Driver Selection

  The Dayton-RS225 woofer was used. It's a nice 8 ohm 8 inch shielded driver. It has very low distortion and a fairly flat frequency response aside from the break-up peak. It looks cool too.

For midrange, a Dayton-RS52 dome midrange was used. Midrange domes tend to have a bad reputation and are usually avoided, but when ZaphAudio's testing on this indicated incredibly low distortion and flat frequency response (coupled with the low cost of $30), I was sold. The RS52 is a two inch aluminum dome midrange with neodymium magnet structure and an aluminum hexagrid cover to protect the dome. This looks pretty cool too.

The Roy Reference's tweeter is commonly found in many high end speakers. A Vifa XT19 (a ring radiator tweeter) was used. This tweeter is interesing in that the dome is secured at two points: the outer edge and an inner phase plug. It was chosen due to its low distortion, as indicated on Mark K's speaker testing page. It has relatively high amounts of second order harmonic distortion, perhaps due to nonlinearities between the two attachment points; however, all other types of distortion are incredibly low. This tweeter is used in speakers costing upwards of $10 000 per speaker; its only issue is its dispersion is rather narrow, even for a 19mm dome. The reason for the three-way system was to widen the dispersion from what I would get if I got a Vifa XT25. This is the bigger brother of the XT19. The performance is very similar except, the XT19 has better off-axis performance. Furthermore it has a small face plate which helps in getting a closer center to center spacing with the midrange dome. 

Crossover Design

A second order electrical filter was used for the woofer crossover and third order filters were used for the midrange and tweeters. This was done to help compensate for the offset between the woofer's acoustic center and that of the midrange and the tweeter.

Woofer Crossover

A second order electrical filter was used. Using the -3dB definition for the crossover point, it is 450 Hz. This was done to optimally mate with the midrange.

Midrange Crossover

The midrange is certainly a very complex crossover in the speaker. It is a 3rd order bandpass filter from 450 Hz to 3500Hz. The 8 ohm resistor is used to attenuate the midrange to match its sound level to the woofer. Most speakers employ an L-Pad network or a series resistor after the crossover, but it was not desired to need to change the crossover values every time I wanted a new attenuation setting; therefore I placed this resistor before the crossover. The disadvantage to this is that the resistor must disappate the power from all frequencies instead of just the treble. I used Mills resistors so hopefully they should stand up to the abuse. I do wonder how this would affect the sound when I turn the speakers up very loud. I'm thinking it will attenuate the treble/midrange as the temperature causes the resistance to rise. All the capacitors are film capacitors; there is a zobel at the end of the crossover, which helps lower the impedance rise at higher frequences.

Tweeter Crossover

The tweeter crossover is also a third order filter. The Vifa XT19 has increasing distortion at lower levels below 2.5KHz, which at that point rise very sharply. As a result, a 2.8KHz crossover point was selected. Perhaps it should have been made higher, but this couldn't be done as the distance between the acoustic centers of the midrange and the tweeter would become greater than one wavelength of the crossover frequency; this can have unintended consequences, such as lobeing. This crossover has a 2.5 ohm resistor for attenuation and a zobel network, as well as a series LCR filter. The zobel prevents the rise in high frequences, and the LCR flattens the impedance peak at resonance to keep the crossover functioning properly and to reduce distortion.

Conclusion and Measurements

The graph to the left is a measured frequency response graph, though done with a poor quality microphone. Everything below 200Hz appears to be an anomaly.

The graph to the left is a simulated frequency response. There is a battle step compensation, which is why there is a rise in frequencies below 1KHz. This evens out in a room when speakers are placed 18 inches away from the back walls. Otherwise, the speakers are very flat.

Overall, these speakers sound great, and the results are rewarding. The treble is among the best, as is the midrange, while the bass is very tight and clean, being augmented by a subwoofer. If anyone builds this speaker or uses parts from it, please contact me and tell how it worked and how it sounds.