Step function tests seem to show all of this very well. Stereophile performs these tests on speakers, but they fall down in not drawing any conclusions from the results.
"This speaker simply can't pass a transient correctly, therefore it's crap, no matter how much a listener may like it." :-) |
The lobing is actually due to having 2 spaced sources of the same signal at the same frequency. It isn't really caused by the crossover, but by the drivers. (Just being pedantic here.)
I don't see directivity patterns (lobing) as a huge problem - provided that the designer is aware of it and communicates with the user.
As long as you know how the speaker is to be used and the proper listening axis, it isn't really very necessary to worry about directivity. (As long as your listening room meets some minimum requirement, which I'll discuss later.)
The ultimate is omni-directional speakers, with Bose being a good example. They design for uniform frequency and power response throughout the room. In doing so, they utterly trash the signal. They can't accurately reproduce a transient.
Siegfried Linkwitz is also very concerned with frequency and power response. But he ignores time-related (phase) elements. His designs are highly engineered, but I doubt they are very accurate. Hmmm... I guess I would say they are good designs, because I respect all the work and thought he's put into them, but I just don't agree with the compromises he has chosen.
Mithat Konar has some interesting comments on the MTM designs. http://www.birotechnology.com/articles/VSTWLA.html
I don't agree with Mithat on a lot of things, but he also puts a lot of thought and effort into his work.
Would it be best to have flat frequency and power response throughout the room? Probably. But there seem to be more important things to take care of first. Pat McGinty said that once you ensure your design passes transients correctly, the frequency and power response fall right into line.
Stereo requires a fairly limited listening position. You have to be equidistant from both speakers. It starts getting into psycho-acoustics, and that's complex and not something I fully understand, but I know a few general things. You need to accurately reproduce the amplitude and time spectrum of the original signal. That requires phase-coherent speakers.
Phase-coherent isn't the best term. Time-aligned isn't either. Minimum-phase? That gets a bit closer I guess. Maybe I'll just stick with phase-coherent... it gets the idea across without having a connotation of a certain method like time-aligned does.
At any rate, accurate loudspeakers don't really need to worry too much about the rest of the room, just the listening position. If there are major problems with the room, then that needs to be solved separately. I you can't remove reflected sound in the room, then you need to get enough delay so you won't smear transients. (I've seen several papers that detail the amount of delay needed to ensure that a reflection is not perceived as part of the original signal.) You also need some amount of attenuation, from distance or damping material in the room.
If you live in the right area, you can take your stereo system outside and hear it without all the room reflections and resonances. This can be quite revealing.
Where there is disagreement is in whether you design to minimize room effects, or allow for them or even exploit them. A lot of time is spent on this. There is probably room for different philosophys here, because some people just may not have a good room to listen in, and designs that work with this may at least give them decent sound, if not accurate sound.
A typical ballroom is very reflective. You can certainly hear the echoes. But, the delay is great enough that you hear them as echoes, not as part of the original sound. (The distance also reduces them in amplitude.)
A small bare room has little delay between direct and reflected sound. It's probably the worst place to listen. Bose is probably correct in this type of room because you're not going to hear transients properly anyway, so you might as well try to have even frequency response and get some part of it right. :-)
Listening outdoors gives you long delay, plus significant attentuation, without being so anechoic that it's disturbing. (I feel that when you're in an anechoic chamber, you can't hear where you are or what type of environment you're in, and this is what is disturbing. It's very unnatural and not like the environments we evolved in or are used to.)
Most people can't listen outdoors, so they need to ensure their listening room is as large as practical, and also well-damped. (Damping or room treatment can be said to increase the apparent size of the room.)
Phase and transient problems seem to be on the "leading edge" of things, with room problems being on the "trailing edge." I feel that it's most important to have an accurate "leading edge," and then you can do whatever you can to improve the "trailing edge." I think Linkwitz explicitly says the trailing edge is most important.
Is phase audible? The studies I've seen haven't insured that there was any "control," in other words there wasn't phase-coherency to begin with. They just took a system with unknown phase properties, then added phase changes to it and asked people if they could hear a difference. There was one Master's thesis on this topic that I saw that was so flawed it's a wonder they let the guy get away with it. I hope to find a more scientific study at some point...
I have found that designs that attempt to insure phase-coherency sound more lifelike than ones that don't. So I'm convinced, even if others aren't. :-) |
Placing the speakers well out in the room should reduce early reflections. (More distance from a reflective surface equals more delay and more attenuation.) That fits in with the idea of wanting to separate the reflection (echo) from the original event so that your ears/brain will hear them as two separate events and not one muddled-up event.
There are definitely people who don't agree with this approach. Some feel that you're going to have reflections, so you should get your speakers close to the walls so that the direct and reflected events blend together. I think Ted Jordan claims that it's best to have speakers mounted in the wall, and I think this is why he says that.
I'm not positive that phase-coherency leads absolutely to pinpoint imaging. The source to microphone to loudspeaker process has some inherent flaws. Binaural and transaural methods attempt to correct this. (With some success IMO.)
I think it's quite possible that phase distortions could result in the type of soundstage that all the reviewers love to describe. "I could hear that the violinist had a rash on his left elbow. Not the right, but the left..." :-)
I have more to learn about this before trying to come to conclusions.
I do remember hearing a remarkable soundstage from a pair of Acoustats a long time ago. But panel speakers are hardly phase-coherent (nor very linear), so there's something more going on here.
I also once heard a demo of signal processing, played on a boombox, where a helicopter moved toward you from the right front toward your rear left, and seemed to pass overhead as it went! It opens up intriguing possibilities, such as listening to a performance in a "virtual" concert hall and being able to choose where in the hall you "sit." |
I don't think I've ever seen a perfect driver, let alone a perfect full-range driver.
Given the limitations of current materials, I think that it's possible to do less damage by using multiple drivers, even though you need x-overs.
The end result seems to be better than what you get with single-driver speakers. Even the most savvy (IMO) of the single-driver designers (Ted Jordan) doesn't really recommend single-drivers. He recommends line arrays of mid-tweeters, along with subwoofers. |
Just to add to the info on vector diagrams, there's a very good explanation of what's going on at the Rane site, particularly in http://www.rane.com/note119.html
The most telling part of that is "The 1st-order case is ideal when summed. It yields a piece of wire. Since the responses are the exact mirror images of each other, they cancel when summed, thus behaving as if neither was there in the first place. Unfortunately, all optimized higher order versions yield flat voltage/power response, group delay or phase shift, but not all at once. Hence, the existence of different alignments and resultant compromises." |
I gave the source as an explanation of vector addition. Then since the thread is about 1st-order crossovers, I included a quote that shows the superiority of 1st-order crossovers. There can be no argument with that. "It yields a piece of wire."
You might argue that some drivers won't tolerate 1st-order crossovers. Ok, that's valid. You can then either look for "better" drivers, or you can compromise with a higher-order crossover.
Lobing? If you're looking for certain types of directivity or power response, then that could be a valid concern. I would certainly look with favor upon a high-order L-R crossover for sound reinforcement use, and this is one of the reasons.
There is much concern with flat frequency and power response. I'm a bit skeptical about their importance. (I feel there are other more important problems. Besides that, to paraphrase Pat McGinty, "Once you solve the transient response, power and frequency response fall right into line."
A number of people have suggested ways to test whether phase coherence is audible, and there are indeed studies with conflicting results.
Rane's suggestion of passing a signal through a 4th-order L-R crossover and then summing the output makes sense. (Linkwitz suggests a similar thing.)
But, in order for the test to be valid, we need to play the summed signal via a transducer with no phase distortions of its own.
A panel speaker would seem to be out, since it has widely spaced sources of the same frequency. In the near field, it smears transients.
Speakers with high-order crossovers are out because they're doing the very thing we're trying to test for and there would be no possibility of a control in the experiment.
Speakers approximating a point source, with 1st-order crossovers might be suitable, but I feel the best bet is headphones. It completely removes any phase distortion due to the room as well.
I do plan on performing this test at some point, but at present I can only say that I find the sound of speakers designed to be time and phase correct to be more realistic than those that are not correct.
Looking at this logically, we can say that a speaker that can pass a transient is better than one that can't. We expect the same from other components in the chain.
The compromises needed to build a speaker that will pass transients is where argument arises. That, or the compromise of building a speaker that won't pass transients. :-)
As for "Infinite Slope," I question whether it is beneficial. What is the phase and time behavior like? Don't sharp filters like these ring? Do most drivers actually need such steep slopes? What kind of load does it present to the amplifier?
The NHT Xd would seem to offer more of an "Infinite Slope" than JosephAudio does. :-)
If one is going to go with steep slopes, then the approach taken by DEQX seems attractive. It corrects some of the problems.
As for the intent of the Rane article, for their purposes the high-order L-R alignment is ideal. They sell such crossovers, so their intent was to sell more of them. However, they certainly cannot say "It yields a piece of wire" about their crossovers.
I did not quote out of context, because I did not change the meaning of what I quoted by quoting only it. As for the author's intent, I don't actually care a fig for his intent. The quote I made stands alone.
The rest of the article deals with: "Are 1st-order crossovers more accurate?" "Yes, but..." The author's intent lies within the "Yes, but..."
Would I choose Rane crossovers for sound reinforcement use? You betcha. I wish I had had them. They offer a superior product for that use. |
There's some disagreement on the benefits of series vs. parallel. I feel series is superior, but I'm not the last word on things. :-)
You can find a fair bit of info on the net about series and parallel crossovers. Rod Elliott has some good info, for example. http://sound.westhost.com/parallel-series.htm
I don't see that you can say that either the W/M or M/T interface is more important. Both will affect the operation of the mid driver, and that's where your ears are most sensitive.
I don't know enough about the Heil to render an opinion, sorry. |
Rod puts a lot of thought into the articles he writes. :-)
I feel series crossovers are better than Rod makes them out to be, but they're not the greatest thing to ever happen to speakers - as some people claim. :-)
Bud had an article that he was making available that contained a summary of his knowledge on series crossovers. However, I was told that this (and other) articles are no longer available. Perhaps someone will make it available on the web so I can read it.
Where I disagreed with Bud (and the current Fried Products) is the claim: "Properly implemented series networks provide superior driver coherence, increased dynamic range and introduce a Doppler effect similar to live music that increases the sense of realism."
"Doppler effect?" Sounds like pseudo-science to me, and it also would qualify as distortion if it exists.
Perhaps it simply strikes me as odd because DiAural were making some unfounded claims about "Doppler" effects, but they were using series crossovers to reduce them. :-)
Nevertheless, I'd like to see some proof for this "Doppler effect" and how it "increases the sense of realism."
|
Well, a single "perfect" driver would be large in diameter so as to achieve good coupling with the air load so that it could reproduce low frequencies well. Unfortunately, that leads to problems with directivity and also possibly smearing of the high frequencies from widely spaced sources of the same signal. (Think panel speakers.)
So we'll go with a smaller driver and increase excursion. Then we start seeing problems with IM.
Since real drivers aren't perfectly stiff, they don't maintain pistonic motion at all frequencies. (Real drivers also have mass, and that changes things too.)
So, a cone driver will "want" to become a smaller cone at higher frequencies, and we start seeing breakup modes. This can be damped to some extent by the surround.
Ted Jordan explicitly allows for this behavior. He claims to control it in his drivers. But he's still using metal cones, and there's going to be a nasty breakup mode. the only question is how well controlled it is.
I suspect that some "full-range" drivers actually use these breakup modes to increase output at high frequencies and provide some impression of treble. Not what you'd call accuracy. :-)
So what's wrong with a multi-way system with drivers that are more likely to be able to display pistonic motion throughout their passband? Crossovers and the physical spacing of the drivers. This causes some problems and makes it harder to provide an accurate "re-assembly" of the waveform at your ear.
1st-order crossovers screw things up less than steeper slope crossovers, but place higher demands upon the drivers. (But still not as high as the demands upon full-range drivers.)
Physical placement is harder to cope with, but you can control some of it, particularly since we're only listening from one point at a time.
It's almost a Catch 22. Real full-range drivers have major problems. To solve those problems we use multiple drivers, which introduces other problems. I submit that the problems with multiple drivers can be solved more easily and more fully than those with full-range drivers, given the current state of technology.
(One possible solution is to use full-range drivers at very low power levels. This works with headphones, but not so well for normal speaker systems.)
There's just no free lunch. :-) |
Gregm> "Why not use a wide-range drivers -- i.e. a 8" + supertweet, then a hefty subwoof"
Well, an 8" driver is generally going to be getting pretty far away from a pure piston at treble frequencies.
Perhaps a smaller mid-woofer crossed at 3-4K to a tweeter, and then a woofer or sub below it, crossed at maybe 100 Hz? |
Srajan doesn't quite grasp the concept either. :-) |
A 2nd-order x-over with the tweeter's leads reversed gives you an in-phase output if you're considering _absolute_ phase (polarity). The 4th-order x-over is similar.
However, you're still seeing a phase shift. If you look at the electrical input and output with something other than a simple sine wave, it becomes obvious that the output is different from the input.
The classic statement I keep repeating is "a 1st-order x-over sums to a piece of wire." This is not true of other standard x-overs.
It's possible to add a delay in an active x-over to put things closer to normal, and this is even easier when the x-over is digital.
Technologies like DEQX hold a lot of promise. It's like having your cake and eating it too. :) |
