MiniDSP XO types

strabismuth

Member
2010-04-13 11:47 pm
Currently using a MiniDSP for active filtering but need advice on XO type select. The unit gives a multitude of possible XO types however based on my research Bessel 2nd order and 3rd order Butterworth are the two I need to decide between. Am I better off using an all Butterworth or Bessel or Bessel on low pass and butterworth on high pass or the reverse is also an option too. I want the sound that will result in less listening fatigue.
 
You should do some additional research. Bessel 2 is one of the worst types for loudspeaker crossovers.

If you want to design or implement a crossover that gets the best out of your loudspeaker you cannot simply dial up some random crossover functions that "sound good" in a textbook or via Google. You need to DESIGN the crossover for the exact loudspeaker drivers, and the cabinet, that you intend to use. To do that, you need to MEASURE the responses or very accurately estimate them from manufacturer data (which is a total crapshoot in most cases, so just forget about MFG data).

Don't want listening fatigue? Then don't play loudspeaker crossover roulette. Do it right!
 
If you are using a digital XO (e.g in a MiniDSP), then the various polynomial types don't really matter.

The issue with most polynomial transfer functions (Butterworth, Bessel, Chebychev, Gaussian, etc.) is that they are simply polynomial approximations of the frequency domain step function (AKA the "brick wall" filter). Being polynomial approximations they can be synthesized into reactive laments (capacitors and inductors). Linkwitz architectures merely add different circuit structures to control impedance and tailor the frequency domain transfer functions. I once made a beautiful crossover using a Gaussian polynomial.. But these are complicated, expensive, heavy (lots of copper), and lossy (lots of copper).

Digital XOs are somewhat different. Particularly FIR filters. Since the coefficients are not necessarily based on a polynomial approximation, it makes little sense to talk about the polynomials.
The really important element of digital filters is that the group delay is not related to the amplitude function, and this makes a huge difference for audio, since you don't get temporal smearing around the crossover regions.

So for your question: Using the MiniDSP I use a 48 dB/octave crossover, and it sounds great. Besides, if you don't like it, you can change it in 5 minutes using your laptop and a USB cable.

The more critical element of the Mini DSP platform is the use of the parametric EQ filters. Check out RoomEQ and get the UMIC (USB mic) from MiniDSP. Test your system and generate eq filters for the room. Makes all the difference!!
 
Last edited:

strabismuth

Member
2010-04-13 11:47 pm
Thanks for the reply. I was initially using 48db/octave XO's, but I feel it was giving me listening fatigue and the image of a central vocal seemed smeared not in phase. I know it must have been a phase issue because out of phase has a non localizing effect, but once I changed to the 3rd order butterworth high pass and 2nd order bessel low pass the image really locked in and the listening fatigue was gone. I am using neo8's and ND140 woofers in a ported enclosure (my own build). I also have notch filter at 12khz, 4khz and 8khz and a gentle roll off eq. They are slightly time aligned 1 step on the tweeters and inverted polarity gave me a higher db reading on my decibel meter at XO point. Neo's have no baffle just top mounted with L brackets.