Originally Posted by mhutchins
But why should that differ between on-axis and off-axis?
Part of the choice for a crossover point is to find not only a good blend between drivers but to avoid where the drivers start to misbehave at the extremes of their frequency response. If you have several octaves of good behavior / smooth frequency response, certainly a shallow crossover can work. But what if you don't have that luxury? Does the off-axis response tend to become more ragged sooner than the on-axis response for most drivers?
Is part of the issue related to the phase response of higher-order crossovers?
Crossover design, like most technical design is an exercise in compromises. Even choosing the drivers is a part of the overall design. In the simplest sense, a loudspeaker filter is used to limit each driver to its appropriate on-axis
usable range. More refined designs, account for the off-axis response and steeper filter slopes can negatively (and sometimes positively) affect the off-axis response. Experience and an understanding the fundamental concepts of speaker design helps to come up with a refined design. This will become more apparent when Scott is able to post some complete measurements and we can post some design examples.
Focusing all design energy, especially at the crossover design phase, on attenuating breakup is misguided. First, choose drivers that can be operated more easily within their limits. Driver selection is part of the design process. Second, you are generally presented with the following design choice:
1. Breakup that is down by 30db on-axis and a smooth off-axis response at crossover with a shallow slope
2. Breakup that is down by 60db on-axis and an abrupt off-axis response at crossover with a steeper slope
My research and experience always push me towards the first choice. I view that as the fundamental crossover design quandry at the crossover stage. Of course, the best way to make that design quandry manageable is to choose drivers, horns, physical layout, baffle shape, etc that allow for positive compromises. Scott could have chosen to go straight from the 4 2226's to the SEOS12 and that would have necessitated steep filters to deal with null issues, but the speaker would be a mess. Steep filters tend to be a blunt tool used to cover up poor choices in the physical design phase. Shallow slopes are a luxury that is sometimes afforded by good choices in driver selection. Obviously there is a balance to be had like with any design.
Also of note, there is sometimes confusion between the electrical and acoustic response slopes. The electrical transfer function slope might only need to be first order to achieve a 4th order acoustic slope in some cases. The acoustic slope is ultimately the target and that is missed by many people, especially when using DSP active crossovers because it is easy to just set "LR4@1khz" but in this case that would result in something like a 6th or 7th order slope when combined with the natural roll off of the midhorn. If there was severe breakup at 1.1khz and you wanted response up to 1khz you might need a slope that steep, but I can guarantee you the off-axis response will be a mess and that will adversely affect the subjective result. Brickwall electrical filters are fool's gold in HiFi (there might be other arenas where it is useful of course).
That seems to peak less than the on-axis simulation from HornResp. HR is not as complete when simulating higher frequencies (not a flaw, just not in its scope).
Looking at that response, what I would do with a DSP is notch out the breakup from 1.8khz to about 4khz with a PEQ. Then apply a LPF filter of an appropriate Q and slope to approximate a textbook slope centered around 1.4khz (maybe just a 6db LPF to get a nice LR2 acoustic response...sims would provide that answer). That should allow for a corresponding textbook acoustic alope from the SEOS12 in that same area. That goal would be to balance the need for excursion control on the DNA360, cleaning up the Delta 10 breakup and very smoothly transitioning directivity from the midhorn to the SEOS horn. Disclaimer: I'm eyeballing this so don't take the specific numbers to mean anything.