Crossover Design Basics

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Now that I have the money to begin another DIY speaker project, I think I'd like to venture off into the world of designing my own 2-way crossover, but I've never done it before. Can anyone suggest some good (preferably great :) sources of information regarding the basics of how to design crossovers (particularly first or second order)? Can anyone also suggest some freeware to aid me in designing the crossovers? Any help would be greatly appreciated.

Regards,
Bryan
 
Bryan,
You didn't specify whether you meant active or passive. The below is mainly oriented towards active, but large chunks of it are appropriate for passive crossovers, too.
First order crossovers are trivial:

F=1/(2*PI*R*C)

Where:
F= frequency
PI= 3.14159
R= resistance in ohms
C= capacitance in Farads (mF= F*10**-6, pF= F*10**-12)

Nuthin' to it. (Except for passive crossovers where you've got to take into account the inductance of the voice coil, etc.)
Okay, now second order...ugh. Can of worms time. Second order crossovers don't sum to unity, i.e. take 2, split it into 1 and 1, now add them back together. 1+1 no longer equals 2. 2nd order crossovers also invert phase.
Third order crossovers are phase correct, and sum pretty close to unity. Not a bad choice.
Forth order is two second order crossover sections, with all the attendant problems.
Fifth and above...don't go there.
That does it for the standard Butterworth crossovers, but only opens the game. There are about ninety-nine other crossover functions out there. Unless you want to have a nervous breakdown, there are only two that need concern you. Linkwitz-Riley and Bessel. Linkwitz-Riley crossovers take the ugliness out of even-order (i.e. 2nd & 4th) crossovers. Bessel function crossovers are nifty because they're minimum phase shift.
*All* crossover functions have tradeoffs, no matter how vehemently some misinformed person insists that *his* crossover is the One And Only. Pass band ripple, sharpness (and control) of slope, and phase shift are the things that we (as audio folks) are most concerned with. M-derived is a good example of the 'brick wall' function--hits the end of the pass band and drops like a stone. Cool, right? Unh-uh. Really, really nasty pass band ripple (plus and minus several dB in the frequencies leading up to the rolloff point), and phase shift problems. That's just one example.
I've never seen a site that gave a good, *comprehensive* cookbook on designing a crossover. The closest I've seen are a few sites that give crossover schematics and the formulas to alter the frequencies to suit your needs.
Try http://www.silk.net/personal/lwright/hwb/hwb for the formulas for the functions listed above. There are generic schematics to show you where the parts go if you're using opamps. Be aware that you can just as easily use transistors or tubes in place of the opamps; use the emitter/source/cathode for the + input.
The first thing to do is select a reasonable frequency for the crossover point. There is as much art as science to this, but it's a good idea to give yourself *at least* one octave before the driver rolls off. For example, if you're using a woofer that's spec'ed out to 500Hz, set your crossover point at 250Hz or below. Why? Because if you wait until the driver's rolling off on its own, you might as well sit out the game entirely and just let the driver drop. (In next semester's advanced class, we'll break this rule just to avoid having the circuitry in there. A commercial example of this was the old Wilson WAMM system which used KEF B-139 drivers down to their natural rolloff point of about 40-50Hz, then brought in these big, honkin' 18" subs at that same frequency. It takes careful planning to do this successfully, not to mention drivers that don't have spurious bumps in their response.)
Proper crossover design takes time, care, skull sweat, and a fair amount of cut and try in the final stages in order to produce first rate results. Note that crossovers, like other audio projects, tend to produce blindness in their owners. If they built it, they think it's perfect...but that's not always the same thing as reality.

Grey
 
Geoff,
At last, a decent page on passive crossovers! (I felt better after noting that it's only been up a month or so...) I have a few quibbles, but they are minor, excepting short shrift given to 3rd & 4th order crossovers (yes, I agree they are more complex, but a fair overview would dictate coverage).
Bryan,
If it's a passive crossover you're contemplating, I fully agree with Rod Elliott that you should consider bi/tri/whatever amping. The advantages are considerable. I just recently went from tri-amp to quad-amp, and am a firm believer in the sonic benefits of at least bi-amping. Besides, it will give you an excuse (as if you needed one) for another project or two.
Should you decide to go with a passive crossover, I would like to throw in two points:
1) Rod Elliott overlooks one fairly obvious possibility in cooling a crossover--mount it where the air flow from the port (assuming a ported enclosure) can blow over it. No, you wouldn't want to mount it *in* the port, for any number of reasons, but directly behind the port wouldn't be a bad idea.
2) He also mentions not mounting drivers eqidistant from cabinet edges. While this is generally good advice, there are a number of ways to take care of problems stemming from edge diffraction--thick felt or high-density foam (at least 1/4", but 1/2" or more preferred) surrounding the tweeter and midrange, rounding over the cabinet edges with a router, making the cabinet front so narrow that it's essentially the same width as the driver(s) so that there's no appreciable delay between the direct-radiated sound and subsidiary waves coming from the edges, etc. etc. etc. There's no conflict in using all these ideas together.
Give us some idea of where you're trying to go, so that the suggestions can be fine-tuned.

Grey
 
Ok, here's what I have to work with. The tweeter I will most likely use can be found here: http://www.partsexpress.com/pe/showdetl.cfm?&DID=7&Product_ID=7350&CATID=49

I don't have the exact page for the woofer, but here are the parameters for it: (Both tweeter and woofer are rated at 8 ohms, but I have no way to actually test this)
Fs 33.5Hz
Re 7.1W
Le 1.28mH
Zo 32.81W
Qts 0.75
Qes 0.96
Vas 46.63L
Sd 0.0214m2
Xmax 3.35mm
Cms 0.00073m/N
Mms 30.84g
BL 6.93

I'd like to make a simple 2nd order Linkwitz-Riley passive crossover for the two. Space is a factor, so the box will more than likely be vented, although I can be flexible about this. This will be my first attempt at designing a crossover, and I've never dealt with any of this before, so most of it is new to me. Although I know how to solder, I need just as much direction with the physical construction of it as I do with the design process. Needless to say, you've got a novice on your hands :) I am yours to be shaped and molded.

Regards,
Bryan
 
Bryan,
Although you can build a printed circuit board for a crossover, there's no shame in doing point to point wiring using terminal/barrier strips (some people call them one, some the other--it can get confusing). For a backboard, you can use masonite, plexiglass, wood, whatever...but not metal. Some people use perf board--that stuff with holes every 1/10". It works well, but I never go past prototype stage with it, as nothing spells a-m-a-t-e-u-r quite so loudly as a perf board in a finished product, even if it sounds wonderful.
It's a good idea to use silicone caulk (standard bathtub caulk--use any color that makes you happy) to glue the parts down to the board. This is for three reasons: one, the vibration of the speaker can cause metal fatigue in the component leads over time; two, the speaker vibration can cause really annoying rattles among loose parts; three, some people feel that it's audible. It's cheap and works well.
Passive crossovers are a nuisance. I positively hate fiddling with them (and it takes a *lot* of fiddling--I meant what I said above--the ESP site said the same thing, using a lot more words).
As for 2nd order Linkwitz-Riley...if you're going to go even order (2nd or 4th), L-R is the way to go. Even order Butterworth crossovers are inferior to L-R, although Butterworth odd order crossovers work decently. Myself, I prefer either first order or third order Bessels, but that's just me being me. Choice of crossover slope and function is part of how you put your signature on your project.
In other words, the mechanical construction ain't all that difficult. Start with perf board--it's cheap, everybody ought to have a half dozen sheets sitting around--then go to a sturdy, clean layout with terminal strips. You'll do just fine.
Back to you.

Grey

P.S.: Oh, and another thing Rod Elliott left out--drivers, like other electronic parts, are manufactured to within certain tolerances. Do *not* be surprised to find that cheaper brands of drivers are +-20% compared to better lines at, say, +-5%. *Very* few companies put +- percentages on their spec sheets for drivers. It really is a good idea to measure the relevant numbers for yourself, because even if you and I were to agree to build 'identical' project speakers, my drivers might spec differently from yours.
A commercial example of good design in a crossover was to be found in the old Rogers LS3/5A. Due to manufacturing variations in the KEF B-110 and T-27 (the woofer and tweeter), they had places on the crossover where you could change little pigtails to bring it all back into tune. And that's for KEF drivers, which were quite high quality, and some of the few I've seen with tolerance specs.
You might find that you need to change part values from left to right channel to make the speakers behave consistently, so don't just test one tweeter, and one woofer, then assume that what works for them will automatically produce identical results in the other channel.
(Yet another reason to biamp...)
 
Well, you've had a lot of tips submitted to you and I really don't want to jump into the passive crossover arena.

BUT . .

I will make a couple of points.

Buy a copy of the "Loudspeaker Design Cookbook" and save some time.

You will NOT have a crossover cooling problem so don't waste your focus here. And, why not just mount the crossover parts to a piece of plywood, jumper it all together and you'll be on your way to a good start. Just make sure you don't mount coils with anything but a brass screw running thru them and mount adjacent coils so that the axes? are not parallel.

I swear, some people on this forum just like to hear themselves talk. Just ignore the babble.

Crossover cooling - give me a fuc*ing break.
 
Heh heh, I like your enthusiasm Bill ;)

Anyways, I'd like to thank everyone for the tips. I will probably pick up a copy of the book somewhere since I've heard nothing but praise and reference to it throughout my internet searching. What I'll probably end up doing is making do with what I have to work with, but I'll just play around and tweak it before I finish the project up (although I have to start it first! ;)). Anyways, once I get it going, I'll let everyone know how it came out.

Regards,
Bryan
 
grataku,
The circuit topology is the same for Linkwitz-Riley as it is for Butterworth, but the component values vary somewhat. For that matter, Bessel is built on the same skeleton, but again with different parts values.
So if someone hands you a schematic of a crossover without parts values on it--until they give you the resistances & capacitances, you don't know which one you're looking at.
The 'brick wall' filters are different, though.

Grey
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.