Carver Amazing Loudspeaker (Platinum Edition)

Dick Olsher, February, 1990

Even to a nontechnical observer, someone without a deep grasp of the germane technical issues, the Amazing Loudspeaker should indeed prove a source of amazement. First of all, there's no box. Don't mistake the back grille for an enclosure—if you pass your hand along the Amazing's behind, you'll realize that the grille is merely a cosmetic cover; you can actually stroke the woofer magnets if you're so inclined. Yet without an enclosure or electronic trickery, this speaker boasts excellent dynamic headroom and true flat bass extension almost to 20Hz. Just think of the woodworking costs inherent in trying to coax such low-end performance from a conventional box speaker. The savings in carpentry have been put toward one heavy-duty ribbon design. The Amazing begins to sound like an incredible bargain at its modest (by high-end standards) asking price. What's the catch? Fundamentally, the answer lies in superior engineering. And, as Bob Carver will readily admit, good engineering isn't inherently any more costly than bad engineering.

It was the promise of greatness that led me to request a review sample from Carver. What eventually arrived was the Amazing's new Platinum Edition. There's also a Silver Edition, a little-brother version with only three woofers per side as opposed to the Platinum's retinue of four 12" woofers. These new Amazings have not been reviewed elsewhere, and differ from the old Amazing in several ways: the ribbon is now pleated for better low-frequency extension and power handling; the new ribbon has more excursion and, to minimize acoustical interference, is one long, continuous design instead of the two shorter segments of the older design.

The elegant fascia of the speaker commands respect and bears more than a passing resemblance to the Apogee designs. However, to characterize the Amazing as a ripoff of the latter would be grossly unfair to Carver. For example, it so happens that the asymmetry of the front baffle is important for smoothing out the bass response of any dipole radiator. A rectangular baffle would have worked, but not as well as an irregular one. So to accuse Carver of mindless imitation in this case makes about as much sense as accusing Ford of copying GM because all of Ford's cars also have four wheels.

The woofer section
More accurately, the subwoofer section, because undeniably the Amazing is capable of subwoofer performance. The 1980s saw the resurgence in several subwoofer designs of the finite baffle, once the low man on the totem pole of baffles. Some of you may recall the Enigma subwoofer, now out of production, and of course more recently the Celestion 6000 subwoofer. In both, electronic equalization is used to compensate for the baffle's LF rolloff down to the woofer's free-air resonant frequency. At frequencies where the average baffle dimension is smaller than half a wavelength, the front-to-back cancellation characteristic of a dipole radiator takes place.

One of the nice features of finite baffles is that the LF rolloff is a gentle, predictable, easily equalizable 6dB/octave down to the driver free-air resonance—below which the response goes to hell at the rate of 18dB/octave. Even so, the bugaboo for such designs remains achieving sufficient dynamic headroom in the deep bass. Push the woofer or woofers extra hard and you run out of voice-coil excursion sooner than you would with that same driver in a box.

The Celestion 6000 is a good case in point. The small-signal response looks terrific with a half-power bass frequency of about 25Hz. But below 40Hz, you'd be lucky to hit 100dB at a realistic listening position even in a small room. Folks, in the deep bass 100dB spls are quite polite.

Bob Carver's innovation lies in his startling reversal of the roles of driver and baffle. Carver's correspondence on this subject is quite eloquent, and the following description is based largely on his writing. Traditionally, the drive-unit alignment is of fairly low Q. (Q stands for "Quality Factor" and is a measure of how sharply defined a resonant peak or dip is.) For example, a woofer Q of 0.42 is considered about optimum for a bass-reflex design. The cabinet, being of high Q, pushes the overall response Q to about 1 in a properly designed system. Bob's insight was to regard the finite baffle as the low-Q or overdamped element in the system and let the high Q of the woofer bring about an overall system Q of 1. Simply put, the idea is to introduce an under-damped or peaky woofer into the baffle and let the peak in the woofer response compensate for the front-to-back bass cancellation of the baffle.

Sounds simple at first blush, but boy, are there technical difficulties lying in the bush, as Bob will only too gladly tell you. He's spent the last two and a half years addicted to the notion of perfecting this planar speaker—until recently to the detriment of Carver the Company and his own personal life. Speaker designers are a breed apart, junkies whose highs are derived from doing battle with the laws of physics. And to judge from my recent encounter (of the third kind) with Bob, the fire still burns bright. His enthusiasm for the Amazing and his untiring drive to improve them clearly go beyond the call of duty.

To underdamp a woofer you have to reduce the size of the magnet. As the magnet size is reduced, the electrical Q of the driver increases, and with it its total Q. The problem is that the efficiency of the woofer is reduced at the same time. The absurd conclusion of this scenario is zero magnet, zero efficiency, and of course zero cone motion. To improve the efficiency of a woofer with a miniature magnet it is therefore necessary to drastically reduce its moving mass. The idea is to try to keep constant the ratio of force to mass, hence a constant woofer acceleration factor. Because most of the moving mass is represented by the cone, that's where you have to cut. What you end up with at the end is a woofer with a gossamer-thin cone and an unbelievably tiny magnet—quite unlike the traditional woofer with its heavy cone and beefy magnet.

So radically different was Carver's woofer design that Tonegen, its Japanese manufacturer, refused at first to produce it. It took a personal visit by Bob to convince them that that was truly what he wanted. The danger, according to Bob, is that there is a tendency on the part of the audiophile who sees this tiny magnet to say, "Ah-ha, Carver has out-cheapied himself and is just screwing us and ripping us off." But it just would not work any other way. Reducing the moving mass, however, is in itself inadequate to confer a decent efficiency; ultimately, one has to resort to a different strategy to achieve a reasonable system sensitivity.

There it is: sensitivity. A term so often confused with efficiency that, in many audiophiles' minds, the two concepts are synonymous. Efficiency is a fundamental driver property, determined essentially by driver parameters, that relates (for example) how efficiently a woofer converts electrical input power to acoustical output power. Sensitivity, on the other hand, is a system parameter. For example, a system with ten inefficient woofers may put out more total power than a system with a single more efficient woofer. Thus, there is combined "strength" in numbers. And from a practical standpoint, most of us could care less about driver efficiency—as long as the bottom line is excellent sensitivity.

Those of you familiar with Small-Thiele theory may be reaching for a pen and pad just about now, getting ready to fire off a letter to the Editor. I know what you're thinking: "How can DO say that? Haven't I seen a design equation somewhere that shows efficiency increasing with box volume"? Well, let me save you the postage. It's only a myth; appearances can sometimes be deceiving. And even Richard Small himself will tell you that driver efficiency is determined almost entirely by driver parameters, not box volume. It can't be any other way; if efficiency did increase with box volume, you'd have infinite efficiency in an infinite baffle. Nonsense. Big, efficient woofers do require big boxes for proper bass alignment, and for that reason big box designs are more efficient. But it's not because of the box; rather, the woofer.

Back to the idea of using multiple woofers to improve system sensitivity. This is not a practical approach for a box speaker. Once you optimize the design for a given woofer to yield maximally flat bass response or whatever, mounting another woofer on the front baffle screws up the bass alignment (the exception is "isobarik" or compound loading). The two woofers in the same box will produce a peaky, non-flat response with reduced bandwidth. The reason is that the acoustic compliance of the woofer pair is doubled, which in turn requires a box volume twice as large. So to keep adding woofers to a box design requires an ever-increasing penalty in terms of box volume. This is not a problem, however, for finite baffle designs.

One of the advantages of the finite baffle is that its bass alignment is unaffected by the number of woofers used—the designer is free to use as many drivers as he pleases to keep pushing the system sensitivity higher and higher. Carver uses four woofers per side in the Platinum Edition because that's all the room there is, but I'm sure he'd have opted for more, space permitting, and of course only if the ribbon sensitivity could keep pace. The end result is a very respectable sensitivity spec of 88dB—good enough in this respect to compete with many box speakers.

The Q of the Amazing's bass alignment is around 1, considerably higher than the 0.5–0.7 range most designers opt for. Don't expect tight, well-defined bass from such a design. This, according to Bob, is just the sort of bass quality he likes—"a warm, rolling bass."

The ribbon
The ribbon assembly flanks the woofer array on the inside edge of the baffle. If you look closely you can see that the planar "voice-coil" consists of four strips of pleated aluminum foil running the length of the assembly. The strips are joined top and bottom to form a continuous loop. The aluminum is glued to a Kapton backing stretched over a particle-board frame and clamped on all sides with a foam suspension. The foil is sandwiched between two layers of magnets located front and back. There's a total of 36' of ceramic magnets per side that provide the magnetic push-pull force for the ribbon. The Kapton is claimed to be very strong, dimensionally stable, and capable of tolerating extreme temperatures—a perfect recipe for a voice-coil former. The glue used is said to provide damping for the internal resonances of the aluminum.

Being a purist, I naturally feel obliged to point out that, strictly speaking, this sort of design is not a ribbon in the classic sense in at least one respect: The foil operates in the peripheral or leakage field of the magnets instead of being suspended directly between the north and south poles. For most folks, Apogee included, this sort of design qualifies as a ribbon, so I'll let it go for now.

The ribbon operates full-range above about 100Hz, with useful output to about 20kHz. There's a second-order crossover network at around 100Hz. However, I have hedged about the details here because the design changed before my eyes as the review progressed (more about that later). While there are no crossovers in the critical midband, there is at least one pot (hedging again) in the signal path. Sample 1 of the Amazing featured an upper-midrange control, while sample 2 had no less than two pots, the additional pot being a high-frequency control. Something unique to Sample 1 were two small foam rectangles mounted top and bottom on the back side of the ribbon. I've dubbed them "training bras" because, according to Bob, their function is to provide some damping for the ribbon until it breaks in—after which they may be discarded (though this is not spelled out in the manual).

The bass resonance for the ribbon is claimed to be 75Hz. In Santa Fe, after an extensive break-in period of over 100 hours, we measured a bass resonance centered at 150Hz. At Santa Fe's 7000' altitude, granted, I would expect the resonant frequency to climb because of the reduced air load; for the ribbon, I would even be willing to concede the possibility of a 25% increase. But a 100% increase cannot be explained by blaming Mother Nature, as Bob is inclined to do. I would rather point an accusing finger at Carver's quality-control procedures.

I'm also worried about the selection of a crossover frequency so darn close to the ribbon resonance. The extreme phase shifts that accompany a resonance make it a bitch to design a network with flat amplitude response in the crossover region. As a rule of thumb, one should choose a crossover point at least an octave away from a major driver resonance. A driver can be modeled as a bandpass filter. In its pass band, removed from the bass resonance and HF rolloff, a driver is known to be minimum-phase and thus much more amenable to conventional filter design. Did Carver succeed? Well, in my own measurements and his, there is a notch in the response between 100 and 150Hz, apparently due to the crossover. Of course, there's the possibility that this is caused by a room mode, but I rather doubt it.

Bob provided me with measurements he made in a hi-fi store at about 4 meters on-axis, comparing the response of the Amazing to that of the Quad ESL-63. Bob figured that since the Quad has no bass crossover, its response would be seamless and would highlight any problems with the Amazing's crossover. With both speakers and the mike in the same physical location, the response curves superimpose very nicely at 75Hz, but there's a 3dB notch in the response of the Amazing relative to that of the Quad, centered at 110Hz.

Virtues of directivity
Conventional box speakers radiate omnidirectionally in the bass. As signal frequency increases, however, the polar response narrows significantly. By the time the tweeter tweets above 5kHz, the response is pretty much confined to a cone with a half angle of 30°. All conventional tweeters beam sooner or later, and may be said to possess increasing directivity as a function of frequency. Looked at as a whole, the directivity of a garden-variety loudspeaker resembles the shape of a pyramid: low-directivity or broad radiation in the bass while taking on an increasingly narrow radiation profile at higher frequencies. In contrast, a dipole radiator has a much narrower radiation pattern, at least through the lower octaves. The pattern resembles a figure eight, with radiation lobes to the front and back and very little side-directed energy. One advantage of the latter pattern is that it significantly reduces early lateral room-boundary reflections and their resultant colorations. There is simply less energy splashed onto the side walls, floor, and ceiling.

I firmly believe that a dipole radiator represents a significant step toward solving the room/loudspeaker interface problem. Early reflections are anathema to accurate reproduction of the soundfield captured by the mikes at the concert hall. These reflections, even if spectrally similar to the direct sound, interfere with it to produce a comb-filter effect that colors perceived timbres. Reflections arriving at the ears during an initial 10ms window—after the direct sound has stopped—are the most troublesome because they are higher in amplitude compared to later arrivals and are not well discriminated against by the ear/brain system. Room treatment is therefore essential to suppress the room's sonic signature. After all, do you want to hear your room's reverberation or the original soundfield?

I've had good results with a dead-end/live-end sort of treatment with the speakers located at the dead end of the room. I'm therefore mystified by Carver's references to the beneficial effects of 6ms reflections when the Amazings are placed 3' from the front wall of the listening room. These reflections are claimed to increase the spaciousness of the soundstage. Well, I would have thought that rear-wall reflections or possibly a delayed rear channel would have done that. But in my opinion, Carver's suggestion is a recipe for destroying the accurate transduction of the original soundfield on the recording. I'm making a tacit assumption here that the soundfield captured during the recording session is worth preserving. This may be a poor assumption in the case of multi-miked pop recordings, in which case the additional adulteration of the room may actually be desirable. In the case of properly executed purist recording techniques, the recorded soundfield is rich with spatial clues and incorporates the right blend of direct/diffuse sound. That's the sort of information I'd like to preserve in my listening room. Unfortunately, this is much more difficult to do than I've so far indicated.

It's a fact of life that most of us listen in semi-reverberant environments. Room treatments are most effective at higher frequencies, say above 5kHz. At much lower frequencies it's very difficult to dissipate sound energy very quickly. As a result, unless you're glued to the loudspeakers, a significant portion of the sound energy at your listening position is due to room reflections. A critical distance may be defined at which the reverb energy just equals the direct sound energy. The basic problem is that this critical distance changes with frequency. It would be highly desirable to keep the ratio of reverb to direct energy constant at the listening seat. But because directivity or beaming typically increases with frequency, this goal is impossible to achieve. Recall that the average speaker is omnidirectional in the bass. It pumps lots of bass energy off-axis; together with the fact that room absorption is ineffective in dissipating bass energy, this assures that reflected energy is bass-rich and treble-shy. As directivity increases, more energy is concentrated on-axis and the direct component of the total sound increases. Finally, above about 5kHz, the absorption of highs by room surfaces, together with the on-axis treble beaming, combine to make the treble sound almost completely direct.

This is an important concept to understand. It explains, for example, why two speakers that measure exactly alike on-axis sound different—in the same room. The answer, of course, is different directivity patterns which change the character of the room reverb. Line-source tweeters like those in the Amazing are by their nature more beamy than a typical dome tweeter. Thus, if you've optimized your system for a dome-based loudspeaker, substitution of a line ribbon at the same position will result in a noticeably brighter balance. Why? Because the direct/reverberant sound ratio starts to increase at a lower frequency for line sources. This predominance of direct sound will emphasize overtone structures in the upper octaves, imbuing timbres with a bright, steely character. The solution is to either move farther back from the speaker or experiment with treble rolloff (footnote 1).

So what response is "correct" for a quasi-line-source design like the Amazing? It would be a definite mistake to equalize the Amazing to be flat on-axis in the near field. I learned this the hard way. I tried it, and had to duck the razor blades the Amazing hurled at the listening seat. Three meters away, the sound was excruciatingly bright. An even bigger blunder would be to equalize the Amazing to be flat at the listening seat. The superimposition of a reverb curve deficient in highs over a flat direct-sound curve naturally results in a listening-seat balance featuring high-frequency rolloff. That's the correct tonal balance. An attempt to flatten this composite curve by jacking up the direct-sound contribution will significantly brighten timbres. Trust me, you'll be reaching for the cotton balls.

Footnote 1: For a more technical exposition of these concepts, see Professor Han's "Frequency Responses in Acoustical Enclosures," AES Preprint 2452(F-3), 1987.—Dick Olsher



Because of treble beaming, the tonal balance of the Amazing changes with listening distance. Generally, at 3m from the front baffle you'll be sitting beyond the critical distance for frequencies up to about 5kHz. To reiterate, the right choice at the listening seat is an upper-octave rolloff to ensure a natural sound balance; one that does not distort the direct/reverberant sound ratio.

It was this sort of experience that led J. Gordon Holt (footnote 2) to embrace the "Down with Flat!" philosophy in earlier issues of the magazine. It was quite puzzling at the time. Jack Hjelm from Audio Research had just finished installing some gigantic tube amplifier in JGH's system to drive one of the Infinity IRS systems (footnote 3). As a final touch he proceeded to equalize the speakers to flat at the listening position using an Ivie hand-held real-time spectrum analyzer. The resultant sound was bright; but, according to Jack, it had to be right; after all, it was flat. Well, sometimes "flat" is not. As soon as the ARC sage left, JGH proceeded to kill the treble.

You should now understand why some argue that the ideal transducer should have constant directivity up to about 5kHz, and decreasing directivity beyond—precisely the opposite sort of behavior exhibited by most loudspeakers. Such a directivity pattern will ensure that the direct/reverberant ratio of sound in the room is fairly constant as a function of frequency. Incidentally, the new Quad ESL-63, as far as I know, has the most uniform directivity of any commercial loudspeaker. The Quad's diaphragm is driven by concentric ring stators with suitable time delay such that the radiation pattern closely resembles that of a point source located some 300mm behind the diaphragm. Most other dipoles start off correctly with a controlled directivity in the bass, but cannot maintain that level of polar response with increasing frequency.

The electronic control box
The Amazing's optional gadget box allows a variety of signal processing. It may be inserted into the tape-monitor loop of an integrated amp or between power amp and preamp. Because the initial intent was to market the Electronic Controller as an integral part of the Platinum Series package, I received one with my first sample of the speaker.

Moving from left to right on the front panel, the first function is the bass Q control. Q can be adjusted from 0.3 to 2.0. Why anyone would want to loosen the bass any further is beyond me, but going the other way, to a Q of about 0.5, proved a worthwhile enhancement. At this setting, the bass character of the Amazing underwent a much-needed transformation in definition, from quivering Jello to reasonable firmness. Pitch definition and bass detail were now much more readily resolvable. It was no longer safe to eat beans while listening to music.

Unfortunately, there's a significant price to pay. The EC box not only squashes dynamics but contributes an earful of solid-state hardness and grain to the mids and treble. The cure in this case is worse than the disease. With the EC in the chain, tube amps sounded solid-state, tube liquidity being well masked. Removal of the EC from the signal path after an extended listen invariably brought forth a sigh of relief.

To the right of the bass Q control are a "High Frequency Trim" that allows you to shelve the treble above 4kHz, a "Sub Bass" synthesizer that will generate or supposedly "restore" sub-bass harmonics to program material deficient in deep-bass energy, a "Gundry Perspective" control that allows you to shelve the upper mids and thus change the apparent orchestral perspective, and finally a "Sonic Hologram" control.

The Sonic Hologram attempts to fix what some consider a basic flaw of conventional stereo. The argument runs something like this: When we listen to an instrument live, each ear receives a single input; call these left and right. Trying to reproduce the same solo instrument via a pair of loudspeakers results in each ear receiving two inputs. The left ear receives a left speaker input, and because of head diffraction there is delayed crosstalk from the right speaker. The situation is similar at the right ear. Sonic Holograpy tries to cancel out these crosstalk signals so that the left ear effectively hears only the left speaker, and vice versa. This sort of argument makes a lot of sense when trying to reproduce a binaural or "dummy head" recording via loudspeakers. Such a recording will only sound right through headphones, and there has been a lot of work done on devising circuits to make binaural recordings compatible with stereo loudspeaker reproduction. However, the logic of the argument breaks down for true-stereophonic recording techniques such as Blumlein. Here, as with other coincident recording techniques that rely on intensity differences for localization, it is precisely this kind of crosstalk generated by two-channel stereo that is relied on to produce natural phase cross-correlation between the ear-input signals. The important point is that head diffraction is operative at all times, generating two ear-input signals even under live listening conditions.

I did experiment with Carver's circuit and found it to make a difference, but only sometimes for the better. Although the action of the Sonic Hologram circuit seemed somewhat unpredictable, nevertheless a couple of generalizations are in order. First, the circuit acted to expand image outlines, providing a blown-up or zoom version of the original spatial outlines. Second, on multi-miked recordings, and in general where the left and right channels contain quite a bit of out-of-phase information, the Sonic Hologram helped to flesh out the soundstage—but at the expense of pinpoint imaging. However, program material recorded with coincident or quasi-coincident techniques, where there is high phase coherency between channels, was adversely affected by the Carver circuit. Here the imaging became unstable, outlines wandering away from their center of gravity within the soundstage. A case in point was my wife Lesley's voice throughout the Lesley Test. The weight of the image shifted from side to side as if an unseen hand was playing games with the channel-balance control.

On the whole, I'd forget about the Electronic Control Box.

Will the real Amazing please stand up?
As this project evolved, it became clear to me that the Amazing was truly a moving target, a fluid design that certainly from my perspective, and inferentially from Bob Carver's, was not altogether final. Santa Fe became a proving ground for the Amazing; several versions of the speaker have been assessed to date. I have to confess, however, that some of the changes along the way were triggered by my thoughts and suggestions. At times I felt as if I were in the design loop for this product. The above scenario is more than vaguely reminiscent of the Bud Fried syndrome: "I'll keep changing it till you're happy." And I'm sure that the initial monkey wrench thrown by Bob into the proceedings was an attempt to manipulate the situation in his favor. But I remain convinced that Bob was subsequently motivated by a passion to perfect the speaker to the detriment of good business practice. There's no question in my mind of Bob's intense commitment to the speaker.

To clarify the situation, let me give you a synopsis of the sequence of events. The first sample (which we received in July 1989), as you'll discover shortly, was beset by serious sonic problems and was on its way to, most probably, a terminal review. Just about then, in early September, as though he had read my mind, Bob phoned me and the fun and games started.

All of the above add up to unusual circumstances indeed. Normally, a manufacturer does not receive this much free consultation. From my perspective, it was a question of wanting to see a promising product succeed. From JA's perspective it was crucial to ensure that the magazine would not get trampled on—our policy is not to allow manufacturers to use Stereophile's facilities and Stereophile's writers' talents for design consultancy purposes, whether paid or unpaid. The ground rules, as JA laid them out, were that if Bob were to visit Santa Fe to redesign the Amazing Loudspeaker, Stereophile would report on all of our experiences with the product; that Carver would take full responsibility for the design; and that a production version of the speaker would be subsequently submitted for testing. It was this final item, however, that proved difficult to pin down. Apparently, the Platinum Edition Amazing Loudspeaker was being sold all along, so that at least some speakers—identical to my first sample—are either in dealers' inventories or in consumers' hands.

Act I
Sample 1, Take 1. Out of the box and before any significant break-in, the sound of the ribbon had a strong metallic flavor to it—as though a large sheet of Reynolds Wrap was being crinkled. After over 100 hours of break-in this coloration largely abated, but did not entirely disappear. There remained a metallic aftertaste, no doubt due to internal resonances in the presence region.

During setup, the recessed terminal cup of the Amazing proved a pain in the butt. It was difficult to fit spade lugs within the recess. However, it proved real easy to strip the plastic shaft of the binding posts provided; be careful not to overtighten these. Lord, why is it so difficult to find expensive loudspeakers with high-quality binding posts?! I also managed to work loose one of the binding-post retaining nuts, in the process breaking a solder joint. After Robert Harley had repaired the connection, I proceeded to tweak the installation.

It proved necessary to back off the front baffle a good 10' and toe-in the speakers considerably before a decent soundstage materialized. I next trimmed the tonal balance using the Upper Midrange Control (UMR). This pot allows a nominal ±3dB of amplitude control over a frequency range of 1–6kHz, with full impact at a center frequency of about 2.5kHz. This exercise turned out to be extremely frustrating. I spent hours trying to get the Amazing to sound right. Using the Lesley Test as program material, it was just impossible to obtain a natural balance. Turning this pot up to where Lesley's upper registers had the right brilliance also elicited sibilance and a metallic transient etch sufficient to tattoo my ears. Shelving the upper mids and presence region down to the point of achieving a tame enough presentation brought about a significant alteration of tonal colors. Rather than preserving a sweet and smooth character through the upper registers of soprano voice or violin overtones, timbres took on a slightly dry and grainy quality.

Footnote 2: In whose ears we trust.—John Atkinson

Footnote 3: See Stereophile, Vol.9 No.4, pp.29–41.—Dick Olsher



There was lots of midrange and treble detail, and it was resolved clearly enough—commensurate with the performance level you'd expect from a fast transducer like a ribbon. So it was all the more disturbing to deal with the level of reproduction below 500Hz or so. On the one hand, you have lots of scintillating treble detail that beckons you to get involved in the music. On the other hand, the lower octaves serve to isolate you from the music. Taken as a whole, the midbass, upper bass, and lower midrange were veiled and muddy. Bass lines were consistently ill-defined and difficult to follow. Because much of a hall's sonic signature resides in this range, it was not surprising that I could not penetrate deep into the soundstage. I just could not get an adequate sense of hall. I could not place myself in the space of the performance.

Every one of my records that I've come to depend on for soundstaging evaluation, the Proprius Laudate! and Cantate Domino for example, failed to flesh out properly through the Amazing. I could make out the leading edge of hall reverb, but its trailing edge became indistinct. It was as though a thick curtain were dropped before my eyes at a crucial moment to obscure the true expansiveness of the hall. The loss of soundstage transparency and immediacy was so obtrusive and disturbing that I had to try something.

One of the things I tried was a number of amplifiers, including Carver's Silver Seven-t monoblocks. The Seven-t turned out to be a sonic disaster, combining the worst attributes of solid-state and tubes. Treble transients were sizzly and steely, the bass heavy and undifferentiated—sort of a bloated thump. The mids were dryish and grainy. And the soundstage—what a mess! Compressed depth to the point of one-dimensionality, and never really coalescing into a unified whole. And to think that Carver had wanted me to bi-wire the Amazing with no less than two pairs of the Seven-ts. The Krell KSA-200 provided the Amazing with a much-needed dose of bass control, but it also emphasized the ribbon's mechanical resonances. The Music Reference RM-9 offered much-needed midrange liquidity without exacerbating the presence-region nasties, but wasn't nearly as accomplished as the Krell in the bass.

The ideal amp for the Amazing (this is true for many ribbon-based designs) would be one that combined solid-state bass control and current drive with tube-like liquidity, textural softness, and imaging. I only wish money could buy such a beast.

In desperation, I decided to try one final trick: a cheap experiment, if you will, since it only involved a dollar's worth of masking tape. I experimented with taping the back side of the ribbon; ¾"-wide tape was just fine for this application, as it neatly fit over the slits in the magnet structure. The optimum pattern turned out to be one where alternating slits were taped shut. The idea was to try and resistively dampen whatever ribbon resonances I could. The really amazing thing was just how effective this idea turned out to be. The ribbon's inherent presence-region resonance around 5–6kHz was not abated, but the difference in the lower octaves was nothing short of dramatic. The upper-bass heaviness largely disappeared. Bass lines became distinct. And, just as important, soundstage transparency shot up to a point where I had no difficulty any longer in delineating hall size—as though the fog that had previously surrounded this region all of a sudden evaporated.

Another surprise was that the veiling was not caused by the woofer section, but rather by the ribbon's bass resonance. This dramatic change for the better is readily apparent from fig.A. (The curves were generated during Bob's subsequent visit to Santa Fe.) The top curve shows a peak of some 14dB centered at 150Hz for the undamped ribbon. With a swatch of silk applied to the back of the ribbon for air-flow resistance, the bottom curve shows the end result. About this time Bob Carver phoned me with a major announcement.

Fig.A Carver Amazing Loudspeaker, nearfield ribbon response with (red) and without damping (blue) (10dB/vertical div.).

Act II
Here I am itching to close out this project, and Bob announces that the current Platinum version of the Amazing, the one I have been listening to, has been discontinued. According to Bob, there are new grilles for the front and back that reduce diffraction effects, as well as a new crossover network. I wonder about Bob's amazing timing, but naturally I want a new sample as soon as possible to close the loop. Bob has a better idea. Since my pair of Amazings was already broken in, why not come out to Santa Fe and install the various upgrades himself?

So, on a Saturday morning in October, there's Bob in my listening room, along with Mark Friedman (Carver's National Sales Manager) and John Atkinson. The grilles are here, but what about the crossover? Bob waves this Rube Goldberg contraption about that looks like an octopus in heat. It turned out later that this "new crossover" was thrown together by Bob that same morning in his motel room using Radio Shack parts and a coil-winder he brought with him. What Bob really wants to do is spend the day experimenting and hopefully end up with something Stereophile likes. We spend part of the morning listening; Bob really likes the concept of damping the ribbon. So we spend part of the afternoon scrounging for various damping-material samples, including nylon stockings, silk cloth, and chiffon. The results of the silk damping experiment (fig.A) are so impressive and Bob so ecstatic that he resolves to incorporate some form of damping in all future production.

Next Bob attempts to build a notch filter to counteract what he measures to be a 7kHz peak in the ribbon response. A new crossover is installed late that evening, and, after some additional listening late into the night, it's clear that the tonal balance still isn't quite right. As Bob leaves, I make it clear to him I expect a new production sample incorporating all of the final changes.

Sample 2 arrived at the end of October 1989. It differed from the first sample in several ways. First, damping was provided on the back side of the ribbon assembly. Bob decided to use a thin gauze-like material called "pre-wrap." This stuff was taped to the back of the baffle with electrical tape; a temporary measure, Bob assured me, until the proper moulding is fabricated to hold the gauze in place. Second, the crossover network was revised to better blend ribbon and woofer. The crossover frequency was pushed slightly higher to compensate for the 2dB loss in ribbon efficiency incurred by the damping material. A 7kHz notch filter was included in the network to kill a major ribbon resonance. A Very-High Frequency (VHF) control was added to give the user some control over frequencies above 10kHz. This, according to Bob, helps in controlling a treble resonance in the range of 10–12kHz. Third, new woofer grilles were installed with larger cutouts on the back side. Fourth, a "High-Altitude" resistor kit, with instructions, was included for adjusting the damping of the woofer section. Finally, this particular pair of Amazings was supposedly already broken-in by Bob prior to shipment.

This pair certainly sounded much smoother right out of the box, so much more so than our broken-in first sample that I began to suspect one of two possibilities: either this was a hand-selected sample unrepresentative of the lot, or just a randomly selected lucky sample. If the latter possibility is the correct one, it highlights QA problems in getting these ribbons to sound alike.

We proceeded to break in the second sample for an additional 50 hours. The first listening session with the new Amazing proved to be more of a positive experience than the first sample ever provided—although I still had some major reservations. Bass definition was decent, although still lacking much impact. There wasn't much sense of punch; bass attack was more like a limp noodle than a whiplike crisp. The amazing thing about the Amazing was that, despite measuring flat into the low 20s, its deep bass simply failed to sound that way. There was always lots of bass, but on organ recordings the Amazing could not generate a convincing bass foundation below about 40Hz.

The "High-Altitude" kit provided by Carver with the Amazing allows for the insertion of either a 1 or 2 ohm resistor in series with the woofer section. According to Carver, this is supposed to lower the woofer Q and thus tighten the bass. I would agree that a series resistor would reduce the quantity of bass, but I don't see how the Q of the response is lowered in this fashion. Substitution of the 1 ohm resistors per Carver's instructions is not straightforward: it forces you to bi-wire the speakers, but since that in itself is not a bad thing, it should not prove a deterrent. The resistors clearly reduced bass output. But, just as clearly, bass quality was not improved. In fact, the midbass got more muddled, taking on more of a one-note bass character. Charlie Haden's double bass on "Lonely Woman" on The Shape of Jazz To Come (Atlantic CD, 1317-2) not only became more anemic, but pitch definition decreased to the point where his lines were hopelessly muddled. This resistor kit is a miscalculation; my kind recommendation would be to give it a hasty burial.

However, the Amazing's imaging was much improved with these samples. Soundstage width and depth were both more than adequate, and the level of transparency through the lower mids and upper bass was quite astonishing when compared with the first sample. Hall reverb was now easy to resolve, and resolution of massed voices was very good. With both pots at their nominal 12 o'clock positions, the upper octaves were now quite listenable. Much of that presence-region scream was absent. The presentation was still overly sibilant, and treble transients were still a tad on the hot side—but tolerably so. The problematic upper mids remained very much so. Timbres through this region just didn't sound right. No combination of UMR and VHF pot settings managed to restore timbral accuracy. The best compromise turned out to be a 12 o'clock pot setting, but I was left with the impression that the Amazing was still hopelessly colored in this region.

A magical moment
At some point during the second session, something very magical happened. It was as though the sky opened up and a hand descended to bless the Amazing. I was diddling with the toe-in angle again, but that wasn't really it. All of a sudden the Amazing got smoother and sweeter at the top, and I found myself suddenly drawn into the music. For the first time, I found myself enjoying these speakers. Things weren't altogether right, but there was enough right here to combine for an enjoyable experience. The upper range of Lesley's voice was still adulterated. But the focus of her vocal outlines was tight, and midrange textures were smooth. An occasional squawk crept in around 1kHz in one of the ribbons. And you also need to know that the ribbon buzzes like a kazoo below 1kHz. The buzzing is normally not audible, being masked by the music. But with pure sinewave test signal it's easy to pick out the accompanying buzz.

But enough bad news for now—the Amazing was beginning to flex its muscles. Record after record was thrown on the Aura turntable (an excellent 'table from Down Under; review forthcoming), with very musical results. Pilar Lorengar as Princess Pamina (The Magic Flute, London OSA-1397) really shone with an intense vocal purity. Mezzo-soprano Janet Baker's voice (Holst: Savitri, Argo ZNF-6) was reproduced with a honey-smooth quality. The husky hue of Cleo Laine's lower registers (Live At Carnegie Hall, RCA LPL1-5015) was perfectly captured by the Amazing. Therese Juel's voice on the Opus 3 Test Record 1 was well-behaved—just slightly erring toward oversibilance. Violin tone was just as captivating. Itzhak Perlman grooves on this sort of music. The Bruch Violin Concerto (EMI ASD 2926) is just the sort of repertoire to give Itzhak freedom to ooze calories. That he does, and the Amazing reproduced his violin tone with sufficient sweetness and suaveness. High-powered orchestral material was reproduced with excellent dynamics. Retrieval of low-level detail was excellent, even when confronted with complex, dense musical passages.

Act IV
It's mid-November. You guessed it. Another phone call from Bob. He's changed the crossover again. As I write this, Bob may be on his way to Santa Fe to install the upgrade. If he does, I'll report the results in a future Follow-Up. And one more thing. Bob followed up on a suggestion of mine to treat the ribbon foil surface with a damping material. Bob used a 3M damping compound which he brushed onto the foil. He found that the treatment yielded, in his words, an instantly broken-in, or "aged," ribbon. The problem is that because of the increase in moving mass, the efficiency drops dramatically. So Bob is thinking of completely overhauling the ribbon design as a long-term project in order to bring the efficiency back up. In a year or so, there may very well emerge "Son of Amazing."

Taking stock
Coupling the Amazing's relatively modest asking price with the level of performance achieved by what has to be viewed as a transitional prototype, there looms the potential for a great speaker. At the outset it should be emphasized that, even at its best, the Amazing possessed an inescapable coloration in the presence region. With this speaker, you'll forever be married to an oversibilant presentation. You can hear the 5kHz peak as an emphasis of surface noise, and it also gives treble transients a slightly zippy or hot character.

On the positive side, the Amazing is capable of exquisite midrange textural smoothness and sumptuous liquidity. Low-level detail resolution is excellent: no nuance is too small for the Amazing. The dynamic range is excellent. The Amazing will play loud without audible compression or congestion. With the right amp, the lower mids take on a convincing authority. Bass extension is very good, with decent impact, but don't expect amazing bass quality and definition. The Amazing's bass performance is very amp-dependent, with a solid-state amp being a requisite for eliciting its full potential. The rub is that a typical solid-state amp is likely to exacerbate the Amazing's presence-region zip. With a lot of sweat and tears in setup, the soundstaging can be quite convincing, with excellent spatial resolution. The damped ribbon is sufficiently transparent to allow one to gaze very deeply into the soundstage and to clearly delineate hall size.

There's some buzzing and grille snapping on large bass transients, but generally I have not found these warts to be annoying because they're masked by the music.

Finally, the big question mark is variability in ribbon quality. Unquestionably, it takes the ribbon a long time to break in. In the interim, the sound is rough and edgy. The controlling factors here do not appear to have been adequately researched by Carver. JA expressed the opinion that Carver's method of clamping the ribbon may not be optimum. That I do not know. But I can tell you that waiting for the ribbon to break in is like having to go through puberty all over again: it's a royal pain in the butt. But when that magical day dawns, the Amazing begins to smile at you and all may be forgiven—unless you've lost your patience along the way.

Finally, I want to make it perfectly clear that, as of now, I do not recommend this speaker. If and when the design stabilizes, and I have a chance to evaluate a true production sample, then a final assessment can be rationally made. I consider it insane to spend any money on what has to be regarded, for now, as an evolving prototype.

Sidebar 1: Specifications

Description: Two-way hybrid with a ribbon operating full-range above about 100Hz, and a patented dipole bass system featuring four 12" woofers in a finite baffle loading. Nominal impedance: 6 ohms. Sensitivity: 88dB/1W/1m. Maximum SPL: 113dB symphonic music, 110dB rock music.
Dimensions: 27.5" W by 10" D by 54" H.
Price: $2195/pair in oak, $2495/pair in black (1990); no longer available (2006). Approximate number of dealers: 170.
Manufacturer: Carver Corporation, Lynnwood, WA (1991); Bob Carver currently heads up the Sunfire Corporation (2006). Web: As of 2006, Carver Audio in Florida offers repairs and updates to the Carver Amazing Loudspeakers: Web:

Sidebar 2: Measurements

As an interested observer during much of Dick's auditioning of the Amazing, and having spent most of a weekend with Bobby the C last September (see my interview with him elsewhere in this issue), I was intrigued to get the final samples of the Amazings into my own listening room for a little measuring fun. Having just installed the DRA Labs MLSSA measuring system in my PC, I felt that this speaker would really give it a workout.

And I wasn't wrong. The Carver Amazing's impedance magnitude and electrical phase, measured with MLSSA, are shown in fig.1. The lower of the two humps is the woofer resonance at 23Hz or so, while the upper is due to the crossover circuit, not to the ribbon suspension resonance. (The impedance of the ribbon alone, not shown, features a smoothly rising impedance below 200Hz.) The sharp peak at 6710Hz is due to the notch filter introduced by Bob Carver in his attempt to reduce the speaker's audible sibilance. All in all, though the impedance doesn't drop below 3.76 ohms, the Amazing's lowish sensitivity means that an amplifier capable of delivering high currents and high voltages will be best suited to drive it.

Fig.1 Carver Amazing Loudspeaker, electrical impedance (top, 2 ohms/vertical div.) and phase (bottom, 90°/vertical div.).

Fig.2 shows the anechoic section of the Amazing's impulse response (calculated from the MLS data), measured at a 48" distance, 36" from the floor (a typical listening height), using a calibrated B&K 4006 microphone. (Both controls were set to their 12 o'clock position, which was how DO had preferred listening to the speaker.) The leading edge of the pulse is well-defined, but there then follows a complicated pattern of ringing. The Amazing's ribbon is most definitely not unresonant (though, of course, the resonances may be due to the airspaces in front of and behind the ribbon that are enclosed by the magnet structure). Fig.3 shows the associated step response, which appears to suggest that while the ribbon (the initial right-angle step) is connected in positive acoustic polarity, the woofers are connected in inverted polarity. The smooth hand-over from one to the other, however, suggests good frequency-domain integration between the units.

Fig.2 Carver Amazing Loudspeaker, impulse response on ribbon axis. (5ms time window, 30kHz bandwidth.)

Fig.3 Carver Amazing Loudspeaker, step response on ribbon axis. (5ms time window, 30kHz bandwidth.)

Performing a Fast Fourier Transform on the discrete time-domain data gives the Amazing's anechoic frequency response. I have shown this from 200Hz upward (footnote 1) to the right of fig.4, which was plotted using third-party graphics software rather than that provided by MLSSA, because I prefer Stereophile's published frequency-response plots to have the same aspect ratio and 50dB scaling as the traditional B&K graphs. Obvious features are the downward trend in the lower midrange, presumably due to proximity effect, with then obvious peaks in the 5–6kHz and 11–13kHz regions. No wonder DO found this speaker to emphasize sibilants. (Bob Carver's notch filter appears from this measurement to be both too narrow and too high in center frequency to significantly reduce the energy in the lower of these two regions.)

I also examined the ±15° lateral off-axis behavior of the ribbon; the response was effectively identical on both the woofer and to the baffle-edge sides, suggesting that, though DO found otherwise, the exact toe-in should be relatively uncritical in producing the best tonal balance. Maybe what the toe-in achieves is to affect the audibility of the time-domain problems rather than those in the frequency domain.

The curve to the left of fig.4 is the response of the topmost of the four woofers, measured in the nearfield; ie, with the microphone almost touching the grille cloth. Without the cancellation effect of the finite baffle loading, the intrinsic response can be seen to be centered on the drive-unit's resonance (though this driver's appears to be a little higher than that suggested by the system impedance).

Fig.4 Carver Amazing Loudspeaker, nearfield woofer response (left) and anechoic response on ribbon axis at 48", with tone controls set to 12:00, corrected for microphone response.

It is impossible to estimate from this nearfield measurement what the speaker's true extension would be in-room. I therefore carried out a spatially averaged, 1/3-octave measurement using an Audio Control Industrial SA-3050A spectrum analyzer with its calibrated microphone (footnote 2). The bass can be seen from fig.5 to be elevated by more than 6dB compared with the treble region, with extension really only down to 30Hz. Certainly in my room, the speaker sounded unmusically "slow," which correlates nicely with this measurement. The midrange, however, does feature an exceptionally smooth response trend, suggesting that DO was right on the money when he talked about "exquisite midrange textural smoothness and sumptuous liquidity" in this region.

But oh, look at the treble. There again is the sibilance, as revealed by the MLSSA measurement. Even with totally different measurement hardware, this excess of energy in the 5kHz and 12.5kHz 1/3-octave bands was apparent on every axis; its audible effect was severe.

Finally, one of the beauties of MLSSA is that its post-processing options allow you to look for resonant behavior in a very effective manner. Fig.6 shows the typical "waterfall" display produced as the system analyzes the decay of the impulse sound emitted by the speaker (shown in fig.2) in discrete time slices. Interestingly, the very audible energy excesses around 5–6kHz and 11–13kHz, though presumably due to resonances—note the persistent small ridges in these regions parallel to the time axis—decay reasonably quickly. But note the very long ridges at 1600Hz (the cursor position) and at 1300Hz, which persist in the case of the 1600Hz resonance right out to the 6ms edge of the graph. Though there are only minor peaks in the amplitude response at these frequencies, these long decay times reveal them to be major resonances. Undoubtedly it was their presence that led Dick to feel that "an occasional squawk crept in around 1kHz."

Fig.5 Carver Amazing Loudspeaker, spatially averaged, 1/3-octave response in JA's listening room.

Though these FFT-derived graphs are unreliable below 500Hz or so, due to the truncation of the time-domain data to remove room effects, note that fig.6 does reveal the exaggerated nature of the lower mids, and hints that the bass energy just keeps on a-comin', as suggested by Dick's auditioning, presumably due to the highish-Q LF alignment.—John Atkinson

Fig.6 Carver Amazing Loudspeaker, cumulative spectral-decay plot at 48" (0.15ms risetime).

Footnote 1: The truncation of the impulse response data before the first room reflection, that from the floor, means that the FFT-derived spectrum has a resolution of approximately 200Hz.—John Atkinson

Footnote 2: See Vol.12 No.10, October 1989, p.166 for how this test is performed and why I feel it gives a good correlation with the subjectively perceived tonal balance of a loudspeaker.—John Atkinson