Table of Contents
Motivation
I am a serious audio enthusiast. I have invested in endless hours in to the pursuit of the ‘perfect’ audio system. A good system requires constant tweaking, audition of equipment, measurements, and most importantly listening tests. If you want a taste of what my system of capable visit my other channel:
https://www.youtube.com/channel/UCDc4YhefZPk6IwYizpWqYRA.
There are many videos where I demo different turntable configurations with high quality analog recordings.
One of the aspects that makes for a great audio system is utilizing the right technologies for doing the right thing. This is especially true with transducers where a signal represented in one form of energy is transformed into another form of energy. This obvious example is a loudspeaker where an electrical signal containing variations in voltage level is converted into an acoustical signal containing variations in air pressure level. In an ideal world, the conversion would be perfect and every nuance of the original electrical waveform would be transferred into the air. Unfortunately, that is far from reality. Conventional loudspeaker drivers contain a voice coil, seated around a permanent magnet. When current passes through the voice coil a magnetic field is produced which interacts with the magnetic field of the permanent magnet and thus causes the diaphragm to vibrate. This produces the sound.
This has several disadvantages. The high mass of the driver means the system is slow to response to transients and any overhang due to its inertia results in distortion of the original signal. The voice coil forms an inductor and it means the load is difficult to drive due to its reactive nature. A loudspeaker of this type projects sounds in (more or less) spherical wavefronts, so the sound intensity falls of in accordance with the inverse square law. This may or may not be a disadvantage depending the on the use case. In a serious listening environment, this is a disadvantage as amplifier power is wasted on sound that is not reaching the listener’s ears and causing spurious acoustic reflections around the room.
To solve this, a driver with a planar form factor can be used. Planer drivers are rare and haven’t really been implemented in many commercial products aside from boutique audiophile speakers and headphones. I hope to change that by showing they can be constructed at home. Drivers are inheritedly low mass, have little to no inductance and as the name suggests, project sound in plane wave fronts. This addresses all the shortcoming of a conventional loudspeaker with only introducing a few small disadvantages.
Broadly speaking, there are two kinds of planar designs: magnetostatic and electrostatic drivers. I won’t bore you with the messy details but what I will say is that magnetostatic drivers have several advantages over their electrostatic counterparts. I chose to build a magnetostatic type for this project.
Driver Design
Planar magnetic drivers consist of a thin diaphragm tensioned and suspended over an array of magnets. Conductive material is attached to the diaphragm in such a way that when a current is passed through it, the conductor will move the presence of magnetic field and in turn move the diaphragm.
The planar design works on the principle of the lorentz force (F = IxB*L) and thus a conductor which ‘snakes’ across the diagram must use an array of magnets. These magnets alternate in polarity such that the current in the conductor, which generates the force, all work in the same direction.
After I worked out the basic design, I had to consider the performance specifications I wanted to meet. Typically, bookshelf speakers have a frequency response of 100Hz – 20kHz (±3dB) and can reach a maximum output of level of 105dB. I thought these were good targets to aim for and provided a guide to determine the details of the driver design. I settled on an overall speaker size of 12″ by 12″. Knowing the active area of the driver and target specification gave me enough information to flesh out the rest of the design.
The first consideration was maximum driver excursion as this determines how loud a speaker can play before the diaphragm starts hitting the magnetic array and distorting. The maximum excursion or Xmax for short, occurs at the low frequency and at the maximum desired output level. Plugging those numbers into an online calculator yields about 1 mm.
This means a gap of at least 1 mm will be needed between the diaphragm and the magnet array for proper operation. However, there is trade off to make here because as the conductor moves further away from the magnetic field, the speaker is less efficient but the extra space allows for low frequencies at higher amplitudes. For this first prototype I felt 1 mm was a good comprise as the magnetic fields are unusably weak beyond about 2 mm.
The entire frequency range is difficult to reproduce in a single driver, even for a low mass planar magnetic driver, so often speaker manufacturers will split up the frequency spectrum over multiple drivers. About 75% of the driver area is dedicated to bass/midrange frequencies (100 Hz~1000 Hz) and 25% to upper midrange and high frequencies (1000 Hz to 20000 Hz). The result is woofer/tweeter combination on a single diaphragm. The frequencies have longer wavelengths and are better suited to wires that are farther apather and vice versa. The difference in area is due to increased area requirement to maintain the same sound pressure level at lower frequencies. As a result, tweeters tend to be far more efficient than woofers. This is addressed later.
Driver Construction
Before any of the electronics can be designed, the driver mustbe constructed and characterized. The first step was building the magnet array. The magnets are the rubber type used to seal fridge doors. These magnets are cut to length and epoxied onto a piece of perforated steel. Perforated steel was chosen to allow the pressure between the diagram and the magnet array to be relieved. The 5mm wide strips are for the woofer section and the 3mm wide strips are for the tweeter section.
After completing the entire magnet array, a corrugated plastic spacer was adhered to the steel. It will support diaphragm and provide the required gap between the mylar and the magnets.
A piece of 12 micron thick mylar was stretched over the entire driver. The tension of the mylar was verified by measuring its resonant frequency. Typically, any sort of driver with a suspension or membrane will have a natural resonant frequency that will act as a mechanical high pass filter with a cutoff frequency close to that of the resonant frequency. In this case, a resonant frequency of around 100 Hz is desired and determines the correct amount of tension.
The most difficult step of the project proved to be the gluing of necessary wires to the diaphragm. The wire used is enameled aluminum wire in both 0.315 mm and 0.172 mm thicknesses. Aluminum was chosen due to its low weight and relatively high electrical conductivity. The thicker green wire is used for the woofer to allow high current handling and the thinner yellow wire is used for the tweeter to minimize the amount of weight. The woofer required three passes of wire in order to hit ~4Ω and the tweeter required two passes to hit ~3.5Ω. 4Ω is required to prevent too much current reaching the speaker. The adhesive is a latex-based epoxy that remains pliable after curing. Shown below is one of the unsuccessful attempts at the gluing the wires.
Crossover
Now that the driver was completed, the next part of the project was creating a passive analog filter called a crossover. A crossover performs two very important tasks in a loudspeaker. The first is splitting up signal coming from the amplifier into frequency bands corresponding to the woofer and the tweeter. The second is providing a small amount of equalization and level-matching so the speaker can maintain a smooth frequency response. The driver’s frequency response was measured using a UMIK-1 calibrated measurement microphone and a piece of software called RoomEQWizard (REW.) The woofer and the tweeter were measured separately. The woofer’s frequency response is in red and the tweeter’s response is in blue.
Every crossover circuit has a crossover frequency wherein the woofer hands off the rest of the frequency spectrum to the tweeter. Ideally, the crossover frequency should lie in a region where both drivers are working optimally so the transition between them is less obvious to the ear. The graph shows below demonstrates that the crossover frequency should be somewhere between 800 Hz and 2000 Hz. After simulating this transition in VituixCAD, the flattest frequency response was found to be achieved at a crossover frequency of 900 Hz.
In essence, a simple crossover is a low pass filter whose output is connected to the woofer and a high pass filter whose output is connected to the tweeter. Put together, the speaker is able to output the full frequency range it was designed for. Combining the responses for the woofer/tweeter using this crossover design yields a smooth frequency response. The graph shown is the output from the simulation and was verified with real-world measurements later.
A circuit diagram was then sketched based on the component values given by VituixCAD. A 3rd order Bessel filter was chosen as the phase response because it is very linear, thus preserving the waveform of the original electrical signal. The Bessel diagram is shown below.
The circuit was soldered together on perfboard with large gold-plated taps serving as connection points for the amplifier, woofer, and tweeter. The inductors were placed at right angles to each other in order to minimize interference.
With the crossover finished, the entire speaker was now complete. A final frequency response measurement confirmed that the simulation matched up very closely with actual speaker performance.
Listening test
Compared to a conventional speaker, listening to the planar magnetic speaker yielded a drastically different audio experience. The first thing that is immediately apparent is the highly directional nature of the sound, especially in the high frequencies but obvious down into the bass frequencies as well. Directly facing the speaker, the sound is loud, clear, and crisp, but listening even a few degrees off-axis causes the perceived loudness to drop off. It is akin to a sonic laser beam rather than the conventional sound of a highly diffused wavefront. The music itself is rendered very precisely and it makes a conventional speaker sound sluggish in comparison. The only drawback was the apparent lack of deep bass and impact, but not much can be expected from a speaker of this size. Below is a sound demo to illustrate the speaker’s sound profile.
Future improvements
To summarize, the final measured specifications of the speaker are as follows:
- Frequency response: 200 Hz – 14 kHz (±3dB)
- Maximum output: 105dB @ 200 Hz
- Crossover frequency: 900Hz
Unfortunately, this falls short of the target response on both ends. The 200 Hz frequency response is likely due to a resonance between the overall speaker housing as some strange peaking occurs around 150 Hz. However, this is not to discount its output below 150 Hz; tones are audible through to 80 Hz. The tweeter falls short of 20 kHz, likely due to the high mass of glue used to attach the wires and a lack of field strength from the magnet array. However, this does not prevent the speaker from generating excellent sound within its range capacity.
In the future, I would like to investigate switching to a DSP designed crossover, integrated a subwoofer for improved bass extension, and a ribbon tweeter for improved high-frequency performance.