Normally crossovers define a separate circuit for each individual driver, called a parallel crossover. I find that a rather shortsighted approach. As in the end you’d like all drivers to cooperate as one and present you with a coherent soundstage. In my opinion the only way to accomplish that is to use a series cross-over, where each driver is also included in the cross-over network of the other(s). In my case the crossover has been designed and optimized with SpeakerWorkshop
. SpeakerWorkshop is a freeware package that allows you to design and simulate loudspeaker enclosure and filters, regardless of the configuration. It hence is more flexible than other software packages that can only offer standard cross-over configurations. The only downside of SpeakerWorkshop is the fact that it needs some getting used to and doesn’t really score points when it comes to being easy-to-use.
As an added advantage a series-crossover will give you steeper filter slopes than a parallel-crossover, a typical 1st order parallel-crossover will attenuate by -6dB whereas the series-crossover will be closer to -9dB. For a 2nd order crossover the actual filter slopes are closer to -15dB versus the -12dB for a parallel-crossover. I’ll be using a combination of both for my filter as that gave the best simulation results and also measured a lot cleaner. The tweeter will feature a 1st order and an l-pad to match the sensitivity of the woofers and flatten out its impedance curve. The woofers will get an extra inductor that will make the filter slope steeper and compensate for baffle step diffraction at the same time. In order to make sure the resonance peak at around 8.5kHz the woofers doesn’t play up I added a zobel network, impedance compensation, so the impedance curve doesn’t rise at higher frequencies. The crossover frequency is set at 2kHz, which turned out to be the best compromise, both from simulations and measurements (MLSSA and 3D waterfall).
The cross-over as built-up on a printed circuit board, all parts are glued to it to avoid vibrations.
The filter itself is constructed using 1.4-mm OFC inductors without a core, polypropylene capacitors and 4-watt metal oxide resistors. I fancy the Evox-Rifa PHE 450 MKP capacitors and make banks using smaller values to get to the value I need. The above photos illustrate this. The 15uf and 10uf capacitors are built using 4.7uf capacitors, the slight variation in capacity does not affect the filter curves much, I actually couldn’t get it to show in either one of the measurements. Resistors are rated at 4-watts and of the MOX, metal-oxide, variety. Unlike wound resistors, these are not made from resistive wire wound onto a round substrate, hence have no, or very little, self inductance.