Crossover Slopes In Context.

[BlueWizard]

Sorry, I'm off on one of my rants again.

A recent post got me thinking about Crossovers, specifically first order crossovers, or 6db/octave crossovers.

Let's put 6db per octave in perspective, but before that, let's put 6 db in perspective.

3db is a slight but noticable change in volume. If you tweak your volume control up or down just enough to clearly perceive a change in loudness, that is about 3db. So 6db is two small tweaks upward or downward of your volume control.

Now let's look at 6db per octave - 1st order crossovers and for this example, let us assume a crossover of 2,000hz. One octave down at 1,000hz is like tweaking your volume control down 2 slight nudges. Dropping another octave to 500hz is like 4 downward nudges. Let's generously say that is about like turning your volume control down from 40% of a turn (4 on a scale of 10) to about 20% volume (2 on a scale of 10).

We are down two octaves and while the sound has diminished, it is still certainly playing and is certainly audible.

Now let's drop one more octave to 250hz (3 octaves or 18db total) which I say for purposes of illustration is about like turning your volume down to ONE (from 4 down to 1). That is pretty quiet, but you can still hear it.

Now let's go the other way; let's move an octave upward from 2khz to 4khz then another octave to 8khz, about like going from 4 on the Vol Cntrl down to 2. Now a third octave up to 16khz, and while the loudness is very low, it can still be heard; it is still at a functional level.

It would seem that if you intend to use a 1st order crossover, you pretty much need two full range speakers; a woofer that goes up very high and a tweeter that goes down very low.

So, briefly let's consider 12db/octave crossovers and ask the general question, how many DB does a speaker have to drop to be consider acoustically out of the circuit? How far down before any sound coming from it become irrelevant?

If you are pushing a speaker beyond its rated frequency response, is it just dropping off on its own, or is there real potential for distortion outside its rated frequency range?

With 12db crossovers, you drop pretty fast; 24db in two octaves and 36db in three. While the phase relationships cause by the crossovers might be better with 6db crossovers, how does one reconcile the extreme frequency range that is forced on the speakers?

Just curious.

Steve/bluewizard

 

[Iain McNeill]

It's a sobering thought isn't it! If you do a vector sum on two sinewaves then one has to be down by 20dB to make less than 1dB difference in the summed total. 1dB difference in level may be hard to hear but I think it could make a difference in the sound of a speaker when it appears as a dip in the response.

This means the audible overlap region in a 1st order XO speaker is 8 octaves!

 

[SY]

That's the least of the worries with 6dB crossovers. Besides the horrific lobing and power handling issues, the acoustic crossover is a mess unless the drivers are so flat that they don't need crossovers in the first place. I've heard some pretty good speakers with simple 6dB electrical crossovers, but they sounded good despite the crossover, not because of it.

Add in the complication of (generally) non-coincident centers and it's a dog's breakfast.

 

[hermanv]

It seems generally accepted that masking occurs at about 40dB. That is to say any signal down 40 db will not be heard in the presence of a 0 db signal.

In multiple driver designs, it's typically the high end of a given driver that makes for a problem. Cone break-up is very audible and objectionable. Some low order crossover designs find it necessary to add a trap at the break up frequency, this causes at least one frequency to be much further down in level than the crossover slope itself would provide.

I notice in speaker reviews that high order crossovers seem to be gaining some favor in commercial designs. They seem to be more necessary in drivers with exotic cone materials (ceramic, metal, kevlar etc).

One advantage of the Linkwitz Riley crossover is that the signal is down 6 db at the crossover frequency instead of the 3 dB down of most conventional designs. This extra 3 dB might be part of the reason that the L-R crossover is well received.

 

[BlueWizard]

Well, one of my questions was answered but the other one remains.

Assuming a typical 12db crossover, how much room do I need to leave when selecting a crossover point?

If my woofer goes from 30 to 3000, and my tweeter goes from say 2000 to 20,000, do I have a workable combination? Can I crossover at 2,500 and still be in the workable range of the speakers?

Someone said, I needed to drop down by 40db to be considered out of the circuit. For the moment, let's just call it 36db or 3 octaves with a 12db/octave crossover.

If this is true, it seems an impossible task to require either speaker to operate 3 octaves above and below the crossover point.

So, to what extent can I count on the natural roll-off of the speaker?

In short, I am trying to determine how to match speakers, woofer/Tweeter or woofer/mid/tweeter to a 12db crossover.

How wide a span do I need on each side of the crossover points to still be within the realistic working range of the speaker?

And what happens if the range is too narrow? Fried speaker? Distortion? ???

Thanks.

steve/bluewizard

 

[Cal Weldon]

If my woofer goes from 30 to 3000, and my tweeter goes from say 2000 to 20,000, do I have a workable combination? Can I crossover at 2,500 and still be in the workable range of the speakers?

Hi Steve,

Some general rules I follow:

1. Cross the woofer as low as possible.
2. Do not cross the tweeter lower than 2X the Fs.

 

[Iain McNeill]

So, to what extent can I count on the natural roll-off of the speaker?

I think this is exactly the point. Pretty much every driver is a 2nd order bandpass. Sometimes the cabinet provides a little Q peaking and sometimes the upper band of the driver is trashy due to breakup, resonance etc etc.

My belief is that the best sounding speakers exploit the natural 2nd order response of the drivers and use a shelving filter if the XO is too low (for high pass) or add a 2nd order on top of the natural 2nd order to achieve LR-4, for example. Of course you have to keep in mind the phase responses of everything because amplitude is only half the picture.

When you don't consider the driver response then it will modify the expected crossover performance and the result is a non-flat summed response at the listening position. If you're XOing your tweeter with a 1st order at 2KHz and fs is at 1KHz then you'll have 1st order response to 1KHz then 3rd order below that. You'll be pumping in bass to the tweeter at -12dB at 500Hz, -24dB at 250Hz. This will increase doppler modulation of the tweeter output and reduce the max acoustic output (headroom) i.e. make it easier to fry the tweeter.
Woofers really don't care what you send them they just have a habit of making nasty noises outside their comfort zone. For instance if the woofer had a resonance/break up mode at 4KHz that provided 10dB of gain then the woofer could become audible again at that frequency with corresponding effects to the sound quality.

In reality, with passive XO's you have to accept the compromise of crossover region width vs crossover complexity. I think the 40dB down requirement is pretty much unachievable as you indicate although it's probably real in careful listening conditions. 20dB down for 1dB total contribution gives you some hope and lets face it, if our speakers were ±1dB we'd be ecstatic right?.

These are a few of the reasons I gave up on analog crossovers some years ago.

 

[hermanv]

Assuming a typical 12db crossover, how much room do I need to leave when selecting a crossover point?

Well that's the trick, crossover design and driver selection are usually done together. Some drivers are more tolerant than others. Look at the manufacturers data and remember that the design will not be ideal.

Originally posted by BlueWizard If my woofer goes from 30 to 3000, and my tweeter goes from say 2000 to 20,000, do I have a workable combination? Can I crossover at 2,500 and still be in the workable range of the speakers?

Someone said, I needed to drop down by 40db to be considered out of the circuit. For the moment, let's just call it 36db or 3 octaves with a 12db/octave crossover.

If this is true, it seems an impossible task to require either speaker to operate 3 octaves above and below the crossover point.

Using 2000 and 3000 as limits and crossing over at 2,500 is not a bad start (Remember to use geometric centers; square root of (X * Y) equals 2450Hz.). You are quite unlikely to achieve the 30-40dB an ideal design would require, compromise, build, listen and adjust. Repeat as needed

Originally posted by BlueWizard So, to what extent can I count on the natural roll-off of the speaker?

See the manufacturers data sheet, most drivers will match published specs and frequency response plots reasonably well. Remember that the enclosure also causes roll-off at the low end.

Originally posted by BlueWizard In short, I am trying to determine how to match speakers, woofer/Tweeter or woofer/mid/tweeter to a 12db crossover.

How wide a span do I need on each side of the crossover points to still be within the realistic working range of the speaker?

This is why two way, three way and four way designs exist, more drivers equals less overlap, more complexity, more costs.

Originally posted by BlueWizard And what happens if the range is too narrow? Fried speaker? Distortion?

No damage will happen to the drivers. However your ears may experience less than perfect sound, possibly painful and ugly sound.

steve/bluewizard: These questions are complex enough that the usual advice for a newbie is to build a kit, learn, decide how much time and effort you want to invest. Next, acquire some tools, such as design software and a method of measuring your designs progress.

 

[BlueWizard]

All very good information and I thank you for it, but I think you are all too deeply into the puzzle.

All I want to know is how much overlap on the raw speakers?

Before I can even start designing a set of speakers, I need to choose the drivers. I've always assumed you designed the crossovers and cabinet to fit the drivers, not chose drivers to fit the crossover.

So, I'm at the starting point, I'm searching through catalogs of speakers, I see speakers that look nice, are priced right, and have reasonable good specs and reputation. That's where I'm at in this hypothetical situation; I'm still looking at the catalog.

For the moment, let's say my goal is a 2-way system with a 12db crossover.

How do I chose the speakers relative to their frequency response?

If I have a 1khz overlap, is that enough? Should I have a 2khz overlap? 3khz? Do I need one octave of response above and below the crossover point? 2 octaves? 3 octaves? Or do Octaves not even matter, maybe is is a matter of raw numbers, 'X' hz above and 'Y' hz below the XO.

So, let's start there; I'm not at the point of designing cabinets or crossovers, I'm on the internet looking through catalogs of speakers trying to find a good match for a woofer and tweeter.

In seeking a good match between a woofer and a tweeter, how much overlap in frequency response do I need?

That is the fundamental question that needs to be answer before anything else can be done.

Steve/bluewizard

 

[Cal Weldon]

How do I chose the speakers relative to their frequency response?[/b][/B]

Originally posted by Cal Weldon
1. Cross the woofer as low as possible.
2. Do not cross the tweeter lower than 2X the Fs.

That is a fundamental starting point. Find drivers that meet that criteria. Example: Don't try and match a 10" woofer with a 1" dome tweeter. Find a tweeter with a low enough Fs to match where you want to cut off the woofer. The woofer cut off can be determined by the beaming and where the cone break up starts.

 

[y8s]

That is a fundamental starting point. Find drivers that meet that criteria. Example: Don't try and match a 10" woofer with a 1" dome tweeter. Find a tweeter with a low enough Fs to match where you want to cut off the woofer. The woofer cut off can be determined by the beaming and where the cone break up starts.

of course when you delve deeper into how low a given tweeter can play, you will probably realize it has distortion issues before you get as low as 2x Fs.

Case in point: Zaph's early measuremetns of the Vifa ring radiator (XT25) which has a 600-something Fs but poor distortion below 2000 Hz.

 

[BlueWizard]

Yes, Cal (Weldon) both your post are good advise and I certainly would not do otherwise, but they don't address to core question.

How do I chose speaker Drivers based on their frequency response alone? If I can't get past this hurdle and if others were not able to get past this hurdle, then no speakers would have ever been built.

Let's take a ridiculous example. I have a woofer with response from 20 to 2000, and I have a tweeter with response from 2000 to 20,000hz.

Well, on paper, conceptually, that would seem ideal, I have two speaker that cover the full response range 20 to 20,000hz. But, of course, we know it IS NOT idea, it is not even good.

So, let's change it; woofer = 20 to 2500 and tweeter = 1500 to 20,000hz, and assuming a 2khz 12db XO. Is that enough, 500 hz above and below the crossover frequency?

Should the overlap be measure in Octaves or raw numbers?

If +-500hz is not enough, then is +-1,000hz enough?

If I can't answer this core question, then I don't see how I can ever choose drivers to be made into a speaker system.

I'm well aware that there are many other factors to consider, but this one very early, very fundamental item needs to be resolved before all other factors can even be considered.

Steve/bluewizard

 

[Iain McNeill]

The short answer, and the one you're not going to like is...it depends. The questions you are asking need to be bounded before anyone can make an intelligent answer. It's a bit like asking "How do I make dinner?" and hoping Emeril will tell you how to make your favorite meal.

I think most people choose drivers based on a prioritized list of parameters. One might want deep bass extension on a woofer, aother might want no beaming in the upper end of drivers range a third might want minimal distortion at the expense of loudness. In the end we pick speakers on their own particular qualities and then decide how the XO can stitch them together best.

If you chose great speakers that had minimal overlap then you'd have to take a high order, complex XO as your compromise. Obviously, the more overlap you can achieve in your driver choice, the easier the XO design gets.

OK I've babbled enough. I'll try and answer your question. Take your example of woofer to 2500Hz, tweeter from 1500Hz. Center frequency is 1936 Hz. You'd need an overall 8th order XO to get both drivers out of the picture by the end of their ranges (20dB down) without having to worry about tuning the XO. Anything less will produce ripples in your amplitude response where the drivers run out of steam but are still needed to produce full summed output.

If you use a 4th order XO, then you'll have to add some peaking to the XO to compensate the woofer for the lack of tweeter output and vice versa.

If you can find drivers you like with more overlap then the XO design task gets easier.

Then again, maybe you like the non-flat response that occurs with the low order XO or maybe it complements your room response nicely, or perhaps the baffle step can compensate for lack of tweeter.

There's so many factors here that there isn't the simple answer that you want.

 

[Cal Weldon]

So, let's change it; woofer = 20 to 2500 and tweeter = 1500 to 20,000hz, and assuming a 2khz 12db XO. Is that enough, 500 hz above and below the crossover frequency?

As Iain pointed out, it depends. You can't make any decision in that regard until you've looked a little further.

When matching drivers for multi way, you really should start with one of them. Let's say you really like a certain woofer. Start with it, realize it's limitations and then find a tweeter that will match that the woofer. If the upper usable limit of that woofer is 2kHz then you want a tweeter with an Fs of no less than 1Khz. As pointed out, that sometimes isn't enough. Sometimes the distortion level of that tweeter is unacceptable and it must be crossed at no less than 3Khz. Well, that eliminates that tweeter, so we have to look further even though the overlap seems to be plenty.

If we are thinking of a 3 way, I always like to start with the midrange. IMHO there's no more critical a driver than the mid. At that point you then match a suitable low end driver according to the other factors and add to that the efficiency. Efficiency on the woofer is important because it's the one driver you never want to pad down. That's just a waste of electricity.

So, I'm not sure I answered your questions or this will be read as more ramble but when matching drivers there are more factors to consider than how much the overlap in frequency has to be. That's not the part you start with. It can help you determine what slope might be needed. I have always made sure there was plenty of overlap so that won't necessarily factor into it.

If you're wondering how much overlap is needed for a certain slope of XO, so that you include the driver roll-off into your XO I can't really tell you. Is that what you are actually asking? If so, I apologize for wasting your time.

 

[BlueWizard]

Thank you and thank you again, but you are still in too deep. I'm not going to get hung up on all these secondary characteristics that I readily admit are vitally important in the final design.

But when I'm looking through a catalog for possibilities, and I see a woofer that goes up to 2000 and a tweeter that goes down to 3000, I'm going to immediately discount that as a reasonable combination. I'm not going to look at T/S parameters or resonate frequency until I've determined I have to workable drivers relative to their frequency response.

This seem the very first thing I need to consider. I've already set parameters for this hypothetical example; 2khz crossover point, 12db/octave XO slope, 2-way system. I'm not saying that is a good combination, I'm saying it serves the illustration nicely.

Again, when I'm simply looking through the catalog, I can't consider these secondary characteristics until I've determined that a pair of driver will work together on the most basic level; which is adequate surplus of frequency response at the crossover.

This seems to me to be a preliminary determination that needs to be made before anything else can be considered.

How much overlap in frequency response do I need at the crossover point?

That is the very first thing I am going to look at when I am choosing speakers. Since it is first, and since every maker of speakers needs to consider it, there must logically be a general rule that applies.

This shouldn't be the hardest question, this should be the most basic, preliminary, and easiest of all questions.

Regardless of all the other complex parameters, I can't even remotely consider a set of drivers until I've resolved this basic question.

This shouldn't be that hard.

Steve/bluewizard

 

[SY]

The catalog frequency response numbers are meaningless. Seriously. You will want to pay attention to the Thiel-Small parameters to see if a woofer meets your targets regarding enclosure size and bass extension, but beyond that, frequency response ranges are just absolute garbage.

You'll be more concerned with how the drivers will stitch together at the crossover point, and that will involve careful examination of measured off-axis responses. If the catalog doesn't provide that, get another catalog. Or a better supplier who WILL give you that info.

Yes, if you want to do a decent design, it is that hard.

 

[forr]

Hi BlueWizard,

---how many DB does a speaker have to drop to be consider acoustically out of the circuit? ---

I did not read every post of this thread but I enjoy to see someone having the same basic preoccupation as me. I think the following may precisely be what you are looking for :
http://www.diyaudio.com/forums/showthread.php?postid=1091254#post1091254
The nice idea (not mine) is that, using a simple setup, you can determine the necessary drop and answer the question yourself.

 

[BlueWizard]

SY, I do appreciate your efforts, but again it ignores the fundamental question. When you talk about cabinet size and bass extension, it assumes I've already picked drivers. And it doesn't matter if the stated catalog frequency response is accurate or not. If I think a pair of drivers might be compatible, I'm going to look at real published frequency response graphs, and if a pair of speakers are not compatible, I'm not going to continue to look at them at all.

Again, if I see a woofer that end at 2000hz and a tweeter that begins at 3000hz, I think I can safely say I don't have a match. If the woofer ends at 2000hz and the tweeter begins at 2000hz, by any stretch of any design parameters, I still don't have a match.

So, where do I have a match at the most fundamental level. If I can't make this determination at this most fundamental level, then I can never move on to the details.

If I have a one octave overlap above and below the crossover, is that enough? Is that enough to start digging deeper in to the other parameters to see if I can build a system? How about 500 hz overlap? How about a uniform half-octave overlap?

There is not point in going on to the other more detailed parameters if I haven't resolved this basic issue.

There must be a general guideline that lets me consider that two drivers might be combined into a possible system.

Again, if I'm considering buying drivers, there is no point in looking at the details, if I haven't or can't resolve this basic issue.

If I don't have sufficient overlap, then I can't even consider the speaker as possible. But then if no one can give me a guideline for overlap, how can I ever get to the next level of design?

Steve/bluewizard

 

[SY]

Woofers don't "end" and tweeters don't "begin" at certain frequencies.

I think you really have it backward and that's what's causing the block for you. You start with what you want the design to do, then begin seeing what combinations of drivers will do it for you. For example, let's say I want a 2 way mini that acts as a point source with limited vertical and wide horizontal dispersion. It must be under 30 liters, have a maximally flat bass rolloff (i.e., Q = 0.707), and an f3 of 80Hz or lower. NOW we have a spec and can start considering drivers. Note that the final spec doesn't consider driver size, crossovers, or anything like that- it's just a listing of what I want the final result to be. Someone else might want a different target (say, wide vertical dispersion and dipolar source, or a riper bass, or narrow horizontal dispersion or...)

With that in hand, I can plug bass driver specs into the TS formulas to narrow down which drivers can be considered. From there, I need to consider the dispersion of the driver and where the breakup modes are (NONE of which is contained in your catalog spec) to see where it needs to be rolled off and how fast. THEN the tweeter can be considered.

It's not a simple numbers game, though you badly want it to be- the frequency response at point A will not be the same as the frequency response at point B, so assigning a scalar value is technically pointless. A speaker is something that creates a spatial sound field, so its three dimensional behavior as a function of frequency must be considered, not just the BS "frequency response" so beloved of creators of line items in catalogs.

 

[BlueWizard]

Sy, I think the problem is...you know too much. I admit I'm trying to make it too simply, but you are trying to make it too complicated.

When was the last time someone walked into a stereo shop and said -

"I want a 2 way mini that acts as a point source with limited vertical and wide horizontal dispersion. It must be under 30 liters, have a maximally flat bass rolloff (i.e., Q = 0.707), and an f3 of 80Hz or lower.

I would say that is a very rare occurrence indeed.

More likely it is - 'That one looks nice. How much is it?'.

You do have a point, even in this hypothetical situation, I likely do have a vision of what it is I intend to built - a 12" standard box 3-way, a tower with two 6" woofers in a 2-way, a tower with 2 woofers in a 3-way, a simple 2-way bookshelf, etc....

But most beginners, other than a general idea of appearance, are not going to be as deep into it as you are telling me. You say model the bass driver...blah blah blah... 'THEN the tweeter can be considered'.

I'm betting that you are instantly going to rule out all tweeters that don't have a sufficiently low frequency response to match your woofer. I'm sure there are many other consideration too. But why look at the other considerations, if the tweeter simply doesn't go low enough.

So, now we have reduce the field down to tweeters who do go down low enough, and among them we are going to consider the other parameters. OK, how low is low enough?

In the other thread pointed to by FORR, it seems that when the response is down 12db we can consider that good enough. For a 6db crossover, we need two octaves of range above and below the crossover. For a 12db crossover, we need one octave above and below the crossover point. For a 24db, we only need a half octave.

So, is that the rule of thumb? If I see two speakers with a one octave overlap, can I then start considering them as possible drivers for a system?

And, I didn't just fall off the turnip truck, I'm well aware that 'response' doesn't begin and end at the rated frequency response. I have see a frequency response graph or two in my life time.

But, again, before I consider anything else, I'm going to look at RATED frequency response, because if I don't have sufficient overlap, then there is no need to even remotely consider the driver combination.

W = ...2khz
T = 3khz...

NO, this won't work.

W = ...2khz
T = 2khz...

No, this won't work.

W = ...3khz
T = 1khz...

Maybe this will work.

W = ...4khz
T = 1khz...

Probably this will work.

But where does 'probably' become 'definitely'? In fact, where does 'maybe' become 'probably'? Somebody must know or no speaker would ever be built.

Is it where the speaker is -12db from its nominal level?

Is it one octave uniformly above and below the crossover?

How can you possibly pick a tweeter in your own example, if you don't know what this guideline is?

Are you going to wrangle with all those technical parameters, then glance at a tweeter and say 'close enough'? I just don't think so. At some point, you're going to have to answer to yourself the question I'm asking. What parameters are you going to use to make that decision?

Regardless of where it occurs in the design process, this question has to be answered, and logically, those who answer are using something to base their decision on.

Steve/bluewizard