Crossover Design Guide

Passive crossover design isn't as easy as a simple CR filter (per that link) because, if you try the calculations, you'll see that the resistors are far too large to be practical.

Which means you have to use inductors and capacitors, CL filters and they come with even more problems because you need to make sure that the inductors won't burn out when you're driving your speakers! Also, to get a better roll-off you need to add more than a single stage and that's an area of mathematics that gives most of us heartburn - complex polynomials. And I don't mean the polynomials are hard, I mean they use complex numbers.

The REAL pain with speaker crossovers is (I hope you're sitting down) every crossover HAS to be designed for a particular speaker. Sure you can get generic ones but they're about pleasant as sticking your finger in your left ear and trying to hum the Star Spangled Banner while standing in a bucket of ice cold water.

If you really want to learn about this stuff and have some degree of math knowledge, I can't recommend Horowitz and Hill's "The art of electronics" now in its third edition. It's not cheap but it will get you to the level where you can understand the discussions I've been having with Matt about Mics and why passive crossovers are bad.

Fortunately there is an easier way with modern Op Amps and class D MOSFET amplifiers. A two-pole Sallen-Key filter will give you a more predictable and accurate rolloff and a lot fewer headaches because you're dealing with components with tolerances of perhaps 5% of rated value (1% or better on the resistors). 

You'll still need *accurate* performance characteristics for your drivers (woofer, [squawker], tweeter) because you're throwing rocks at rainbows otherwise. I put the mid-range in [] for a reason. My experience with them has been "m'eh" at best. You see even the speaker isn't a constant impedance so that loads the whole network differently dependant on the wave it's trying to produce. For this reason, the big designers were the only people really able to get an accurate design because there is a small amount of experimentation and deviation from the ideal mathematical design: and that costs money.

But it depends if you're looking for accuracy or something that sounds pleasant.

I have a pair of Castle tower speakers here. These things are monsters, bi-powered and have concrete bases. But the original owner blew the low-end driver sky high so I just dropped in a pair of £50 car speakers. Sacre bleu! They actually sound very nice for music, background radio and even television.

Matt has had a crack at this but he used a passive network which won't give the same performance as a well-designed Bessell or Butterworth low and high pass to drive two duplex channels with each speaker getting its own class D amp.  

Here's some information on Sallen-Key and a bandpass (in case you wanted to drive a mid-range).

But what is the Class D amp and why do I rave over them? 

Well Class D is typically a MOSFET (or perhaps IGBJT) because it needs to switch on and off very rapidly. The signal is converted to PWM at with a centre frequency at some ludicrous level from 250 KHz to well over a 1 MHz! (That's up in the AM band.)

But when you pass the resulting signal through a simple LC filter (wait... didn't I say they were bad?) and what pops out is a lovely clean signal (all that HF switching noise is filtered below what we can hear).

But Class D has another feature compared to Class A, B, AB or C which is this. It's efficient. I mean REALLY efficient. To explain would take us off the centre of your discussion, but all you need to know is that they are a great way to drive a Bi or Tri-powered. Sure, it means more cables but the result is cheaper overall and uses less power so you can drive it with a battery.