[In September 2009 Jim Thiel died of cancer. As audiophiles we will miss him and his philosophical drive to improve speaker performance. The Boston Audio Society was fortunate to have heard and learned from him. His company, Thiel Audio (www.ThielAudio.com), continues his work. DJW]

Meeting Feature: Jim Thiel 
Jim Thiel (JT): For 30 years I have been trying to find out why loudspeakers in rooms sound like loudspeakers. I like to think of loudspeaker performance in four areas.

• Tonal Accuracy
• Clarity
• Spacial Realism
• Dynamic Realism

Some 25 to 30 years ago it was quite common to have loudspeakers that were very good in one of those, and very, very limited and even poor in the others. It is not so much like that anymore.

Tonal Accuracy
Evenness of reproduction across the frequency range — flat frequency response. Usually it's specified as ±3dB, which is a huge range. Even ±1dB does not guarantee a tonally accurate speaker. Drivers produce sound by vibrating and moving the air. If the diaphragm moves as a unit it is called piston motion. All diaphragms do this at low frequencies.

At higher frequencies different parts start vibrating independently, causing peaks and dips. It rings like a bell and produces output through time when it shouldn't, covering up details that should be audible. Take a typical 4-inch driver. In the past it was made of paper or plastic and maintained pistonic action to about 1 kHz. Ours are made of aluminum, which is a much more rigid material and maintains pistonic accuracy to twice that frequency.

You have to compensate for other factors and if it is not done well you can get a sound that sounds like aluminum. In the Thiel model CS7.2 speaker the diaphragm is made of three layers: aluminum on the front (which is what you see), then a layer of polystyrene that has been cast into a certain shape, and behind that another layer of aluminum. That [structure] is much more rigid and as a consequence operates as a piston up to 7 to 8 kHz. At 8 kHz it is breaking up, and even though that is far above the crossover point, there is still a little bit of imperfection.

One of the things I am most excited about in the Thiel model 3.7 is the new diaphragm geometry, which is basically flat but with ridges. They are oriented from the center outward and give it a great degree of strength. If you tried to grab the diaphragm and bend it in that [radial] direction, [you’ll find that] it's extremely strong. It acts like little I-beams. This makes the diaphragm weaker in the other direction, so we use a very large voice coil so it is driven midway [radially] and that provides the strength around the circumference. This allows it to operate as a piston all the way up to 20 kHz. Throughout the entire audible range there are no resonances or stored energy or mechanisms where the vibrational behavior of the diaphragm will alter the tonal characteristics of the sound. This is something that is satisfying to me and very exciting. When you take transient measurements on this [driver] it's like something you've dreamed about for years and never thought you'd see.

There is another benefit from the flat diaphragm. We have been mounting the tweeter coaxially with the midrange for years. If you mount a tweeter in a conventional cone with a steep slope you solve some problems but cause other ones (cavity resonances). With the three-layer construction we can make a rigid flat diaphragm that otherwise would be very weak. We use a similar geometry in the woofer and also a large voice coil for a wider frequency range. The cavity in conventional woofers is not a problem in itself, but a little bit of the energy from the other drivers is going to fall into the cavity and cause a reflection. The result is that our octave-averaged frequency response is ±0.5dB and even the [more detailed] averaged response is well within ±1dB. Part of the reason is that the drivers are so well behaved over their range.

Clarity
Another aspect is reproducing soft passages as well as loud. I've talked about transient behavior affecting details. A lot has to do with the enclosure. The primary job of the enclosure is to keep the energy from the back side of the speaker from entering the room and appearing as distortion. In our speakers we use 1-inch MDF with hardwood veneer on the inside as well as the outside for more rigidity.

We use a lot of internal bracing and the front panel is two inches thick (even in our least expensive models). Other speakers we make use a 4-inch thick baffle. The 7.2 has a 3-inch thick front baffle made from a marble polymer composite casting weighing 70lbs.

In the Thiel model 3.7 the sidewalls are curved plywood. We actually curve the plywood panels before they are glued together. Plywood is also a stronger material than MDF. The top is formed aluminum, curved for more strength. There are no parallel walls in the speaker. This speaker has an aluminum front panel. Aluminum is 34x stronger than wood, which allows us to mount the drivers rigidly. Even if we made the baffle two feet thick of wood, the driver could still move a little bit.

Question: What measurements do you perform to measure clarity?

JT: We take what is called a cumulative spectral decay plot. You excite the speaker at all frequencies and then see how it vibrates over time. Ideally you would see a pulse that would decay rapidly and uniformly. If you have cabinet resonances you'll see ringing. Also you get audible effects from the types of capacitors used. 

Joe deMarinis (JdM): What is it about the capacitors that do harm?

JT: Dielectric absorption — some of the energy gets stored in the capacitor and released later.

David Griesinger (DG): I'd like to know what kind of music you use to evaluate dielectrics.

JT: Short, quickly changing transients — how clear they are — or material that has a lot of low-level subtlety.

Spatial Realism
(imaging)
The sound should not sound like it's coming from a speaker. One factor is that in conventional speakers the acoustic centers of the drivers are different, so the sound arrives at different times, unlike in real life. One technique we use is time-coherence by building a sloping speaker-mounting baffle. Another way is coincident mounting, where the acoustic position of the tweeter is the same as the acoustic position of the midrange or woofer.

JdM: What about the difference in mass between the two diaphragms?

JT: There is a misconception about mass. Heavy mass doesn't slow the response of the driver at all. It only means weaker output. If I made this diaphragm out of a 0.5” thick material that weighs 2lbs, it will hardly move at all, but its transient response is the same. It seems counterintuitive. What mass affects is sensitivity or efficiency, which can be dealt with by magnet strength and other means.

We use 6dB/octave slopes [in our crossovers] because that (and only that) allows perfect phase and transient response. You put a square wave in, you get a square wave out. It gives you an aspect of realism you can't get any other way. The disadvantage is that because of the gradual cutoffs the drivers have to handle an extended frequency range.

David Moran: You must be the only speaker designer in the world to still use first-order crossovers, at least through the entire line.

Micha Schattner: What about lobing when the drivers are operating together over a wide range?

JT: We use a relatively low crossover frequency to minimize that. It keeps the wavelength at least as long as the driver spacing. That's another benefit of mounting the tweeter inside the midrange — there's absolutely no lobing.

[Talking about dispersion:] With sharp crossovers there's a sharp discontinuity in the radiation pattern as you cross over from a large driver to a smaller one. It goes from narrow to wide and causes a dip in the reverberant field. With 6dB/octave crossovers the transition is very gradual.

[Discussion of connecting drivers out of phase]
DG: If you have a 2nd-order crossover you have to connect the drives out of phase, otherwise you get a cancellation at the crossover point. With 3rd-order crossovers you don't have that problem [because the drivers are always 90º out of phase. DBH] and you can get very good phase response. I have to say your approach is the best if you can make it work.

JT: The Thiel model 3.7 [about $13,000/pr] uses an electrical crossover. Our cheaper models use a mechanical crossover, which has a cost advantage. I've spent a lot of time trying to make the radiation smooth and wide to make the listener position less critical. The problem is that the interaction with nearby walls is stronger than with a speaker having a narrow radiation pattern. So it's nicer not to have them near a wall. Question: How is your design process affected by the potential room?

JT: I don't pay attention. I'm designing what I want in the direct sound, which is what you hear first in any room. I measure anechoically (either with a small chamber or outdoors or with gated FFT measurements). I design for a 4Ω impedance, where better amplifiers produce more power [than at 8Ω]. I also design for a flat impedance curve that is not very reactive [reactive loads might adversely affect some amplifiers. DJW]. Sensitivity is 90dB/2.83V/m so it's easy to drive.

Dynamics
Dynamics are related to distortion. We've done a lot to reduce distortion. Almost all dynamic drivers use a system of a long voice coil and a short magnet gap. The reason is that when the coil leaves the gap you get tremendous distortion — as much as 1000%. How can you get more than 100% distortion? It's a ratio of wanted signal to unwanted signal, and if there is very little wanted signal...

We use a short coil and a long magnetic gap, which has 1/10th the distortion. We also do things in our magnet structures with copper rings and copper sleeves that greatly reduce the inductance and the nonlinearity of the drivers, which decreases the distortion and increases the dynamic range of the clean output a lot.

Question: What has computer-aided design allowed you do to that you couldn't do otherwise?

JT: We use it a lot, but I can't think of anything it enables you to do that you couldn't do anyway. It does save a lot of time.

Enjoy the Music.com highly encourages our readers to join the Boston Audio Society by clicking here).

This article is copyrighted © by the author or the Boston Audio Society. It is posted on Enjoy the Music.com with their permission, and with all rights reserved.