[Roger A. Modjeski]

Thanks for the clarification. IP backwards is PI= Propriety Information. Have you seen that one? if not, I claim it.

What property of BCS that is done in a speaker level crossover that cannot be done at line level before the amplifiers?

Could you enlighten me on how this is accomplished in speaker crossovers and how much does speaker placement in the home affect it?

[JoshK]

I am not claiming it can't be done in a passive, just that off the shelf active crossovers don't generally provide for this provision.

[Scott F.]

Here is a very good article or two that covers some of the advantages of active crossovers written by John over at The Tube CAD Journal.
Active Crossovers
Tube Based Crossovers

While I agree with Turbo regarding the DEQ and its hundreds of parts, the typical line level crossover is actually very simple. A simple first order, line level high pass (set at 100Hz) is nothing more than 10kR in series and a .1592uF to ground. Here you can get by with 1/4watt (or less) resistors and a small voltage cap (that depends on the type of circuit, tube or SS, yada yada).

Although the majority of active crossovers I've owned over the years have had fixed slopes, the Pioneer SF series came with (not only) adjustable XO points but it comes with adjustable slopes (6,12,18db). This is by far the most flexible XO I've owned. Though I haven't played with the new DEQ's, the other pro-sound XO's I've owned have not performed well mainly due to (as others have alluded) cheap parts, crappy power supplies and overly complicated designs that incorporate things like time delays, phase adjustments, limiters and a whole host of nonsense geared towards the pro-sound market. Compared to the rest of the XO's I've owned, the Pioneer is top shelf (especially after I tweaked the power supply a bit). Next would be my Audio Research EC3 tubed XO (fixed at 12db slopes). Next would have been my old Sony TA4300 (fixed slopes 12db). All of these are 'vintage' as they were designed and built in the 70's.

It's truly is unfortunate that designers and manufacturers have overlooked this market. The guys that currently make and sell active XO's have made it extremely difficult and expensive for the end user to change XO points and slopes (the DEQ being an exception of course). I often ask myself why these guys didn't (if anything else) pick up a copy of the Pioneer SF850 schematic and see just how simple it is to design a XO with adjustable points and slopes. After the basic filter the only other thing in these are simple gain stages. This could be accomplished as a simple add on board (with switching) and swapping (or removing) a couple of parts in an existing preamp design. Thats how simple it could be.

Heck for that matter (and I've stated this many other times), build a pre-amp that has a built in XO. You take your standard line out and build a high pass, fixed (or variable) filter into it then build a second active gain stage for the low pass filter. Your line stage could be tubed and the low pass gain stage could be a simple opamp (tube or FET). That gives you (at the very minimum) the ability level match the lows with the highs.

When it comes to BSC, I don't know. Sure, a speaker design could come up with some notch filter to correct for the nasty things a particular driver does when mounted in a baffle but then again, you could apply some wool felt to the baffle to knock down the diffraction, that or move the driver off to one side of the panel or both.

I realize what I just typed was a very simplistic view of a complex set of issues (and resulting compromises) so don't flame me too hard. Just remember, everything in audio is a compromise. For me, I'll take the compromises of an active XO over a passive one every time. Much of what you guys are hearing with the Emerald Physics I've been trying to push for the last ten or so years. Get rid of the passive crossovers and free your speakers. Trouble is, the vintage XO's I use have become hideously expensive in the last decade. Couple this to the fact that there are very few mfgrs that build a decent sounding, affordable and flexible active XO that its no wonder there are so few of us doing active XO's.

As I view it, active XO's are probably better suited to the DIY crowd rather than the consuming masses of audiophiles. The average guy armed with an active XO and raw speakers in a cabinet is likely to destroy something (a driver) or not set up a speaker per the mfgrs instructions. That same consumer will likely turn his sights on the mfgr and rip him a new one on the audio forums. That alone can spell death for mfgr.

Turbo (Ryan), one of these days you need to bring over a couple of pairs of your speakers, one with and one without a XO and we'll hook them up and tri-amp them. When you hear the triamped version, it will definitely open your eyes to more design possibilities.

...just my $.02

[NewBuyer]

...Here is a link to what is going on: http://www.freepatentsonline.com/7085389.html

Scroll down about 20% of the page and you will find the schematic that the patent is based on. This 3 way crossover employs 5 caps, 3 inductors, 4 resistors and 2 iron core transformers.

Something they don't mention and something I have not seen discussed is off axis performance. That's what the person next to you is hearing. Depending on the relation of wavelength (speed of sound/freqency) to cone diameter it may not be wise to quickly (120 dB/octave) change from one driver to the next. Measured on axis there will not be a problem. Measured off axis there will be sharp peeks or dips as the woofer transitions to the mid and as the mid transitions to the tweeter.

Thanks very much Roger for the link to the patent! Much better for the technology explanation, rather than just the less specific implementation info etc contained in my links.

Regarding off-axis performance, the Joseph Audio line has the widest and most startlingly convincing sweet spot of any speakers I've ever heard - in my view this is in fact a distinguishing aspect of these speakers. So to me (just as a listener), the 120db/octave crossover slope (and resulting non-ringing implementation of metal cone drivers) brings nothing but convincing audio improvements, with sound quality surpassing all other passive speakers I've ever heard. I really hope more designers will be able to simply hear it for themselves - especially so that this technology can thus be further explored, along with active implementations if possible.

In fact to my ears, this particular "infinite slope" passive implementation surpasses the Linn Aktiv approach in resulting sound quality as well.

However, if I could ever find an active speaker that sounds even remotely close to as good as the Josephs, I'd probably upgrade and sell my amp, in order to simplify my system and make it more "family-friendly".

[Roger A. Modjeski]

Scott F.,

Thanks for your sensible approach to what is becoming a hot topic. If someone tries to flame you I'll be right there with a fire extinguisher.

I started this topic because it is my belief that money spent on exotic cables, power cords and conditioners might better be applied to a second amp and crossover. For the same money, the improvement will be many fold that of a speaker cable or interconnect.

I am not speaking to the DIY crowd here. A lot of them already know line level crossovers are the way to go. The audience I hope to reach is the one that is spending thousands of dollars on cables etc. in the hope of getting a little more from their system. I get several calls a week that ask for some little tweak to get a little more sonic improvement.

Many respected speakers have 6 dB/octave crossovers. These are the easy ones to convert. Just think of all the fun you can have swapping around twice as many amplifiers. It's way more fun than playing with cables.

I agree that most of the crossovers available are for pro sound/sound reinforcement. While they do their job they don't sound so good. Pro-sound equipment is dirt cheap compared to audiophile gear.

As to driver safety consider the following. Most tweeters have far lower power handling capacity along with lower power requirements than woofers. A listener might have to use a 100 watt amplifier to power the woofer but could easily power the tweeter with 10 watts. He now has the option of having that 10 watt amp be something really nice like a singled ended triode with no feedback. Tweeters don't need high damping amplifiers but woofers certainly.

Let's consider another problem of internal crossovers. Since most solid state and some tube amps have damping factors over 10 a speaker designer is most likely going to design with that in mind. Once you have a damping factor or 10 or more there won't be much change in overall frequency response or woofer control from one amp to another. The moment you start using no-feedback amps with low damping factors e.g. 2 or 3 or 4 you are no longer listening to what the designer of your speaker was listening to. Unfortunately he can't do much about it as he would have to re-design the crossover for a damping of 2 and again for 3 and again for 4... see where this is going?

Richard Vandersteen, one of the more sensible and honest speaker makers, has often told me that he designs his speaker with the expectation that there will be some series resistance (1/damping + cable resistance) that the speaker will see. He has chosen 1 ohm and I agree that's a good number.

Likewise I design my amplifiers (the ones with feedback) for a damping factor near 10 which presents 0.8 ohms to an eight ohm speaker. Adding a few tenths of an ohm from the cables and we have one ohm. Both of us had chosen our numbers quite some time before knowing that we had chosen the same number.

Sure you still need that 100 watt amp for the woofer with its feedback required to get good woofer control (damping) but you don't have to listen to it's midrange and treble which might not be so pleasing.

Now if a cable comes loose or something horrible happens that produces a very very loud sound you are protected by the fact that the amplifier powers are commensurate with the driver's ability to handle their full power. How many tweeters have gone up in smoke when a 100+ watt amp goes nuts.


Keep in mind that I trust the speaker maker will give his customer the simple crossover points and slopes. I find it a little like going to buy a car, asking what kind of engine is under the hood and having him say..."I'm not gonna tell you"

Let me know if Turbo accepts your invitation. I would like to be there too.

[*Scotty*]

It is quite possible that BSC is a solution looking for a problem in many cases. Here is a link to a good technical discussion of the impact that the choice of baffle size and driver location have on the speakers response curve. http://sound.westhost.com/bafflestep.htm   As far as I am concerned I would like to see better implementation of electronic crossovers and their incorporation into more speaker designs.
I differ from Rogers position on their impact on the cost of the loudspeaker however. The same parameters that apply to a no compromise preamp design also apply to an electronic
crossover. Obviously electronic crossovers can be done inexpensively but the result will be no better than a mass-fi product which has no place in a Hi Fidelity system.
 The losses of information from an inexpensive electronic crossover will be on par with with speaker level crossovers. It becomes a pick your poison situation.
Scotty

[Vapor Audio]

Instead of a quote fest, Roger and Scott - you're missing the key here.  99.5% of crossovers used in commercial speakers are NOT just slopes set at specific frequencies.  Unless you design some it's hard to realize just how much response shaping goes into a crossover design to work with specific drivers.  Like Josh mentions, Baffle step comp is response shaping - and it's not obvious from looking at a schematic.  Normally BSC is done simply by 'oversizing' your woofer inductor a bit.  Notch filters are another form of response shaping, but they're anything but simple ... series or parallel, and determining the right filter Q, and keeping it all in sync with the phase and step.  You can even do things like ladder step delay filters to time align drivers with a passive crossover. 

All that passive line-level crossovers do is set a slope at a specific frequency, maybe with adjustable Q, but that's simply not enough power to achieve anything of value with most drivers.  The majority of what I do with crossovers can't be duplicated with a passive line-level unit. 

So we're right back to the DEQ, that's what I'm getting at ... ignore passive units, they're only pertinent when talking about about well-behaved paper/poly woofers and tweeters with perfectly flat response ON YOUR BAFFLE, which rarely if ever happens.

Scott, cone break-ups and diffraction are not to be confused.  Felt does nothing for a cone break-up, and if the break-up is from anything other than a paper/poly cone, you better notch it out    Even my GR M-130x's sound better with their break-up essentially notched out.  I say essentially because I don't use a notch, I just adjust the Q of my Zobel to bring down the level of their breakup significantly more.  I should throw these Seas W16 mag cones and a Fountek JP3 in a box and watch your ears bleed with you apply nothing but a XXdb slope at XXXXhz 

Roger, Transfer Functions are the only way for somebody to truly reproduce my passive network with an active one.  I've never done a network that's simply 24db at 2000hz, there's always MUCH more to it than that. 

I'm sure I missed some points since I didn't quote, sorry ... feel free to make them again 

[Vapor Audio]

Oh, and BTW I have 2 of these I use for prototyping ... but they only get me in the ballpark.

http://www.audioxpress.com/bksprods/products/k-vcb100.htm

They're pretty slick, and sound fine when listening to something through them ... but it's the 'extras' in a passive network that put the icing on the finished design.

[Jon L]

The same parameters that apply to a no compromise preamp design also apply to an electronic crossover.

Absolutely.  I've tried at least 5 or 6 active crossovers, and barring the Krell crosssover (which I don't want to afford), the Bryston is a great example of a flexible, pure-sounding active crossover.  It does not use IC's and uses discrete components with non-switch-mode PS.  Even the Bryston starts to deviate from absolute purity once the gain controls are used, so I bypass those. 

Having looked at many passive crossovers, I think people may be giving way too much credit to the thought and effort that goes into these things.  Even if a lot of thought goes into them, the various "solutions" to problems like BSC doesn't necessarily sound better subjectively.  I once achieved better subjective sound quality by removing BSC circuit from my speaker, so YMMV as always..

[Vapor Audio]

Having looked at many passive crossovers, I think people may be giving way too much credit to the thought and effort that goes into these things.  Even if a lot of thought goes into them, the various "solutions" to problems like BSC doesn't necessarily sound better subjectively.  I once achieved better subjective sound quality by removing BSC circuit from my speaker, so YMMV as always..

Take a look at my waveguide design crossover from RMAF last year and say not much went into it, or ask Bob at SP    Not trying to be a smart-ass, just sayin ...

I've never seen a speaker with a separate BSC circuit, how did you remove it?

Many speakers build in too much baffle step comp, textbook is 6db ... but in real rooms 3db usually is more appropriate. 

[Roger A. Modjeski]

Dear Turbo,

Please post a list that shows that 99.5 % of speaker crossovers are more than simple slopes.

If you can realize your transfer function at the speaker level then you should be able to also create the same transfer function at line level. Then we could enjoy the reduction of intermodulation and other problems amplifiers face when asked to handle full range signals and we can select amplifiers which have particularly good sound at different power levels.

[Roger A. Modjeski]

The same parameters that apply to a no compromise preamp design also apply to an electronic crossover.

Absolutely.  I've tried at least 5 or 6 active crossovers, and barring the Krell crosssover (which I don't want to afford), the Bryston is a great example of a flexible, pure-sounding active crossover.  It does not use IC's and uses discrete components with non-switch-mode PS.  Even the Bryston starts to deviate from absolute purity once the gain controls are used, so I bypass those. 

Having looked at many passive crossovers, I think people may be giving way too much credit to the thought and effort that goes into these things.  Even if a lot of thought goes into them, the various "solutions" to problems like BSC doesn't necessarily sound better subjectively.  I once achieved better subjective sound quality by removing BSC circuit from my speaker, so YMMV as always..

Thanks to Jon L for sharing his experience with electronic crossovers. I also agree that too much is made of speaker level crossovers. While we can assess the performance of a crossover or preamp by taking it in and out of the system, we cannot do the same with the speaker crossover that is in the box.  Let's have the ability to assess these passive speaker crossovers by comparing them to line level crossovers.

[Duke]

In my experience, a good "crossover" often is doing some equalizing as well, even if the circuit topology looks simple.  Cases where a textbook active crossover would meet the designer's goals probably exist, but not for me.  Turbo's number may not be exact, but I agree with the point he's making.

I have a copy of "Loudspeaker Recipes:  Book 1" by Vance Dickason, which goes through the design process for four simple, straightforward two-way systems.  He starts out with drivers that "on paper" go together well, but really getting it right is not a simple process.  Despite being the author of the "Loudspeaker Design Cookbook", Dickason's final circuits do not look like cookbook circuits at all. 

One might then ask, why not just make the effort to choose drivers that would work well with a simple active crossover?  The short answer is that there are tradeoffs at every turn, and fixating on this requirement would force many designers (myself included) to make major tradeoffs in areas that we believe are more important.

Also, note that effective implementation of an active crossover, analog or digital, is not trivial.  Active crossovers are a tool, and the results will be no better than the capabilities of the man using the tool.  They are not a panacea - while active (especially DSP) crossovers can often address issues that are not readily addressed with a passive circuit, they do not magically solve the problems of designing a quality loudspeaker using real-world drivers. 

Duke

[Vapor Audio]

Dear Turbo,

Please post a list that shows that 99.5 % of speaker crossovers are more than simple slopes.

If you can realize your transfer function at the speaker level then you should be able to also create the same transfer function at line level. Then we could enjoy the reduction of intermodulation and other problems amplifiers face when asked to handle full range signals and we can select amplifiers which have particularly good sound at different power levels.

Certainly you realize that essentially every crossover has some level matching.  So right off the bat you'll either have to add an attenuator, or the much worse but probably more often taken route of using amp gain (if it has one) as a volume control.  I'd say that covers the 99.5% estimate.

It's not about if you can do the same thing at line-level, any transfer function can be achieved at line-level.  But adding BSC, notch filters, or other response shaping and equalization adds complexity and components as well.  I'm speaking in the context of AudioCircle, and it's fair to say designers here don't use cut-n-dry slopes with level matching in their speakers.  The designs here are more complex and more thoroughly exectuted in the crossover.  Anything using a cone with any sort of break-up is going to require a complex transfer function, as will any irregularities in the FR of woofers or tweeters.  But sure, it can all be done line-level.

The point I keep trying to make is essentially everyone in the speaker industry does it one way for their own reasons, but you're advocating a different method.  For that reason the burden of proof lies on you, not to spell out theoretical advantages, but to show us the proof.  I think this site and myself included is very open minded, we're always looking for a better way.  I have yet to see a product that can replace crossovers and realize the advantages you state, which are both audible benefits and cheaper to implement.  If it did exist I'd be the first person in line to give it a run.

[Vapor Audio]

In my experience, a good "crossover" often is doing some equalizing as well, even if the circuit topology looks simple.  Cases where a textbook active crossover would meet the designer's goals probably exist, but not for me.  Turbo's number may not be exact, but I agree with the point he's making.

I have a copy of "Loudspeaker Recipes:  Book 1" by Vance Dickason, which goes through the design process for four simple, straightforward two-way systems.  He starts out with drivers that "on paper" go together well, but really getting it right is not a simple process.  Despite being the author of the "Loudspeaker Design Cookbook", Dickason's final circuits do not look like cookbook circuits at all. 

One might then ask, why not just make the effort to choose drivers that would work well with a simple active crossover?  The short answer is that there are tradeoffs at every turn, and fixating on this requirement would force many designers (myself included) to make major tradeoffs in areas that we believe are more important.

Also, note that effective implementation of an active crossover, analog or digital, is not trivial.  Active crossovers are a tool, and the results will be no better than the capabilities of the man using the tool.  They are not a panacea - while active (especially DSP) crossovers can often address issues that are not readily addressed with a passive circuit, they do not magically solve the problems of designing a quality loudspeaker using real-world drivers. 

Duke

Well said Duke, and I think one of the best examples that I'm sure you'll understand is how to deal with phase and step response in your crossover.  Good designers put a lot of effort into phase and step responses, especially with getting proper summing at and near xover points ... not doing so can destroy your staging, and cause all sorts of FR aberrations.  But how are those dealt with at line-level?  I honestly don't know the answer, or even if they can be without some sort of DSP.

Lots of real-world drivers can work with simple slopes and produce decent results, but I've yet to see one that can't be improved with tweaking to take the network beyond just a simple slope. 

[doug s.]

Dear Turbo,

Please post a list that shows that 99.5 % of speaker crossovers are more than simple slopes.

If you can realize your transfer function at the speaker level then you should be able to also create the same transfer function at line level. Then we could enjoy the reduction of intermodulation and other problems amplifiers face when asked to handle full range signals and we can select amplifiers which have particularly good sound at different power levels.

i am w/roger here.  whether it can be done cheaper, i am not sure.  certainly at the same cost, or not much more of a premium.  

but, i think turbo has a point, for mfr's - manufacturing speakers w/outboard electronic crossovers makes for the need for end-user input - which can result in worse sound, as well as better sound.  for tinkerers, it's great, but for those who expect a plug & play product, it could be less-than-great...

doug s.

[Duke]

Thanks, Turbo. 

At the risk of oversimplifying, designing loudspeakers is pretty easy; its doing a really good one that's the trick. 

[Roger A. Modjeski]

Dear Turbo,

Please post a list that shows that 99.5 % of speaker crossovers are more than simple slopes.

If you can realize your transfer function at the speaker level then you should be able to also create the same transfer function at line level. Then we could enjoy the reduction of intermodulation and other problems amplifiers face when asked to handle full range signals and we can select amplifiers which have particularly good sound at different power levels.

Certainly you realize that essentially every crossover has some level matching.  So right off the bat you'll either have to add an attenuator, or the much worse but probably more often taken route of using amp gain (if it has one) as a volume control.  I'd say that covers the 99.5% estimate.

It's not about if you can do the same thing at line-level, any transfer function can be achieved at line-level.  But adding BSC, notch filters, or other response shaping and equalization adds complexity and components as well.  I'm speaking in the context of AudioCircle, and it's fair to say designers here don't use cut-n-dry slopes with level matching in their speakers.  The designs here are more complex and more thoroughly exectuted in the crossover.  Anything using a cone with any sort of break-up is going to require a complex transfer function, as will any irregularities in the FR of woofers or tweeters.  But sure, it can all be done line-level.

The point I keep trying to make is essentially everyone in the speaker industry does it one way for their own reasons, but you're advocating a different method.  For that reason the burden of proof lies on you, not to spell out theoretical advantages, but to show us the proof.  I think this site and myself included is very open minded, we're always looking for a better way.  I have yet to see a product that can replace crossovers and realize the advantages you state, which are both audible benefits and cheaper to implement.  If it did exist I'd be the first person in line to give it a run.

Turbo,

Are you saying that you can provide attenuation for a driver that is better than a volume control preceding the amp that would drive that speaker in a multi amped system? There is no need for amp gain. Just turn down the one that is too high. Most systems have too much overall gain as evidenced by systems where the volume control cannot be turned past noon without requiring earplugs. A little attenuation here and there would be a blessing.

Level matching in a crossover implies turning audio power into heat. Are you proposing that is a good thing to do?

I don't see how you have covered 99.5% I would still like to see the list.

Are you able to create notch filters, response shaping and equalization without adding components?

How do you compensate for cone breakup in a crossover?

One last question. When the whole system is put together what would be the most advantageous setting of the volume control that produces the loudest desired level?

[Duke]

In my opinion, it only makes sense to solve those problems that are of audible significance. 

As to whether level-matching is accomplished by a line-level potentiometer or a speaker-level resistor network, I don't think there's an audibly significant difference.  I take advantage of the extra efficiency in one of my drivers to generate a smoother impedance curve, which is of audible significance with SET and OTL tube amps.

Response-shaping can be implemented actively or passively, and if done well either will produce good results.  Cone break-up is addressed by good system design, and it's not obvious to me that one type of crossover has an advantage over the other in this regard (well, DSP might have some advantages here, but not that much if the system design is good to begin with).

While the finest system I heard at RMAF 2007 and CES 2008 (the Sonicweld Pulserod) used a hotrodded DSP crossover, achieving that system's level of transparency is not trivial and so I am not sure that it would be cost-effective in most situations.

In my opinion, the primary hurdles to good loudspeaker performance lie with their acoustic properties, and the acoustic properties of the types of speakers normally used for home audio are not something that you can significantly change by switching to an active crossover.  Acoustic problems call for acoustic solutions.  There may be some room for improvement in using an active crossover (at what additional expense?), but in my opinion that's not where the biggest problems lie and that's not where the biggest problems are going to be solved. 

[JLM]

Not a "lab" sort of reply, but to give perspective to what we're discussing:

A few years ago I auditioned Paradigm Studio 20s (typical small two-way standmount, $800/pair) versus Paradigm Active 20s (almost the identical speaker in active form with the amps built onto the back of the cabinet, $1600/pair).  It was truly an audio epiphany.  The differences were staggering, jaw-dropping, and mind-bending.  Dynamics were greatly enhanced, frequency response seemed flatter, and bass extension was unbelievable.   The differences were not wasted on bystanders in the shop that day either.  I've only had three or four audio epiphanies in my 30+ years of listening and this was the greatest by far.

As mentioned above, all this helps to explain the attraction to single driver speakers (another one of my epiphanies).  Although single drivers side step all the crossover issues, they are still in need of BSC and zobels.  My solution (after wifey gets back to work) is to use a Behringer DEQ2496 as DAC, EQ, BSC, and zobel.  This would allow the near ideal connection between my monoblocks and drivers.