In this section, a description of my first 3 way loudspeakers system follows. The system is based on a time-aligned loudspeaker arrangement and the filler driver principle presented by Bang & Olufsen. The document is split up into two subsections for improved readability. First a description of the filter is presented and then the documentation of the enclosure design follows.

Filter design

When designing a filter for a three way loudspeaker system, several issues must be considered. First of all the loudspeaker driver units should overlap each other by a sufficient amount of bandwidth, unless filters of high orders, that provide steep frequency rooloff, are used. In my opinion, it is nescessary that the loudspeakers have a constant sound pressure output in the desired frequency range at least until attenuated 12dB by the filter beyond the crossover frequency. As little as 1 dB SPL difference is noticeable in certain frequency ranges. Particularly in the presence area (1-5 kHz) the perception is very distingtive. On this basis I have chosen to use speaker units that overlap each other by minimum one octave. Choice of units/filter The units that I have used are : Vifa PL22WR0908 as a bass driver, Vifa PL11WG0908 as a midrange driver unit and the Scanspeak tweeter:D2905/95000. I don't have much experience with different loudspeaker unit and these were exclusively chosen on the basis of the datasheets. All of the drivers have relatively constant SPL outputs in the frequency bands where they should be used and the midrange and bass loudspeaker both have DC resistances of 5.9 Ohm. Choice of crossover frequencies The choice of crossover frequencies was made considering the qualities of the different driver units, combined with my knowledge on perception. During the different designs I tried out, I experienced that if the crossover frequency between the tweeter and midrange is too low, the tweeter will dominate the presence-area. This is undesireable because the dome-tweeter distorts relatively much at midrange frequencies due to the lower rolloff. I decided to use 2. order Butterworth filters.

A rule of thumb is that the midrange driver should represent the entire part of the audio band, and the tweeter and bass driver should only add detail to the sound image. This means that the midrange driver in an ideal case should represent the entire audio band, which is obviously not possible. The crossover frequency between the midrange driver and the tweeter was on this basis and the frequency response chosen to be 3400Hz. The crossover frequency between the midrange and bass driver, was chosen considering that the midrange driver works constanctly downwards to about 100Hz, and the bass driver represents the sound image upward linearly to about 1700Hz. By choosing a crossover frequency of 400Hz, problems with unlinearity around the crossover frequency because of the drivers, are prevented. This means that the bass driver represents the entire low frequency audio band, from 20-400Hz. However, the downside to this choice of relatively high crossover frequency is, that there will appear some distortion in the upper bass area, caused by the doppler effect. Impedance Compensation The use of components for the filter was considered carefully. As can be seen on the datasheet for the bass driver, this unit has an impedans of 8 Ohm nominally, and an ugly impedance peak around the resonance frequency, about 26Hz. Off course the amplitude of this peak is decreased by the use of a enclosure (which increases the mechanical loss), but still, it is a frequency region which is difficult to control by the amplifier. For impedance smoothing of the bass driver I decided to use a very large voice coil of 10 mH and a high voltage capacitor (250V AC). Besides that, a Zobel network was implemented to ensure that the impedance for the driver would remain relatively constant around the DC resistance as the frequence increases. The midrange speaker provides no particular load for the amplifier, although it rises to about 13 ohm around the upper crossover frequency. This problem was solved using the Zobel network aswell. The tweeter has a DC resistance of 4.7 bOhm, and differs from the others (5.9 Ohms). Since the tweeter has a higher sensitivity than both the midrange and the bass driver it requires attenuation and therefore a 1.2 Ohm resistor was simply connected in series with the tweeter. The tweeter has a sensitivity of 90dB/W/1m, which is 5dB more than the two other drivers. Therefore I additionally included an L-pad network to provide the proper attenuation of this driver. The L-pad also provides an impendance compensation, smoothing the impedance at the upper frequencies. To see the filter diagram click here

Enclosure design

During the work with different kinds of loudspeakers enclosures, the main tool for design of the optimal enclosure has been the simulator tool LSPCad Lite, which is a freeware tool from a swedish software company. It is a light version of a more advanced simulator, but it enables the user to try out different enclosure sizes for: vented systems, closed systems and bandpass systems. tool can be downloaded at http://www.occasionaudio.com.au/products.htm In my design of the enclosure I wanted a system that would work from around 30Hz to 20kHz, so at first I tried out different vented enclosures constructions, to make the system play deeep bass. I used the simulator tool, and designed a three-way system, that met these demands. The system I ended up with had a bass enclosure of around 60 liters in volume, the midrange enclosure was 10 liters in volume, and the tweeter was from birth mounted in a sealed enclosure. This combination of enclosures resulted in a system that played relatively constant from around 24Hz to 20kHz, but during the following listening tests, I was not pleased with the representation of bass/lower midrange. I thought that it was 'blurry' and unrefined, though deep. Instead I tried closing the bass port for the bass enclosure. This improved the transient response of the system, but then the bass was not deep enough (only around 50Hz constant according to the simulations). However, when the choice had to be made between the sealed and the closed enclosure the closed enclosure was definitely preferable.

The design of the enclosure shape has been considered regarding diffraction and standing waves. The shape of the bass and midrange enclosure is that of a pyramid. This means that in theory, there are no parallel sides in the enclosure, thereby minimizing the amount of standing waves in the enclosure. Hereby, the problem about axial modes (standing waves) between two surfaces is avoided. The problem with difraction is solved by placing the driver units asymmetrically on the baffle and rounding the baffle corners. Besides this, the acoustic center line of the speaker is placed in a horizontal, almost 90 degree angle (85 degrees) to ensure maximum uniformity and consistency of the sound image. This is done in an attempt to acoustically align the loudspeakers, providing the same arrival time at the listening position. The midrange and bass enclosure is made of 25mm MDF. The tweeter enclosure should only provide a fixed position of the loudspeaker, since it is mounted in a sealed enclosure. For filling material, I used regular sound absorbing loudspeaker materials(rockwool). The bass enclosure was filled around 50%, the midrange enclosure around 30%. An interesting feature of the tweeter enclosure is that the radiation axis of the tweeter can be adjusted, simply by turning the entire tweeter enclosure towards the listener. This proved to be useful during integration with the listening room.