teaching demonstration used the human voice, and indeed most singing and talking voices are grossly changed, to sound all wrong, by polarity inversion. But these huge sonic errors also make the sound all wrong when applied to any sound that naturally blows out instead of sucking in, from trumpets to cannon shots. And these huge sonic errors also make the sound all wrong for all transients and percussive sounds (from triangle to piano to plucked strings to drums), since the initial attack transient is always asymmetrical, and therefore sounds wrong when the absolute phase polarity is inverted.
      The Hyperion HPS-968 sounds hugely wrong throughout the midrange, and sounds wrong and weird in all the ways that your own voice just sounded wrong and weird when you said "Pop!" while inhaling instead of correctly exhaling. Through the 968, the polarity inversion of the midrange, which is the heart of the total sound for most voices and many instruments, makes all voices and all these instruments sound like ghostly hollow, weakly wheezing phantoms that are far away, instead of solid, three dimensional entities that are immediately present, forcefully dynamic, and believably real.
      In the 968, the wrongness of the midrange driver polarity inversion is made even more obvious by the fact that the tweeter, handling the spectrally adjacent and aurally prominent upper midrange and trebles, operates in the correct absolute phase polarity. Thus, not only is the 968's midrange hugely wrong sounding from its polarity inversion, but there is also in the 968 a huge disparity and conflict between the midrange with its inverted polarity sonic qualities and the adjacent upper midrange and trebles with their correct polarity sonic qualities. Thus, for example, the main portion of a singer's voice will sound weakly ghostly and far away, while her sibilants (primarily treble energy) incongruously sound forcefully solid and immediately up front.
      Now, phase and polarity problems are exhibited by many loudspeakers, particularly those which employ higher order crossovers. So competent loudspeaker reviewers ought to be aware of this issue, and be sensitized to hearing and being able to identify the very peculiar (indeed unique) sonic problems caused by inverted phase polarity. As you heard yourself in your own voice, the sonic differences caused by phase polarity inversion can be huge, and very peculiar and weird sounding, hence easily identifiable by any reviewer who has any clue about what's happening in the product he's reviewing. Moreover, even though these phase polarity sonic problems can be partially masked in many other loudspeakers by the fact that their drivers operate in phase-scrambling cone breakup mode over much of the range, these same phase polarity sonic problems are glaringly obvious in the 968, precisely because the 968's midrange driver is so superb, being so transparent and, more importantly, so pistonically accurate over most of its range. So you'd think that every competent reviewer would notice and specifically identify this glaringly obvious inverted phase polarity problem in the 968's midrange.
      Why does the 968 invert the phase polarity of its midrange driver? The manufacturer claims that the 98's "new crossover design achieves extreme phase accuracy", a claim which was copied verbatim in one review of the 968, without any reviewer comment, critique, or correction. But our trained ears clearly tell us that this claim cannot be true, since the very unique and very weird sonic qualities of phase polarity inversion are so audibly obvious in the 968's midrange.
      The wiring from 968 input terminal through the crossover to the midrange driver seems to be in the correct polarity, so how then could the phase polarity become inverted? As noted above, the 968 runs its midrange driver full range at its low frequency end, so its crossover eschews the usual series capacitor in the signal path to the midrange driver, the capacitor which in other conventional loudspeakers acts as a high pass filter, rolling off input energy below the 150 Hz crossover point (the 968 woofers are rolled off above this 150 Hz crossover point by a series choke, but there is no complementary series capacitor to roll off frequencies below this 150 Hz crossover point for the midrange driver).
      But the 968 design engineer has inserted a series notch filter, in the form of a parallel LC tank circuit, in the crossover path to the midrange driver, to equalize the midrange driver's response, notching out electrical energy fed to the midrange driver (to offset a driver resonance). One of our tests suggested that this notch filter is tuned at about 55 Hz, but the manufacturer says this tank circuit is tuned to about 500 Hz (to combat and offset the buzzing resonance noted above). Regardless of which is the case, the majority of the midrange driver's output lies above this tank circuit's tuned frequency. Certainly, all those portions of the midrange driver's output to which human hearing is the most sensitive, and most sensitive to phase correctness, lie above a 500 Hz tuned frequency, and a fortiori above a 55 Hz tuned frequency.
      Now the real fly in the ointment, the real zinger, is the fact that a parallel LC tank circuit warps phase response, changing the phase response a full 180 degrees (which is complete inversion) from low frequencies to high frequencies. Thus, this midrange driver might well be in perfect phase synch with the woofer at DC (reflecting the way the wiring is connected), and at very low frequencies, below the tuned frequency of this LC tank notch filter (either 55 Hz or 500 Hz). But, thanks to this LC tank circuit in series with the midrange driver, all of the frequencies fed to the midrange driver above this tuned frequency (above either 55 Hz or 500 Hz) will be phase shifted to be substantially 180 degrees out of phase with the midrange driver's output at DC and very low frequencies (reflecting the way the wiring is connected).
      Since either the entire useful output range (150 Hz to 3000 Hz) of this midrange driver, or certainly the most audibly prominent portion of that range (500 Hz to 3000 Hz), lies above the tuned frequency of this LC tank circuit, this means that the entire useful range radiated by this midrange driver, or certainly the most audibly prominent portion of that range, will be substantially 180 degrees off in phase, i.e. will be substantially inverted in phase polarity, relative to the way that this midrange driver (and the woofers and tweeter) are electrically connected by the wiring. In sum, the 968 midrange driver radiates in substantially inverted phase polarity over most of its range, and over the most audibly important portion of its range, both absolutely and also relative to the other (woofer and tweeter) drivers, probably thanks to the phase shifting by this LC tank circuit. We suspect that the 968 design engineer (in China, we believe) is simply not aware of this phase shifting property of LC tank circuits, and also must have neglected to actually measure the phase response of his loudspeaker system design, before making that claim that "the new crossover design achieves extreme phase accuracy".
      Fortunately, we were able to develop a simple workaround for you, which makes the 968 sound very good through its midranges.
      Thanks to the fact that the 968 has 3 separate input terminals for its 3 driver ranges, it's easy for you to simply invert the polarity of the wire jumper you use to feed the midrange driver input. Voila! The midrange driver's output is now substantially in the correct phase polarity, over most of its range, and over the most audibly important portion of its range. And of course the sound of the 968's midrange is totally transformed. Try again saying "Pop!" while inhaling, and then say "Pop!" normally and correctly, while exhaling. That's how big a difference you'll hear, and that's the kind of sonic differences and kind of improved sonic reality you'll hear, from the 968's midrange.
      With its midrange driver now playing in substantially correct phase polarity, the 968 comes to life, and sounds much more vivid, real, and seamlessly cohesive for the whole spectrum, with the superb qualities of this unique midrange driver now clearly evident for the first time, and also blending cohesively with the tweeter for the first time. The images portrayed on a stereo or surround stage are now solid, three dimensional, and believably real, instead of being hollow ghostly phantoms.
      Our simple workaround, of inverting the wiring polarity to the midrange input terminal, is in theory not perfect, because in theory there could be an amplitude response notch at the 968's crossover frequency to the tweeter (3000 Hz). That's because the crossover filters, for both the midrange driver (low pass) and for the tweeter (high pass), are electrically second order, and pure second order crossovers cause a cancellation notch at the crossover frequency, when both drivers are in the same phase, as they now substantially are. However, in actual practice the sonic benefits of having the 968's superb midrange driver radiating in the correct phase polarity for most or all of its range far outweighed any cancellation notch that might have developed. We have a hunch that, in practice, there is actually only a minimal cancellation notch problem at the 3000 Hz crossover frequency, so it is in practice a negligible new problem. What might make it negligible in practice? Even though the 968's electrical crossover network is second order here, the drivers themselves surely have additional phase shifts at 3000 Hz from their nearby mechanical or acoustical rolloff points. These added phase shifts, added to the phase shifts of the of the pure electrical second order crossover network, would make the phase cancellation less 'perfect', less complete, less deep, hence less problematically audible.
      Speaking of phase polarity, it's also important for you to connect the 968's woofer polarity so that it matches the polarity of the lowest frequencies put out by the midrange, so that the two driver ranges correctly dovetail together. If they are connected in opposite polarity for the spectral region (around 150 Hz) where they overlap, then they will destructively interfere and cancel each other, rather than correctly adding together. This creates a suckout hole in the warmth and upper bass regions, cheating you of natural bass and warmth energy. This destructive cancellation is made worse by the fact that, in the present 968 design that eschews a high pass filter for the midrange driver, the midrange driver runs full range down to its lowest possible frequencies, and thus has a broader spectral overlap with the woofer, hence a broader cancellation region. What is yet worse, when we deliberately tried this polarity mismatch, to discover what it would sound like, we found that this destructive interference also has an ugly, peculiar, modulated phase quality, similar to the phasey distortion effect called flanging that is intentionally imposed on some recordings (made famous by Toni Fisher's hit The Big Hurt).
      How do you make sure that you connect the 968 woofer in the same polarity as the midrange driver's lowest frequencies? The simple answer and solution is to actually try connecting the woofer polarity both ways, while listening to say a plucked jazz bass (after you have correctly set up the inverted midrange driver terminal connection), and pick the woofer polarity that audibly gives you more energy in the upper bass and warmth regions (it will also be audibly more coherent, focused, and dynamic energy, instead of being phasey, hollow sounding energy from destructive interference).
      The complex and detailed answer gets a bit messy. Here goes. The 968 feeds its midrange driver via that LC tank circuit which warps the phase 180 degrees, from the lowest frequencies to the upper frequencies (the latter constituting the majority of the range) handled by this midrange driver. Since we want the majority of the range handled by this midrange driver to be in the correct absolute phase polarity, and certainly the upper frequencies where the ear/brain is most sensitive, we have to invert the polarity of our connection to the midrange driver input terminals. But this means that the lowest frequencies output by the midrange driver will now be inverted in absolute phase polarity. So, to match this, you should make the woofer terminal connection such that the woofer actually radiates in inverted absolute phase polarity.
      Is an audible problem created by playing the 968 woofer in inverted absolute polarity? Not to worry. Luckily, human hearing is relatively insensitive to absolute polarity at low bass frequencies (unlike midrange frequencies, where human hearing is very sensitive to both phase and polarity), and in the 968 the woofer is limited to frequencies below 150 Hz, so having its actual radiation be inverted in absolute polarity is not a significant aural issue. Moreover, most sonic waveforms are quite symmetrical at frequencies below 150 Hz (i.e. when band limited to 150 Hz, as they are for this woofer), so the woofer waveform itself stays substantially the same if turned upside down, i.e if its polarity is inverted, so there is actually no difference in the waveform itself to perhaps then be heard, even if human hearing were to be sensitive to hearing absolute polarity at low bass frequencies. That's the opposite of upper frequency waveforms (e.g. the midrange driver's), which do look, and then sound, very different when they are turned upside down by their polarity being inverted, as they are for the 968 midrange driver.
      How do you connect your cable to the 968 woofer terminals, in order to obtain this inverted absolute polarity of actual radiation? Here's where things get messy. The manufacturer designed the 968 with the woofer drivers themselves actually wired internally to the cabinet terminals in inverted polarity, so all early 968 production units, including our review units, have this internal inversion. For these 968s you should connect your cables correctly to the woofer cabinet terminals (red to red, black to black), in order to obtain the woofer polarity inversion we desire here (to match and correctly add to the inverted polarity of our midrange driver at its very low frequencies, produced by our deliberately inverted midrange terminal connection). But then later the designer evidently changed his mind, so for later production 968s they changed the internal wiring, such that the woofer drivers themselves are actually wired correctly internally to the cabinet terminals. So, for these later 968s, you should connect your cables wrongly to the woofer cabinet terminals (red to black, black to red), in order to obtain the woofer polarity inversion we desire here.
      The manufacturer's own records apparently fail to correctly show the serial number at which they made this production change in 968 woofer polarity, which is an unfortunate oversight, in view of the profound sonic difference this makes (his records show that he made his polarity change at serial number 4021572, but our review sample is serial number 4021637, a later serial number, and its woofer polarity is still inverted, like earlier production). But fortunately you don't need to know the serial number. To discover which way your particular 968 woofers were wired, you can simply perform the sonic test we developed in our product analysis, as discussed above, and connect your woofer terminals based on our sonic test.
      Note that the 968, as designed, manufactured, and sold in all those earlier serial number units, actually contains two severe phase problems and phase conflicts, in two distinct spectral regions (both problems being highly audible and objectionable in our testing) - if you connect all terminals as indicated by the manufacturer, instead of following our special instructions here, derived from our research on this product. Firstly, the 968 midrange driver is in substantially inverted phase polarity for most of its range, with the attendant sonic problems discussed above, including a polarity conflict with the adjacent tweeter that is in correct phase polarity. Secondly, the 968 midrange driver is also polarity mismatched, in the low portion of its spectral range, with the adjacent woofer, causing that destructive interference discussed above, with the attendant sonic problems including a weak amplitude hole in the upper bass and warmth region, plus that peculiar phasey distortion there. This second polarity conflict occurs because, if you connect all terminals as indicated by the manufacturer, the woofer will be inverted, whereas the midrange driver will be in correct polarity for that lowest portion of its range that dovetails with and overlaps the woofer.
      Again, you don't have to worry about these complex details, nor do you even have to check the serial number of a 968, if you follow our simple advice to merely first connect the 968's midrange driver terminals in inverted polarity, and then literally try both polarities on the woofer connection, to hear which gives you more power and more coherent dynamics in the upper bass and warmth  regions.

Sonic Degradation by Tweeter Resistors

      The second sonic weakness of the Hyperion HPS-968 is only half a weakness, since you can fix it so easily, as can the manufacturer. The 968 tweeter is outfitted with an attenuator, so you can easily change the tweeter level to match the differing high frequency absorption characteristics of different rooms and different room furnishings. To Hyperion's credit, they elected to make this attenuator a switched device, with a rotary switch selecting among different discrete resistors, instead of using the cheaper rheostat employed by many other loudspeakers. Unfortunately, the particular discrete resistors selected by Hyperion for this attenuator are not sonically optimum.
      On our high resolution lab system, it was easy to hear that the highest level setting for the tweeter was not merely the brightest, but was also substantially more transparent and focused, articulate than all the other settings. All the other four lower level settings sounded similar to one another and similarly degraded sonically, sounding more veiled and fuzzier, less focused than the brightest setting. The fact that all the four other settings sounded similarly degraded, even though their brightness quantity was different, reassured us that we had not been misled into favoring the brightest setting simply because it was the brightest (hence making treble details the loudest).
      From this listening experiment alone, we could scientifically infer the design of this attenuator, even before opening the loudspeaker to look at it or looking at a schematic. We could scientifically infer that sonically substandard resistors were affecting the tweeter signal path in all four lower settings, but that the highest setting completely omitted this sonically substandard resistor at this location (instead of the attenuator design having the same type of resistor at all 5 locations, with merely different resistance values). That's why the 968's treble transparency and articulation got better at the highest level rotary switch setting, and only at this setting. Sure enough, when we later opened the 968 head unit, there were four resistors connected to the four lower switch positions, but no resistor at all similarly connected to the fifth, highest position.
      Clearly, the manufacturer should change to a power resistor of higher sonic fidelity, for these four circuit locations. When we later saw the crossover schematic, we noticed that there is also another similar resistor in series with the tweeter (this resistor was buried in the potting compound, so we couldn't see it previously), and that the four resistors identified previously as sonic culprits are actually in shunt across the tweeter, as part of the attenuation network. This tells us that passive parts used even in shunt are still actually in the signal path (contrary to popular opinion), and can still degrade sonics (they do so by imperfectly, distortedly bleeding a portion of the signal to ground, thereby leaving an imperfect, distorted remainder at their top node, and this top node is indeed in the

(Continued on page 158)