Celestion 3000 loudspeaker Page 3
Continued listening confirmed the 3000's schizophrenic personality. At levels much lower than I usually play music, the sound was quite pleasant. But approach replay levels in the high '80s, where normally music in the listening room starts to cook, and the sound, as well as being bright, became both unlistenably hard in the treble and confused in the lower midrange. Thinking that this might be a basic incompatibility between the loudspeakers and the Mark Levinson No.20.5s, I replaced them with an Audio Research Classic 60. The sound with the speakers powered by four pairs of 6550 tubes was somewhat less aggressive in the highs than with the solid-state amps, but the basic problematic sound quality remained.
Celestion's 3000, at least to judge from the sound of these first review samples, is a "quietspeaker" rather than a loudspeaker. This conclusion caused me a great deal of heartsearching, as Ken Kessler, in the January 1990 issue of HFN/RR, had raved about the 3000's euphonic quality and the way it could convey music's "power and scale without compression." Was I listening to the same speakers? Certainly the 3000s at the 1990 Winter CES had sounded bright, but they seemed free of the loudness threshold above which I felt driven from the room.
It seemed time to do some measurements to try to track down what was going on. And as you'll read later, I found that one of the review speakers generated large and audible amounts of intermodulation distortion once the playback level got much above 80dB. Coupled with the bright balance, this undoubtedly correlated with the unpleasant hardness I was experiencing. It seemed fair to Celestion to inform them of my findings; this problem didn't seem something that they would not have noticed and it therefore might not be typical. I therefore requested that they send a completely new pair of loudspeakers and put the review on hold while I busied myself with other projects.
When the replacements arrived, I reversed my normal routine and measured the speakers before listening to them. If the measurements were identical, then I would know that the first pair had been typical after all. But if the measurements were significantly different, then I should, in effect, start listening afresh.
And that, in fact, was the case. As you will see later, the new samples had lower levels of intermodulation distortion; in addition, there was more top-octave energy apparent, so it was time to sit down for a second period of serious listening.
Without anything like the first pair's threshold limitation on loudness, I was able to play recordings at levels where the music was allowed to sing. And at such levels, peaks reaching 96–98dB, it became apparent that the 3000 allowed a wealth of recorded detail to be heard. (Higher in level than that and the ribbon could be heard to buzz a little; I thought this was the amplifier clipping, but the Classic 60 was only putting out 11V RMS, which is within its limitations.) Partly this detail was due to the speaker's tonal balance, which was still rather mid-treble–forward, but there was also a freedom from the kind of low-level confusion that afflicts, for example, many soft-dome tweeters.
It was very easy to hear recorded ambience and reverberant decays; conversely, background noises on recordings were also rendered more audible. On the superb recording of Gershwin's "Someone to Watch Over Me" on Sting's A&M 3" CD (also available in the UK as a normal-size CD single), you can clearly hear on the Celestion 3000s that there is a faint rhythmic rustling noise accompanying the piano and voice. Perhaps there was originally a brushed-snare drum accompaniment that was later deleted from the mix, its only legacy being faint noises escaping from Mr. Sumner's foldback headphones and faithfully reproduced by these Celestions.
This transparency more readily revealed ambient backgrounds on recordings that I thought I knew well, a greater sense of space, a "bigger" sound, being the result. Though the conventional wisdom holds that speakers sited near the rear wall will throw less image depth than speakers out in the room, due to the early reflections from the wall confusing the ear's decoding of reverberant information, this didn't seem to be true with the 3000s, perhaps because they reproduced so much more of that low-level reverberation. The depth test on the Chesky Test CD (Chesky JD37) consists of male voice and tambourine recorded further and further away from the microphone. The voice could be heard to recede in a lifelike manner, though the tambourine was significantly closer than the voice, perhaps due to the speaker's rather forward treble accentuating the sound of its jingles. The relative positions of the flute and piano on Stereophile's Poem LP were also well-decoded by the 3000s, the dome of ambience around and behind the instruments being very audible without obscuring their direct sounds.
Laterally, imaging was both stable and precise, again surprising in view of the close rear-wall placement. While the speaker doesn't have the holographic imaging of, say, the Spica Angelus, it's still pretty good in this respect. The Chesky height test was less accurately reproduced; though the test signal could be heard to move above the plane of the speakers, it was extremely unstable and smeared toward the middle of the soundstage. This is undoubtedly due to reflections of the sound from the rear wall. Moving the speaker further out into the room to improve things made the sound too lean overall, however.