[konut]

The 'So you think all cables are pretty much the same' thread has caused me to revisit the SP Tech site. While there I was reminded, while perusing the various white papers and explainitory pages, how Bob's excellent explainations about all things technical mirrored my own subjective impressions. This reminded me to ask about the subject of this thread.

While employed as a sound man, many years ago, I obsereved that the major sound reinforcement companies all used active Xovers in their touring PA systems. This caused me to purchase a JBL active, with plug in cards for various frequencies, that made my own modest PA sound much cleaner and transparent. After getting out of the sound business, I purchased a pair of Magnaplanar Tympany 1D 2 way planar speaker which came with the option of an external passive line level Xover. This also made quite an improvement over the stock passive power level Xover. Lately, I've integrated a Marchand XM46 passive line level Xover, 24db per octave @ 70hz, to cross my sub to my mains. This has worked very well besting both the included Xover in the sub and the inexpensive Behringer active Xover I own. I was wondering if you've ever considered this kind of arrangement with your speaker designs? I realise that this kind of  implimentation has both benifits and liabilities.

PROS: 1. Direct coupling of drivers to amp
            a. Since only part of the frequency spectrum is being amplified, 1 amp per driver, this results in greater headroom(transients!)
            b. Easier impedance load for the amp
            c. Offers specific amp tailored to specific driver
         2. No active circuitry
            a. No power supply
            b. Eliminates the need for opamps or discreet amplification
            c. Simpler circuit design with less componentry
         3. Cheaper component costs

CONS: 1. Requires twice as many amps
          2. More interconnects
          3. More complicated wiring
          4. Insertion loss

In a thread awhile back, I had a back and forth with David Ellis about the same points. He made the accurate assertion that, compared to years ago, passive power level components, inductors, capacitors, etc, had improved markedly in quality and that little was to be gained by going line level. I would be curious as to your thoughts on the subject, Bob. If you have addressed this topic, either on your website or in a thread, please post a link. It is not my intension to   

[Aether Audio]

konut,

Thanks for writing!

This is a very interesting development.  Check out my comments a while back on the subject:

http://www.audiocircle.com/index.php?topic=35532.0

I was not aware that Marchand made a PASSIVE crossover.  That could be really good.  Of course, it depends on the component and internal wiring quality of the device.  Personally, I would think that potentially, the best implementation of such a device would include active buffer stages both before and after the passive crossover section.  Using such would insure that the input and output impedances "seen" by the passive section would be constant and not influenced by the outboard gear connected to the unit.  If very high quality buffer stages were used, the device would/should be essentially transparent.

But even without such buffering, as long as the driving gear and amplifiers used were well designed with low output and sufficiently high input impedances (respectively), it seems to me the Marchand could be just the "ticket." 

In the end, especially now that were using the superb Mundorf crossover parts, the difference between a passive line-level and a passive power-level crossover might be a bit of a wash with respect to the final sonic results.  Going with the "power-level" version you still have the option of passive bi-amping, so as a standard offering it only makes sense to continue doing so - from a marketing perspective.

Still...it is intriguing to ponder the results of connecting the output of those power amps directly to the drivers.  There's no question that a measurable difference with respect to damping factor would result.  What sonic difference that would make is of some question, but it sure would make for an interesting experiment.

As a little side note, many years ago I tried a little experiment that may shed some light.  My partner and I were in the beginning stages of R&D on our present products.  He had a pair of Infinity Kappa 8s which to be honest, I wasn't all that impressed with.  I had a Crown Macro-Reference power amp that we were using as part of our experimentation.  I made a modification to the amp and the results were clearly audible and significantly improved the sound of the Kappas.

The Crown sported a damping factor of 20,000 - an extremely high level by anybody's standards.  Crown achieved this by splitting the negative feedback loop into two sections.  For RF stability, they tapped the feedback from before the standard output terminator circuit (coil,cap & resistor) that's usually in series with the output terminals.  This loop operated from about 7KHz on up.  The second loop was tapped directly off of the output terminals and operated from 7KHz on down to DC.  That's how they achieved such a high D.F. - no terminator circuit in the path to reduce it.

Well...ol' "braniac Bob" thought..."What if I extend that L.F. loop down the length of the speaker cables and tap right off of the loudspeaker terminals.  Now mind you, this was fraught with potential hazards.  While the amp wouldn't oscillate and fry because the internal H.F. loop was still solidly connected, if the L.F. loop "opened up" (due to a bad connection) the amp would loose all DC stability.  The moment that were to happen, the output would "slam" to one of the DC rail voltages and dump its full DC current capability through the speaker.  Can you say "vaporized woofer voice-coil?

So...being the "smart guy" that I am, I made darn sure that the loop was solidly connected via an ohmmeter before we turned on the amp.  There was no "woofer cone launched into the room," so all was well.  Then we listened.  WOW!  I expected the bass to get better but... much to my surprise the midrange really opened up too and became far less fatiguing than it had been before.  Obviously this was the result of achieving a very high D.F. at/near the drivers/internal crossover.  A similar result would be achieved (I would think) by using high D.F. amps and low impedance speaker cables in the line-level passive (or maybe active) crossover arrangement.

Case in point:  You never know until you try. aa

-Bob

[konut]

Thanks for the quick reply. You make points that I had failed to. If tou have never tried this approach, it could be a real ear opener for you. OTOH your point about the better Mundorf parts may indeed make this a wash. I had never considered a the buffered idea, but by being careful with matching components, in the manner you describe, I have gotten great results, naked.  The points you make in the link are exactly my experiences as well re active Xovers. Heres the link to the Marchand.
http://www.marchandelec.com/xm46.html
I'm sure with your backround you could refine this even further. I know you're busy, but I really hope you can make the time to investigate this idea. I'm not 100% sure but I think it would improve your already stellar products. Up to now my touting of this approach has met with polite yawns for the most part.

[Aether Audio]

konut,

Thanks for the link.  I noticed that they specify the input/output impedance as being 1K/10K-Ohm, respectively.  I see no problem with the input impedance as I suspect most sources could drive that with no problem.  The only thing I see that may become an issue is the output impedance. 

Some power amps have a rated input impedance of 20K-Ohm or higher.  That would be like wiring a power-level crossover intended for a 4-Ohm load to an 8-Ohm speaker.  Doubling the impedance "could" raise the low-pass crossover frequency and lower the high-pass.  If both amps used have the same input impedance it might be a "wash" with respect to the actual crossover frequency.  But combined... this could change the effective "Q" of the network.  The likelyhood is that a "notch" or dip in the frequency response at the X-O frequency would occur. 

The whole issue becomes more complicated if either of the amps have input attenuators.  Using those to match woofer/tweeter levels (as would be necessary in our case)  would really open a can of worms as the attenuators typically alter the input impedance seen by the upstream driving component.  If you knew exactly what the final input impedances were going to be, then I guess one could have Marchand build the thing to that spec.  From a marketing standpoint though, that just wouldn't work.  We'd have to have something that was stable into any impedance that would commonly be encountered.

That's why I mentioned the "buffered" implementation.  A buffered version of that device would essentially be immune to the above conditions.  Now...I'll bet Marchand would be glad to design one - especially for an OEM application.  Time may tell. aa

-Bob

[konut]

Would the buffering be passive as well? Or would it require an active circuit? If passive, would that increase insertion loss?

[Aether Audio]

konut,

In EE terms, typically "buffered" means active.  One might tend to think that adding active circuits would just become another source for distortion.  The fact is, in this case the opposite may be true.

First of all, the output buffer's input impedance could be set to something a little lower than the average outboard device's input (20K-ohm) impedance.  That would not only lower noise sensitivity but also reduce the values of the coils needed.  Doing so might make it possible to do away with the ferrite cores Marchand is using.

Second, the input buffer's input impedance could be set at a stable 20K-ohm or so and that would reduce the load on the CDP/preamp etc. that is driving it.  Doing so would reduce the potential for that device's output stage to generate any distortion.

If "unity gain" buffers were used on the input and output, those circuits could potentially be very linear and low distortion.  Gain X Bandwidth = Product (GBP) is a term used in EE to describe the limits of an active device.  Basically, it's a tradeoff of one for the other.  If you want wide bandwidth, it comes at the expense of less available gain (amplification).  If you need a lot of gain...you have to give up some bandwidth to get it.  In the case above, "unity" means a gain of X 1 or no gain at all.  Voltage in = voltage out.  If you want gain, you buy a preamp.  With no gain being used, the buffer sections could have very wide bandwidth/excellent phase characteristics and very low noise.

Now all we need is for someone to build us one.

-Bob

[konut]

This all sounds great. What type of circuit topology would the "unity gain" buffers require? To simplify things, could you power it with a (SLA) battery? How much more complicated would a balanced version be? As for someone makings us one, I don't think theres enough demand for this type of product unless someone like you could demonstrate its efficacy and promulgate it in conjunction with your exisiting products.

[Aether Audio]

konut,

Typically, these days most devices (preamps, CDPs, etc) use opamp ICs as buffers and for preliminary gain stages.  There are companies that build their own "operational amplifiers" from discrete transistors, but usually that's pretty hard to justify.  There are opamp ICs out there now by companies such as Signetics and Linear Technologies, etc., that offer extremely good performance - better than most discrete circuits could ever hope to achieve.

A unity gain stage couldn't be easier to build from these ICs either.  For a non-inverting circuit there's not much more to it than simply taking the output of the chip and directly routing it back to its inverting input.  Changing the input impedance seen by the previous stage may require the use of a couple of resistors, but that's about it.  Virtually all of the modern opamps are internally compensated for stability so even that's not needed in the external circuitry.

A balanced version utilizing an idealized "instrumentation amplifier" approach (and that not even really necessary) requires and extra stage in parallel to handle the inverted half of the input, but most opamps come as dual (2 stages on one chip) or even quad packages, so that's not a big deal either.  For a balanced device all one would need is either 4 dual chips or 2 quad ones.  For best isolation between inputs and outputs I'd recommend 4 duals, but either could potentially work just fine.

When I said that "all we needed is for someone to build us one," I actually meant SP Tech.  This would be an application specific device as making the crossover frequency user adjustable would add a fair amount of complexity.  In fact, I'm not exactly sure how you could pull that off.  In order to change frequency you could always use a switch to engage different combinations of capacitors, but that wouldn't really be ideal as doing so would alter the "Q" of the filter.  In order for the Q to remain constant you would have to change values of inductors too.  That gets pretty hairy and is not really practical. 

So...the thing would have to be designed to one frequency and who's gonna buy that?  I guess you could make it to use plug-in modules to change frequency, but that's not very pretty or convenient either.  The upshot is that an OEM would have to spec it and have it built specifically for their application.  To do that you typically have to purchase "X" number of units up front and then hope you sell them before long.  I can't see us purchasing 50 or 100 of these things and then have them sitting around collecting dust...maybe selling a system using them once every 3 months or so.  Unfortunately, that's the ugly side of "real world" business economics.  But who knows?...maybe some rich audiophile will purchase a custom made system from us and want the thing so bad that he'll ante up to have a batch built.   I'm not holding my breath though.

-Bob

PS.  I forgot.  Yes, the thing could literally be operated from a pair of 9V (15V would be better, but they're not commonly made) batteries.

[konut]

There are a lot of audiophiles that walk away at the mention of opamps. They've gotten a bad rap because, most of the time, the implementation is poor. I agree with you 100%, that correctly used, they are virtually transparent.

The only way I see any kind of demand for this kind of thing is for a device that is built specifically for the integration of a subwoofer at 70-80hz. There is plenty of resistance to that approach as well, though people that finally try it are usually blown away by how much the midrange opens up when their woofer is relieved of the low freq duties. That's how I use my XM46, with a powered sub, and the results are both seamless and transparent. This approach beats the pants off anything costing less than $2500, at which point the pro digital active xovers, with well designed analog outs, get into the fray with tremendous versatility and adjustability.There is this mistaken notion that they have to squeeze every last ounce of performance out of their overworked woofer before crossing over to the sub.

While I've got your ear, and since I'm the OP, I'm going to veer off topic.
I've become a fan of wide range drivers, sometimes referred to as full range. This is another case of people trying to maintain some kind of "sonic purity" by trying to make one driver cover 20-20,000hz. It is my contention that just as long as one avoids putting an xover near the midrange, one can avoid any phase, impedance, and time anomalies that plague so many speakers. Specifically, if one were to design a system where a subwoofer was a given, at 70-80hz, and xover to a tweeter at about 4000hz, where virtually all fundamentals of instruments cease, one could reap the benefits of a more 'unified' midrange which, of course, is where most of the music lives. As with any other approach  there are drawbacks to this kind of arrangement.

1. With a speaker designed to cut off at 80hz, its practically unusable without a sub, lessening its appeal.
2. A driver designed to work from 80-4000hz has limitations imposed. Specifically, to be reasonably efficient at 80hz it must have a fair diameter otherwise distortion, compression, excursion limits would occur. The larger the diameter, the more directional it would be at 4000hz.

Would it be possible to use a 1st order network, at 4000hz, to ameliorate the limitations of the mid driver to blend well and retain reasonable dispersion characteristics?

An xover that high would have much less power demands on it and, in this case, would benefit less from a line level network.

[Aether Audio]

konut,

I see your logic in the single driver approach and in a “better world,” crossing over at 4Khz might just be the ticket.  But…other than the directivity drawback you mentioned there is yet another.  This all involves the tradeoffs every designer is forced to contend with.  In the case you’ve outlined, the larger diaphragm has a secondary and more severe issue.  I will explain.

If a diaphragm of sufficient diameter, such that it can provide a usable bandwidth to 80 Hz while also providing a reasonably high sensitivity, were to be sufficiently rigid as to act as a “perfect piston,” it would not simultaneously achieve a linear frequency extension to 4KHz.  The materials used and the construction required to achieve “perfect piston” performance would preclude this, as the mass of the assembly would be too great.  A perfect piston driver of that diameter would reach what is termed its “mass cutoff” frequency well below 4KHz.  This is the frequency at which the mass is too great to permit the acceleration required to extend bandwidth beyond at a linear rate.  The mass cutoff is the frequency point where the frequency response begins to drop off – usually at a rate of 12dB/octave or even greater.

Therefore, the only way to extend the frequency response of a driver of such a diameter is to “sacrifice” the requirement of perfect piston behavior.  This means that the cone of the driver must be flexible such that the center (where the voice-coil attaches) is free to move somewhat independently of the rest of the cone.  Under operation, such a device exhibits the equivalent of a “mechanical crossover” wherein as the frequency reproduced is increased, the acoustic origin of that frequency moves ever gradually more to the center of the diaphragm.  So you see, there is really no “free lunch” as you still have a crossover involved, albeit a mechanical rather than electrical one.

The real downside to this is that this “mechanical crossover” is subject to many more degrees in its “freedom of movement” or action.  If you study the behavior of membranes being excited by impulse stimuli, you will discover that exceedingly complex modes of resonance and energy storage/release occur.  It is for this very reason we all go to such lengths to make our speaker enclosures as rigid and well damped as possible.  The total of these complex resonances are very “a-harmonic” with respect to the music and if neglected in design, impart a very non-musical coloration to the reproduced signal.  Any transducer, be it a large panel or a cone, that is designed to reproduce a wide range of frequencies with moderate to high efficiency is prone to this source of distortion.  In fact, they represent a “worst case scenario” for it.

Now…this is not to say that there aren’t effective remedies to minimize these effects.  Obviously, there are many examples of designs that achieve varying degrees of success in that regard.  The designer has various options regarding the balancing of rigidity, mass and damping to help in his efforts. 

At the end of the day though, it is our opinion that all such efforts require a degree of compromise with respect to the ideal behavior exhibited by that of the perfect piston driver.  It is for this very reason that we use the aluminum cone drivers in our products.  We would rather use an electrical crossover that is far more linear and exhibits far fewer degrees of action than that of the mechanical crossover outlined above.  Combined with our waveguide technology that permits such extremely low crossover frequencies to the tweeter, we believe we have developed a truly optimized solution to the many problems plaguing more conventional designs. 

Our approach not only avoids the above-mentioned source of cone break-up distortion from the low frequency driver, but also those of dispersion and compression resulting from the same.  Although considerably more complex on several levels, we feel the effort is well worth the trouble.

-Bob

PS.  Sorry…yes, that would definitely be possible – if you can find a tweeter that has the same sensitivity as the woofer and offers a flat response to at least 1 KHz.  Good luck.

[konut]

Thanks for bringing my fantasy to an end.  A most thorough explanation of why this approach is flawed. I really appreciate you taking the the time to answer my many questions and postulations. Hopefully one day I'll be able to repay you when I'm able to afford one of your outstanding products.   

[ooheadsoo]

I've enjoyed reading this thread.  Konut, keep asking questions while you've got Bob's ear    I take it they can't work carbon fiber in such a capacity yet.

[konut]

Well, the posts to view ratio is pretty good so I guess this thread hasn't been just for my own edification after all.    Its funny you mention carbon fiber. The Visaton B200s in my Omega Aperiodic 8s utilize carbon fiber blended with banana leaf fiber. The cabinet Louis builds for these is a constrained layer construction so resonances are very well controlled.There are many that love this driver in an open baffle run full range, bass supplemented, EQed, and tweetered and probably other ways I can't even imagine. I hope its not bad form to mention another manufactures product in Bobs circle. If so, I apologise. I imagine that as materials science improves, formulations will be fabricated that will address some of the weaknesses that Bob has brought up, one day bringing this kind of design to fruition.

[ooheadsoo]

Is that why the visatons are so dark in color?  I have a pair of them myself, in an open baffle.  Not bad for the price.  Anyway, back on topic.

[Aether Audio]

konut,

Thanks for bringing my fantasy to an end.

Hey…I’m not out to rain on anybody’s parade here or slam other designs.  I was speaking in terms of the absolute.  As I said, there are many designs that use an approach similar to your idea, and according to popular opinion, they work quite well.  In fact, since our waveguide takes our tweeter down to 600-700Hz, we could have easily used that very concept – with a different woofer and (probably) tweeter.  Truth be told, in all likelihood it would have even made the crossover design easier too.

But, for the very reasons I outlined previously, we chose not to.  You have to remember that across the loudspeaker market, most consumers have been comparing the performance of products – one to another.  Across the board most have more in common than not and therefore sound alike in many ways.  The definition of “insanity” is doing the same thing over and over and expecting different results.  We had reviewed these issues long before we committed to design and concluded that if we truly wanted to improve on the state-of-the-art, we had to do something different…hence the waveguide. 

For us, the problem is that to this day so few have really heard the results of our products and those that are aware (but haven’t heard) can’t really imagine that our stuff is all that much different.  Not to be bragging, but we’re not really on a level playing field with most of the competition, yet there are only a handful in this world that really know for sure.  The upshot is that unless you do something a little more extreme as we have, you’re going to get results similar to what has come before.  Certainly, that’s not necessarily bad though as there have been many speakers over the years that are considered to be quite good that have used the heretofore commonly used techniques.  It’s all a matter of degree.

The cornerstone of the problem for most designers has always been the fact that they can’t cross the tweeter over much below 2KHz and the woofer/midrange generates cone break-up distortions in the 500 – 5Khz band.  Using a midrange just “pushes” those byproducts up to a higher frequency, but then most designers feel the need to make use of the advantage it offers by reducing power demands on the tweeter, so consequently they raise the crossover frequency to the tweeter.  By doing so they loose a lot of the potential they would have otherwise gained with respect to reduced cone distortion.  In retrospect, I guess that’s still less offensive than a screaming tweeter, so it seems their reasoning is valid.  Even still, the use of a midrange is clearly not a panacea with regards to cone distortion and it adds complexity to both the crossover and the vertical polar response.  Oh, and you can bet that the midrange driver that can compete with our 1” dome tweeter/waveguide, with regards to speed and low cone distortion…is rare indeed.

The requirements for a cone-based driver to get around the above issues and provide a “nearly” full range response are many.  The Accuton ceramic drivers appear to have come the closest, but according to Linkwitz: http://www.linkwitzlab.com/mid_dist.htm they too have energy storage issues. 

I would suspect that some of this may be in the mechanical assembly rather than the cone, but I could be wrong.  If I am correct, that would be good as it is more easily corrected.  If I am wrong, that is a problem as we are right back to where we started.

How would the “ideal” quasi-full range driver be constructed?  Well, first of all we need to start with a 6-8” driver for decent L.F. extension, that has an incredibly powerful magnet/motor assembly.  That would initially provide very high sensitivity and low Qes/Qts.  Then we would need to “drag” the sensitivity down by using a moving assembly that had sufficient mass to raise Qts and lower Fs such that it provides usable bass.  Part of this mass would be used to create the extremely rigid “perfect piston” cone.  The mass cutoff frequency would then be around 10KHz, so ceramic or even diamond cones would probably be required.  We have to use mass to lower Fs (which also lowers sensitivity) because there’s a limit to how “sloppy” we can make the suspension compliance (which lowers Fs as well).  If it’s too loose, the cone won’t be properly supported and its voice-coil will eventually rub against the magnet assembly and burn out.  Otherwise, we could somewhat reduce the mass requirement (by increasing compliance) and have an even higher efficiency driver.  There are two areas that presently represent the bottleneck.  First, the availability of sufficiently strong magnets is a limitation.  Neodymium may be sufficient but the size/amount needed would be quite expensive.  Second, making diamond cones that size is a little difficult.

I take it they can't work carbon fiber in such a capacity yet.

Ooheadsoo,

I suppose carbon graphite (or virtually any other commonly used material) could be used if made sufficiently rigid.  The issue is back to the old problem…up goes the mass so down goes the sensitivity – using affordable/available magnet technology, that is.

I really appreciate you taking the time to answer my many questions and postulations.

No problem.  I usually write these essays in the morning while I’m getting my caffeine levels back up.  If it’s helping then it’s a better use of my time than zoning out to Fox News.  There’s nothing uplifting there these days anyway.

Hopefully one day I'll be able to repay you when I'm able to afford one of your outstanding products.

That day may be closer than you think.  You’ll all find out what I mean in just a few days…so stay tuned. aa

-Bob