Hi all, I was forwarded this thread by Michael Wolff, and since my speakers are being discussed here, thought it would be appropriate to throw in my 2c. Many of the above posts are correct. It is indeed possible to get good bass out of small drivers, and at the same time, as the old saying goes, there is just no substitute for cubic inches (in this case, driver displacement and cabinet volume). The answer to the original question depends almost entirely on the priorities of an individual listener. What is considered a "realistic" volume level and "deep" bass is totally different from one person to the next. Many would consider 90dB at 30Hz to be plenty deep and loud, because the vast majority of listeners will never exceed this. (It's louder and deeper than you think.) Others may not be happy unless they can achieve 110dB at 15Hz! These two things are so far apart as to make discussion utterly meaningless without first defining what you are trying to achieve. Many of the knocks on the excursion levels of small drivers are substantially solved by the new ultra-linear motor systems developed in the last five or ten years by several manufacturers ("Symmetric Drive" by Scan-Speak, etc). This results in a drastic reduction in IM distortion, and the performance achieved by the Ultimate Monitor would not be possible without this. It still can't break the laws of physics, but if used within its linear excursion limits, will give extraordinary performance regardless of volume level or frequency. The issue of the extra power required by the BOMB is more of a problem for the amplifier than for the speaker. The reason for this is that there is surprisingly little energy (on a continuous basis) in the deep bass. But when it appears, it can make very heavy transient demands on the power amplifier. The BOMB has a maximum boost of 10dB at 24Hz, which equals a factor of 10 in amplifier power. That is why I recommend relatively high power amplifiers for use with the BOMB. It's not unreasonable-- 100wpc into 4 Ohms is plenty for most applications, provided it's high quality-- but I wouldn't dream of trying to pair this system with a 3W SET amp. The real tradeoff here is that you lose nearly that same 10dB in peak output at the loudspeaker end, in order to stay within the linear excursion limits. As bad as it sounds on paper, this is not a limitation for the vast majority of listeners. Those who have heard the UM/BOMB system would likely agree not only that the system obliterates every preconceived notion they have about "small speakers", but furthermore, that for "reasonable" listening levels in "normal" sized rooms, they deliver all the volume and bass that most people will ever use, and then some. Will they deliver 110dB at 15 Hz? Not a chance. But that's why they were designed to integrate extremely well with the REL subs and similar sub-bass units. The bottom line, with this issue and many others, is that there are always compromises. Every time you decide you want to improve performance in one area, you give up something elsewhere. Adding a column of 15" woofers is a great way to make more bass, and it's also a great way to (m)uck up the entire system. There are a hundred things that are of absolutely critical importance in the design of a high-performance loudspeaker, and deep bass extension is one of them. In the design of the Ultimate Monitor, the goal was always to achieve extraordinary "real world" performance-- that is, to make a small speaker that has virtually no serious flaws when used in average-sized rooms at average volume levels. This is a far, far harder task than is generally realized, and drives the design compromises in directions that most speakers don't take. It's quite simple to design a 3-way ported box speaker that has a wide frequency response, if that's all you're after. It's another thing entirely to make it sound like real music. I would encourage anyone interested in the technical aspects of speaker design to read the two articles under "tech notes" on our website, http://www.audiomachina.com. They will (hopefully) give a better understanding of the tradeoffs and necessities in any loudspeaker design that aspires to "high fidelity". Best Regards, Karl Schuemann AudioMachina |
Hi Sean,
To respond to your main points:
1. It is a mistake to make a blanket assumption that increased excursion per se is a serious problem, provided that the linear limits are not exceeded. While in an ideal world, excursion would be kept to zero, that obviously would require a driver with infinite surface area. The point I tried very hard to make in my post, and which was apparently lost, is that there are many, many other issues in loudspeaker design, all of which are extremely important. While it would certainly be nice to minimize excursion, that automatically requires doing others things that, in the final result, are far more detrimental to the overall performance of the speaker. I will repeat: the recent development of highly linear motor systems has not eliminated, but HAS substantially mitigated the problems caused by high driver excursion. The proof is in the listening.
2. Regarding power handling, you did not read what I wrote about the BOMB very carefully. I pointed out that while there is a 10dB boost at the amplifier, there is also nearly a 10dB loss in maximum output at the speaker. It is actually somewhat less than this, but the implication is that in the final analysis, the speaker does not receive significantly more power with the BOMB than it could take without it. In fact, on a lot of music, the argument could be made that the average power delivered to the speaker is actually more limited with the BOMB in place, due to the linear excursion limits of the drivers. In addition, thermal compression effects are far more important on an instantaneous basis than on a long-term one. In other words, when one is speaking of thermal dynamic compression, it is not a significant issue to dissipate heat from the driver chassis into the outside world. Certainly the aluminum baffle helps in this regard, but the primary concern here is with effects that are measured in nanoseconds, which the baffle doesn't help in the slightest.
3. Again, as I tried very hard to point out in my original post, it is useless to discuss "deep" and "loud" without having defined those terms, because it varies so much from one listener to the next. The UM/BOMB system, properly set up in an "average" room of, say, 4000 cubic feet, will be able to achieve an average continuous SPL of 90dB on typical rock/pop/jazz/folk recordings, extending flat to about 30Hz and being 3dB down at about 25Hz, without audible strain. Now, I will make the assertion that very few people who value their hearing will ever exceed this level. It is MUCH louder than most sane individuals ever listen. However, audiophilia is a diverse crowd, and there are certainly exceptions. Again, as I pointed out in my original post, that is why the UM is designed to integrate so well with high-quality subs such as the RELs. The combination of a pair of UM's and Stentor III's, when properly integrated, is something to behold.
To Eldartford:
If you have read our website, it is clearly stated that the Linkwitz Transform was developed over twenty years ago by Dr. Siegfried Linkwitz. And I am eternally grateful to him for having published it, for it is by far the most elegant solution I have ever seen to this particular problem. The BOMB is most emphatically NOT a "band-aid", as you seem to imply, and has nothing to do whatsoever with having chosen "too small a box" (it is, in fact, exactly the right size), but rather is an integral part of achieving a level of performance that is well outside of the experience of most audiophiles.
Secondly, your statement about the ease with which the baffle can be made on a CNC machine betrays a complete ignorance about manufacturing techniques and production costs. As someone who has over a decade of experience with state-of-the-art CNC production machining and several decades of experience in advanced composite construction, I can tell you categorically that you have no idea whatsoever what you are talking about. If I had any sense at all when it came to minimizing cost (and making a profit), I never would have chosen this manufacturing approach. I would have banged out the enclosure from CNC-routed MDF or Corian or phenolic laminates, or laminated it from pieces of solid wood, or at the very least made it from extrusions and flat slabs of aluminum ala Krell, which would have been 1/10 the effort and cost (this is not an exaggeration, believe me). But this speaker was never meant to be a "me too" design. It was intended as an all-out assault on the state of the art, and its success is in large part due to the extraordinary effort put into the cabinet.
To Metralla:
Thanks so much for the kind words. It was indeed an honor to have Stan Ricker in the room, and a delight to hear his comments. One of the highlights of CES for me, to be sure.
Karl Schuemann
AudioMachina |
Hi Eldartford,
Basically, it is possible to describe the low-frequency acoustical rolloff (transfer function) of a sealed-box speaker as a second-order high-pass electrical filter. This is an "equivalence" in the mathematical sense. Because of this, it is also possible to create the inverse (transfer function) of the speaker's rolloff, in the form of an electrical filter, and apply it anywhere in the reproduction chain.
Now, this is very hard to "get your mind around" on first glance, but the result of appyling this inverse filter ANYWHERE in the chain (in our case, typically it is prior to the power amp), is identical. In other words, you have to consider the mathematical product of the transfer functions of the BOMB, amplifier, and UM all at once.
What this achieves is not only to totally flatten the rolloff out, but to exactly cancel its associated phase shifts as well. This is simply the mathematical result of multiplying a transfer function by its inverse: unity. In other words, it is a virtually perfect solution in both frequency and time domains, within the limits of tolerances on the speaker drivers and electrical components.
Now, if that is all you did, you would end up with perfectly flat response to DC and infinite driver excursion. So one must also insert a new rolloff (at a lower frequency) into the transfer function as well. The Linkwitz Transform accomplishes these two tasks in a single step, by what is known as a pole/zero transformation, using a single op-amp section. The new rolloff can be chosen for any frequency and system Q, but in this case was optimized for the best compromise between driver excursion, music SPL's, transient performance, and frequency extension. It has a -3dB point (anechoic) of 32Hz and a system Q of 0.7. This is about all that can be done with drivers of this size if one wishes to retain reasonable output capability. Fortunately, it gives a very satisfying overall result for typical listeners in typical rooms.
Best Regards,
Karl Schuemann
AudioMachina |
Eldartford,
One more comment, about your statement that "it costs too much". While it is undoubtedly a very expensive system, that is not the whole story. In my opinion, and that of many others who have heard it, it significantly outperforms any of the well-known floorstanders in the range of $20-30K, when it comes to actually recreating the full and complete illusion of "live music" in a typical listening room at typical listening levels. Viewed in this light, and if one is able to close one's eyes and simply listen, it is actually quite a bargain. Not inexpensive by any means, but still a bargain compared to other choices which cost much more but don't deliver as much musical satisfaction.
It is only the obsession with size so prevalent in the USA that keeps many people from being able to take it seriously. Somehow, they feel that their dollars should be spent buying units of mass or volume, not units of quality. I don't share that view. In fact, many of the best qualities of this system are attributable precisely to the fact that it was made as small and simple as possible. The fact that it is expensive on a $/lb or $/cuft basis is merely a byproduct of the necessities of its design philosophy. It says nothing about its $/performance ratio.
While many will question the sanity of anyone who would spend $10K+ on a speaker system (see Sean's post above), there are nonetheless many people who are willing to spend that kind of money, simply for the joy it gives them in return. We all share the same obsession; it only varies by degree. Any outsider would still shake his/her head in wonder that we aren't perfectly happy with a Bose(R) system like everyone else has.
Best Regards,
Karl Schuemann
AudioMachina |
Sean,
Sorry, I was unclear about the "loss" of 10dB at the speaker. I should have said "reduction" instead. In other words, the average music SPL that the speaker is able to deliver is almost 10dB less when used with the BOMB. This is, of course, due to the extra excursion needed with the boost. In other words, you cannot turn the volume knob up as high with the BOMBs in the system, because the drivers are now excursion-limited to a greater degree than previously. This naturally has the effect of simultaneously limiting the (average) allowable power input to the speaker from the amp. So while the spectral distribution of power is shifted, its maximum allowable value is not significantly increased overall, and may even be reduced depending on the program material. Of course, the overall system efficiency is reduced also, again depending on the program material. Hope this makes sense.
I mean no offense by my use of the term "sane". Trust me, I've always been a bit crazy myself, and have been known to dial the volume to 11 on more than a few occasions. I was only trying to make the point that not many people listen at 90dB average SPL. Most of them don't even own SPL meters, so wouldn't even know. I trust that you have one and have used it on a variety of program material, and so know what levels you listen at. (If you truly exceed 90dB average regularly, the physician in me would advise you to turn it down a little, for your ears' sake. This means you are likely running peaks in excess of 110dB, and there's no question about the long-term effects of that practice. Again, no insult intended, merely caution.)
I would also guess that if you took a wide sample of audiophiles, the vast majority of them would show a continuous average SPL at the chair of 80 +/-5 dB. I certainly don't know many folks who exceed this regularly.
Sheesh, being compared to Bose 901's. Now that really hurts.
Best Regards,
Karl Schuemann
AudioMachina |
Sean,
When you ask if I've tried them actively crossed over, I assume you mean to a sub on the low end. The answer is that I haven't. It is my opinion, having played with this stuff for a long time, that putting any crossover on the low side of the midrange drivers, anywhere but in the low bass, causes more problems than it solves. The transition from woofer to midrange is always audible.
Many people (and I would guess that you would be in this camp) would rather run subs with active crossovers on both high and low sides, primarily for the reason we have been discussing: limiting excursion in the midrange drivers. But there is another way that is superior, in my opinion.
First of all, I will state that a subwoofer should be crossed over as low as humanly possible, for all kinds of reasons. 100Hz is way too high, and 80Hz is pushing it. Given that this is the case (and I would be very surprised to hear anyone disagree with this, if they've had experience in this area and have good ears), then it makes far more sense to let the speaker itself provide the high-pass function. In the case of the UM, its natural rolloff is essentially equivalent to a 12 dB/octave Butterworth at just under 70 Hz. Of course, this only works with a sealed-box monitor, which is self-limiting in the deep bass, but of course that is exactly what I intended from the start.
This approach is exactly equivalent to an electrical solution, but with the advantage of no added electronics and their inevitable colorations. It allows perfect compatibility with the RELs and other similar subs which are designed for exactly this approach. In my experience, this is the only way of connecting a subwoofer that actually achieves the goals of making the transition sonically invisible and causing no sonic degradation to the main monitors.
Best Regards,
Karl Schuemann
AudioMachina |
C'mon guys, I was kidding, honest. It just struck me as funny, given the status of 901's as the reigning poster child for how not to make a high fidelity speaker.
Best Regards,
Karl Schuemann
AudioMachina |
And to Eldartford:
I was too quick to jump on you in my last post, got my hackles up I suppose. I don't like to think of myself as a pontificator, but I do sometimes get a bit hardheaded when it comes to trying to get people to see what I'm saying. My apologies, and I agree with your assessment that we are only differing in how we name the elements. I think of the series and parallel forms as being equivalent, since the equations are the same, but I must admit that my calling it a "woofer" inductor could raise some eyebrows.
Peace,
Karl |
Sean,
I haven't tried active crossovers, no. It's not really designed for this, because as you probably know, series crossovers only use one set of binding posts. There's no such thing as biamping. So there's only one set of posts on the speaker, and no way to easily add another one.
One of the interesting things about series crossovers is that they work like sh** unless you have everything, and I mean everything, exactly right. This means both drivers flat in frequency response, impedance, and phase, along with perfect time alignment. But once you accomplish this, it's simply a matter of punching the numbers on a calculator to get the correct values for C and L, and it will work perfectly every time. This was a revelation the first time it actually happened, but it makes perfect sense when you think about it. That, in a nutshell, is why almost everyone uses parallel crossovers. They're easy to "fudge" if you haven't done your homework. Not so with a series.
This is a long way of saying that the "other stuff" is by far the most important, not the primary C and L. And as such, there isn't much benefit to making them active. All the usual problems with passive crossovers, such as impedance and phase anomalies, have already been dealt with, and since it is only a two-way with a high crossover point, it can use small, high quality C and L, which effectively eliminates (or at least minimizes) the other argument for active crossovers.
Eldartford,
The marketing side of this project really hasn't started in earnest, but hopefully next year. So there are no dealers yet, but we do make demo systems available to people if they're interested enough to pay the shipping, and offer very good pricing on the demo systems if they decide to keep them.
Best Regards,
Karl Schuemann
AudioMachina |
Sean,
I'm aware of the advantages of actives, but again it becomes an issue of the overall design compromise. As few people would be interested in a $15k/pr monitor, only a tiny fraction of them would even take a glance if it required biamping. Besides, using really high quality parts makes a huge difference, and keeping it simple helps perhaps even more.
The tweeter power handling (from LF leakage) in a series crossover is a function of the DCR of the woofer inductor. The tweeter attenuation "shelves" at some point, and the lower the DCR, the lower that point is.
The power handling in our case isn't even close to being an issue. The inductor is small with very low DCR, the crossover point is fairly high, the tweeter is padded down, and on top of all that, the response is designed to fall like a brick below the tweeter resonance, which is more than two octaves below the crossover point. So in this particular case, it is more than bombproof. But that isn't always the case.
Best Regards,
Karl Schuemann
AudioMachina |
The terms "series" and "parallel" describe the basic topology of the circuit, and refer to the connection arrangement of the DRIVERS. Look at the diagram link from Sean above for the first-order "series", and note the difference in driver connection compared to a "parallel".
But regardless of whether it's "series" or "parallel", the main inductor is still in series with the woofer, and the main capacitor is still in series with the tweeter. If you don't understand how this can be, it's time to do some homework. I'll give you a hint to get you started: Always think about where the CURRENT flows at a given frequency, because that's what makes the driver move. It follows the path of least resistance, just like water.
Best Regards,
Karl Schuemann
AudioMachina |
Eldartford,
If you want to learn a new way of thinking, perhaps you would do well to listen to what I said, and not continue to insist on thinking about things the way you have for 40 years. I will give you another hint, and perhaps this time you will listen and actually do some thinking.
In a simple two-way parallel crossover, there are two possible paths which the current may follow. At low frequencies, it follows primarily one path (the one through the woofer), and at high frequencies, it follows primarily the other path (the one through the tweeter). In the crossover region, it goes some through one and some through the other.
In a simple two-way series crossover (such as that shown on our website), there are four possible paths which the current may follow. I leave it up to you to (1) figure out what those four paths are, (2) figure out which two paths it primarily follows, one at low frequencies and one at high frequencies, and (3) compare those two paths to the two paths in the parallel case.
Hopefully, after you have done so, you will begin to understand what I was trying to tell you. The analogy between electrons and water is an old one, but still invaluable.
As far as your question of "which" element goes with "which" driver in a series crossover, the short answer is: they don't. Both elements affect both drivers, as a cursory glance at the circuit will tell you.
Karl Schuemann
AudioMachina |
Eldartford,
I'm not pontificating at all. It's just that things aren't as cut-and-dried as you would like to assume. The crossover point in a series crossover is a function of both the inductor and capacitor. If you want, you can change them both, and yet still keep the crossover point the same. If you don't believe me, look it up. It's not that hard to find.
In order to carry this conversation further, you are going to have to dedicate some effort to learning the actual math behind the problem, starting with the concept of transfer functions. It is in most college junior-level electrical circuits classes.
The advantages of series crossovers are real, but again require some mathematical background to understand. The two that in my mind are the most beneficial are:
1. It guarantees a constant-voltage transfer function. This is theoretically possible, but by no means guaranteed, by the parallel.
2. It is essentially insensitive to tolerance variations in both the drivers and reactive elements, thus giving a much better chance of good performance in the real world compared to the parallel.
To this can be added the additional benefit, stated in one of my earlier posts, that the series topology forces you to do everything exactly right. While this is also theoretically possible in the parallel, for all practical purposes it NEVER happens in the real world. Almost every speaker you look at with a parallel network has grossly obvious and serious flaws, simply because the designer didn't want to spend the effort necessary to make it perfect. Thus it is a final "test" of the design, ensuring that there is no "fudging" going on and compromising the end result.
What I'm saying is that if a parallel is done exactly right, there is no reason, at least theoretically, it can't be as good as a series. But in the REAL world, it never is.
Karl Schuemann
AudioMachina |
inpepinnovations:
That, unfortunately, is not a useful contribution to this discussion. Of course semantics mean something, but when it comes to filter design, math rules, whether you like it or not. If you were to put a hundred filter design experts, none of whom spoke the same language, into a room with a chalkboard, they would still be able to communicate perfectly well on the subject of filter theory. But ask a hundred English majors who know nothing of filter theory to design you a crossover, and see how far you get.
I understand Eldartford's explanation perfectly, but what you haven't figured out is that my way of thinking is a very useful intuitive approach to figuring out any electrical circuit, not just crossovers. As such, it was my hope that Eldartford would grasp it and come to see that series and parallel are two sides of the same coin, to mix a metaphor.
It is not just an issue of semantics; it is an important equivalence rooted in (gasp! horror!) mathematics.
Karl Schuemann
AudioMachina |
Eldartford,
My apologies again for yesterday's outbreaks, it was a pretty rough day and I'm not the best at being calm anyway. A good night's rest helps a lot.
To try to explain what I was after:
In the parallel case, you have two separate filters, LR and CR, which both have Butterworth responses by default (maximally flat, Q=0.7). You can change L or C, and it will change the corner frequency but not the Q of the filter, and it does not affect the other filter at all. On the other hand, it does affect the voltage sum of the two drivers, which affects the summed frequency response.
In the series case, you have the equivalent of an LCR loop, which is a resonant circuit. (It's not the simplest form of LCR, but it's a loop nonetheless.) Now if you stick to Butterworth values for L and C, you have an equivalence to the parallel case, and everything works the same. However, in an LCR loop, the resonant frequency is singular and is determined by the product of L and C. In addition, the loop has a resonant Q which is determined by the ratio between L and C. What this means is that you can double one, and halve the other, and end up with the same resonant frequency but a different Q. So it doesn't behave the same as a parallel except in the case where you use standard Butterworth values. Also in contrast to the parallel network, the series network by its very nature maintains a constant voltage sum across the drivers, which maintains flat frequency response despite the variations in the components.
Of course, there are a lot of assumptions built into all of this, such as equal resistive drivers, equal amplitude and phase response, constant voltage source, etc., none of which are really achieved in the real world. That is why I view the series as being superior to the parallel, because it automatically minimizes the effects of at least some of these "non-perfect" conditions.
Best Regards,
Karl |
Here is a link which has a downloadable .doc file (as a .zip) by John Kreskovsky, which is an excellent primer on series crossovers. Note that the damping which I referred to as "Q" is called "zeta" here.
http://www.geocities.com/kreskovs/Series-1.html
Best Regards,
Karl |
