a lot of use before I measured them.

I set up the new samples of the P3ESR on massive single-pillar Celestion stands, using the same system I'd used for most of my auditioning of the earlier samples: dCS Puccini SACD player, Simaudio Moon Evolution P-8 preamplifier, and Classé CTM-600 monoblocks, all connected with AudioQuest Wild interconnects and speaker cables. The new Harbeths were as good as I remembered the old ones sounding, with an uncolored midrange, smooth highs, and well-defined if somewhat lightweight lows. The upper bass offered good bloom and didn't sound any different, and I could drive this new pair to high levels without the sound changing character or anything failing.

The problem encountered with sample 0472R must, therefore, have been a one-off event, and I can reaffirm my strong recommendation for this immaculate-sounding minimonitor.—John Atkinson

Harbeth Audio Ltd.
US distributor: Fidelis AV
14 E. Broadway (Route 102)
Derry, NH 03038
(603) 437-4769

claud's picture

Might the apparently upgraded P3ESR-XD merit remeasuring of its off-axis response?

hemingway's picture

I would also be very curious to see how the XD update has affected measured and subjective performance (especially measured).