Within a few minutes of firing up Harbeth’s HLP-3ES2 mini-monitor, I smiled and thought, “I know this sound. It’s every bit as good as I remember.” For several years in the nineties, I used a pair of the original HL-P3s in my film work. Replacing LS3/5as, they displayed far more neutral tonal balance, greater timbral accuracy, much lower coloration, better bass, and wider range. When I had just finished a project with several months to go before the next and no place for the speakers to go except storage, I finally gave in to a musician friend who, after carefully auditioning several small speakers (including LS3/5as), had been begging me to sell them to him. Come the next project, Harbeth was in the process of changing U.S. distribution. I never did get around to replacing the P3s.
Meanwhile, since 1999 I’ve reviewed three Spendor mini-monitors and Stirling Broadcast’s resurrected LS3/5a (in Issues 119, 143, 166, and 182). As this might suggest, I’ve had something of a longstanding romance with mini-monitors, especially when used in the applications for which they were originally intended: high?accuracy reproducers in settings too small to accommodate larger speakers, spaces that typically cannot support low bass and where very loud playback isn’t required. But I wouldn’t choose them as main speakers for normal and larger listening rooms because their limitations become more difficult or impossible to overlook. And since optimal performance requires stand-mounting away from boundaries, their small size doesn’t really even save all that much space. Still, there’s something about the sheer impossibility of the problem and the ingenuity of the various solutions that appeals to me.
Designed by Harbeth’s owner Alan Shaw and introduced in 1990, the HL-P3 and its successors were by no means the first mini-monitors in the wake of the LS3/5a. But I believe they were the first to build directly upon the BBC research that went into the LS3/5a and upon the 3/5a itself, right down to lifting one of its hat?tricks, a little boost in the upper bass to suggest more bottom?end than is actually there.
Slightly larger than the LS3/5a, the ES2 is otherwise similar, being a sealed two?way with SEAS drivers instead of KEFs. This is, in fact, the second revision of the original P3, the first being the ES, the differences mainly in the crossover, treatments for the cone edges, and a rounded cabinet?edge in the ES2. Impedance is complex, but nominally 6 ohms (treat it as 4 with tube amplifiers). Sensitivity remains a low 83dB, recommended minimum power 25 watts—a mite optimistic for a normal?sized room. In my plus?2600?cubic?foot space, I managed to make a high?current 110Wpc amplifier clip before the speaker gave out (this may have owed partly to the ES2’s complex impedance). For most of the evaluations I used the reference McIntosh MC?402 (400Wpc) and Quad 909 (140Wpc). Think a good 50 watts minimum and don’t be afraid to use 100.
For this speaker’s something of a little giant when it comes to loudness, doing a commendable, even an impressive job reproducing the dynamics of Richard Goode’s Beethoven sonatas (Nonesuch) and placing the piano in the room. It also played the Kings College Advent service on Argo LP, large organ and all, to levels too loud to talk over. But when the big stuff, deep stuff, powerful stuff came along—organ, piano music in its lower registers—the sound was perched right on the edge of stress and strain, where I found it preferable to ease back.
Yet Shaw is such a gifted designer that the ES2’s low end manages things none of the other minis does. For one thing, that upper?bass boost is to my ears subtler and better integrated than the LS3/5a’s, which always sounded a little coarse to me. For another, the ES2 is good enough to suggest a bit of the low?end ambience of recorded venues, more difficult than reproducing top?end air. Finally, it will actually reproduce bass drums and other low?end instruments, not room?fillingly, but with more than a mere suggestion that they are actually there. And string bass is handled convincingly, without the bloom of a larger speaker, but with superb articulation and harmonic integrity backed by surprising strength and even some power. Ray Brown on This One’s for Blanton (45?rpm vinyl reissue) will show you what I’m talking about. So will Harmonia Mundi USA’s Eroica, Andrew Manze urging his medium?sized orchestra to play in the style of the early classical period, with sonorities lean and sec. But the Harbeths clearly reveal that these are modern, not period instruments, double?basses and cellos coming through with sufficient warmth and foundation to balance the spectrum. And the virtuoso timpanist has to be heard to be believed in the Marcia funebre, his attack reproduced with stunning clarity, ferocity, and even some size by these little buggers.
In other words, while Shaw hasn’t rewritten the laws of physics when it comes to bass performance and loudness capability, the ES2 is the only mini-monitor I could live more or less happily with as primary speakers if I had to. And because the bass is so well behaved, this is one mini-monitor that really will mate well with a good subwoofer (the LS3/5a has always been the very devil in this regard).
The tweeter is that rarity, a metal dome that doesn’t ring or otherwise constantly point to itself. One of my notes reads, “The highs—clean, clear, pretty characterless, practically perfectly balanced.” This is not just because the tweeter doesn’t rise in the manner typical, say, of so many non?BBC?oriented British speakers these last twenty years, but because it is very extended yet of a piece with the mid/low?end driver. I am reliably informed that Shaw expends fanatical effort in all his designs in getting the drivers to dovetail coherently at their crossover points. It shows, but experiencing this coherence requires on?axis listening and stands that bring the tweeter to ear level.
Some people find the LS3/5a’s highs more detailed. I disagree. They may sound more detailed because its top end has a few peaks that accentuate detail. Put the same peaks into the Harbeth with an equalizer and voila!—there’s the same detail, the differences being: (1) that it’s bogus, and (2) that you can switch it out. The ES2 reproduces details in truthful proportion to what else is on the recording.
Which brings us to the midrange. I’ve already mentioned its neutrality and freedom from coloration, with timbral accuracy superb by any standard. But there is a small anomaly, a tiny rise in the 1kHz–2kHz range. With most program material, you might not even notice it, though pink noise reveals it. And unlike the LS3/5a’s similar but bigger rise in the same region, the ES2 never gets nasal. Instead, it provides a bit of extra presence?projection that makes for a very open sound. But it also makes the speaker sound, again in common with quite a number of mini-monitors, a bit more midrange-y than it might otherwise. (Spendor’s S3/5R is dead flat in the same region.)
One recording that lets you hear this anomaly immediately is the LP of Doris Day’s “Over the Rainbow” from Hooray for Hollywood. Day’s voice is recorded with glorious presence, richness, and warmth, but on the Harbeths it can sound slightly too forward, as in a little bright, especially as it approaches its loudness limitations. Pulling back the 1.2kHz band on the McIntosh C46 preamplifier’s equalizer to around 10–11 o’clock makes her voice sound just right. On the other hand, play Shelby Lynne’s new tribute to Dusty Springfield, Just a Little Lovin’, and she sounds too distant EQ’ed this way, just right bypassed.
In other words, I don’t want to overemphasize this anomaly; it’s not noxious and it truly is subtle. Moreover, from the top of the upper?bass through the highest octave, fewer than ten percent of speakers on the market, I’d guess, are as flat and thus as accurate in frequency?response as the ES2. It really earns its “monitor” moniker.
The ES2 images more or less as all these subcompacts do: with near holographic precision that belies its small size, albeit with the usual reduction in image size and scale, less so here than with most. The Harbeth is more open and somewhat bigger sounding than the others, less boxy too. And because Shaw has paid his usual careful attention to reducing diffraction effects, you won’t hear the typical beyond?the?speaker?boundaries soundstaging artifacts, beloved of audiophiles but hardly accurate.
Is there an issue of value here? Of the five mini-monitors I’ve reviewed, the cheapest is $1499/pair, the costliest these Harbeths at $1895/pr. And I don’t know of comparably priced larger speakers that equal the sheer accuracy of the best mini-monitors over their admittedly restricted frequency range. But my colleagues have written about several quite good larger speakers in this price range that, in addition to playing louder and responding deeper, will also suggest a more life-size scale in their reproduction. Which matters more to you: high?accuracy, restricted frequency and dynamic range, reduced scale versus deeper bass, greater loudness, a bigger presentation at some sacrifice to ultimate tonal accuracy? Not a decision I can make for you . . .
Meanwhile, the HLP-3ES2 sufficiently pushes the boundaries of mini-monitor performance, while reducing the compromises imposed by the small size, that I’ve made a decision of my own: I bought the review pair, and this time I will not make the mistake of selling them any time soon.
Interview with Alan Shaw
PS: It's rumored that you never really liked the LS3/5a and that one reason you made the HL-P3 is to improve upon the original design.
AS: Before answering that, I’d like to say that I have the greatest respect for the LS3/5a, especially when you consider when it was designed. But after the euphoria of becoming a BBC?licensed supplier around 1988, I didn’t really listen to the 3/5a for a couple of years. Then I made recordings of my daughter and was quite disappointed how colored she sounded on it in the presence area. Far, far too much energy. So I designed the P3 to reduce that overall intensity.
PS: One of your LS3/5a criticisms concerned its inability to stay in spec over time.
AS: The performance of the LS3/5a bass unit dramatically changes with time, and after about twenty years a peak of around 5–10dB centered on 1300Hz is very typical. It’s related to chemical degradation in the rubber surrounds used in the original 3/5 (PVC in the later—and all Harbeth—computer?optimized ones).
PS: You seem to have used some of the LS3/5a’s “tricks,” however, like the upper?bass boost to suggest weight and warmth.
AS: To my way of thinking this was mandatory to stand any chance of selling the P3. But the interesting thing about the bass hump is it’s a hump only to the eye on the graph. In reality, it’s a perfectly executed Butterworth second?order alignment. What’s more interesting is what follows the hump: a shallow contouring of the drive from 120Hz to about 1000Hz. The way the ear works, this 125Hz stands out against the general contour and this is what gives the apparent weight. The voltage plots on our Web site show what’s going on in detail.
PS: I also hear a bit of a rise in the presence region, albeit rather more subtly applied.
AS:Yes, a little, but that’s essential. It lifts the perceived loudness of the speaker, it draws attention away from the “sound of the cabinet,” it enhances stereo imaging, and it brings the soundstage forward, clear of the box as it were.
PS: Why don’t you use your RADIAL compound in the P3?
AS: The P3 predates RADIAL, and tooling?up for a Radial driver would cost around fifteen to twenty thousand dollars. Since the performance of the P3 is well known and liked, especially in professional circles, and the speaker continues to sell very well, this can hardly be a priority for a company the size of ours.
PS: Can you tell us something of how you go about designing speakers? I know you use a combination of exacting measurements and intensive listening. When you start listening, what are you adjusting? The crossover, the cabinet, the parts? What are you listening for?
AS: This really is a first class question that I’ve not been asked before. I need a little time to give you an answer.
In about a week, he wrote me the following reply:
AS: My heroes were all at the BBC. In a quasi?government, engineering-based organization like the BBC, where public money is being spent, there is no room for pure subjectivism. There is a hierarchy that ensures solutions are 100% engineering-based, repeatable, on budget, documented, thoroughly critiqued, and then published. So it was with the BBC’s work on loudspeakers in the Sixties. The beauty of the prose, the elegance of the arguments, the simplicity, the logical progression of ideas and concepts, and perhaps above all the approachability of the subject to even the schoolboy gripped me then as it does now.
So I work in a way which gives me the documentation to illustrate to peers the design process, including all the meanderings, the dead?ends, the time-wasting but which slowly builds toward the final design. I record my observations in meticulous detail in logbooks, accompanied by the appropriate computer graphs, annotated for my ongoing learning—very deliberate and painfully slow. I have these scrapbooks for every model. The Monitor 40.1, just finished, fills about 150 pages. I hope that long after I’ve gone, these will be of use to future designers.
The actual process is this: First, perfect the drive units. Absolutely mandatory. Fix all drive?unit issues at source, mechanically; do not rely on electrical fixes in the crossover, as they always leave a sonic signature, the cure often worse than the illness. Second, build the box to the final size. Third, mount drivers in the box. Once the drivers are mounted, take careful frequency?response measurements over a wide arc, process them, and feed them into HALNet, our own loudspeaker crossover?design simulator.
Then, take a break for a day or two! I have been designing by simulator for nearly twenty years now, and I have great confidence in the model versus the actual, but—big but—while the built circuit measures exactly as the simulation predicts, it does not tell you anything at all about how it sounds.
This is the really challenging part: How to balance what the simulator tells you is a good frequency?response with what measures well in-room with what your ears tell you sounds “right.” You have to work all three together and you have to guard against being pulled by your ears into something that sounds very seductive but measures terrible or—more usually—sounds terrible but measures great.
It offends me if the measured response is not flat, or knowingly deviant from flat. There has to be, in my mind, justification for shading the system response, as the simulator can give you a dead flat response in minutes, so why not just use it? But this is the core of the job, which transforms the task from product design into an all?consuming 3-D chess challenge.
Experience tells me that two sorts of audible issues loom out of extended listening: Those you can—eventually—attribute to some wiggles in the measurable, hence simulated, response, and those you can’t. Those we call “colorations.” The very word sends a shiver down my spine. It’s not a hard-engineering word. It smacks of failure: of uncharted sonic turf between real science and emotions. But once my subconscious locks onto a coloration (perhaps fifty?plus hours listening in to the design)—I dread this stage—I find myself going round in circles for weeks pushing the response here, pulling it there, moving the crossover up or down in level or frequency, more listening, more eureka moments late at night which the next day aren’t. Every time I swear that I’ll find a hard-engineering path from this random phase to a solution, but every time I resort to (a) going over the notebooks in ever more detail hunting for clues, (b) slowing down, taking the pressure off to get a result, and (c) trusting my ears above the test equipment. It comes good eventually! I should add here that I have no interest in fancy components—standard polyester caps and wire?wound resistors on good quality fiberglass boards are all you need to get a great sound.
As to what am I listening for, I’m listening for coloration that breaks down the illusion of “being there.” For me, speech/vocal quality is the real arbiter because the human voice?box just doesn’t produce the sort of colorations that speakers do. It’s soft, wet, highly damped tissue and it can’t produce spitty, gritty, beaky, wiry, quaky, hollow sound—all those are speaker colorations. Because of its emotional content, music is less revealing of coloration than speech and voice.
SPECS & PRICING
Harbeth HLP-3ES2 mini-monitor loudspeakers
Drivers: custom 4.33" polymer midrange/woofer, 0.75" aluminum?dome tweeter
Frequency response: 75Hz–20kHz +/?3dB
Nominal impedance: 6 ohms
Dimensions: 12" x 7.4" x 7.8"
Weight: 13 lbs.
Fidelis AV (U.S. Distributor)