2 Way Loudspeaker SystemThis loudspeaker system is based on loudspeaker units from Vifa.
Loudspeaker System features:
The choice of loudspeakersI am a big fan of danish loudspeaker drivers. Vifa, Dynaudio and Scanspeak loudspeakers are in my opinion characterized by a very clean sound. Unfortunatly, Dynaudio drivers are no longer available for the amateur diy people, and Scanspeak are quite expensive, so these were left out.
The choice fell on Vifa PL18WO09-08 which I found a test on performed by Linkwitz. The test was performed using multitone test signals. I found that it was good for implementation in a 2 way system due to the wide frequency range. The THD measures are quite good although it was not the best in the test, but considering the test participants which are all high priced drivers, I think it did pretty good.
The EnclosureIt was decided to use the 4-sided pyramid design for this loudspeaker enclousure aswell as for the three way project, because it minimizes the amount of standing waves due to the few amount of parallel surfaces. The top and bottom of the enclosure are the only surfaces where room modes may appear, but in order to avoid this possible deterioration, the enclosure was filled with dense rockwool. PL18 has an acoustic equivalent volume of 24 liters and therefore, an inner cavity of approximately this volume will give a critical damping. A volume of 27 liters (+ the virtually increased volume due to the rockwool will overdampen the acoustical system slightly). Below, you can see plots of the mathematical models that I have used for simulating the response of the system, in the on-axis position.
In order to meet the demand about the enclosure volume of 27 liters, I used the enclosure illustrated on the
Schematic drawing of the enclosure.
A decent apperance was nescessary because the loudspeaker systems were to be located in a living room and therefore the enclosure was made by a professional carpenter. The total price was 1500 dkr, which I think was a very fair price given the amount of work. (angled corners, assembly, milled down holes for the loudspeakers and the spray painting job and including all materials)
The final result!.
The Filter NetworkFirst, the third order Butterworth filter was implemented. It consisted of an RLC resonator for smoothing the bass resonance frequency region, a Zobel was implemented for both the bass and the tweeter. Besides attenuation of about 3 dBV was included in the tweeter network. The designed/implemented network is illustrated below.
Filter diagram of the third order parallel filter network.
This filter network was modelled using Matlab and the it proved to be beneficial to reverse the polarity of the tweeter. The electrical and acoustical response is modelled below.
Acoustical response of parallel filter.
Electrical response of the parallel filter.
After having listened to this design for some time, we agreed that the sound was very nice and had very detailed. However, the sound was a little thin for certain genres of music; simply because the tweeter was too loud relative to the bass. It is not apparent on the above simulated responses, but in that respect, notice that these responses are based on relatively simple models, without baffle considerations.
Furthermore, the models are based on perfectly time-aligned arrangement of the tweeter and bass driver, which is ofcourse only reality at a certain position in front of the speakers. In the meantime, I had experimented with serial networks and learned that they are very good for certain designs, involving low Q loudspeaker drivers.
Below, you can see the series network with impedance smoothing circuitry and the simulated acoustical and electrical response.
Electrical diagram of the serial network.
Electrical response of the serial network.
Acoustical response of the serial network.
It's not a baffle step compensation based on the dimensions of the baffle, but one thing is for sure: the serial network sounds better and requires fewer components. The downside to the solution is that the out of passband attenuation is smaller, whereby the THD is probably larger than the parallel network solution.