Dirty little questions...

[Bent]

1. What is baffle step compensation?
2. In steeper order crossover (4th order and higher), how is it that "bumps" and stuff like phase angle response don't get in the way of good sounding speakers?
3. Why did Paradigm increase their crossover frequency in their Monitor line to 2100 Hz from 1800 Hz when going from V2 to V3 (were they cooking tweeters, or did the manage to make their mid-woofs get better dispersion?
4. How come I always thoughgt a 6" mid woof was better than an 8" in any two-way, but am changing my mind?

I'll think of more after a couple of Rums.

[GrahamT]

I'm still a newb but I'll try to help a fellow Canuck. I just hope I dont get any wrong and incur the wrath of Evil Twin.

1. Baffle step loss is caused by the drivers firing into half space at high frequencies and then transitioning to full space at the lower frequencies. This happens as the wavelengths exceed the width of the baffle. There is a theoretical 6dB difference in output. BSC usually attenuates the upper end response of the midrange by the desired amount. Usually 3-6 dB depending on the speaker placement. There are a few ways to compensate.

2. Dont know what phase angle response is.

3. Are you sure? I have both v2 and v3 Monitors and I thought they were both at 1.8 kHz. I think if they raised the crossover point it may have been to save money on the crossover (they use wee iron core inductors and Bennic electrolytics, ewww) or because maybe the new bullet dust caps improved dispersion as you said. I doubt that the tweeters were frying, if it was possible I would have done it . I actually had two v3 tweets replaced. They both used ferro fluid.

4. 6s have better dispersion and higher break up modes usually. Also the C-C distance is smaller. The dispersion and breakups are not a big issue with some 8s if you crossover low and steep like Jon did in the M8a project.

[Bent]

# 2, I don't know what I really meant - Phase angle response is likely not the right term.
I mean that as the order of Xover increases, the drivers are more often "not in phase" whit each other, causing even more lobing then if just using a 1st order x-over
(or am I waaaayyyyy out to lunch?)

# 3, oops, I was off a bit - it seems they shot for 2000 Hz in V3, (I can confirm that my V2 Mini's are crossed at 1800 Hz, according to their spec sheet from 2 years ago.

I'd love a set of M8a's if I liked the look of the HiVi driver, maybe I will in the future, but not yet...

Edit here... (I really tend to slur my typos when I' ad a couple...)

[ThomasW]

In plain English, baffle step compensation is a form of EQ. If a speaker is placed out in the room away from the walls, there's no free bass boost from boundry loading with the wall. BSC lowers the output of the upper frequencies and makes it sound like there's more bass output.

HERE's a link to a good basic explaination

[JonMarsh]

Ah, a philosopher.

Well, I don't know that I have any real "answers" for you, but perhaps I can shed light on a couple or so.

1) Baffle step compensation is providing an electrical filtering of the response between the upper bass and midrange to take into account that at higher frequencies, a midwoofer on a speaker front panel acts like it's radiating into a half space environment- i.e., all the sound output is going forward of the baffle, but as the frequency drops and the wavelength gets longer, it wraps around the baffle, radiating in all directions. The actual baffle step effect then causes the output to be about 6 dB higher on axis in the upper range where the driver is radiating into a half space, compared with the lower range where it's radiating into a full space. The "compensation" is to roll off the upper midrange response by a compensating amount. For a typical baffle width of 12" or so, that will mean decreasing the upper range output from 200 Hz to about 800 Hz by 5-6 dB. Sometimes full baffle step compensation isn't used, because speakers are often used in proximity to the wall, which provides some compensation.

Look at the transfer function curve for the M8ta release candidate crossover which I posted today. You'll see this filter effect, before the system gets to the actual crossover frequency at 1 kHz.

2) Bumps or phase angle stuff CAN get in the way of accurate sound, but not just in high order crossovers. High order ones are typically EASIER to get everything right with, if you know what you're doing. The driver overlap is in a much narrower range, which means the tweeter doesn't have to go as low, and the woofer doesn't have to go as high. If you figure an above average performing 1" dome tweeter ought to be at -20 dB by 500 Hz, if you have a 1st order crossover (6 dB/octave), the crossover frequency will have to be 3.5 to 4 kHz to meet that criteria. Which means the midwoofer will have to be clean to over 8 kHz, say 10-12 kHz. I don't think I have to go on very long about how unlikely that is...

OTOH, with one of my strange little cauer-elliptic crossovers which emulate an 8th order Linkwitz-Riley network, (again, see the post in the Extremis thread), that tweeter may be able to be used with a 1 kHz crossover and be down -50dB at 500 Hz. That's easier on the tweeter, and only requires the midwoofer be fairly clean to 1400 or 1500 Hz, where it's already down -30 dB. Now, that's a lot easier to find...

The trick is you still have to pay attention to the transition band design, so that the phase of the output from the drivers sums correctly for smooth response.


One other advantage to this low crossover frequency, is that it's easy to meet wavelength criteria for driver spacing, so you tend to have a much wider dispersion lobe in the crossover region. What kind of wavelength criteria? Well, what I use is a simple relationship- the driver spacing should be no more than 13568/Fx, where Fx is the crossover frequency. Let's go back to that 3.5 kHz first order network- center to center driver spacing meeting that criteria would be 3.88". Ouch. How are we going to do that between our 7" midwoofer and 4" diameter frame tweeter? OK, let's look at my M8 two way system with the high order crossover and 1250 Hz crossover frequency- hmmm, that works out to 10-7/8"; which is quite a bit more feasible. Looking at those Paradigms, 2100 Hz is 6-1/2". That's doable with a 7" woofer, but just barely. It's a little higher than I would want to see.

3) Don't know. Don't know about their crossovers, components, etc. Might have been for reliability, if they have a lowish order crossover. I doubt they got their midwoofers to "improve" the dispersion. BTW, an 8" midwoofer will have pretty good off axis response up to about 1250 Hz, a 7" to about 1400, a 5" to about 2000.

4). It all depends... the problem with most 6" mid woofers is radiation resistance. They don't have much. I.E., cone area times Xmax isn't so hot. Also, because of the lower radiation resistance, and smaller magnets, they generally can't have very high Xmax or efficiency suffers too much. A lot of 8" drivers have 5-6 mm Xmax; most 6" are only about 3-4 mm. The better/best 8" midwoofers, like the Datyon RS, are 7 mm. Because it's only the cone area that radiates, 6-1/2" woofers have about half the Sd (radiating area) of 8's. So, it takes two 6 or 7" midwoofers to equal the radiating area of one 8, but then they may have as little as 1/2 the Xmax, which again means lower output, and/or higher distortion.

If you have a conventional crossover slopes with a 2 kHz or higher crossover frequency, even 7" don't work that great, and 8's are at a real disadvantage. If you have a high slope network at a lower frequency, like 1200-1400 Hz (the Linkwitz Orion uses 1440 Hz), then you can have very good midrange performance, including off axis, with an 8" midwoofer and the right crossover.

Most GOOD 7" drivers are almost as expensive as 8" midwoofers, so using 2 to get the performance of an 8 isn't generally cost effective. The Dayton RS180, which we're talking about on the MTM thread, has very good Xmax and distortion numbers, decent frequency response, and a pretty low price- about $30 each, less than half of what most decent 6-1/2" drivers cost. They're something of a bargain. The Dayton 7" is shown in this picture, of some MTM speakers I'm working on (see the MTM thread).




~Jon

[taz13]

A true turkey track, have a few rum (bacardi, palm or the captains spiced) sleep fast and go play with big sparks. gotta love it!
8O :rofl:
Jon beautiful pic, just love that boxes finish :T :T
After reading your explaination, is that why Adire advertises there new Extremis as a 6" but in reality is a 7"?

[Saurav]

Is your cauer-elliptic crossover published anywhere? Is it something that would be easy to adapt to different driver/crossover points? Or is this your IP and not publicly available information? Just curious

[cjd]

Suarav, you need to read some of the other posts here - I think the Arvo thread, and now the Extremis thread, both detail a lot of info on the CE filters, philosophy... lotta brain dumps in there!

[JonMarsh]

A true turkey track, have a few rum (bacardi, palm or the captains spiced) sleep fast and go play with big sparks. gotta love it!
8O :rofl:
Jon beautiful pic, just love that boxes finish :T :T
After reading your explaination, is that why Adire advertises there new Extremis as a 6" but in reality is a 7"?

Well, could be, but they don't advertise there 12's as an 11, or their 15's as 13's, so where's the logic?

Is your cauer-elliptic crossover published anywhere? Is it something that would be easy to adapt to different driver/crossover points? Or is this your IP and not publicly available information? Just curious

I first wrote an article on an "Unorthodox Eight Inch Two Way" which was published in three parts in the Autumn of 2003 in AudioXpress magazine; it uses that topology. In the thread on the Extremis 6, I explain and showed schematics and transfer functions, in page 3 of the thread.

Explanation of my use of Cauer-elliptic Filter in crossovers- Extremis Thread

After all the recent questions and chatting wtih THomasW, while I'm out there there next week I'll put up another sticky thread just about the philosophy and reasoning behind this crossover design approach, as I implement it in systems like the original M8, the Arvo Part, the M8ta, the Extremis two way's being designed (yes, two versions, one with a tweeter with a sane price, one with a tweeter with an insane price), and the MTM's currently being developed for HT and music. Then I won't have to answer the same questions a dozen times in a dozen places...

~Jon

[GrahamT]

Then I won't have to answer the same questions a dozen times in a dozen places...

Good idea. There is very little info out there about the application of elliptics for crossovers. The only stuff I can find is from your posts and from Jason Caudra's site.

I didn't know you were doing another Extremis design. What tweeter are you thinking of using? Believe it or not, that extreme version with the deep box is really growing on me, very unique shape.

As for the name of the Extremis6.8 I dont think Adire is trying to be deceptive. It is about the same size as this driver, Paradigm claims it's 6 1/2". You can see it fits well in the routered out recess for the Extremis here.

http://clubpolk.polkaudio.com/forum/attachment.php?s=&postid=248710

[JonMarsh]

Yes, but by Adire's own specs, it's a 7" diameter frame. But hey- what's it really matter in the big picture?


Well, I was thinking if you have an Extremis Woofer, it would also be fun to have an Extreme tweeter, and there aren't many dome tweeters more extreme than the SEAS Millenium Excel (on the right).





These have lots of Xmax as tweeters go, low distortion, and besides, they look cool, right? Just not everyone wants to pony up the $159 each, understandably. Well, I have a set, with no designated home, since my Hales are still going into the M8ta, so why not try these too?

The combo should be pretty killer, though not necessarily cost effective... I've been criticized before about combining megabuck tweeters with midpriced midwoofers, though the standard pricing for Extremis isn't exactly low, either. It's just that recently we've gotten some really nice midwoofers that perform very well below 1500 Hz, and deserve (perhaps) mating up with a tweeter that might normally be outside their league. This combination might get by with just a modified LR-4 network, given the healthy low end capabilities of the Millenium Excel. I'd probably use a Bullock compromise alignment, (see “Loudspeaker-Crossover Systems: An Optimal Choice,” J. Audio Eng. Soc, vol. 30,) though, as it sounds better than a conventional 4th order L-R due to the flatter power response. (4th order L-R, while flat on axis coincident to the tweeter, has a relatively narrow lobe in the crossover region, and the summed power response over a 30-40 degree window around the driver axis has a 3 dB dip).

We'll see....

[Saurav]

Thanks a lot. Hadn't seen that thread. Had seen many references to this magic crossover in many other threads

[GrahamT]

Those sure are pretty. The Seas is $160 in the US and almost $260 CDN. Ouch! The LPG 26T is about $40 US, and it is about $43 CDN.

[Dennis H]

Cauer question

Jon, let's assume for a moment you were programming a digital XO. I know you hate 'em but let's pretend. Let's also pretend we have perfect drivers so the electrical response equals the acoustical response. Would you get a "better" transfer function - including phase tracking and group delay as well as magnitude - using straight LR6 or LR8 filters or would you do better building a Cauer response with 4th order filters plus parametric notches in the stopbands?

[JonMarsh]

Yep. That's why I don't ALWAYS do the way over the top solutions; sometimes I try to improse a few cost constraints...

[Dennis H]

Bump, in case Jon missed my question above. I'm still trying to get a handle on this whole Cauer thing. Is the advantage just fewer components than an LR6 or LR8 with equivalent early rolloff, or are there real advantages in the transfer function?

[JonMarsh]

Jon, let's assume for a moment you were programming a digital XO. I know you hate 'em but let's pretend. Let's also pretend we have perfect drivers so the electrical response equals the acoustical response. Would you get a "better" transfer function - including phase tracking and group delay as well as magnitude - using straight LR6 or LR8 filters or would you do better building a Cauer response with 4th order filters plus parametric notches in the stopbands?

Dennis, as far as what counts, IMO, it's the first 30-48 dB of crossover transistion band, and the amplitude and phase reposnse within that region. Realistically, it may be only the first 25 dB, but I'm a little bit of a luddite, and so a skosh conservative.

BTW, your post didn't show up before my earlier response earlier today- don't know what gives with the internet relays.

Now, in principle, there is no filter performance advantage to the conept of a cauer-elliptic filter over the staight LR-6 or LR-8 if implemented in a digital crossover or other active crossover- you could even make the argument that since the cauer implementation which I uses gets funny at -50 to -60 dB, you'd be better off with something that doesn't do that...

For now, let's not dwell on the point that the parasitics of conentional passive crossover components (DCR especially) often means they get squirrelly below -25 or -30 dB. A typical problem might be a 2nd order HP network for a tweeter, that at the frequency where DCR=XL the attenuation of the inductor stops, and the circuit reverts to a 1st order network.

Let's look at the real world, though, in spite of the fact that I'm not selling product here, old habits die hard.

A passive ladder implementation of an LR-8 crossover would take a WHOLE lot of components, if you could get it to work at all. Frankly, before computer optimization, getting a 4th order L-R to work reliably with minimum component count was touch and go, unless you zobeled everthing to death to get the impedance "text book", then do a text book calc. Now, for a two way it's not too bad- but just try doing a 4th order L-R bandpass in a three way network. That's one heck of a lot of parts, just on the midrange crossover. Granted, typical practice might be to design close to the driver roll off, and rely on the driver transfer function to give you one or two of the poles or zero's you need- but then you have sonic side effects that sometimes aren't too nice.

So, being able to emulate an LR-8 with only one more reactive component per network than an LR-4 is a pretty useful thing- in a two way system, it saves six reactive components, and results in about half the insertion loss compared with trying to build a conventional passive LR-8- which realistically, would be a nightmare with standard approaches.

That's why Joseph Audio uses a very different approach also, with tranformer coupled reactive cancellation to get a high order transfer function (so called infinite slope crossovers, designed by Richard Modiferri). A conventional ladder passive filter with that degree of attenuation isn't feasible.

It's likely this technique could be used to develope a transistion band behavior with even steeper response, but as you do this, the Q at the corner, and it's sensitivity to component and load impedance tolerances becomes fairly high. I think, personally, the LR-6 or LR-8 emulation is the best tradeoff for sonics, simplicity, and stability of performance.

But hey, that's just my opinion- it would be fun if there were other folks messing around with this...

[Dennis H]

Thanks, Jon. As usual, that makes perfect sense. You need to work on that obfuscation thing if you want to be a real "guru" that nobody can understand.

[Saurav]

Thanks a lot, Jon. Now I'm trying to understand this and thinking out loud at the same time...

The difference I see with a 'straight' 4th order filter is the extra cap, which goes in parallel with the coil for the low pass, and series with the coil for the high pass. That's the only additional component, right? Or did I miss something totally?

If that is correct, could you tell us some guidelines for how to figure out the component values? For instance, are the other components the expected values for a 4th order filter, and it's the extra cap that does everything that makes this filter special? Or are there other rules of thumb for getting component values in the ballpark? I know the final tweaking will obviously require measurements/listening, so I guess my questions are more about where to start. In theory I could give Speaker Workshop a target curve and ask it to optimize the network, but that seems to be the weakest part of the software (or I just haven't figured out how to use it correctly). I often get totally wacky values out of the optimizer.

My biggest reason for being interested in this is insertion loss. As long as I retain my attachment to my SET amp, every dB I can save in the XO is very valuable to me Which directly conflicts with my desire to have high order slopes on my drivers, especially the ribbon. So your filter design is probably as close as I can get to a magic bullet for my application.

And once again, thanks a lot, this is invaluable information.

[Hank]

it would be fun if there were other folks messing around with this...

Oh if I only had the equipment and time...but I'm going to buy the Behringer mic and preamp soon, so maybe there's hope.
I've said it before (more than once) and I'll say it again: the only time I've been uber-impressed - just stopped in my tracks - by a speaker in a home, store, show, etc, was at the '03 CES in the Joseph Audion room. I've been saying it was his implementation of that Millenium tweeter, but that over-the-top crossover must have had something to do with the seamless sound. That was the most I've been impressed with a tweeter, but the total sound of that big speaker just made me think there's hope of getting accurate sound reproduction from a box.

[JonMarsh]

The difference I see with a 'straight' 4th order filter is the extra cap, which goes in parallel with the coil for the low pass, and series with the coil for the high pass. That's the only additional component, right? Or did I miss something totally?

If that is correct, could you tell us some guidelines for how to figure out the component values? For instance, are the other components the expected values for a 4th order filter, and it's the extra cap that does everything that makes this filter special? Or are there other rules of thumb for getting component values in the ballpark?

Hello Saurav,

I'm not familiar with your level of knowledge or experience, so I'll probably have to qualify my answer. That, and ask a few questions.

Do you understand the basic concepts of s-plane math and how to calculate the transfer function of a filter? Are you familiar with the standard filter classes used for crossovers (Butterworth, Bessel, Linkwitz Riley, Bullock modified all pass), and the issues/techniques implementing active or passive versions of these? Have you ever impelmented and optimized a 4th order LR bandpass crossover for the midrange in a three way system, using measured data for amplitude, phase, and Z?

Or do you have PC software for speaker design and a measuring setup? Can you measure your driver's SPL magnitude, phase, and impedance on your baffles?

IF you didn't answer yes to most of these questions, then you're probably not ready to work with this crossover type on your own without experiencing some hair pulling problems...

High order pasive crossovers, even conventional 4th order L-R, can be a bit tricky, especially for the new comer to the practice. A fouth order L-R filter can be implemented in the active case by cascading two Buttworth 2nd order filters with the same tuning frequency- that doesn't work in the passive case, becasue all the components in the filter interact with each other, and the load impedance, changing the tuning becuase the pole frequencies are changed by the variations in loading. Then, when you take it to the next step, of a 4th order L-R bandpass, things get really interesting.

One other necessary point- in many cases the goal is NOT to make an electrical filter of "X" target transfer function, but to make a net acoustical response of the combination of the driver and the electrical filter- unless the driver is dead flat in the crossover region, this is ALWAYS the design methodology you have to use. Obviously, you can't do this unless you have good data on what the driver is doing.

The reason that these questions are relevant is because all filters interact with the load impedance- and high order filters interact even more. To answer your quesiton above, no, you can't make an 8th order or 6th order L-R filter by taking a textbook fouth order filter, and adding one component. All of the componnt values have to be tuned for the transfer function target and the impedance target, and getting the Q just right in the transistion region means have a bit higher precision in component values, too- forget using 10% tolerance parts. Also, while you can get sort of in the ball park using the target optimizer section of programs like LspCAD, in the case of the cauer elliptic filter, I'm also tuning for stop band voltage ripple as well as load impedance; I still find myself doing most of the tweeaking of component values to driver data manually, and rarely using the optimizer these days. For instance, if I have the ratio of the three caps and two inductors in the high pass version just right, the impedance curve will be quite flat from just above the crossover frequency- and even if there's a tweeter pertubation on axis which the optimizer would try to correct by mis-tuning the network, my experience is that it sounds better if I optimize the filter transfer function for flat impedance, then fix anything fixable with the tweeter acoustically (diffraction control - see my AudioXpress article - speaker input Z plots, and diffraction testing and treatments.) Maybe this is where the artistry comes in, in addition to the science.


BTW, there's another practical requirement for doing your own high grade crossovers- you need to have a good LCR meter. Absolute must. Many times you'll need component values which are not standard off the shelf values. With inductors, that's not hard- just buy the next largest off the shelf value and unwind to the right inductance- OR - wind your own. But even with good formulas, winding a precision air core inductor is a bit iffy; there's no substitute for measurement and verification. (Who ever said DIY was about saving money was lying- unless you're just building a pre-designed kit)
In the case of capacitors, you have to parallel parts; rarely will you need more than 2-3% precision, but it's a good idea to use your copy of LspCAD 6 to do a component sensitivity analysis and make sure things will be OK with the worst case spread of tolerances.

Last, what are your goals for DIY, Saurav? To complete a system which you listen to for some time before considering upgrading, or to develop a real mastery of the techniques and skills for design, and become the next DIY Dennis Murphy? If the former, the wise route might be to partner up with someone with similar intersts but more electrical design knowledge, or find someone like Mr. Murphy to do the crossover design for you- he's done so for a lot of other folks, like Ellis Audio, Jim Salk, etc.

If the latter, you probably have a lot of studying ahead of you. Have you been to AudioDIYCentral?

Audio DIY Central

They have a lot of good references.

A book which might be good for you is Ray Alden's "Speaker Building 201"

Speaker Building 201

He's a high school teacher that actually teaches speaker building as part of science, and the book is lucidly written. Included are chapters on using design software and doing measurements.

While you're on Audio DIY Central, be sure you hit this page:


Dave Dal Farra's and John K's reference list


And check out Bullock and White's home page- lots of useful info and free programs.


Bullock and White


OK, last hint...

Since you're working on a three way, keep in mind you need to think in terms of what the natural response of each driver is, and how you use the network to optimize that in the system- without using TOO much network and too many parts.

For example, I don't do a lot of three ways these days (used to, but that's another story), but obviously, the Arvo Part is a three way, and funny enough, the driver responses sum pretty well without even having a crossover- and this is a dipole speaker! However, my point it, don't use any more crossover than you have to- in the case of the Arvo, it retains the cauer-elliptic on the midwoofer to tweeter transistion, because it's needed for the same reasons as in a more conventional two way system- but for the rest, the acoustical design of the system was done to minimize the amount of crossover support needed. Think about that... speaker design is more than picking a few different drivers that you think are interesting in and of themselves, and trying to figure out how to shoehorn them together...

[JonMarsh]

Oh if I only had the equipment and time...but I'm going to buy the Behringer mic and preamp soon, so maybe there's hope.
I've said it before (more than once) and I'll say it again: the only time I've been uber-impressed - just stopped in my tracks - by a speaker in a home, store, show, etc, was at the '03 CES in the Joseph Audion room. I've been saying it was his implementation of that Millenium tweeter, but that over-the-top crossover must have had something to do with the seamless sound. That was the most I've been impressed with a tweeter, but the total sound of that big speaker just made me think there's hope of getting accurate sound reproduction from a box.

Those were the Joseph Audio Peal's, right? Dual 8" woofers, and MT with Excel components. I'm sure they sound pretty righteous... but to hit the ease and SPL's you're looking for, they'd need a mini X SLAMM, say, dual Seas 10", and an MTM setup with the 7" and the Millenium tweeter. That would "rock".... :B

Oh, that would be a lot like the Mini-Maxx that Brian Bunge and I are going to build, hmmm.. ?

[Saurav]

Wow... wonderful post. Thanks a lot.

I'm not familiar with your level of knowledge or experience

It's always safe to assume I know nothing, since that's usually pretty close to the truth.

IF you didn't answer yes to most of these questions, then you're probably not ready to work with this crossover type on your own without experiencing some hair pulling problems...

That's what it looks like. I'm definitely not into this with the goal of becoming a speaker designer myself, at most I'll probably do 2 or 3 speakers in my life. Thanks for putting things into perspective. It looks like this is quite a bit beyond anything I should be attempting right now. FWIW, my answer would be 'Yes' to everything except actually having done it before. And a qualified yes to the s-plane math - been away from that for a long time.

One other necessary point- in many cases the goal is NOT to make an electrical filter of "X" target transfer function, but to make a net acoustical response of the combination of the driver and the electrical filter- unless the driver is dead flat in the crossover region, this is ALWAYS the design methodology you have to use. Obviously, you can't do this unless you have good data on what the driver is doing.

Understood. SW optimizes networks with a driver in it, and based on what I've seen it do, I think it tries to hit the target acoustic slope.

Thanks a lot for the references too. Buying a book or two and doing some reading is long overdue

[TacoD]

Hello Saurav,

I'm not familiar with your level of knowledge or experience, so I'll probably have to qualify my answer. That, and ask a few questions.

[..]

Wow, thanks for writing everything down. I always learn a lot from your posts, but I am caching up. Can you give some reference to filter techniques.

I also came across a "subtractive" serie-filter, what the heck is that?

[JonMarsh]

The second link to the Audio DIY Central forum is a page with a lot of good crossover filter design references.

Subtractive filters are types which define one function, such as the high pass, and then subtract it from the full range to derive the low pass. The above references some papers on that topic.

Series filters are just that- often just first order, occasionally higher, with the reactive elements in series for one driver shunting the other driver.

Often times to get acceptable magnitude or power summing, these networks have to be modified from the "ideal", because, in general, the drivers aren't ideal!

I don't think I've heard of a series subtractive crossover; if you can reference where the term was used, I'd be curious to see- it may be a misapplication of terminology. But maybe not.

The well known (or well reputed) diaural crossover is in reality just a type of series crossover; another well known series crossover in DIY circles is the Acoustic Reality one created by Peter Thomsen published a couple of years ago...





This one modifies the "classic" concept with the additional inductor in series with the woofer to provide baffle step compensation. It can be modified further with zobels on the woofer or tweeter, or additional padding on the tweeter. It takes a pretty robust driver set to make this crossover work well. For example, the cheapest drivers that I think might work OK in this configuration might be a Peerless 850439 combined with a SEAS 27TDFC, nominal crossover frequency around 2 kHz. Still, I think it would start to exhibit some "stress" at higher playback levels compared to a higher order network. But the higher order network would cost more money...

If anyone's interested, I'll take the driver data I'm getting this weekend and do some measurements on the 850439 and see what this would look like...

(WHAT AM I THINKING!?!? another two way 6" speaker? JUST what the world needs!)

Actually, I have a friend back east that doesn't listen to music very loud; for her, this might be perfect.

[TacoD]

I am familiar with series filters and used those filters with an additional coil for compensating the BSC. What you said it is important that you have drivers with smooth FR. I had the most succes with Scanspeak / Vifa, although not perfect from your point of view (pistonic range and such), but they sound so nice ...

I saw those subtractive series-filters in a reviews of Kharma speakers (another Dutchman) and I had no clue what they mean with it. I can't find a review right now only the following link (with a lot of buzz about series cross-overs, but that isn't my point, just the first sentence)