Search the Blog

Have a technical or support question?

Hit the "SUPPORT" tab on the right margin of this page to email our support team directly!

Contact Us
Featured Product

Blog Index
The journal that this archive was targeting has been deleted. Please update your configuration.
Jan 13 2015

The Reference: Designing the Crossovers and Voicing the Loudspeakers

With the release of the new Reference Series last year, we raised the bar on our already lofty standards of audio reproduction. I know that sounds like marketing jibber-jabber, but the truth is, we're all very proud of how amazing The Reference line sounds. This was accomplished a number of different ways, not the least of which was the painstaking method we selected crossover components and voiced the loudspeakers themselves.

A Brief Note About Crossovers

Since different sized speakers are better at different jobs, but a single loudspeaker cabinet is required to reproduce the entire audio spectrum, crossovers are need to make sure the correct signal goes to the correct speaker or driver. There are two types of crossovers – active and passive.

Passive crossovers do their work using only the power supplied by the audio signal and passive electrical components and as such are fixed – the crossover point is designed by the engineers at the time the speaker cabinet is developed.

Active crossovers are electrical circuits that use some level of processing to get the job done. The subwoofer setting on your AV receiver is a type of active crossover.

There are almost as many different opinions about crossover networks and how they should be implemented as there are music lovers. At the end of the day, how you enjoy what you hear is pretty much subjective, but there is a great deal of research and engineering that goes on long before you get a chance to listen to a pair of KEF speakers.

Designing The Reference Crossover Networks

Each component in the signal path adds some level of distortion (meaning the original sound is changed somehow). This is an unyielding fact of physics. Therefore, extensive objective testing of the components to be used in the crossovers was done to identify potential components with the lowest distortion. Surprisingly, what our engineers found was that there was very little correlation between the cost, spec, or physical size of a component and the amount of distortion it introduced. So rather than just picking or eliminating components based on those criteria, an enormous variety of components were tested to find the best match between the drivers (and the other components). One $0.05 resistor can do a great deal of damage to the fidelity of a crossover (and therefore the loudspeaker), so it's extremely important to be thorough and get it right. Having our own in-house and highly qualified engineering department goes a long way toward making sure we've gotten it right.

In the Figure A below, you can see the measured THD (Total Harmonic Distortion) from a 20V input signal for three different inductors that were all similar in cost and resistance but were from different manufacturers and were constructed differently.

Figure A .

In Figure B below you can see the measured distortion (from the identical signal as above) for three capacitors, all with equal capacitance values but each from a different manufacturer with different construction.

Figure B.

It's easy to see from these graphs that there's a lot more to choosing a component than it would first appear. While listening may be subjective, science is objective and the brilliance of the sound of The Reference lies in the meeting of the two.

It's not just that components singularly can make a difference, but how each selected component interacts with every other component in the circuit can have a huge impact on the performance of a crossover network. Even though testing weeded out a lot of components that weren't up to the task, even more were weeded out once our engineers started seeing how each component interacted with every other component. In Figure C (below) you can see some components that passed the distortion screening on their own failed msierably when they became part of a circuit.

To avoid distortion caused by vibrations within the individual components, mastic was applied to the capacitors (due to their construction, capacitors are very susceptible to disruption caused by vibration) and the crossovers were split into two separate sections, each mounted away from each other in the cabinet.

It's this attention to every minute detail of design that makes us so proud of our products.

Voicing the Loudspeakers

Just like the crossovers, the drivers were also designed from the measured responses of the individual drivers within the final enclosure. Once this was done, hard, subjective science made room for the soft, objective science of the listener.

This was done over the course of several months and was carried out by a large number of KEF personnel, all with different areas of expertise and with different musical tastes. Concerns raised by any of these listeners were then investigated in the lab using different measuring techniques to try to uncover the underlying cause of the concern. More often than not there was a direct correlation between the objective data provided by the listener and the subjective data uncovered in the lab through rigorous measurement.

Rather than simply re-voicing a driver based on concerns and issues raised through the listening process, our engineers were able to dig into the underlying reason for the concern and correct it there, rather than simply applying what could be considered an audio Band-Aid on a problem. This method is extremely time-consuming and expensive, but the end result has been absolutely well worth it.

Gratifyingly, as this systematic procedure progressed the preferred subjective performance merged with the balance that gave the best objective measurements, in particular the lowest distortion and the smoothest and flattest on and off axis frequency response. - The Reference White Paper

Jack Sharkey for KEF

Oct 21 2014

Tangerine Waveguide: It Does Way More Than You Think

The picture of the swamp at left was taken with a fish-eye lens. I was looking for one of those funny fish-eye pictures with a dog's face. You know, where his nose looks ginormous and his eyes are pleading at you not to make fun of him, but I couldn't find any that were licensed for re-use, so you'll have to deal with the swamp picture.

In photography, a fish-eye lens uses a principle known as . Snell's Window is the circular area of light above the water that a person (or presumably a fish) underwater views through a "cone" of light which is caused by light refracting through the water. The area outside the cone of light is either dark or a murky reflection of objects on the bottom of the body of water.

For our purposes, the fish-eye lens distorts the otherwise linear light waves passing through the lens making them appear convex, or non-linear. To use this analogy for a tweeter, we can basically view the distortion of the light waves as similar to the non-linearities of frequency responses across the face of the tweeter.

The parallel between a fish-eye lens in photography and our Tangerine Waveguide basically stops there, but the comparison is a good way to get your head around the audio science we're going to explore next.

Let's Talk About Surface Normal Velocity

As usual, when it comes to all things audio, in order to understand one thing, we have to know about other things first. Without getting involved in things like tensor calculus and covariant derivatives, let's just say that the surface normal velocity toward the dome perimeter is smaller than the surface normal velocity at the center of the dome.

Translation: The center of the dome moves faster than the outside of the dome.

This is because the curved nature of the shape of the dome causes the perimeter to move slower (and at an angle) as compared to the center. A perfect surface normal velocity for a tweeter would be the same over the entire dome surface, but this is not possible because the dome surface would have to stretch. Tweeter domes that stretch don't sound very good at all.

Because the Tangerine Waveguide corrects for the non-ideal dome motion, sensitivity at the top end of the audio band is increased. Dispersion is also improved because KEF's engineers have been able to shape the fins and channels of the waveguide to control the expansion of the soundwave into the horn (waveguide).

And Another Thing: Compression Drivers

The next time you go to a show in a large hall, or your neighbors call for the kids on their bullhorn, you are listening to a compression driver. Spraying your wife by using your thumb over a garden hose is an example of a) bad decision-making; b) a compression driver.

Basically, a compression driver is a small diaphragm loudspeaker that attaches to a horn (the part you see) which is a duct that radiates the sound into the surrounding air. Because the area of the diaphragm of the actual driver is larger than the throat of the horn, high sound pressure levels are created very efficiently (about 9-10 times more efficient than standard cone loudspeakers; the more efficient a speaker the less power you need to create sufficient sound levels).

This is a Celestion compression driver. The diaphragm is facing up; the threaded part connects to the radial horn (shown below).

Definition: A phase plug is a body having an input side of multiple channels for receiving acoustic waves and an output side of multiple channels for transmitting acoustic waves.

KEF's Tangerine Waveguide is basically a radial channel phase plug that creates a slight compression loading effect which results in slightly higher output gain. This extra gain from the waveguide is seen from around 7kHz to around 15kHz. Above 15kHz the effects are minimal (but so is the audio information). The extra gain also helps to mitigate the affects of the peak of the tweeter caused by mechanical break-up. This all results in more efficient high-end with greater dispersion and better fidelity.

Seen below is the Tangerine Waveguide on the tweeter of a brand-new Reference Uni-Q.

All of this and it will protect the fragile tweeter dome from Junior's curious fingers when you're not around!

For more information on KEF's Tangerine Waveguide, see .

Jack Sharkey for KEF

Sep 04 2014

KEF's Z-Flex Surround: Smoother HF Through Science

Every component and sub-assembly in a driver affects the overall performance of the driver. That's a pretty basic and logical statement, but when you really consider what that actually means, you realize how complicated speaker design is.

Consider the surround. At first glance it's basically the bit that holds the speaker cone in place, but the surround also moves, and with that movement come resonances and unwanted vibrations that can alter or change the sounds a speaker is trying to replicate.

In the cut-away of a bass driver (right), a typical "half-roll" surround (#3) is shown. Typically made of butyl rubber or other foam material, the half-roll surround, when properly designed, can perform quite well. However, KEF's use of Finite Element Analysis (FEA) has allowed our engineers to discover other possibilities that result in a better listening experience when listening to music through our products.

The biggest benefit from the Z-Flex surround is with the high-frequency response of our Uni-Q Driver Array.

For optimal performance the Uni-Q tweeter is placed in a perfectly smooth waveguide, which in the case of a Uni-Q would be made up from the mid-range driver and the surround. In an ideal situation, the high-frequency driver produces a smooth (hemispherical) 'point-source' radiation pattern, resulting in a tweeter with higher sensitivity than a tweeter placed in a flat baffle (the front edge of your speaker cabinet). Previous Uni-Q designs have used the conventional half-roll design, but the half-roll design, by the nature of its shape, compromises the ideal waveguide for the tweeter.

With the half-roll design there is an abrupt discontinuity in the waveguide at the transition from the cone to the surround which causes a secondary radiation. That secondary discontinuity, or diffraction, tends to muddy or smear the sound from the tweeter. Figure 1 shows the diffraction from a half-roll surround (in the area between the tweeter and the pink arc about halfway through the graph). Figure 2 shows the absence of the diffraction with KEF's Z-Flex surround. The absence of that diffraction results in smoother high-frequency response and dispersion.

Figure 2. High-frequency response with KEF's Z-Flex surround.

In order to get the true experience from a musical performance or soundtrack, there's a lot that needs to be considered when designing (or purchasing) a speaker. And one thing to consider is that the surround is way more than just the rubber bit that holds the speaker cone in place.

Jack Sharkey for KEF

Jul 30 2014

Why So Called "Coaxial" Speakers Aren't Uni-Q

First, a few definitions:

  • Coaxial: Geometrically speaking, two (or more) three-dimensional forms that share a common axis. Triaxial would specifically refer to three three-dimensional shapes sharing the same axis
  • Paraxial: Refers to two (or more) three-dimensional forms that lie in close approximation to each other and that form a small angle between each other
  • Point Source: A single localized source not perceptibly distinguishable from other sources
  • Directivity: Referred to as 'Q,' the measure of the radiation pattern from a speaker. A loudspeaker with a high degree of directivity (narrow dispersion pattern) is said to have a high Q.

Now that we've got that out of the way, back in the day, the Holy Grail of audio enthusiasts was the coaxial speaker, and if you were really advanced, the triaxial speaker.

The photograph at left shows a very cool "triaxial" speaker from a 1955 Pilot Console Hi-Fi Set . The speaker is from University Sound which was an early division of the Altec Company. The advertising copy listed the speaker as "diffaxial," with three drivers mounted concentrically to each other. The size of the HF driver is pretty substantial, as the LF driver looks to be about 12" in diameter.

As coaxial speakers evolved in the years since this behemoth was first introduced, one of the main problems to developing a truly coaxial speaker was the sheer size required to make a decent HF unit. One look at the magnet in the Pilot/University Sound speaker and we can see that even with all of that space, the speaker was still not truly coaxial. Just sticking an HF unit in the center of an MF unit doesn't solve the problem of point sourcing and phase coherency.

Right about now you might be wondering why you should even care about any of this.

Right Now I'm Wondering Why I Should Even Care About Any of This

To truly replicate a musical performance, the concept of acoustical point source needs to be taken into account. When you speak, regardless of whether you use your lowest tone or your highest tone, all of the sound (basically) comes from one point: your mouth. To take this one step further, a recording engineer will generally use only one microphone to record your voice. A sound generated from a single source point is recorded by an instrument (the microphone) that is a single point receiver: All of the frequency components of the sound (in this case your voice), come from one source and are then captured by one receiver.

A recorded speaking voice isn't a particular challenge for most speakers as the human voice is limited to a rather small range of frequencies (think about the speaker in your phone).

That is unless you want the playback to sound as much like the original as is technically possible.

  • Male singing voice typically ranges from about 100Hz to about 800Hz
  • Female singing voice typically ranges from about 250Hz to a little over 1kHz
  • Our speaking voices are far more limited. My regular speaking voice sits right around 200Hz but when I say words with T s and S s in them I can measure components that go well into the 4kHz range

To reproduce sounds as faithfully as possible, two (or more) drive units are required. Large, heavy speakers do a great job reproducing deep bass but just can't cut it for higher range instruments and sounds (like flutes and cymbals). Conversely, drivers that are quick and light enough to reproduce high frequencies generally begin to distort as soon as they are asked to reproduce bass sounds.

Let's take a listen to Yo-Yo Ma and Bobby McFerrin (along with guests Marc O'Connor on violin and Edgar Meyer on contra bass) performing Hush Little Baby. Not only is this a fabulous performance of a simple song, but it will also help illustrate the concept of single point sourcing.

There are several ways to look at single point sourcing.

From the audience's perspective, the stage is an array of single point sources (the cello, McFerrin's voice, the contra-bass and the violin) that all work together to produce a single, cohesive musical performance. Add the reverberation of the room and the ambiance of the audience and you get the full experience of a live performance.

From a recording engineer's point-of-view, there are also multiple single-point sources. At the 2:31 mark we can see that there are four separate microphones used to record the performers on stage, (there are three stage mics on boom stands and one handheld microphone). There are also mics placed in other strategic spots on the stage and around the theater to capture the ambiance of the space. These sounds, all captured by single-point receivers (the microphones) are then mixed together resulting in a recording that replicates (if everything is done right) what the audience heard. The tricky part is when you take the recording home and try to replicate it on your home stereo, and the purest way to do so is with a loudspeaker that is in and of itself a point source.

Now let's look at the individual components for a moment to illustrate just how hard it is for a speaker system to replicate this performance:

  • The contra bass is producing frequencies from around 50Hz to around 130Hz while McFerrin's whistle is producing frequencies around 2kHz
  • When McFerrin scats (vamps), his voice is around 600Hz and when he sings he's between 300 and 500Hz
  • The cello is around 500Hz and the violin is up and over 2kHz in some spots
  • Add harmonics and the ambiance that makes a performance "real" and you've got frequencies around 3.8kHZ (voice harmonics), 5kHz (violin), and 7kHz and 9kHz (McFerrin's whistling),

Even in a very simply orchestrated piece, that's a lot of work for a loudspeaker!

In the audio industry it is not new knowledge that to achieve single point source reproduction the acoustic centers of the MF and HF drive units should be in the same place. But as our 1955 hi-fi showed earlier, sheer physical size was an issue, and developing the technology to achieve single point source reproduction was not easy.

To get around this, loudspeakers are designed with the HF mounted above the MF, or with both drive units mounted side-by-side (paraxial). The problem with this is that the sounds come from two distinct locations. Even though McFerrin's whistle, scat singing and regular singing all originally came from the same place, they would be handled by different drive units in a loudspeaker. It's easy to see that if one drive unit is mounted to the left or right, or above or below, another drive unit trying to replicate the same performance, the sound comes from two places, and inherently will not arrive at the same time at the listener's ear.

With one drive mounted slightly ahead of the other as seen in the picture of a conventional coaxial speaker to the left, the sound may come from almost the same physical location on one axis, but phasing becomes an issue if the drive units are not aligned in space properly on all axes.

Both of these issues can cause phase incoherency or confusion that gets processed by our brains unnaturally, resulting in performance reproduction that isn't life-like.

The answer to this problem was found in a magnet made from a combination of neodymium, iron and boron that is ten times stronger than a standard ferrite magnet. The strength of this new magnet material allowed us to build a HF unit that was small enough to fit inside a standard-sized MF drive unit, thus allowing our engineers to find the precise placement point of the two drivers so that they were completely coincident with each other. This 'coincidence' allows all of the upper and lower frequencies of a single performance to actually emanate from the same place in space.

Put technically, the bass and treble units are time-aligned in all directions in the three-dimensional plane, not just in one axis as with vertically separated units like the speaker shown above. This mis-alignment, known as the vertical interference pattern , can result in a very tight 'sweet-spot' ( high Q ) or area of high-quality sound reproduction that is often less than 10 degrees above or below the principle axis.

Translation: Only one person gets to hear the high-quality sound and everybody else gets to hear a mish-mosh of phase incoherency.

Also, in a conventional design in addition to the unwanted distortion and audio confusion, there is an energy dip at the HF/MF crossover point which results in further resonant distortion in the listening area. By being able to precisely time-align our drivers in all directions from the axis, Uni-Q technology completely eliminates both of those problems.

Because it is mounted at the center of the MF unit, the HF unit's directivity is also governed by the angle of the MF unit's cone – the HF directivity is the same as the MF directivity. This means that as the listener moves away from the main axis, the output of both the HF and the MF reduce at about the same rate, increasing the tonal balance and the stereo imaging throughout the listening area.

To quote the engineers who figured this all out: " From a listener's perspective, the combination of the matched directivity and precise time alignment in all directions gives significantly improved stereo imaging over a wide listening area, the realism of which is enhanced by the even balance of the reverberant energy within the listening room ."

The picture at right shows the time-aligned and concentric single point source Uni-Q used in our Blade speaker. Hit the link for more detailed information on our Uni-Q technology .

Jack Sharkey for KEF .

Jun 25 2014

Better HF Through KEF's Stiffened Dome Technology

Conventional wisdom has held for years that the optimal shape for a tweeter was a flat baffle with no waveguide or element between the listener and the baffle. Using Finite Element Analysis (FEA) KEF’s engineers have concluded that, like most conventional wisdom, the conventional wisdom in this case needs to be adjusted.

The graph on the left ( Figure 1 ) is the output of a short pulse a sent through a simple dome tweeter. A high-frequency transient signal is clearly shown in the area corresponding to the tweeter’s off-axis response. The graph on the right shows the output of a short pulse reproduced in a KEF tweeter with our Optimum Dome technology and Waveguide geometry and it clearly shows that the off-axis transient has been eliminated. This translates to the listener as a wider sound field (wider high-frequency dispersion) due to the tweeter's consistent off-axis response. This means that a listener is not chained to a “sweet-spot” but can hear clear and articulate high-frequency reproduction in a wider area in the listening room. For more detail on this please visit .

Having achieved an optimum shape for our tweeter drivers, KEF’s engineers were then tasked with turning their attention to the mechanical construction of our tweeters. Rigidity is the key to consistent, distortion-free response in a tweeter dome, but fast response is required to replicate musical tones and signals in the upper frequencies.

For example, certain percussive instruments like cymbals and bells can vibrate at upwards of 16,000 times per second (16kHz). The upper range of human hearing is generally considered to be 20kHz (after years of being around loud live music I now top off at around 14kHz).

This means that a tweeter must be able to accelerate from 0 cycles per second to 16,000 cycles per second (in the case of a small cymbal or chime); this is nearly equivalent to 1000G (1000 times the force of gravity). A fighter pilot, with a proper G-suit and other aids in the cockpit, can withstand up to 9G before blacking out. This is why aluminum is used in our tweeters: it is extremely light (low density) and stiff (high rigidity), so it is able to respond at the acceleration rate needed to replicate high frequency sounds while maintaining its shape. But even aluminum has limitations under the extreme stress of such high acceleration rates.

Elliptical & Spherical: The best of both worlds. Over the years, we have learned that there are in fact two optimal shapes for a tweeter dome:

  1. Elliptical
  2. Spherical

The elliptical shape resists the acceleration forces and remains rigid, offering a mechanical solution to the problem, while a spherical shape offers the best acoustical response. As in almost all things scientific, there is a trade-off (you can’t get something for nothing) and compromises have to be made.

The challenge for us was to produce a tweeter that didn’t compromise its acoustic capabilities across its sonic range.

In 2006, we combined both shapes as we prototyped our first Stiffened Dome geometry: A brilliantly simple (and patented) solution to the mechanical and acoustic challenges faced when trying to design a tweeter that is both acoustically and physically stable. The spherical shape is superimposed over the elliptical shape so that a triangle is formed where the two shapes meet along the edges of the dome, which is typically the weakest and most prone to losing its shape to due acceleration forces.

By combining the two shapes we have eliminated deformity in the tweeter dome at frequencies up to 38kHz while maintaining the pure high-frequency response that KEF speakers are renowned for.

While others may claim to be developing this technology for the first time, we have been perfecting our Stiffened Dome technology for nearly a decade, and we're pretty pleased with the results.