top of page

Seemless crossover transition

   Crossovers are now set at 170/800/3000 Hz. They don't introduce any phase shift or frequency change in horizontal plane. That's because drivers are linearised perfectly two octaves below and above transition frequencies before the crossover kicks in. So the sum is also perfectly linear for the drivers which are linear in phase and frequency. This can be realised only with FIR filters. Filters are 48 dB/Octave. And like everything this is also for a reason. Shallower filters means that the region around crossover frequency is wider where both drivers outputs same frequency with different amplitudes. That brakes the whole concept of acoustical point sources working around its first dipole peak. The minimum baffle sizes are designed for specific frequency ranges so the drivers can't work in very broad range.

   Steeper FIR filters than 48 dB/Octave have very long pre and post ringing. Of course it is only for the vertical offset angles from listening height around 1.10 m. In horizontal plane there is no such effect. So the reflections from the ceiling and the floor can bring delayed attenuated pre-ringing artefacts. How much it is audible? Hard to say, but when i test system with 24/48/96 dB/Octave crossovers and switch between them there is a difference in sound. Going from 24 to 48 sound improves, but going from 48 to 96 is hard to say, it is just different. And that is impossible to tell does pre/post ringing or other factors kicks in here. Because not only ringing changes when i change the slope. So i chose to stay safely at 48dB/Octave where at least ringing should not be a problem. And it seems that 48 dB/Octave for a 170Hz is a maximum i can go according to this master's thesis http://lib.tkk.fi/Dipl/2008/urn011933.pdf. Quick note that in mentioned paper strangely they call taps number - the order of the filter. So you have to recalculate it to dB/Octave. Also to do that you have to know sampling rate which is not mentioned. But i have calculated from the graphs that is 44100Hz. It is not completely scientific approach paper, but it is good enough to look at. So actually ringing is not so audible as many imagine and here is another interesting article about that http://www.bodziosoftware.com.au/Pre_Post_Ringing_IR_And_Pulses.pdf

bottom of page