View Full Version : A Basic Guide To Crossovers

06-07-2007, 02:44 PM
Hey guys, I keep linking to this post I created on another forum. It's a basic tutorial on crossovers.

The original link is here, and has a bit more information added from some VERY knowledgeable people.


Maybe I can get a sticky on this, as it comes up VERY often.

There have been a lot of questions of late on the differences between active and passive crossovers. I'm sure with the influx of people on this board there's some confusion over what's what when crossovers are mentioned and how they're employed.

So, while this is in no way a comprehensive tutorial on crossovers, it's a beginning. Especially for those just cutting their teeth on audio.

Feel free to add, as hopefully this will grow, and feel free to correct me for anything I've said.

Okay, as promised here goes.


First, this whole topic can be confusing in the respect that there is so much debate on this subject. Some guys swear by 24 db Linkwitz-Riley filters, while another guy can write an entire doctoral thesis for his PhD on how the 6 db Norman Bates filter is superior. So what does it all mean, 6 db, 24 db, Linkwitz-Riley, Butterworth, L-Pad, Zobel, blah, blah, blah, blah, blah.....?????????

Let's start simple. High Pass, Low Pass and Band Pass. I just need to mention this just to be sure everybody stays on the same page, and even understands the very, very, very, very, very, basics. High Pass means only frequencies above the crossover point, or sometimes referred to as the cutoff frequency, are played. For example, a 2000 hz high pass will only play frequencies above 2000 hz. Low Pass means only frequencies below the crossover point are played. For example, a 2000 hz low pass means only frequencies below 2000 hz will be played. BandPass plays frequencies between two points by utilizing both a high pass and low pass in the same filter network. So, for example you could have a midrange driver only playing 200 hz to 4000 hz.

***** Note - Before everybody and their grandmothers jump on me, let me cover myself. Because I say that a high pass will only play frequencies above a certain point, does not mean I don't understand that the frequencies below that point are still played, just cut me some slack. I'll get to that point later.

Next are slopes! Slope is the name associated with the numbers, 6 db, 12 db, 18 db, and 24 db. These are the most common, and I'm sure you've all seen them. What they actually mean, is "X" db drop off after the cutoff frequency per octave. So, to translate, let's take a 2000 hz "High Pass" I said above that a 2000 hz "High Pass" will only play frequencies above 2000 hz. That's not entirely true, as you may have gathered by my "Note" above. Okay, first you need to understand what an "Octave" is. An Octave is half, or double of a given frequency. 1000 hz to 2000 hz is an octave, as is 5000 hz to 10000 hz. So a 6 db per octave slope, cut at 2000 hz will yield a 6 db dropoff of output at 1000 hz, and 12 dbs by 500 hz, etc. Now understand that a 3 db gain is double the output, and 3 db loss is half the output. A "Low Pass" at 2000 hz using a 6 db slope would yield a 6 db drop of output by 4000 hz. Clear as mud???? Good!

Why are these numbers important? Well let's take a tweeter for example. You'll toodle around on this site and see many people talk about "Fs" or "Resonant Frequency". As a general rule that we use around here, we like to say that when picking tweeters, use double the "Fs" of the tweeter to determine lowest crossover point for a 12 db slope. This is in direct conjunction with how much power you can expect your tweeter to handle. So for example we'll take a tweeter with an Fs of 1500 hz. This tells us that we can safely use this tweeter at 3000 hz with a 12 db per octave slope. Now anybody that's been around speakers knows that it's not the high frequencies that blow speakers. It's the low ones. Just look at subwoofers. You might be able to get that 10" sub to play 20 hz, but you might start hearing some nasty clunky sounds when you try to get that sub to play 20 hz with double is rated power. The same can be said for a tweeter. So how do we get the tweeter to play closer to its Fs? Raise the slope of the crossover. A 24 db drop, when a 3 db drop is half the output is considerable. I'm not doing the math. You do it if you actually want to figure out how many halfs of halfs that is. Anyway, because 24 dbs is a lot more than 12 db, you can drive your tweeter closer to it's Fs. Same can be said of midbass drivers. You want your 7" midbass driver to play flat to 60 hz, but it's got an Fs of 40 hz. Well, you'd probably better use a 24 db slope.

Getting muddier, and muddier, isn't it??

Now on to Linkwitz-Riley, Butterworth, etc.

You'll usually see a name associated with a crossover. It's usually Linkwitz-Riley, or Butterworth, or even Bessel, or some fairly exotic ones. So, a crossover is a crossover right? WRONG!! With the case of passive crossovers, not knowing the difference can be very detrimental if you're trying to design one. Let's talk about what happens at the crossover point. We'll use our 2000 hz example again. If you have one driver playing 2000 hz, and another driver playing 2000 hz, and both are putting out the same amount of output, that would be effectively doubling the output, correct? Double the output is a 3 db gain. So, if you have one speaker playing flat to 2000 hz, and the other is playing flat to 2000 hz, then at 2000 hz, you'll have a summing of frequencies and effectively a 3 db gain in output. Maybe not so desireable, or maybe it is. Depends on the drivers and the application of the drivers. Well, this is actually the basic design behind a 12 db Butterworth crossover. It's simple and uses one crossover point to calculate the values for the components of your network. Only hitch is the 3 db gain at the crossover point. So, how do we get rid of that 3 db gain. Well, maybe by offsetting the frequencies a bit. Let's say 1850 hz for the low pass, and 2150 for the high pass (these are arbitrary numbers). Since they're not technically overlapping at this point, then the summing of the frequencies is only happening AFTER the output has started to go down. (I use calculators to determine this stuff so please don't ask me to do any math). With the right calculations, you can determine what sums will create a flat response. This is what a Linkwitz-Riley filter does. A Bessel filter will yield a 1.2 gain, and if you want to do the math, you can probably create your own filter network and apply your name to it. Then you can write a long diatribe about how great your filter that has a .82564345 db gain is, and how it will revolutionize the audio world.

"Man, are you sure this is mud, it smells more like....

Now to put this all together.

There are some common statements on this board. "Go active, the results are better". "I'm new to car audio, and am unsure about tuning an active setup". "Passives are way too hard to get right, you're better off going active." etc, etc, etc....

So, let's take a car application. You have a choice. That shiney new Diamond Hex set looks really sweet and you can get a great deal on it, or you can do it the hard way, and get an active crossover, some new shiney Seas drivers, and another amp. Well, with all this information you now have, please tell me how Diamond Audio knows how their components will be installed in your vehicle. Did they use a 12 db butterworth filter in their crossover? Doesn't that mean there's a 3 db gain at the crossover point. If I put these in kicks, on-axis, will that give me a nasty peak at the crossover point? If I put the midbass in the doors though, maybe the fact they're off-axis will be compensated for a bit by the butterworth design....

This is why we're advocates of active processing. It doesn't matter what kind of filter it is. As long as you can change it and experiment on the fly to get what sounds best. There's no math that needs to get things all muddled up.

However, if you really want to design a passive crossover go for it. Remember home audio uses the passive with great success. Their environment, however, is MUCH more controlled than ours.

I'm done... I know I mentioned L-Pads, and Zobels, and a lot of "blahs", but for now this will have to do.

06-07-2007, 02:45 PM
Okay, now that you’ve had some time to absorb all the previous information, I’m going to step it up a little bit.

Let’s first look at the challenges we face in building any kind of speaker system.

First is level matching. It’s very difficult to find a tweeter and a woofer that have the exact same sensitivities, and will operate at the same volume given the same power.

Second is impedance variations. As you move through a speakers frequency response the impedance varies. As you get towards the upper end of response the impedance starts to rise considerably. This is demonstrated in the following graph. This a factor of inductance of the voice coil.


As you can see, the impedance is rising the higher you go. This causes a drop in output. A Zobel will level that out, giving you a flatter response.


Third is spikes and dips that cause your response to be something other than flat.

Let’s start in order here. Level matching is done using an L-Pad. Generally, the tweeter is much more sensitive than the woofer. So the tweeter needs to be attenuated. The L-Pad circuit can be introduced into your passive crossover. It can be done 2 ways. First is fixed. If you know the amount of attenuation, you can get the appropriate resistors and build the network right into the crossover. The problem with this is if you don’t get it right, you need to completely redo it. The second way is to use a variable L-Pad. Parts Express has these and in a car audio application using passive crossovers they are a MUST in my opinion.

http://www.partsexpress.com/webpage....Group_ID=1 96

The drawback is you need to build around the L-Pad and mount it somewhere that is accessible. However, it does allow you to adjust your tweeter to compensate for different mounting locations. These are what upstage kits, like CDT use.

Next is the Zobel network. The Zobel network compensates for the rise in impedance due to voice coil inductance. Pretty much enough said.

Third, and this is where crossovers start to get REALLY tricky, and even I’m a little fuzzy in this area. Depending on your baffle, mounting locations, varying distances of voice coils, your response will be something other than flat. If your baffle is too narrow, (as in a tower speaker) your mid drivers can experience a spike in response of around 6db starting at around 100 hz and flattening off at around 1k hz. This is called baffle step. You use baffle step compensation to, well, “compensate” for it. I’m not going to go into it a whole lot, and baffle step is just one example of how a speaker reacts to it’s environment. I will say that there is a compensation circuit that will cover just about every peak, and dip.

What does this all mean? It means that you can drive yourself CRAZY trying to build the perfect crossover for any application. There’s a point of diminishing returns when a crossover just becomes too over-engineered. It’s safe to say that a decently built crossover will contain 3 parts, the actual crossover network, the L-Pad, and the Zobel. The other networks don’t really come into play until you NEED them. Baffle step can be fairly accurately predicted, but other “notch” type filters are usually used to compensate for something unpredicted and unwanted. It’s possible to “notch” out a peak caused by a Butterworth type crossover at the crossover point, but it’s just better, if it’s a problem to pick a different alignment.

LEAP is great, as are a lot of other programs out there. LEAP allows you to build your crossover ‘to the enclosure’ for the best possible results. LEAP can only be effective when the surrounding variables of a driver are controlled. For example, the difference between a well built speaker enclosure, and a door frame. LEAP can predict and design a crossover to compensate for the way the speaker will react within a given enclosure by simply inputting the dimensions. LEAP is going to “leap” out of your computer and kick the snot out of you if you try to enter door frame parameters.

So, if you are looking to go the passive crossover route, you can pretty easily design a crossover as long as you know which speakers you’re going to be using. You add your three essential items and voila you have a working crossover. However, based on the above information that’s all you’ll get. You won’t get any kind of compensation for the way the speakers will react in the car. And believe me, they will react BADLY.

So, what do we do to compensate for these unknown and highly erratic, unpredictable peaks and dips in our frequency response? We EQ them out. An equalizer will do everything that a notch filter will do, and is effectively an “Active Notch Filter”. We couple the EQ with an Active Crossover, and you get a system that is highly adjustable and can correct for the many, less than desirable, effects your car will have on your system. Add Time Alignment and you can achieve results comparable to a cheap home audio setup (given that you spend $1000.00 on high quality speakers).

By going active, we’re essentially, trying to take the “car” out of the “car audio”. Active processing is also used in home audio and is highly desired by many audiophiles out there. However, in my opinion, the difference between a well built tower with a passive crossover and an active setup in home audio is not nearly as dramatic as an active to passive setup in car audio.