The name “tweeter” is derived from the high pitched sounds made by some small birds. The terms is contrasted to the low “woofs” made by bigger dogs (woofers).

Sounds whose wavelength is smaller than 17 cm (above 2 Hz) are important inter alia for the definition of the tone color of various musical instruments. Speaker units that can produce such high frequency sounds, ie. tweeters, have existed since the 1920s but for a long time there was no real need for tweeters – in fact, many listeners preferred a limited frequency band. The true need for tweeters came later along with LPs and FM radio, argues the head designer of Gradient speakers, Jorma Salmi, in the below article. Having a tweeter in the 1950s was like having a subwoofer or super-tweeter today. In this article Salmi deals with the history of tweeters, the theory HF reproduction, the pros and cons of the most common operating principles: cone tweeters, dome tweeters, ribbon tweeters, AMT tweeters, electrostatic tweeters, ionic tweeters, piezzo tweeters, and horn tweeters.

What then is the best tweeter type?
In Salmi’s opinion there is no unequivocal answer to this question. A speaker can be excellent irrespectively of which tweeter type have been used. What is clear is that however good and expensive the tweeter is it does not warrant a good speaker. The midrange still remains crucial for the speaker’s performance. The tweeter must fit into the whole. Since none of the operating principles is transcendent, maybe the ideal one has not yet been invented. In the meantime one can always take an old principle and try to improve it with the modern techniques. Are modern cone tweeters the next hit?

The whole article is below.

The word “tweeter”, which refers to the treble element of speakers, describes chirping of a small bird. But what is the history, theory and operating principles of the tweeter in speakers?

The tweeter range usually refers to the upper decade of the audio frequency range, i.e. frequencies above 2kHz. The wavelength is therefore 17cm or less. This is good to keep in mind, because in acoustics all dimensions are related to the wavelength.

The fundamental frequencies of few musical instruments reach the treble range defined above, but the harmonics of the fundamental frequencies do. The initial tones of musical instruments also contain high frequencies, which, along with the distribution of harmonics, help to identify the instrument. Therefore, the reproduction of treble frequencies is necessary.

A bit of history

It has been possible to make a speaker (or speaker element) capable of reproducing treble frequencies since at least the 1920s. There was just no direct need for it for a long time.

Harry F. Olson reported in his book Elements of Acoustical Engineering (van Nostrand 1947) two listening experiments: “A six-piece ‘dance band’ was playing in a large living room, of course with acoustic instruments at that time. Between the orchestra and the listeners was an acoustic low-pass filter made of perforated plates, which could be switched off by turning the plates 90 degrees with the help of levers. The cut-off frequency of the filter was 5 kHz. There was a curtain between the perforated plates and the listeners – so it was a blind test. The filter was switched on and off every 30 seconds. The listeners were asked which they liked better.”

The second experiment tested the audio equipment of the time (Chinn and Eisenberg 1945). The program sources were a cable connection directly from the studio and a turntable. The playback range could be selected from 150 Hz-4 kHz, 80 Hz-6 kHz or 40 Hz-10 kHz. The listeners had to choose their favorite of these.

The results of the experiments were not surprising in themselves. With live music, full bandwidth was invariably considered better, but with hi-fi equipment the result was the opposite. The narrowest or at most middle option was considered the best. Olson thought that this was due to two reasons: people were so used to a limited bandwidth that this was how it “should” be. A wider frequency band revealed distortion and noise from the playback equipment, which was felt to be unpleasant.

When and why did the reproduction of the treble range become necessary?
Reproducing the full audio bandwidth became worth pursuing with the emergence of LP record and FM radio in the 1950s. The treble end of a LP record already contained more than just noise and crackle. It was a great leap forward compared to 78 discs.

The technology and advantages of FM radio transmissions were already known in the 1930s, but transmissions did not really begin until after World War II. Germany was a pioneer in this. The allied countries had appropriated Germany’s old medium and long wave broadcasting frequencies, and so the Germans had to look for new frequencies. This meant switching to ultra-short waves and at the same time using frequency modulation (FM). The FM broadcasting standard set the upper limit of the audio band at 15kHz, which was “absolutely sufficient for music reproduction”. The difference was considerable compared to 4.5kHz for AM radio (5kHz in the USA, where the station spacing was 10kHz).

As early as the 1950s, better German table radios began to have separate tweeters, even electrostatic ones (in tube radios, sufficient polarization voltage was already available and the audio signal was taken before the output transformer). In the rest of the world, FM radio broadcasts progressed at a different pace, with Finland starting broadcasting in 1953, the United States only really in the 1960s, and Australia after 15 years. At that time, of course, ready-made multi-way loudspeakers were already available – including quite good ones – but hobbyists were more independent than today.

Dynamic principle

Almost all tweeters follow the dynamic operation principle.

In 1925, in their article “Notes on the Development of a New Type of Hornless Loudspeaker”, Messrs. Chester W. Rice and Edward W. Kellogg described the dynamic loudspeaker element as it is still known today. It had a fixed magnetic circuit with a voice coil in the air gap. The sound-producing membrane was light, rigid and conical in shape. The voice coil was attached to the base of the cone. The cone was suspended flexibly from both the bottom and the upper edge. The size of the cone was small relative to the wavelength at the frequencies at which it was primarily intended to be used. The lowest reproducible frequency corresponded to the resonant frequency of the element, so it operated in a mass-controlled region. Rice and Kellogg also mentioned that the element is good to mount either on a faceplate or in a housing to avoid acoustic short-circuiting at low frequencies, and that at higher frequencies, where the cone is no longer small relative to the wavelength, it begins to direct the sound.

Briefly on the underlying theory

The force exerted on the cone of a dynamic loudspeaker element is F=Bli, where B is the magnetic flux density in the air gap, l=the length of the voice coil wire in the air gap, i=the current flowing through the voice coil. The resonant frequency fs of the loudspeaker element is determined by the elasticity of the suspension and the mass of the moving parts.

Let us assume that the current I in the voice coil is constant regardless of the frequency. Then the force F acting on the cone is also constant. Below the resonant frequency fs, the movement of the cone is resisted by its suspension (spring). We say that we are then operating in a flexibility-controlled region. A constant force is used to achieve a constant displacement (for example, the operation of a spring balance is based on this). The displacement is therefore constant regardless of the frequency below fs.

Above the resonant frequency, the inertia of the mass of the moving parts becomes a determining factor. Then we are operating in a mass-controlled region. From the basics of physics, we remember that F=ma. Since the mass m of the moving parts is constant – as was agreed, as well as F – the acceleration a must also be constant regardless of the frequency. This means that above fs, the displacement of the cone decreases by a quarter when the frequency doubles.

A cone that is small in size relative to its wavelength has poor impedance matching with air. But as the frequency increases, the wavelength decreases and the cone, as it were, increases in size (relative to the wavelength, of course). Thus, the resistive part of its radiation impedance also increases as a function of frequency. This means that in a mass-controlled region with constant current, the sound power radiated by the speaker element is constant regardless of frequency. This is an important observation that Rice and Kellogg also realized at the time.

However, the radiation resistance of the cone does not increase indefinitely with frequency. Namely, above the frequency (f1) at which the wavelength is approximately 1.5 times the diameter of the cone, the radiation resistance becomes constant. The sound power radiated by the element begins to decrease. At the same time, the cone begins to direct the sound and, when viewed directly from the axis, the sound pressure it produces does not decrease.

In summary: (i) below the frequency fs the frequency response decreases by 12dB/octave, (ii) between the frequencies fs and f1 the response is flat, (iii) above the frequency f1 the power response decreases by 12dB/octave and (iv) the free-field response on the axis does not decrease but even increases.

So that’s the theory. In practice, things are a little different.

In practice

The cone of the driver is not infinitely stiff. At higher frequencies, only the middle part of the cone vibrates and its effective diameter thus decreases. The voice coil is not a pure resistance, but also has an inductive component. The speaker is controlled by a voltage source instead of a power / current???? source and thus at high frequencies the inductance of the voice coil reduces the current. By modifying the properties of the cone, it is possible to change the frequency f1 to a place where the cone size does not match. If only part of the cone vibrates, its effective mass is smaller. This sets a new parameter for the operation of the mass-controlled region.

The inductance of the voice coil can be used to equalize the upper end of the frequency range.

These methods have been used in widebandwidth elements, when a single speaker element is intended to cover most of the audio frequency range. However, there are problems with the effective size of the cone decreasing as the frequency increases. In practice, this does not happen smoothly.

If a flat free-field response and a power response that decreases nicely towards high frequencies are considered to be the characteristics of an ideal speaker, an ideal speaker cannot be realized with a single speaker element. For this reason, speaker elements of different sizes are needed and multi-way speakers have been used.

Based on the above, it can be stated that by changing the dimensions of the driver, it can be scaled to operate in the desired frequency range: f1 (and also fs) are moved to the desired location.

Cone tweeters

The first widely used tweeters were cone tweeters. The body was closed at the back, so it was a “closed-box speaker”. The cone diameter was typically 35-70 mm and the voice coil diameter 10-12mm. The resonant frequency was usually around 1kHz. The quality of such an element is determined almost entirely by the behavior of the cone. Some cone tweeters were equipped with a “mechanical crossover”: the voice coil was glued to the base of the cone with a suitably elastic adhesive and the dust cup with hard adhesive directly to the end of the voice coil body. Thus, at the lower end of the treble range, the entire cone vibrates, as the frequency increases, the middle of the cone and at the highest frequencies only the dust cup. At the upper end of the treble range, it was actually a 12mm dome tweeter! No adhesive is infinitely hard (stiff) and as the frequency increases further, only the voice coil vibrates and the dust cup no longer produces sound. The upper limit frequency of the element has been reached.

Since the voice coil of such an element is small and light, its thermal power handling is low, only a few watts. That is why a high-pass filter is always needed.

The design of a cone tweeter seems reasonable, but with the production technology of the 50s, the dispersion was large. The best specimens were even excellent.

The cone material used was paper pulp, but not always. Audax’s TW8B has achieved a cult status among enthusiasts. It had an exponential cone made of 0.05 mm thick aluminum, with a diameter of 5cm. The mass of the moving parts was 0.55g. The upper limit frequency was stated to be 40 kHz. It was therefore a super tweeter of its time, usable from about 4 kHz upwards. The free-field response was quite good when viewed directly from the axis, but quite uneven at the corners. Due to its low internal damping, the aluminum cone cannot be made to “shrink” nicely as the frequency increases. The TW8B was in production from the early 1960s until the 1980s.

Dome tweeters

In a dome tweeter, the sound-producing membrane is a dome glued to the end of the voice coil. Its shape is intended solely for structural rigidity. The dome diameter is usually the same as the voice coil. The suspension is a narrow edge fold around the voice coil attachment point, often made of the same material as the dome itself.

Dome tweeters began to become popular in the 1960s and today they are by far the most common tweeter type. There are very good reasons for this. The dome tweeter is simple in design and has only few parts. It is easy and cheap to manufacture. The quality is consistent.

When the dome tweeter came onto the market, sales pitches praised its superiority over its cone-equipped competitor. Its wide radiation pattern was particularly praised: “You can see it with your own eyes, when the dome bulges out from the front plate.” However, only one side of this is true. The 25mm dome is indeed small compared to most cone tweeters, so in a certain area in the lower treble it radiates more widely (whether that is desirable or not is a completely different matter).

In the case of a 25mm dome, f1 is about 9kHz, while for a 50mm cone, f1 is 4.5kHz. But since the diameter of the voice coil is the same as the dome, the effective diameter of the source does not decrease with increasing frequency. This is why the power response drops off rapidly above f1. The most problematic thing, however, is that below f1 the power response is almost horizontal. At frequencies above 10kHz, the radiation pattern of a 25mm dome is narrower than that of a good cone tweeter.

The directional properties of a dome tweeter can be improved by using a horn adapter, or a deflector (waveguide). At the highest frequencies, the beam cannot be made wider, but at low frequencies it can be narrowed, and the end result is thus more balanced and easier to match with the driver reproducing midrange sounds.

The dome of dome tweeters is made of several different materials. There is an eternal debate about their mutual superiority and there seems to be no clear answer to this. By changing the material, some individual property can be improved, but it is done at the expense of another. This way, compromises are only made in another way.

The voice coil of a dome tweeter is also small and cannot withstand high current. The wire diameter is typically around 0.1mm. The voice coil can withstand only a few watts of power and its thermal time constant is around a second. If the coil is wound from copper wire, at a temperature of 100 degrees its resistance has increased by a third. At the same voltage, the voice coil current is lower. This is called thermal compression. One way to cool the voice coil is to use a magnetic fluid (ferrofluid). The air gap is filled with a magnetic fluid, which conducts heat much better than air. The ferrofluid thus transfers heat efficiently from the voice coil to the pole plates of the magnetic circuit. With the help of ferrofluid, the thermal time constant of the voice coil can be increased threefold.

One of the first dome tweeters on the market was Celestion’s HF1300, of which there were many different versions from the 1950s onwards. Unlike its modern relatives, the diaphragm was not a dome but a shallow cone with the tip pointing outwards. The diameter of the voice coil was 19mm and the rim was quite wide, the total diameter of the moving parts was 38mm. The diaphragm was rigid, made of phenolic resin-impregnated textile. The upper end of the element was modified by a diffuser. The practical upper limit frequency was 14kHz and a supertweeter was sometimes used with it.

Ribbon tweeters

E. Gerlach developed the ribbon microphone in Germany in 1924. The Englishman Stanley Kelly was one of the first to “translate” it into a loudspeaker and develop it into a commercial product about fifty years ago. Production was started by Decca Special Products.

The ribbon loudspeaker diaphragm is a ribbon made of conductive material placed in the air gap of the magnetic circuit, which therefore performs the function of a voice coil. The ribbon loudspeaker is the only one, besides the electrostatic loudspeaker, in which every point of the sound-producing diaphragm is even attempted to be driven with the same force and in the same phase. The effective length and width of the ribbon are therefore constant regardless of frequency.

The ribbon of the Decca/Kelly “London” tweeter was made of aluminum, 55mm long and 8.5mm wide. The thickness was 0.01mm and the mass was only 4.65mg. The resistance of the ribbon was 0.02 ohms, so it could not be driven directly by an amplifier. Sometimes a transformer was needed. The “London” was equipped with a horn, but a version without a horn was also available, which was used as a super tweeter.

A few observations can be made about the numbers listed above. In order for an 8.5mm wide ribbon to fit into the air gap, the gap must be at least 9mm wide. This makes the magnetic circuit quite disadvantageous. 97% of the magnetic circuit of a ribbon tweeter is wasted, only about three percent of the total flux can be directed to the air gap.

Another obvious problem is the electrical connections of the ends of the aluminum ribbon to the secondary winding of the transformer. The contact resistance must be much lower than the resistance of the ribbon itself. The “London” ribbon tweeter could be used from 1kHz upwards. The upper limit frequency was around 30kHz.

The directional characteristics of a ribbon tweeter equipped with a horn are largely determined by the geometry of the horn. Without a horn, they come directly from the dimensions of the ribbon. The ribbon is usually mounted vertically, which results in a wide radiation pattern in the horizontal plane. In the vertical plane, it is significantly narrower and a change in listening height clearly affects the audibility of the upper trebles.

Ribbon tweeters have made a new appearance in recent years. Neodymium magnetic material has made it possible to build a previously large magnetic circuit in a significantly smaller space and at a lower cost.

Air Motion Transformer

In the 1970s, Dr. Oskar Heil developed a tweeter element resembling a ribbon tweeter in the United States. The diaphragm is made of plastic and has parallel conductors metallized into it at regular intervals. The diaphragm is then folded so that a conductor runs along the side of each pleat. The current flow is arranged so that its direction is opposite in adjacent pleats. The poles of the magnetic circuit come to the front and back of the diaphragm. The poles must therefore be acoustically transparent.

The operation is quite simple: during the first half-cycle of the current, the pleats on the front of the diaphragm narrow and pump air out. On the back, the pleats spread and suck air in. During the second half-cycle, the exact opposite happens. In relation to the movement of the conductors, more sound is extracted from the diaphragm in this way than by moving it back and forth.

The resistance of the “voice coil” is around 6 ohms, so no transformer is needed.

As in a ribbon tweeter, the diaphragm is usually narrow and long. For example, in the case of the ESS Amt 1B, the diaphragm is 3cm x 10.5cm. The directional patterns are therefore different in the horizontal and vertical planes and the effective size of the diaphragm does not change with frequency. Over the past two decades, this “accordion tweeter” was revived for the same reasons as the ribbon tweeter.

Other operating principles

In an electrostatic speaker (speaker element), an electrically charged diaphragm is moved between stator plates by an electric field. At least in principle, the same force is applied to every point of the diaphragm. The achievable displacement is very small, so the diaphragm surface area must be made large. This results in a narrow radiation beam, especially at the upper end of the frequency range. The radiation pattern can be modified by bending the diaphragm mechanically (e.g. Martin-Logan) or electrically (e.g. Quad ESL-63) or by dividing the diaphragm into parts (Quad ESL, 2.5-way solution). An electrostatic loudspeaker requires a polarization voltage between the transformer and the speaker to operate; the voltage from a standard audio amplifier is not sufficient to drive it. The principle has been known for a very long time. Arthur Janszen’s electrostatic tweeters were the first commercial products of this type in the 1950s.

Above, a charged diaphragm was moved by an electric field. No diaphragm is needed if air particles are moved directly by the in a field. The particles just need to be ionized first to make them obey. This is how an ion speaker (also known as a plasma speaker) works. This principle was already known in the late 19th century, but it was not until 1946 that the Frenchman Siegfried Klein introduced such a tweeter. The best-known commercial version was the English IonoFane. The maximum sound pressure level obtained from an ion tweeter is low, which is why it is always equipped with a horn.

A piezoelectric crystal or ceramic changes its shape when it is driven by a voltage. When a cone or horn is connected to such a device, a piezo tweeter is created. Tweeter elements that work in this way have traditionally been overlooked by hi-fi enthusiasts, but for example, the prestigious Dahlquist DQ-10 speaker featured a Motorola piezo tweeter horn.

Horn tweeters

When high sensitivity and/or high sound pressure levels are required, a horn tweeter is the best solution. The horn improves the impedance matching to the air and can increase the efficiency or at least the sensitivity of the actual sound source – it works on any principle – with the help of the horn. The radiation pattern can be tailored to the desired shape with the shape of the horn. Horn tweeters have a bad reputation in some circles. They talk about the unpleasant “horn sound”. This is probably because there have been too many unsuccessful practical implementations. For example, it is impossible to make a strongly directional tweeter horn work seamlessly together with a midrange that radiates significantly more widely. In that case, it would be better to use the horn also in the midrange range, or choose a horn so that it also handles most of the midrange.

So which is the best?

It is impossible to give a clear indisputable answer to the question. There are excellent speakers out there that use tweeters that operate on all of the above-mentioned principles. However, it is clear that a “good and expensive” tweeter – no matter what principle it works on – does not make a good speaker. The midrange is and remains as the most important in a speaker and the speaker either stands or falls on it. The tweeter must be suitable for the rest of the whole. Since no principle is superior to the others, it can also be thought that the right and final one has not yet been found. Instead, some forgotten structure is always dug up from the mothballs. This happens in periods of a few decades. Now, would it be time for a cone tweeter made with modern technology to appear again?

PS. The word tweeter comes from the word “discant” (discantus), which means a sound or melody above other sounds and melodies. The word “woofer” for the bass element is onomatopoeic, like tweeter, and describes the raucous barking of a dog. Wof,wof!