Selecting Crossover Frequencies for 3-way build
 

This is a newbie question so I'm really looking for a nice "101" intro information source I guess... but here's what I did:

A while back I bought a 2-way custom built cabinet that has a 12" driver and a small bullet tweeter. It had a crossover in it that has a 2nd order circuit with a 3.5kHz crossover point.

The maker of this cab now does a 3-way design with the same driver & tweeter and has added a 6.5" midrange driver. His website says the midrange is 800Hz to 3.5kHz so I got a 3-way crossover that has those crossover frequencies. The baffle of the box I bought had plenty of room for the 6.5" and the internal volume allowed me to build a box-in-a-box with the correct volume sealed off beind the 6.5" driver.

Well, it's all wired up and sounds really good, which I recognize is due largely to luck.

But it got me thinking about why crossover at 800Hz? If the 12" can handle up to 4kHz with reltavily flat response curve and the 6.5" specs claim it's fairly flat down to around 120-150Hz, how was that 800Hz frequency determined?

I ordered a book online by David Weems on speaker building but would love to find some online info on how one goes about determining crossover points for bass guitar cabs.

Any thoughts or guidance would be most appreciated.

Cheers.

Because of dispersion. The 12 might reach 3 or 4khz, but only on center, it won't spread those frequencies in a wide arc. The larger the cone, the lower in frequency it will start to "beam", meaning it's upper frequencies become concentrated in a narrow beam on center with the speaker, but don't get spread widely anymore.

The 12 could still have a good spread a ways higher than 800hz, but, another reason is sometimes to take down the rougher, more "peaky" response many woofers have towards the top end of their range, and hand it off to the next driver, which would likely play smoother in that region.

It takes way more powerful coils to cross over lower which raises he weight and expense a lot.

I look at the response curves (both on- and off-axis), as well as power handling (both thermal and mechanical) when coming up with a ballpark starting point for a crossover. From there, I trial-and-error it until I'm happy with the results. In my experience, crossover design is always a juggling of tradeoffs. Well so is the rest of speaker design, so nothing new there.

My guess is that your cab's designer found the 6.5" midrange does a better job than the 12" woofer in the region from 800 Hz to 3.5 kHz, but below 800 Hz the 12" woofer does a better job. A "better job" would probably be more desirable frequency response both on- and off-axis. For one thing, the smaller cone diameter gives you better dispersion; a 12" woofer is going to be beaming pretty badly up at 3.5 kHz.

Now the 6.5" midrange driver's curve may have looked better south of 800 Hz as well, but power handling considerations could have made the designer decide not to push it any lower than that.

http://www.usspeaker.com/faital%20pro%2012pr300-1.htm

Notice how the frequency response varies when standing directly in front (0 degrees) to standing off to the side (45 degrees). It's a very in-exact science, but if you can trust what the manufactures tell you, the mid range makes up for the muddiness that you'll hear standing anywhere but directly in front of the cab. The upper range of the woofer is, (as BFM loved to say) moot.

WOW! Some great info here, thank you. The off-axis info is mind blowing.

@DukeLeJeune - when you say trial and error, do you have some way to vary the crossover point easily or are you swapping in and out capacitors and inductors?

First "modelling", and then building & tweaking (swapping in and out capacitors, resistors, and inductors).

To start off, I run a series of measurements. Then using those measurements, I use a computer program to "model" the response with various possible crossover filter topologies and component values. Once the modelled response looks good, I build that crossover and measure the system response. The actual measured response is almost never exactly what the modelling program predicted, so I tweak the component values until I like the way the measured response looks. Then, I start listening to it. If something doesn't sound right, I have to figure out what it is. The hard part comes when the measurement looks good but something still doesn't sound right. Then, I have to figure out how to change my measurement technique so the measurement corresponds with the subjective impression, and that way I can figure out needs fixing. Armed with that information, I go back to the modelling program... lather, rinse, repeat.

Wow, way deeper than I anticipated. When you measure the actual system response, are you inputing a sine wave signal that starts low (20hz maybe) and then goes all the way out of hearing range (20kHz)? I did something similar when I setup my control room for my little studio, and could see some value in it but wonder how you measure it?

Also, on a whim, I played a little through just the 6.5" speaker, which presumably is only giving me 800hz to 3.5kHz and it was pretty lame sounding and quiet. But in the whole system it sounded fine. Is that typical?

This is just so fascinating to me. Thanks for all the input!!!

That's typical. That speaker is just catching the upper harmonics of your bass, which without the beefier lower frequencies, will sound like that. I hear the same thing through active biamping and shutting off the low amp.

BTW, Duke there is among the best in the business.......pay attention.:)

Good to know. I'm learning loads and have to confess I might get hooked and actually try building my own. Heck, I did the built-ins from scratch for the master closet, how hard can a big wooden box be (he says tongue in cheek).

I had to take a couple circuits classes when I was working towards my mechanical engineering class. If they'd been talking about this sort of stuff, I might have done a bit better :)

I'm definitely starting to sense that :)

I use a time-gated sweep to get a quasi-anechoic measurement down to 400 Hz or so. Rarely do I need to go lower, but if so, I use other techniques.

A time-gated system is computer-controlled and works like this: The microphone turns for long enough to capture the direct sound from the speaker, and then turns off before the first reflection arrives. That way you only pick up the sound from the speaker, not the room. Once the wavelengths get too long, the microphone isn't on for long enough to capture enough of the waveform to be useful. This is a wonderful tool, because otherwise it's a lot more trouble to capture the sound of the speaker without the room intruding.

At lower frequencies I leave the "window" open long enough to allow some room reflections into the data, and have learned how to recognize them in my room so that I can disregard their effects. Still lower in frequency I switch to close-miking the woofer cone, so that the close-up nearfield response from the cone pretty much drowns out the room response. That being said, enclosure modelling programs and manufacturer's published curves are usually accurate enough that I don't need to actually measure down low unless I'm doing a crossover down there.

Be careful, my friend. This stuff is highly addictive. I started building speakers as a hobby in 1979, turned pro in 2005, and at any given time I have a half-dozen ideas that I want to try. The hard part is figuring out which ones a) the people will want to buy; and b) I'll actually make money on.

Nah, Duke just talks a good game.

Huh... that totally makes sense. When I did my control room, I of course was only concerned with the room in particular standing bass waves, but of course you wouldn't want your testing to include data from the room; fascinating.

So for a neophyte, without the extensive equipment you have, what the easiest/best way to experiment with crossover frequencies? For instance, I'd love to hear what my cabinet would sound like with 500Hz/2.5kHz vs. 1kHz/4kHz. Do you keep a bunch of inductors and capacitors on hand?

I take the active 3-way crossover out of my PA rack, and do bi or tri amping to dial in a "sweet spot". Could use that info as a starting point for a passive, but I usually just gig with the active system.......I'm a hack.:p

Meaning I can't play upright ... (yet:smug:).

On electric, the woofer and mid are all I need. Plenty of bassamps have 2-way biamping to do that. If you want to build a proper speaker anybody can use with any amp, you must master the passive crossover. I'm currently up to a couple of simple filters.:)

So try one of these. Used to be on the TB wiki - must have missed it.

That is a type of "beam blocker". It does work, but it does so by taming down the output of your speaker on axis to better match the off axis. Maybe not a bad idea for some guitar players.:)

What that would do to us on bass is make our loudspeakers quietspeakers. The crossover + small cone mid keeps the dispersion without taking a hit in spl.



.....as explained by BFM.

It is pretty cool how/why those things work...just a piece of foam. It essentially turns the highs into a "virtual" 3" speaker.

Anyone who doesn't know-

http://www.ti.com/tool/filterpro

Fantastic tool for desining pretty much any filter you could want, provided you understand what it is you want in the first place.

It is a whole lot easier to use an active 3-way crossover when building your prototype.
This bypasses the huge expense of various inductors and capacitors for different configurations.

Used with an RTA, the response can be measured on and off axis quickly and no additional cost over the active crossover and amp.
When the final frequencies are determined, measure the actual driver impedance at that frequency and calculate your passive crossover values.

The active crossover I'm familiar with was in a "big" PA system. It was fed with the +4dB signal then sent into the various power amps. Would a similar setup be needed here? With a separate amp for each speaker? or are there crossovers built to split the signal AFTER the power amp?