July 2001

It is a generally accepted notion that the choice of loudspeakers is the most personal when it comes to the assembly of an audio system. Loudspeakers come in more different flavors, and with more widely varying design philosophies, than arguably all other component categories. For the unsuspecting customer who is suddenly confronted with hyped-up marketing propaganda and seemingly scientific white papers, the entire subject can quickly degenerate into a volatile mine field. Whom do you believe? To add confusion, loudspeakers based on conflicting or diametrically opposed design philosophies may sound equally appealing. In short, the fascinating subject of loudspeaker design is over-ripe for endless pickings, disagreements and multiple opinions.

Part 1 presented the "Tao of Speaker Design" according to Kevin Voecks of Revel Loudspeakers. Today Jim Thiel of Thiel Audio adds his observations. Already a picture emerges. Loudspeaker design is a skillful negotiation among certain trade-offs. They become active the moment a particular approach -- any approach -- is chosen.

The Thiel-er picture

Thiel Audio has been in business since 1977 and is located in Lexington, Kentucky. Today the company manufacturers 11 different loudspeaker models that sport the "CS" nomenclature (and model-specific numbers) to indicate "Coherent Source." This is in reference and tribute to the first-order time-and-phase-coherent design philosophy that spawned Thiel. Thiel Audio is one of the oldest high-end loudspeaker manufacturers in the US, as well as one of the very few championing this particular design approach. Other US firms in the first-order vein are Vandersteen, Dunlavy Audio Labs, Meadowlark Audio and Green Mountain Audio -- a small handful when compared to the more common second-order-and-up firms.

As a shallow-slope designer who strongly believes in the importance of signal fidelity, Jim Thiel naturally can't agree with Kevin Voecks' earlier statement that time-and-phase accuracy is inaudible and therefore the least important consideration in loudspeaker design. However, Thiel does agree with Voecks on the majority of his engineering priorities. He even goes as far as stating that time-and-phase fidelity is not as important as, for example, accurate frequency response or smooth off-axis behavior. He simply maintains that fidelity to the original signal not only in the frequency and amplitude but also phase and time domains is important and audible. He is comfortable that in a non-sighted comparison, he could pick out any speaker that was not time-and-phase correct.

This immediately begged my question as to what particular sonic characteristics would allow such a clear differentiation. To explain one of its aspects -- improved image accuracy -- Thiel pointed to how our ear/brain mechanism naturally localizes sounds via the difference in time arrival of aural events that reach one ear earlier than the other. This requires that speakers reproduce, say, the single tone of a clarinet in such a fashion as to deliver the complex structure of fundamental and harmonics at the ear all at the same time. If there were delay between the fundamental and any of its overtones, the image would appear less precisely defined and incorrectly sized.

By implication, this means that all other speakers scramble the signal in the time domain and force the ear/brain to rely instead on a less natural and more primitive means of localization: the loudness differential between left and right speakers. We all know how this works. A monaural recording reproduces all sounds equally loudly through both speakers. The sonic-event horizon is confined to a small space midway between the speakers. Highly accurate systems compact this space into a tight cluster of sounds all piled atop each other. The phenomenon of the left-to-right soundstage only occurs because two speakers don’t reproduce most sounds at identical levels. The more this channel imbalance favors one speaker, the closer toward it a particular sound will seem to originate. If only one speaker reproduces a sound, it will seem to emerge directly from that speaker.

Proponents of properly implemented first-order designs argue that the added work of translating spatially scrambled signals into any semblance of realism causes subconscious exhaustion in our ear/brain decoder. This manifests as eventual listener fatigue. Common and readily identifiable causes of listener fatigue (to mention a few) are unsuitably high playback levels, exaggerated treble and undamped ringing from metal drivers. The fatigue caused by time-and-phase anomalies is much harder to identify. We're simply not conscious of how our brain analyzes, discounts or integrates such "reminders of artifice" into a satisfactory reconstruction of the audio signal. In fact, the relative insidiousness of these playback errors means they turn obnoxious, in an indirect way, only over the long haul. Without a clear "Aha, that's what's wrong with my system" insight, listening sessions may simply shorten. The apparent cause may be a subtle discomfort, a wandering of attention or perhaps a diminished sense of enjoyment and relaxation that listening to the system originally bestowed. Needless to say, an appreciation for the subtle benefits of designs that avoid this extra decoding on the part of our sensory system won't reveal themselves in short-term dealer demonstrations.

Thiel claims that especially with minimally miked recordings, the first-order superiority of soundstaging precision (between and outside the speakers as well as in the depth perception) is far from subtle. He also hears clearly improved percussive transient attacks and an overall enhanced realism that makes distinctions between "is this real or reproduced" ever more difficult.

Not every first-order design acts that way

When talking first-order networks, an important distinction needs to be made between purely electrical first-order crossovers and synthesized variants. It is an often-overlooked fact that raw drivers connected directly to an input signal will exhibit natural or mechanical roll-off. In fact, common tweeters usually feature a second-order mechanical high-pass function. Simply adding the single cap of an electrical first-order filter results in an acoustical third-order response. While this could be touted as a first-order network to claim the benefits of single-part simplicity (after all, electrically it is a first-order filter), this is not how Jim Thiel designs his speakers. His crossovers are synthesized to create a true first-order acoustical function. To avoid lengthy shop talk, he simply states that the crossover will do whatever is required to compensate for the natural driver roll-off and create a time-and-phase coherent response at the listener's ear.

To deliver this time-coherent signal intact at the listener's ear also requires an unalterable relationship between speaker height and listener height. This is one reason why stand-mounted first-order designs are rare. Unless the designer can somehow force you to purchase his own stand, he won't know how high or low you will position the speaker in relationship to your ears. Leaving this parameter to chance undermines the inherent benefits of Thiel's compensated-filter-network design and sloped-baffle physical time alignment.

Coincident first-order drivers can go places

The advent of home theater, with its wall- or stand-mounted rear-channel speakers, forced Jim Thiel to explore coincident driver technology. The physical alignment of a driver within a driver remains constant regardless of placement. (Richard Vandersteen uses a coaxial driver in his center-channel speaker as well.) Originally pioneered by KEF and Tannoy, coaxial or dual-concentric drivers aren't new. Coaxial drivers that are suitable for the specific wide bandwidth requirements of shallow-slope networks are. Having his own in-house driver design and manufacturing department proved priceless when Jim Thiel embraced home theater. He originally felt compelled to get into the costly parallel business of transducer manufacture for other reasons.

In 1995, his CS7 demanded a very special 12" high-excursion woofer, which none of his suppliers could fabricate to his specifications. Embracing the added complexity and required funding to start a second business, Thiel has exploited his self-reliance on driver design and fabrication ever since. Every new model from 1995 on has benefited from custom Thiel drivers rather than off-the-shelf units or modified OEM units that before required suppliers to re-engineer stock models. As an example, Thiel cited tweeters that for his applications needed to be flat to 500Hz. Asked what kind of adaptations a company like Dynaudio or Scan-Speak would have to implement to lower tweeter response this dramatically, Thiel named as the first requirement the lowering of the tweeter's resonant frequency. This could be accomplished by venting the pole piece into a damped rear chamber. A softened suspension would go yet further. To exercise broader frequency range, the driver would now require increased linear excursion, more robust lead-in wires (that wouldn't break under the broader range of motion) and a longer gap, necessitating a larger magnet. Needless to say, an affordable tweeter perfectly adequate for higher-order applications just turned rather more complex and expensive. Ç'est la vie in windy trade-off city!

Dual-concentric or coincident drivers place a tweeter inside the dust cap of a midrange/woofer. This creates a minor horn-loading or wave-guide effect for high-frequency radiation. Thiel optimized the geometry of his surrounding mid/woofer diaphragm for the tweeter by making it shallow. To compensate for the incurred loss of rigidity of a shallower profile, the company bonds a second layer of cast polystyrene to its inside. In Thiel's top coincident transducers, the construction actually turns tri-layer, with an inner and outer layer of aluminum sandwiching the inner polystyrene core. In another version of its dual-concentric driver, Thiel uses a mechanical rather than electrical crossover. This eliminates not only the network proper but also the second magnet structure, which together account for the majority of driver cost. Everything else being equal, using this particular driver allows Thiel to give three-way performance for two-way cost in models where this driver is mated to a separate bass driver.

Metal to the cone

Thiel Audio championed metal-driver technology as early as 1992, when their model CS3.6 (still in today's lineup) introduced an aluminum woofer. Painted black, the CS3.6's woofer (and a companion version for the CS1.5) eluded most people as a metal driver. Asked why he prefers metal diaphragms, Jim Thiel cited a number of reasons. Cone/dome resonances are easily pushed well outside the pass band in which the driver is to be used. Metal diaphragms conduct heat and can be used as integral heat sinks to equalize changes in voice-coil temperature and minimize thermal non-linearities. Aluminum also features higher compression strength than, say, treated paper. This makes for clearer, crisper attacks when a tone first bursts out of silence.

The trade-off lies in diminished self-damping properties that are common for more lossy transducers. When metal drivers ring, they do so to the nasty tune of 10dB peaks versus the 3-4dB spikes of non-metal versions. To eliminate this annoyance, Thiel implements notch filters that produce the inverse of the resonant peak to cancel it. Since these notch filters are very precisely set to specific frequencies, they must rely on high consistency from driver to driver to work properly. This is yet another reason for Thiel's preference -- the consistency and thus predictability of aluminum, from transducer to transducer, is much higher than with any other material. Contrary to certain opinions, the addition of notch filters is said to not only not undermine the time or phase alignment, but in fact help establish it in the first place.

Power

Thiel speakers have long suffered a reputation for being power hungry. Jim Thiel admitted that most of his speakers sport 86-87dB sensitivities and are 4-ohm loads. He could easily transform them into 8-ohm loads, but he prefers to avoid the concomitant loss of sensitivity by 3dB. He also uses 4-ohm loading since most well-designed solid-state amplifiers will double their output power into such loads. Why purchase this amp capability only to leave it unused?

When asked what compromises would be involved in raising the sensitivity of his designs, Jim Thiel pointed to the requirements of reducing his drivers' moving mass (diaphragms) and/or increasing their motors' magnetic strength. The problem with the latter, especially with tweeters, is that saturation already tends to occur with existing designs. A larger magnet simply becomes redundant. A lighter woofer cone won't be as rigid, will have less power handling and might over-damp the bass. As a speaker designer, Jim Thiel is knowingly in the trade-off business. He prefers to sacrifice a couple of dBs of sensitivity for the greater output and bass extension capabilities he wins in return. In fact, he's proud to claim that Thiel speakers in general offer greater dynamic range and lower bass extension than equivalent models by competitors.

Another argument on Thiel's behalf

Asked why well-designed two-way speakers seem to always outperform more complex multi-driver designs in the areas of soundstaging and coherence, Jim Thiel offered an easy and logical explanation. (That it happens to justify his shallow-slope philosophy only adds further elegance in this context.) A miniature 4" two-way with a 5kHz crossover point acts, for all intents and purposes, as a crossoverless speaker well into the treble. The de facto absence of a crossover means no screw-ups in the time and phase domains (the same holds true for headphones that are full-range, single-driver designs). The moment a third driver adds another crossover point (this time clearly audible because it usually occurs around 500Hz, where human hearing is very acute), the imaging and coherence of our former 4" two-way suffer. No longer is time-and-phase alignment in the crossover region maintained. Thiel claims that the addition instead of a first-order crossover remains utterly inaudible. His three-way designs are said to be as coherent as minimonitors and soundstage just as well when set up properly to optimize the time alignment.

Distortion, or looking behind the pretty face

Another big concern of Jim Thiel is distortion. He explained that most exotic and expensive drivers spend the majority of their money on esoteric diaphragm materials. End users get to admire them in exchange for their hard-earned dollars and feel content. Thiel holds that with their actual motor assemblies, such exotic drivers tend to skimp on the inside, where casual onlookers will never notice. He attributes the significantly reduced distortion in his designs (less distortion by a factor of 10 to 20) to the optimized short-coil/long-gap architecture of his motors. A short voice coil inside a long magnetic gap always remains under full control regardless of excursion. It always operates in a constant and linear magnetic field, thus increasing usable non-distorted dynamic range. High-output low bass notes can cause voice-coil currents to momentarily create their own magnetic fields. Such temporary and secondary fields oppose the stationary field of the actual magnet to cause a momentary loss or partial demagnetization of its radiated field strength. Thiel likened this effect to pushing against a non-stationary object. To prevent such counter-productive action and expand the bandwidth of linearity in his bass drivers, Thiel uses solid copper rings around their pole pieces. These act like massive inductors around AC current and counteract the secondary magnetic reactions. The end result is enhanced low-frequency performance and accuracy for his proprietary transducers.

To chop in your own shop doesn't make you a chop shop

Jim Thiel's woodworking brother, Tom, was part of the home team when the fledgling company still operated out of a garage. Relying on a consistently growing and more sophisticated in-house cabinet shop as the years progressed enabled the company to build a well-deserved reputation for superior cabinetry. "It looks as good as a Thiel" shortly became a similar measure of relevance as "it drives as well as a Beemer." In modern corporate parlance, this all-under-one-roof model of running a business goes by the name of vertical integration. In the case of Thiel, it means to design your own drivers, fabricate your own drivers, build your own cabinets, then assemble, test, package, warehouse and ship everything. In the world of high-end speaker builders, this is fairly uncommon. Even a large speaker company with as well-funded an infrastructure as Revel subcontracts cabinet manufacture to outside facilities. Most smaller companies simply operate like job shops -- drivers and cabinets are outsourced (in some cases, crossover assembly is too) and only final assembly occurs in-house.

Conclusions

As the position of an engineering-driven firm, Thiel's perspective on elemental loudspeaker design mimics Revel's. Frequency-domain honesty, uniform and wide dispersion for optimal power response and a reduction of distortion are considered key aspects by both. The primary difference in philosophy lies with the relative importance placed on fidelity to time and phase. Thiel claims it's audible and has staked its nearly 25-year history on it.

Who is right? Listen to Revel and Thiel speakers. If you like both, neither is wrong. Whether that makes either of them right, I'll leave that for you to ponder. That very dilemma is why this series was called "The Tao of Speaker Design." The concept of Tao is big enough to contain within itself all polar opposites. But this certainly doesn't appease our minds. They're habitually derailed by paradox. However, paradoxically is how life works in general -- so why not in audio too? Here's to madness -- and variety!

More?

The next contribution to this series has been promised by MartinLogan. Keep your speaker leads plugged in until then.

...Srajan Ebaen
srajan@soundstage.com

• The Y-Files Current Issue