|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
72.73.83.124
In Reply to: Does a loudspeaker's time coherence matter ? posted by bunni on February 22, 2007 at 17:36:31:
...is it possible?I chose Vandersteens because, as Floyd Toole has written, the idea of time-alignment has "considerable engineering appeal." I used them for 4 years (I wrote 3 years in another thread, but I reconsidered; it's actually 4). But I've stopped using them. In the same paragraph, Floyd Toole also wrote that there isn't much evidence to support putting a high priority on time-alignment.
So is it possible? It's possible to get excellent step response, but there are tradeoffs. But there are fundamental problems that I can't think of an answer too.
1. You have to use first-order crossovers because they have a phase shift that doesn't depend on frequency. And first-order crossovers have a low slope
2. Time-alignment is achieved by offsetting the drivers by an amount equal to the phase shift. BUT THIS CAN ONLY BE ACHIEVED AT A SINGLE FREQUENCY.
3. It can also only be achieved at wavelengths (eg, frequencies) where the required offset is of roughly the same dimention as the loudspeaker: 3 inches, yes. 10 feet, no.
* Points 1 and 2 mean that you get interference effects (or, more accurately, incomplete addition, of signals coming from the different drivers, which makes it hard for a speaker to have good frequency response (though it's possible to do a very good job at a fixed distance, and if you look at the Stereophile measurements you'll se that Dunlavy managed to solve this problem, somehow.
That said, some speakers--like Theils, Vandy's, and a few others--come much closer than most other designs. which helps to illustrate an important point: IT'S A MATTER OF DEGREE. No speaker is perfectly time-aligned, but some do it better than others. I've come to believe that the most important thing is to have one driver hand off to the other in a well-ordered way.
One other point:
* First-order crossovers usually mean using drivers outside the range where they are at their best. This can mean distortion and dispersion problems--which mean that it doesn't work as well in a room as it does in an anechoic chamber.
Does it matter? The answer depends on what matters to you. If you trust well-conducted, rigorous subjective listening tests, you MUST conclude that other things matter much more--like (as someone else said) frequency response, dispersion, and distortion. But we're all different, we value different things, and we hear (or think we hear) different things.
Every manufacturer, of anything, requires a niche to sell into. Time-alignment is a very effective marketing niche (though I am NOT suggesting that the companies advocating time-alignment are insincere). It may matter to you, but the evidence suggests that it doesn't matter to most folks.
Follow Ups:
You certainly make some good points Jim, thank you.
Regarding Toole's opinion: For example, his papers about integrating subs with the main speakers tell us to place several of them in various locations and distances from you around the home-theater/family room. This shows he cares more for obtaining a smooth amplitude response (which is better than an irregular response of course), than he does transient response, which is one thing that time coherence always improves, by definition.
I have to say that in your #2 above, your second sentence is a common viewpoint that is simply wrong. What you wrote is actually true of high-order crossovers, not of first-order crossovers. Because of the way first-order crossovers change the phase (the time delay) as they crossover, the result is a CONSTANT time delay at ALL frequencies between the two drivers. This means then that there is no RELATIVE time delay occurring between those two drivers. Higher-order crossovers produce time delays between the two drivers that is always changing at every frequency, and thus can be 'right' at only one frequency. This is shown in the math of crossover filters. I can point to the 'filter transfer function' equations in an electrical engineering book, but can't do the derivative (calculus) that shows the time-delay changing with frequency!! But this is called a 'non-constant group delay.'
Can you clarify what you mean by your #3?
> > Because of the way first-order crossovers change the phase (the time delay) as they crossover, the result is a CONSTANT time delay at ALL frequencies between the two drivers. < <I'll answer this, but I'm going to be very measured because--though what I said makes sense to me, others who are clearly knowledgeable have disagreed with me in this thread. I am not a speaker builder or an engineer. I'm trained as a physicist and I think in terms of first principles--which when naively applied sometimes are too simple to apply to real-world problems. Furthermore, I've been out of research for more than a decade and earned my PhD more nearly 15 years ago--so I'm rusty. So until I can get to the point where I understand--and either agree with or can refute--their points of view I'm not going to write much more on this subject.
My understanding--which might be wrong--is that there is a constant, 90 degree PHASE DIFFERENCE between the drivers in a first-order crossover configuration. But 90 degrees of phase is equal to 1/4 of a wavelength--NOT a fixed time difference (because the speed of sound is frequency/wavelength independent the time it takes for sound to travel a quarter of a wavelength depends on HOW LONG that wavelength is). So, for example, 1/4 or a wavelength at 1 kHz is not the same distance as 1/4 wavelength at 1.1 kHz. If you offset the drivers so that they they emit a 1kHz signal so that it reaches the ear at precisely the same time, then, it seems to me, a 1.1 khz signal will have a small time offset relative to perfect time alignment.
As for my ponit #3, it's so obvious to me that if you don't understand I must have it wrong, and I'm not being facetious. I simply mean that if (say) a 100Hz signal reaches the loudspeaker from the amplifier intact, 1/4 wavelength--90 degrees of phase--is, what, two and a half feet, roughly?
So there you go. Hopefully I haven't humiliated myself too much.
"Thanks Jim. I should started by saying "there is a constant phase difference between the two drivers, which means there is no relative phase shift between the two drivers at all frequencies. Thus, the first-order circuit does not disturb the original phase differences of the fundamental and harmonics".
To think of this as time differences: For any given frequency, the complete first-order circuit injects a constant 1/4-wave-period of time difference BETWEEN the two drivers. More on this below... Thus, with that relative time difference between mid and tweeter remaining constant, a simple first-order circuit does not disturb the original time relationships of the fundamentals and harmonics.
What keeps a simple simple first-order crossover from working as it should?
Non-linear drivers (those with distortion and resonances)
Drivers with limited bandwidths
Drivers with tilted responses
Drivers with varying impedance curves (varying at each frequency, and with stroke)
Cabinet resonances, reflections and diffractions
Imperfect inductors and capacitors in the crossover circuit.
As with any man-made device, there is no perfect driver or cabinet or crossover part, only 'better'.
What should be made clear about a simple first-order crossover circuit used on a perfect mid and perfect tweeter is:
The upper range of the mid driver gets a varying offset (away from you = delay) as we go up the scale.
This changing amount of delay begins at zero down in the bass, and reaches +45 degrees at the crossover point, for a single inductor in series with the mid. We can also represent that 45 degrees as 1/8th of that wave's period, if we want to calculate the actual 'time delay' at that crossover frequency.
The bottom end of the tweeter gets a varying offset (towards you = advance) as we go down the scale.
This changing amount of advance begins at zero in the ultra-high frequencies, and reaches -45 degrees (or 1/8th of the wave's period) at the crossover point, for a single capacitor in series with the tweeter.
But how can we 'advance' time? We can't, so for this analysis, we are simply setting t=0 at say, 50kHz, and not worrying about anything else except what the plus and minus signs tell us relative to that arbitrary zero point up at 50kHz. What is really happening is that as we go down the scale from 50kHz, that tweeter's capacitor is injecting less and less delay, which can then be thought of as more and more of an 'advance'... I hope that's clear!
At the actual crossover point, the mid is delayed by 1/8-period of that crossover frequency, and the tweeter is advanced by that same 1/8-period amount, for a difference of 1/4 period.
As we go down the scale, the mid's delay goes back to zero, while the tweeter's advance levels off at 1/4-wave period advance, for a difference of 1/4 period.
As we go up the scale, the tweeter's advance goes to zero, while the mid's delay levels off at 1/4-wave-period delay, for a difference of 1/4 period.
At all other frequencies, the total difference between the two is always a constant 1/4-wave period.
Thus no RELATIVE timing change occurs between the two.
What then happens is that the outputs from this mid and tweeter sum together at your ear as if they could be replaced by ONE single driver that you and the measurement mic would believe was doing all the work at all the frequencies, with no time delays at any frequency.
In contrast, high-order crossovers inject advance/delay differences (varying phase differences) between these same two drivers, which means that there are 'frequency-dependant phase shifts' between the two drivers. Your imaginary 'one-driver source' is moving farther away or closer to you, depending on the frequency.
Any sine-wave's period is 1/(its frequency) = time
Example for a 40Hz sine wave: period = 1/40Hz = 1/40th of a second. 1/4 period of 40hz = 1/160th second or '90 degrees'.
Thanks again.
Nice job. Phase is not easy to describe in plain English, much less phase between two drivers.I would simply add that the points you made about non-perfect devices applies to all systems, not just first order. First order is just more difficult WRT taking the compromises into account.
For those who might like to see a graph, I've linked to a first order Butterworth response of theoretically ideal drivers, 90db each. Red is lowpass, green is highpass and blue is summed response.
This is the minimum-phase response, no excess-phase.
..the world would be an even better place.
I'm sorry, but it's hard not to respond to this. All of this is absurd:"1. You have to use first-order crossovers because they have a phase shift that doesn't depend on frequency. And first-order crossovers have a low slope"
Low slope, yes. The first sentence has no bearing on real drivers, but in addition there are crossovers, both passive and active and not DSP, that are transient-perfect.
The electrical transfer function of the crossover is immaterial except as it relates to how it's transfer function couples with the transfer function of the driver itself to yield the acoustic crossover from the system. Any system that uses a true first order electrical XO will in no way result in a first order acoustic output from the driver.
First order acoustic crossovers (BW1) from real drivers (2nd order bandpass devices) are minimum-phase with a group delay that is fully dependent on frequency. The only crossovers using real drivers that are not are DSP generated linear-phase types. The crossover sections of non-DSP BW1 crossovers maintain a constant relative phase relationship of 90 degrees at all points (phase quadrature), but there is a phase rotation in the summed response due to the bandpass nature of the drivers.
"2. Time-alignment is achieved by offsetting the drivers by an amount equal to the phase shift. BUT THIS CAN ONLY BE ACHIEVED AT A SINGLE FREQUENCY."
This is just absurd. Time alignment has nothing to do with any single frequency. It has to with the location of the absolute acoustic center and is relative to all frequencies.
"3. It can also only be achieved at wavelengths (eg, frequencies) where the required offset is of roughly the same dimention as the loudspeaker: 3 inches, yes. 10 feet, no."
Totally absurd.
dlr
There is only one frequency at which (to pick an offset at random) 3" is equivalent to 90 degrees of phase.
nt
..
No, it has nothing whatsoever to do with the time alignment of two drivers. Your comment about a single frequency in that context shows a total lack of understanding or the principles involved.
"No speaker is perfectly time-aligned"Well, a full-range single panel electrostatic speaker would be perfectly time aligned, would it not?
Quad introduced delay lines in the Quad 63 ??
Yes they did to simulate an expanding sphere but this would not preclude phase/time coherence as all frequencies are being delayed equally.
.
- This signature is two channel only -
And why not? If you disagree please state why.
You assume the full range, single panel electrostatic's diaphragm is pistonic, along with certain other requirements for time and phase accuracy. These are design goals which may or may not be fully realized. It likewise is not necessarily true that a first order crossover speaker with a stepped baffle is time and phase coherent.If you look at some of the step response graphs from Sphile, you will see that some of the multi way Meadowlark and Thiel speakers have much better right triangle shapes than, say, this Quad does.
However, the Quad is still better in terms of these graphs, for obvious reasons, than a high order crossover speaker. Phase shift is not an entirely black and white phenomenon - most designs concerned with it try to minimize it over the midrange frequencies.
> The Quad's impulse response on the mid-panel axis (fig.6) suggests a time-coherent presentation, with some high-frequency ringing evident. The step response (fig.7) has an almost perfect right-triangle shape, disturbed by what must be a reflection of some kind about 300;us after the initial arrival of the step, and again some high-frequency perturbations. These show up as ridges of delayed energy at 8kHz and above in the cumulative spectral-decay plot (fig.8).
snip
[a little off the topic because he's speaking of the measurements in general] > You can find my 13-year-old measurements of the ESL-63 online. Though the measurements were performed with completely different hardware to those for the '989, those graphs are almost identical to and as enigmatic as are these measurements of the ESL-989. All I can say is that the reasons for this speaker's undoubtedly superb sound quality are not readily apparent from its measurements. I hope to explore this subject in more depth in a follow-up.
http://www.stereophile.com/floorloudspeakers/720/index7.html
- This signature is two channel only -
My mistake. I was writing about multi-driver speakers. I should have been clearer.
At one time weren't you listening to them? I thought they were Mackies.
> > How do these [Mackies] handle the crossover issues? < <Short answer: I don't know. Mackie claims "onboard Active crossovers (which eliminates the need for external processing and eliminates phase anomalies), and time-correction circuitry, so the highs and lows reach the listener at the same time." But I haven't made any careful time-domain measurements on this. Certainly the use of active crossovers provides more flexibility than passive designs, but this isn't a subject I know much about. And a lot of people would tell you that active crossovers have their own problems.
I suspect that the (subjective, to me) superiority of the Mackies has more to do with a flatter in-room response and the tweeter waveguide, which helps to match the dispersion of the tweeter to that of the mid-range driver near the crossover point. But who knows? I feel as though I should add that I view these speakers as transient, though that might be due to a prejudice on my part. They are highly (and well) engineered, built to a price point in ugly (but very solid) little boxes, and without a lot of loveliness or charm. To me, their superiority to my old Vandersteens (which were Stereophile class B, full-range) is obvious, yet I find myself characterizing them to people as "making fewer mistakes"--which is to say that, though they are clarly superior to the Vandy's (at least in terms of what I value in a loudspeaker), the Mackies are not easy to love, for me at least.
The better pro gear features no-nonsense engineering at a high level, avoiding some of the mistakes made by a lot of audiophile gear--which makes them better than most audiophile stuff. But I currently believe that audiophile gear has more potential for exactly the same reason that it makes more mistakes: It's less beholden to engineering orthodoxy.
If one wants a good pair of speakers that won't mess things up, the Mackies are a very good choice, IMO.
Best,
Measurements here show pretty good time-domain performance--along with a lot else about the measured performance of these speakers. (Notice the circa +/- 2dB frequency response 30 DEGREES OFF AXIS.
Jim,the measurements in your link show the Mackie tweeter is inverted in polarity, arriving about 4 inches before the mid-band tones of their woofer, with considerable ringing just above 20kHz. This is not great time-domain behavior, even though I have been told that the reviewer, Mike Klasco, knows a great deal about speaker design. Maybe he didn't want to upset the Mackie company, who knows? What you are seeing in that impulse response is a speaker that is smoothly changing the phase between woofer and tweeter as the frequency goes up the scale. This is why the amplitude response is quite smooth and flat. Time coherent design can produce the same amplitude response, but with much more sharply-defined transient response. But this would have cost Mackie much more for the next-level-better drivers. Considering their retail pricing, that would have crushed their sales to the home recordist. I'm trying to talk a fellow musician here, Bruce Becvar (Windham Hill Artist) into some time/phase coherent designs as he re-vamps his home studio. I'll let you know his opinion on them if and when he makes up his mind, as this has been ongoing for the past few years, LOL.
.
This post is made possible by the generous support of people like you and our sponsors: