|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
24.117.206.83
Is any company making these any more? And why?
Follow Ups:
Hello to all,Time coherence is best described as all frequencies getting to your ear at the same moment as each was supposed to/as recorded. There are no unwanted delays across the spectrum. The 'group delay' is constant, for the group of frequencies that we call 'audible'.
For this, a two-way's woofer and tweeter must move at the same moment on all frequencies they share, and also continue on with that same timing individually as far as possible, down into the lowest bass and into the highest highs. Every tone, high or low arrives in the same temporal order as it was recorded. On a `scope, we are preserving the shape of the waveform.
Time delays in speakers are imposed by four things:
- the speaker's crossover-circuit design
- by non-pistonic drivers
- by the low-frequency tuning of any raw driver, woofer or tweeter
- by the locations of the drivers relative to one's ear.One indication of time-coherence (but not the sole one) can be seen in a speaker's specifications:
Each driver on its own, with its particular crossover circuit, has a frequency response at the listening position with an acoustic rolloff of six dB per octave (= a first-order slope) above (below) its crossover frequency.More rapid rates of acoustic rolloff, such as 12, 18 or 24 dB/oct, inject time-delays that are also different at each frequency. These varying delays thus cannot be compensated for by just moving a tweeter back nor via standard DSP calculations.
Another indication of time coherence is found when examining the impulse or step-responses of woofer and tweeter separately. At the listening position, the signal's very first step-upwards on the `scope, from either woofer or tweeter, must BEGIN at Time = 0.
Note-- here we need not be concerned with 'when' each driver's final peak occurs on a `scope, if each is operating as a perfect or nearly-perfect piston. For in such drivers, those max peaks will happen when they are 'supposed to', if each driver's output started upwards from zero at the same moment (again, at your ear). When combined, they will then also produce a single-shot square wave at best, or a right-triangle response at worst, depending how far down the scale you measure-- in other words, how wide the step/pulse across the face of the 1scope. Furthermore, each driver's 'peak' loudness on a step/pulse input must be the same, when it finally does occur, whether it's from the woofer's slower rise time to full loudness or a tweeter's faster one.
Which are all devilishly hard to measure accurately, because of:
- the interference of floor reflections arriving late at the microphone (which we don't 'hear' in the same way on music)
- the phase shift above 2kHz in all affordable measurement microphones
- the fact we need to examine the very bottom of the signal, where our S/N ratio is the poorest, to see exactly where each driver's signal first left the horizontal graticule of the `scope.
- cabinet-surface reflections, which do not show up clearly on a `scope or MLSSA file.
- the step/pulse signal must sound VERY LOUD to show up on a `scope, which threatens driver excursions and power handlings. This is an SPL/energy vs width-of-pulse issue, not about a sufficient S/N ratio.Speakers can have complicated first-order crossover circuits and produce time-coherent output. These circuits are complex in order to 'fix' drivers that do not have perfect pistons and for each driver's mechanical low-frequency cabinet tuning (woofer or tweeter).
Other notes:
One can have phase coherence without time coherence (but not the other way around). All this means is, at each crossover point, two drivers are separated by 1/2 of the wave period (of the crossover frequency) or 1 full wave period, instead of Zero (which is another indicator for time-coherence).One-way speakers have mechanical phase shifts, both high and low.
Speakers with only a capacitor to the tweeter/nothing on the woofer, have varying mechanical time-delays both high and low.
Panel speakers can have varying time-delays both high and low, for several reasons.I hope this helps!
Best regards,
Roy Johnson
Designer
Green Mountain Audio
Edits: 03/09/12
Roy said:
"Time delays in speakers are imposed by four things:
- the speaker's crossover-circuit design
- by non-pistonic drivers
- by the low-frequency tuning of any raw driver, woofer or tweeter
- by the locations of the drivers relative to one's ear."
You did leave out one very important cause of time delay that every designer has to account for in proper engineering so as to establish smooth driver integration at and through the crossover region in a multi way system :
Each driver's native or minimum phase response
Non pistonic and non point source behavior relative to cone breakup or beaming problems are usually dealt with by selecting the proper crossover point and driver for the intended function. It is not a significant player in establishing phase and time coherent speakers if the proper steps are taken up front. Not sure what you mean by "low frequency tuning" and its impact. However, as we all know, every driver (especially moving coil drivers) exhibit capacitive and inductive behavior when stimulated by time varying voltage (a music signal, for example). The resulting storage of electrical and mechanical energy in every transducer results in what is commonly referred to as a native or minimum phase response. If one is to achieve a well integrated pair of drivers at crossover, this inherent phase behavior must be taken into account - otherwise, achieving a substantial reverse null at crossover (on the order of 30-40 db indicating excellent driver integration) will be impossible.
Roy also said:
"More rapid rates of acoustic rolloff, such as 12, 18 or 24 dB/oct, inject time-delays that are also different at each frequency."
You seem to be implying that steeper roll offs result in "time delays" that can't be compensated for where this doesn't occur with first order crossovers??? If that is what you are trying to say, you're incorrect. All passive components in crossovers introduce acoustic phase changes that vary with frequency. The simple delayed discharge of current through a capacitor and resistor in series is ample evidence of that. The fundamental difference between 1st order and higher order crossovers is higher group delay at the crossover point and the level of energy storage contributing to an undamped or overdamped impulse response. More importantly, contrary to what you're implying, crossover induced "time delays" are a straw man problem. With any properly executed crossover, acoustic phase difference between drivers moving from the lowest frequency portion to the highest about a particular crossover point should be constant (in quadrature with odd ordered networks, and either 180 degrees or in phase with even ordered networks). Group delay and its audibility are another issue not really related to phase and time "coherence". All crossover networks - 1st order through Nth order - suffer from group delay. Widely acknowledged studies have been published discussing the degree of audibility with respect to group delay and frequency. That issue is separate and distinct from the traditional issues of time coherence and phase coherence in speaker design.
Hi Villa,
Thank you for your input and questions.
You wrote " ":
"Non pistonic and non point source behavior relative to cone breakup or beaming problems are usually dealt with by selecting the proper crossover point and driver for the intended function."
And I agree. However, those two problems of cone breakup and beaming remain detrimental to what we hear, because of their need for higher-order crossovers.
"Not sure what you mean by "low frequency tuning" and its impact."
In any dynamic driver, as we proceed down the scale, the driver's acoustic position begins to drift backwards in time-- moving away from us the lower we go. The cabinet-tuning creates the low-frequency resonance of any driver, woofer or tweeter, and different tunings affect both the frequency of the onset of delay and the ultimate amount of the time delay, which occurs down at resonance. In other words, the LF-tuning affects the LF group delay, starting at ~3x higher than the resonance frequency.
"However, as we all know, every driver (especially moving coil drivers) exhibit capacitive and inductive behavior ... result[ing] in what is commonly referred to as a native or minimum phase response. If one is to achieve a well integrated pair of drivers at crossover, this inherent phase behavior must be taken into account - otherwise, achieving a substantial reverse null at crossover (on the order of 30-40 db indicating excellent driver integration) will be impossible."
First, please note that no such cancellation occurs on a first-order type of speaker.
Other readers here may not know the sources of those changes in a raw driver's phase response and in its impedance curve. When those causes are understood, each can be offset or avoided. One result would be a very wide-band raw driver, with a smooth response and low distortion before any crossover is added. Only then can it be crossed-over over with a simple circuit.
"You seem to be implying that steeper roll offs result in "time delays" that can't be compensated for where this doesn't occur with first order crossovers??? If that is what you are trying to say, you're incorrect."
I am correct. Please read on, thanks.
"All passive components in crossovers introduce acoustic phase changes that vary with frequency. The simple delayed discharge of current through a capacitor and resistor in series is ample evidence of that. The fundamental difference between 1st order and higher order crossovers is higher group delay at the crossover point and the level of energy storage contributing to an undamped or overdamped impulse response."
Please know that this is a common mistake made in every electrical-engineering textbook I've seen, because the mathematics behind it was not presented from scratch using Time as the variable, not Frequency. What you'd find is that what appear to be damping issues are actually artifacts created by time delays that vary with frequency. There is an excellent AES paper on exactly this subject at
http://www.aes.org/e-lib/browse.cfm?elib=2618
One does not need a math degree to understand this paper, either. It's the best presentation I've seen on the subject.
"More importantly, contrary to what you're implying, crossover induced "time delays" are a straw man problem. With any properly executed crossover, acoustic phase difference between drivers moving from the lowest frequency portion to the highest about a particular crossover point should be constant (in quadrature with odd ordered networks, and either 180 degrees or in phase with even ordered networks). Group delay and its audibility are another issue not really related to phase and time "coherence". All crossover networks - 1st order through Nth order - suffer from group delay. Widely acknowledged studies have been published discussing the degree of audibility with respect to group delay and frequency. That issue is separate and distinct from the traditional issues of time coherence and phase coherence in speaker design."
I am not quite sure what you meant by your last sentence. I will point out the studies conclude that "We don't hear it." or "Time coherence does not seem to make much difference." That is only an indication that the studies were poorly designed, in my and our customers' and retailers' experience.
Regardless, if you think time coherence is a straw-man problem-- well, anyone is entitled to an opinion of course.
I do know that you have been given wrong information if you think group delay is not intimately related to time coherence. The math is presented quite well in many AES papers, and is unassailable, and is also in that link above.
The nice aspect about using a first-order crossover circuit is that the phase-shift DIFFERENCE between the low-pass and the high-pass filters is a CONSTANT 90 degrees at ALL frequencies, not varying with frequency as in higher-order circuits, and not just "about a particular crossover point" as you wrote above.
That word "Constant" is the clue behind the benefits of using first-order networks. Why this one word matters is far too lengthy to present here, but one can see the math for it fully laid out in many AES papers including that link above, so it's definitely not 'Roy's opinion'. And in it, you'd come to see how time-delays that do vary with frequency are then not easily corrected by any means. So, its best to avoid those types of circuits by the design choices made for the rest of the speaker system.
I hope this helps, in a small way at least!
Best regards,
Roy
Oh boy...where to begin....Firstly, with respect to "reverse nulls", Mr. Johnson said -
"First, please note that no such cancellation occurs on a first-order type of speaker. "
Not sure what kind of first order speakers you've built but reverse null in speaker designing parlance means reversing the leads of one of the driver pairs and examining the drop off in amplitude across the entire crossover region. And believe it or not Roy, it happens with 1st order designs too. And if you just stop and think for a moment, your taking one vector out of top two quadrants and placing it in the bottom. Each leg is no longer operating in quadrature. Yes, the null isn't as severe as with an even ordered network but the null DOES EXIST. If you still doubt this, look at one of the most popular DIY speaker project documentations on the net here:http://www.zaphaudio.com/WaveguideTMM-minimalist-FR-reversenull.gif
http://www.zaphaudio.com/WaveguideTMM-perfection-FR-reversenull.gif
Roy also said -
"Please know that this is a common mistake made in every electrical-engineering textbook I've seen, because the mathematics behind it was not presented from scratch using Time as the variable, not Frequency. What you'd find is that what appear to be damping issues are actually artifacts created by time delays that vary with frequency. There is an excellent AES paper on exactly this subject at"Apparently you did not notice that I used the term "acoustic phase". Do you understand what acoustic phase implies?
Roy also asserts:"The nice aspect about using a first-order crossover circuit is that the phase-shift DIFFERENCE between the low-pass and the high-pass filters is a CONSTANT 90 degrees at ALL frequencies, not varying with frequency as in higher-order circuits, and not just "about a particular crossover point" as you wrote above.
That word "Constant" is the clue behind the benefits of using first-order networks."
Every capacitive or inductive influence whether associated with the transducer's moving mass or a component in the crossover produces a change in ACOUSTIC PHASE. Acoustic phase shift by implication means an apparent time shift since the sinusoidal is represented by amplitude variations over time and the spatial phase shift is inextricably related to time shift. This is electrical engineering 101. I don't think anyone needs a lecture on the distinctions between time and frequency. The AES paper you cited is ancient history and does not disagree with anything I've stated. On the contrary, you don't seem to understand what group delay in the context of a crossover truly represents. For a simple illustration, I'd direct your attention here:http://www.rane.com/pdf/ranenotes/Linkwitz_Riley_Crossovers_Primer.pdf
Group delay, as noted in the link above is the rate of phase change per incremental change in frequency and is represented in units of time. However this is not a measure of time or phase separation between the two driver outputs but rather a total time/phase shift of their sum from one frequency to the next since both outputs experience phase shift that is equal to one another at each frequency.
The focus in a crossover as far as "time coherence" is concerned is the melding together of wavefronts from the disparate sources so that they appear to hit the target AT THE ESSENTIALLY THE SAME TIME with a constant phase between them. You are asserting that group delay somehow creates PERCEPTIBLE TIME DELAYS between the driver's wavefronts and IT DOES NOT. In a crossover that takes into account each driver's minimum phase behavior, the rate of phase rotation between the two wavefronts IS HELD CONSTANT. IT DOESN'T MATTER IF IT'S 1ST ORDER (90 DEGREES) OR 8TH ORDER. Each driver experiences the same phase shift which may not be constant with frequency but they both maintain a constant phase between them from one end of the spectrum to the other. The phase plots shown in the link bear that out and the text also states on page 11:
"The resultant vector
is back in phase with the original input signal. So, not
only are the outputs in phase with each other (for all
frequencies), they are also in phase with the input (at
the crossover frequency)."to describe a textbook LR crossover - one of the most popular crossover types in existence. And if you've built/modeled a typical L-R crossover - your own phase plots would bear that out.
What is not constant with the higher order networks is group delay. That's linear with the first order network but not linear with the others. This apparently is where you are getting confused. If you want further clarification, please read the RANE link above. And again, more modern studies (within the last 10-15 years) have found that humans can reliably detect group delay on the order of a few milliseconds at around 1khz. Fourth order Linkwitz Riley crossovers exhibit phase shift per unit frequency below that threshold. Your concerns for "time delays" in higher order crossovers is indeed a STRAWMAN argument.
As to the issue of transient response, again - that's electrical engineering 101. As you increase the number of poles and zeros in a circuit, the number of energy storage points is increased and instability (harmonics) is introduced. You can call them "artifacts" but Linear Control theory very explicitly defines the transfer functions and the various levels of Q associated with the resulting resonances.
This is old hat to me - something I learned about 25 years ago. You might be correct and both I and Rane Corporation among many others might be wrong but I highly doubt it.
Edits: 03/12/12 03/12/12 03/13/12
Hi Villa,
I think we agree more than disagree. And I probably could have been more clear in my response.
You wrote:
"Not sure what kind of first order speakers you've built [visit our website] but reverse null in speaker designing parlance means reversing the leads of one of the driver pairs and examining the drop off in amplitude across the entire crossover region. And believe it or not Roy, it happens with 1st order designs too. And if you just stop and think for a moment, your taking one vector out of top two quadrants and placing it in the bottom. Each leg is no longer operating in quadrature. [WRONG, sorry- see below, thanks] Yes, the null isn't as severe as with an even ordered network but the null DOES EXIST."
You claimed 30-40dB cancellation, Villa. This is not even remotely close to true on a well-done first-order speaker design. In Figure 13a of that Rane PDF of yours, please note how the addition of 180 degrees to one of its two 0.707-amplitude vectors (which how one is supposed to represent a polarity reversal of one driver), results in two vectors still separated by 90 degrees, not the 180 required for cancellation. Again, perhaps you were misinformed.
You wrote:
"Group delay, as noted in the link above is the rate of phase change per incremental change in frequency and is represented in units of time. However this is not a measure of time or phase separation between the two driver outputs but rather a total time/phase shift of their sum from one frequency to the next since both outputs experience phase shift that is equal to one another at each frequency."
And later, you added:
"What is not constant with the higher order networks is group delay. That's linear with the first order network but not linear with the others. This apparently is where you are getting confused."
No, I think I failed to communicate properly, as I do agree with these statements.
You wrote:
"If you want further clarification, please read the RANE link above. And again, more modern studies (within the last 10-15 years) have found that humans can reliably detect group delay on the order of a few milliseconds at around 1khz."
Here, our studies show we can detect frequency-varying time delay on the order of microseconds, not milliseconds, over most of the tone range.
Don't believe me, fine. But everyone we know hears what we do when the adjustable tweeter on our Eos two-way model is moved back less than an 1/8th of an inch. Which is less than 10 microseconds difference in arrival-time. Or move the mid-driver in our 3-way Calypso back and forth less than 1/4". But for one to hear those differences, many other variables and problems must first be fixed or avoided, topics which we thoroughly present on our speaker's design-concept papers, especially on our flagship model(s).
I probably could have expressed myself more clearly, as you are exactly right-- in higher-order crossovers, the delay in the highs is not the same as the delay in the lows, and in-between the delay is constantly changing. I hope that is now clear enough for both of us!
This difference in delay between 'when the lows arrive' and 'when the highs arrive', a delay that is also constantly changing as we move up or down the scale, has also been found important by other speaker companies including Thiel, Dunlavy, Vandersteen and Soundlabs. And of course, very many more have not.
Why is a time delay that varies from low to high audible?
Partly because the sound of any musical instrument or voice consists of a fundamental tone plus its harmonics which span many octaves. A varying delay between the lows and highs thus changes the sound of an instrument or voice, making it much less natural in timbre and in other qualities, at least when one knows the real sounds.
A varying time delay between lows and highs also affects the sound of transients from instruments and voices, because the sharp, leading edges of their complex musical waveforms arrives much sooner than their lower tones- sibilance and 'coarseness' are exaggerated in some speakers. Detail is exaggerated- too much 'picking' on the guitar is often heard. And we have heard many examples where it has yielded speaker that can't play rock music nor 'poor recordings'.
A varying delay we hear detach the lower-range 'pulse' of the rhythm from the drive of the mid-band instruments and voices. The beat seems sluggish from that low-frequency delay.
A varying time-delay also hides the depth of the soundfield-- dissociating instruments with different tone ranges from being 'in the same studio'. One example is hearing the sounds from the tweeter physically separated upwards from sounds in the middle range (the famous 'rising treble image' which many report).
You wrote:
"As to the issue of transient response, again - that's electrical engineering 101. As you increase the number of poles and zeros in a circuit, the number of energy storage points is increased and instability (harmonics) is introduced. You can call them "artifacts" but Linear Control theory very explicitly defines the transfer functions and the various levels of Q associated with the resulting resonances."
I am not sure why you are bringing this up, as this is not an issue in first-order speaker designs. The audible artifacts I am concerned with include the ones I list above.
I do think you would quite enjoy that AES paper if you read through it, as it does clearly show (mathematically and with `scope photos), how the varying time delays look exactly like ringing, warping, or tilting of a square wave. Now, that is not to say these are not 'handled' by many engineers as if they were actual ringing, etc. It is the most-clear paper on the subject I have seen.
Thank you for your input. I hope you see we do agree on many points. However, I find that the execution of design in most all first-order speakers is not very thorough-- one indication is their use of complex first-order crossover circuits.
Best regards,
Roy
Roy Johnson said " I failed to communicate properly"No Roy, you communicated very well - citing an outdated AES paper from 1975 and clearly stating:
"The nice aspect about using a first-order crossover circuit is that the phase-shift DIFFERENCE between the low-pass and the high-pass filters is a CONSTANT 90 degrees at ALL frequencies, not varying with frequency as in higher-order circuits, and not just "about a particular crossover point" as you wrote above."
As shown in Rane Corporation's "Crossovers For Dummies" tutorial, the phase shift difference between the drivers is CONSTANT in the first order AND THE HIGHER ORDER CROSSOVERS. You obviously made a major error that unfortunately is common in this industry - even among "experts" but especially among those who are trying to promote "advantages" of 1st order crossover networks.
What is even more alarming though is your refusal to acknowledge that first order networks do exhibit reverse nulls.
Roy Johnson said:
"First, please note that no such cancellation occurs on a first-order type of speaker. "
This last blunder of yours is rather inexcusable for someone who claims to be a professional in the loudspeaker business and I wish for your sake that you'd retract your original statement or you will likely face a lot of future ridicule if you continue to endorse it.
Roy said -"You claimed 30-40dB cancellation, Villa. This is not even remotely close to true on a well-done first-order speaker design. In Figure 13a of that Rane PDF of yours, please note how the addition of 180 degrees to one of its two 0.707-amplitude vectors (which how one is supposed to represent a polarity reversal of one driver), results in two vectors still separated by 90 degrees, not the 180 required for cancellation. Again, perhaps you were misinformed."
To which I say - WHEN THE VECTOR SWITCHES FROM THE UPPER PAIR OF QUADRANTS TO THE LOWER PAIR - THE POSITIVE GOING PORTION OF THE SWITCHED VECTOR IS NOW GOING NEGATIVE. YES THE TWO VECTORS ARE STILL RUNNING IN QUADRATURE - IN A DIFFERENT PART OF THE VECTOR PLOT. BUT NOW ONE HAS A NEGATIVE GOING COMPONENT WITH RESPECT TO THE OTHER VECTOR. THIS CREATES CANCELLATION (NULL).
Added Edit:
[To make it clearer for you and potentially others who might not have a solid background in electrical engineering - discard the vector plot and consider the original input sinusoidal wave. One driver's output is peaking at the 45 degree mark where the other is peaking at the 135 degree mark. They are in true "quadrature" or occurring with a 90 degree phase between them. If you add 180 degrees of phase to one of the outputs, evaluating the amplitude of the original signal at 315 degrees phase produces a negative amplitude (corresponding to acoustic rarefaction instead of acoustic pressure). The two drivers are no longer operating in true "quadrature" with respect to the listening target or original signal and THAT is what counts. Yes, they are 90 degrees apart but now one is negative and the other is positive. This holds true whether you add 180 degrees of phase or subtract it - you still wind up with cancellation. (hope that helps)]
I can't believe someone with industry experience is actually making these statements - particularly when they are shown concrete data from a respected industry figure that disputes what they are trying to say. And no, I never explicitly stated or suggested that a 1st order network was going to produce a 30 or 40 db null - the focus with that statement was with even ordered networks suggesting that reaching a 30-40db null without proper phase integration would be "impossible". Perhaps the next post from you will be a slippery backtrack suggesting that "no such cancellations" meant lesser cancellations are possible. If that's the case, I'll respond here by saying I was only implying the degree of null that is possible with a well balanced crossover - not singling out 1st order networks which appear to be your favorite kind.Please don't suggest that we agree on anything. We clearly don't. You made errant assertions - one of which you're slightly backing away from. But you are still worlds away from the about face that is needed to correct obvious mistakes on your part that someone who builds speakers for a living should NOT be making. First order crossovers are very rare in today's high end speaker market. Anyone who has lived with a first order speaker and the inherent beaming problems off axis can likely attest to why that is so. And those that tout their superiority by citing bogus claims of "time coherence" and "phase coherence" will always attempt to ignore these and other obvious problems with first order designs. I'm only attempting to set the record straight on the ACTUAL differences between the topologies rather than the wished for or "believed".
Edits: 03/14/12 03/14/12
Hi Villa,
I appreciate that you think I am wrong. but I am not backing away from anything.
Let us look one more time at that same vector example-- in which the upper vector was rotated by 180 degrees, to be pointing at a 225-degree angle (45 + 180), and the other remains at 315 degrees-- just as you also wrote.
Now, you and I agree they are still in quadrature, which means there is still 90 degrees between them (315 - 225 = 90 degrees).
And let each have an amplitude of 0.707, compared to having a value of One (which would be 'full output'). For the benefit of others, when each vector is drawn .707 units long, that is each driver "being -3dB at this crossover frequency".
When these two vectors are added, the resultant vector points straight down with full magnitude.
And you agree.
However, if 'cancellation' was occurring, the magnitude of that resultant vector would be Zero- it would not point anywhere-- it would not exist on the graph, as it would have a length of Zero, which is a 'point'.
But the magnitude of the resultant vector is ONE, which demonstrates how there will be NO cancellation in output when the tweeter is reversed in polarity from the woofer or vice versa, on a first-order type of speaker. Acoustic tests here confirm this.
Vector addition applets show this:
http://comp.uark.edu/~jgeabana/java/VectorCalc.html
Draw one vector at 225 degrees (into the lower left quadrant).
Draw the other vector at 315 degrees (into the fourth or lower right quadrant). Click 'add'.
On to your other point, to clarify about what happens when the phase difference between woofer and tweeter remains a constant 90 degrees-- this leads to the group delay remaining constant.
With higher-order crossovers, the constant phase difference of 180, 270 or even 360 degrees leads to the group delay NOT remaining constant, which was the point I wanted to originally make. Sorry for any confusion caused by not making that more clear at the beginning.
And to your last point:
Please note that first-order designs do not have beaming or cancellation problems off-axis. The math for that is based on pure sinewave tones measured at one point in space, and yet music is not pure sinewaves, nor do we listen at a single point. Therefore, while the math predicts what you wrote, the result on music is not what you claim.
By the way, that AES paper is 'outdated'? By that, you mean its math is no longer accurate? Wow.
Best regards,
Roy
Man Roy,
You seem to be a nice guy from what I've read. Please, give up while you're ahead!
Roy said - "Let us look one more time at that same vector example-- in which the upper vector was rotated by 180 degrees, to be pointing at a 225-degree angle (45 + 180), and the other remains at 315 degrees-- just as you also wrote."
No Roy, I didn't say that both vectors get advanced when one of the speakers terminals are reversed. I only advanced one of the vectors (from 135 to 315 degrees). Advancing both vectors does nothing and would not be considered the execution of a reverse null in any way, shape or form.
Please, you seem to be regressing with each additional post and I'm beginning to feel embarrassed for you. Let's just agree to disagree, ok? I have cited the documentation by way of published measured data (John Krutke) and the published work of a highly respected corporation (former members of the original Phase Linear Group)to support what I've been saying. The more this is rehashed - the more ridiculous we both look. If anyone has any doubts as to what I've been saying, they don't need to purchase an old AES paper or membership to AES - just review the web links I provided. Have a nice day Roy and go build some boxes!
:)
Hi Villa,
Thanks again for your comments. I did not advance both vectors. I began with one at +45 degrees and the other at -45 degrees (or 315 degrees). Then I added 180 degrees to that first vector, which rotates it to 225 degrees (45 + 180). The resultant vector is as I described: straight down, with full magnitude. There is no cancellation.
I appreciate the information Rane Corp. put on the internet, but as with many of the math relationships underlying time coherence, there are a few mistakes made in their characterizations of what's going on. This comes about, in my experience, because the textbooks for Electrical Engineering used in universities present the math developed for continuous steady-state tones (Frequencies) rather than presenting the underlying ones which show what happens as Time flows along from T=0.
This is done because pure-sinewave math is easier, but mostly because time-domain math (used in transient theory) proves unnecessary to solving most circuit-design problems. What is done to that math to make it simpler is to get rid of Time as a variable, by assuming the sinewave(s) have no beginning or end (unlike music).
One difficulty which then arises is the use of 'phase' to describe the relationship between waves. 'Phase' is a relative term-- used compare what's happening in one wave versus another at ANY point in time (since Time is not a variable).
But when the time-domain is allowed to remain part of the math, then one must consider what happens from Time = 0 onwards, and that leads to Absolute relationships (Relative now only to T = 0) to describe waves, not a Relative one between waves.
And that math is harder to describe, to teach, and for students to grasp (especially when it appears 'they don't need to learn this stuff'). Electrical Engineering has always been a set of tough courses, and this time-domain math I saw discarded in the early 1970's from EE textbooks, to help make room for the emerging digital circuits coursework.
Best regards,
Roy
Roy, peeped your profile, not that I didn't know who you are. I have a Micro Seiki BL91 TT, also, with a Sumiko arm. One of two tables I use. Just sayin. Cheers.
They're still out there. Active speakers from Scandinavia and Britian. Passive single driver planar or full range conventional driver speakers. Vandersteen never abandoned this, has he?
Hi Jim,
I don't know of any active speakers that are time-coherent. If you know, please share, as I am always curious.
FYI, single-driver (cone-type) 'full-range' speakers have time delays that vary with frequency- at the upper end by cone breakups and whizzer-cone operation, and at the low end by whatever bass-alignment is used, from infinite-baffle, rear-horn, sealed or ported.
Best regards,
Roy
Meridian qualify?
Not to my knowledge, Jim.
Best,
Roy
Davey's post hit the nail on the head. There's no magic bullet that eliminates frequency combing in ANY design. Move about the target listening area with ANY system and you will experience constructive and destructive interference - even with a single full range driver (every driver beams because no one has yet invented a point source loudspeaker). You can design for phase coherence over a limited frequency range and limited target listening area range but that's it. Manufacturers stopped touting these benefits long ago because proper design these days assumes a high degree of phase integration between drivers in a multi driver system. This implies that the wavefronts of the two drivers at the crossover point will be hitting the listening target at the same time and that the amplitudes of both drivers will be balanced with one another so as to produce a flat frequency response not only on axis but at various listening heights. The ironic thing is - most low order crossovers that are part of systems being touted as "phase and time aligned" are some of the worst phase and time aligned speakers available. Off axis performance is horrible because the crossover point is spread over a very wide band of frequencies. The use of the terms "phase and time aligned" these days says more about the user's lack of knowledge than whether or not the speaker is properly engineered.
NT
Single driver speakers cannot be truly transient accurate because they have rolloffs at their extremes.
A multi-way transient accurate speaker done right can have better pulse response than a single driver speaker.
Although "crossoverless", single driver speakers are not "phase shiftless".
Cheers,
Presto
my acoustats ran from about 30hz to 20Khz without rolloff (on-axis of course). They were actually 2 single drivers vertically oriented. I say two single drivers because both were driven full-range.
Thanks for insightful info on the constraints of a single driver design. So they are still not the best mouse trap available.........
Regards
PS
PS:
My belief is that with speakers there are many ways to skin a cat. Every way has it's own strengths and weaknesses. Electrostatics are nice because of their good impulse response, but when you need to cross in a woofer at 100-150Hz you're into crossover territory again. Single drivers are neat and minimalistic, and if horn loaded can provide decent bass response but I will always prefer a good quality tweeter to the high-end of most full-rangers. Human voice and certain instruments on a good full range is fantastic! The impulse response of the full ranger is really only "less than perfect" at the frequency extremes, and it's still a lot better than what a 4th order acoustic response looks like impulse wise. Horns? That's a whole other kettle of fish.
My belief is that a good speaker is one that makes your listening sessions longer and longer and makes you want to get more music. Lesser speakers leave you wanting more.
If you can walk out of your audio room saying "Damn, that was good!" then I think that is the ultimate test of a speaker system!
Transient accuracy is a much debated design metric. It's fun to learn about and really fun (hard) to achieve. But I am not convinced that it's as dramatically audible as some suggest, and I am not yet convinced it alone results in being able to hear "absolute polarity" changes, if such a phenomenon is even audible.
Some say it is, other say no. The jury is still out for me!
Cheers,
Presto
That is the whole idea of high end audio design is loudspeakers
that are time aligned and phase coherent.
Even DIY guys design with this in mind
Also, All of your huge line arrays and sound companies time align
drivers, and cabinets, and subs, and make sure phase is right.
This is nothing new here..
My Fried C6 and O6 are but not made any more.
phase coherent and time aligned? If not, why produce such speakers? Wont it become a particular sound liked only by a certain section of audiophiles? I like speakers mentioned in this post a lot. Perhaps I find them more 'musical'. Dont know for sure though. 'Chronic' audiophiles could perhaps enlighten us.
Cheers
Bill
If a speaker is "phase coherent and time aligned" it would then be, by definition, transient accurate meaning it has near-perfect pulse response. Of course, there will always be a small amount of overshoot (acoustic overhang) and with digital filters there will be pre-ringing for FIR and post-ringing for IIR.
But truly? This is more of a design mantra than a specific sound. How a transient accurate speaker is voice is still dependent on choices made by the designer.
Cheers,
Presto
Any given instrument is time aligned and phase coherent. However, in an orchestra different instruments can be at very different distances from you. For example, the first violins may be spread out over 10- 20 feet. Though they are playing the same notes, the sound will arrive at different phases from the various instruments depending on their distances. Since sound travels at about 1000 ft/sec, there will also be very small differences in time of arrival. I think (wild guess) that the phase coherence of any given instrument is important, but the difference among the various instruments in the same section doesn't matter much. Not that you can do anything about it!
Somewhere on the web I once found a university study of the audibilty of phase coherence and time alignment. As I recall, there were about 20 listeners who listened blind- they didn't know how much out of pahse or time alignment the different drivers were for each evaluation. The results were very clear-cut. In this study either time alignment or phase coherence were judged to sound significantly better than the absence of both, but you didn't need to have both at the same time. Either one did the trick, and both was no better.
Joe
a physical distance between the source and listener / microphone, does NOT alter he original signal as all frequencies are delayed equally.
Thus a signal like a square wave if produced can travel across a room and still be a square wave.
A loudspeakers phase response on he other hand is the deviation relative to the input signal, phase shift here prevents a speaker from reproducing a square wave etc (following the input signal).
Yes, it's that elusive concept which differentiates delay from group delay. Fun stuff! :)
Cheers,
Presto
After living with Dunlavy SC IVs for many years (only recently replaced after many, extended listening sessions with several high end speakers) I discovered I need my speakers to be phase coherent and time aligned so I got Reference 3A Grand Veenas. The theory for doing so is verified by the sound in my listening room.
Technically, what is the difference between "time alignment" and "phase coherence"?What observable property diagnoses either one of these? Say with test signals & a A/D converter & computer, how would you know one vs another?
What is the simplest canonical physical representation of each or the absence of them?
I've never really understood this at technical level.
Edits: 02/21/12
Dr. Chaos,
Good question.
One simple way to visualize the difference begins by looking at the sine wave coming from each driver individually, at the crossover frequency.
On a phase-coherent speaker that is not time-coherent, these two identical waves line up over each other peak-to-valley-- as they are 'in phase'.
But if you go back to the beginning, you will see one wave (the tweeter's) began one full wavelength sooner than did the woofer's. And therefore the woofer's copy of that wave will also stop one full wavelength late.
A time-coherent speaker means that both waves started and stopped at the same moment. And this automatically gives phase coherence, because the two waves will then overlap everywhere. The opposite is not necessarily true-- phase coherence never implies time-coherence.
Best regards,
Roy Johnson
Designer
Green Mountain Audio
Roy, Hi. Feel free to correct me if you think me wrong. Meridian makes active speakers with active crossovers wherein they can adjust the output of the drivers so their waveforms coalesce in time. I think there are several manufacturers of active speakers that do this, hence, my reference to Scandinavian and British manufacturers.
Doesn't Vandersteen and didn't Dunlavy do this passively by staggering their drivers so the voice coils of their drivers, for lack of a better description, align for time coherence at the listening position?
Thanks
Hi Jim,
Vandersteen, Dunlavy, Thiel, and my designs stagger the drivers so their acoustic centers line up.
When Meridian and others attempt this, they are only concerned about being 'in phase' at the crossover point, which means the two drivers are either one-half wave period or one full-wave period out of time-synchronization at the crossover frequency.
And that means the sine wave coming from the tweeter at the crossover point is starting sooner by one-half or one-full wave period from the same wave coming from the woofer (which thus stops late by one-half or one-full wave period). This is all any of these active-speaker companies attempt to do- program in a fixed time delay to make up for not staggering the drivers from front to rear.
Now, after adding that time delay to the tweeter electrically (and/or mechanically), to obtain exactly that half- or full-wave period offset, then the timing between all tones is still changing as we move up and down the scale. So it is not a time-coherent design.
Any such speaker is then only 'phase correct' or 'phase coherent', but not time coherent-- where all tones start at Time = 0. Of course, the phrase "Phase Coherent" sounds worthwhile and certainly quite technical-- making it good for marketing and for impressing reviewers.
Meridian's work is shown in Stereophile's review of their flagship a few years back, right when it first came out. And an interview with their designer at that time, also in Stereophile and perhaps in the same issue, quotes him as saying "We couldn't hear the difference, so we gave up doing any time-domain/time-coherent correction' (I am paraphrasing). I seem to recall that the tweeter or perhaps the mid-driver in that flagship still runs in reverse polarity- shown in Stereophile measurements.
Best,
Roy
Thank you for the detail. Does any speaker to your knowledge electronically delay the output of one driver so that it would coalesce with another driver?
Not that I know-- and to be specific here, I mean by not having a time-offset that changes with frequency.
However, many powered speakers do program in a fixed time-delay for the tweeter's signal, so the speaker cabinet's front face does not have to step back to line up the tweeter's acoustic center with that of the woofer's. Cheaper to make a flat-front cabinet, and "Digital Time-Delay Correction" sounds impressive for marketing purposes.
Best regards,
Roy
I wouldn't know if there is a "digital" time correction or not. Perhaps, you do? Regardless, there is and has been a methodology of time delay in an active crossover in lieu of physically staggering conventional drivers for the same purpose of uniform time arrival of two or more drivers, if I read you correctly. I can tell you that this was being done before there was digital anything. Cheers
You are right, if I recall correctly.
The problem remains that what was/is being done is to provide a fixed amount of delay for the tweeter's entire tone range, such as 0.3 milliseconds. For my purposes, the time-delay correction would need to vary at each frequency- from 0.0 seconds in the ultra highs to ~5 milliseconds in the bass, smoothly changing as we go down the scale.
The math for calculating such a change is well known-- I don't recall anyone implementing it though. If someone has, the phrase 'constant group delay' would identify such a design. Rough calculations I performed many years ago seemed to indicate the required number-crunching would consume a very large computer. Though I am far from being any sort of a computer expert, perhaps that remains the case against its implementation.
Best,
Roy
You are already there with your designs, of course. All the best to you.
is
"Man is the only animal that blushes - or needs to" Mark Twain
VMPS = Veritone Minimum Phase Speakers
Actually, many speakers are "phase-coherent" and "time-aligned." It's just marketing-speak that describes a multi-way speaker where the drivers have been physically and/or electrically time-aligned to remove the offsets. Acoustic crossovers of any "order" can be realized as "phase coherent."
Large transducer speakers (like Magnepan's) can't be called phase-coherent under any circumstances because the created comb-filter effects result in phase-shifts that can't be corrected.However, I think you might be confusing "phase-coherent" with "linear-phase" or "time-perfect" or "can reproduce a square wave" or "constant group-delay" or some other definitions I've seen. Those are a different category and can only be achieved using first-order analog acoustic crossovers, and/or DSP solutions, and then only at a single point in the measuring space.
Whether we can hear waveform distortion that those systems "solve" is debatable. Many claim these type of systems are categorically better....but generally there are many other variables that yielded an apples/oranges evaluation.
Cheers,
Dave.
Edits: 02/20/12
"Those are a different category and can only be achieved using first-order analog acoustic crossovers, and/or DSP solutions, and then only at a single point in the measuring space."
While that is true for hifi loudspeakers, the ones in the link are a newish horn invention (pat pending)used in larger scale sound , they can reproduce a square wave, over a broad band and not just in one location and they have higher order passive crossovers. While they are horn speakers, there is no trace of horn sound.
Not only that but the speakers have no self interference and as a result they only have one radiation lobe and not a pattern of lobes and nulls in the polar patterns.
They measure and act as if they only had one wide band driver.
Here is a big giant one you can listen to with headphones;
http://www.youtube.com/watch?v=5MOG_sPejGA
Best,
Tom Danley
Better question would be why isn't everyone making them?
.
except those who thought they could.
P
As I slowly slip into the dark cesspool of audiophalia neurosis. . . .
My speaker building site
After you have lived with phase/time coherency it's hard to go back.
Oz
Don't worry about avoiding temptation. As you grow older, it will avoid you.
- Winston Churchill
I have a system here (WMTMW arrays) with DSP-based active IIR crossovers (4th order) with "phase correction" that turns the textbook "LR4 transient response" (aka mess) into near perfect transient response, except of course for the post-ringing caused by the use of IIR filters.
I can say this much, being able to literally toggle between the two completely (theoretically different) systems, the changes are subtle at best. The fact the amplitude response is identical helps to ensure there is no confusing amplitude variations with changes in the time domain. I am not thumbing your nose at your speakers or your hearing accuity - but it is possible much of the "goodness" you attribute to transient accuracy COULD just be a case of you having a damned good speaker that is pleasantly voiced. I have had some very very fantastic sounding LR4 (acoustic) designs via both passive, active and DSP based filters - and LR4 has a transient shape that looks like a truck drove over it, with a large negative going initial impulse followed by considerable overshoot. Negative impulse, from all drivers with the + to + and - to -. Very interesting math that leads to all things positive and "in phase" to a negative impulse! Enter the strange land of group delay...
This kind of a/b testing is only possible with DSP based systems, but it can give one a feeling about the magnitude of these kinds of things. I am still trying to see if/how transient accurate speakers have this profound affect on ones ability to detect absolute polarity changes. I have yet to hear this phenomenon myself and freely admit it.
My ears are only 18c gold, after all.
Cheers,
Presto
And at only one spot in space.
And that's were your ears should be. Audio is a lonely endeavour...
And at only one spot in space.
So what? I only listen in that spot.
Oz
Don't worry about avoiding temptation. As you grow older, it will avoid you.
- Winston Churchill
Not trying to bust your chops just the way it goes. My loudspeakers are also time aligned similar to yours.
Mine are in phase everywhere in my room.
The joy of owning active speakers with a phase-adjustable xover.
nt
As I slowly slip into the dark cesspool of audiophalia neurosis. . . .
My speaker building site
I don't think so, I know so.
PS: May be I should mention that my speakers use co-axial drivers with a maximum timing difference of <6 microsecs between woofer and tweet.
Edits: 02/21/12
I think what Pjay meant is that you can't adjust "phase" and correct for group delay. Unless you have a 1st order acoustic, filler driver or subtractive delay design, you can't get transient accurate speakers using a Behringer or similar crossover unless it has group delay aka phase correction. Phase adjustment is not the same as "phase correction".A 4th order LR acoustic design boasts all signals being in phase for all frequencies, but the filter does not have constant group delay. The transient response looks like the mess below. Some say this is audible - but as scary as it looks, many can't hear it at all.
The graph below represents a square wave input into the Thuneau Allocator. The output is the sum of a 2-way Linkwitz Riley (LR4) crossover. The output is shown with and without the Arbitrator phase correction module enabled. This is not measured with a mic, but the signal coming directly out of the crossover, aka, it's a theoretical model. This is why one needs to account for driver response and get the correct acoustic response coming out of the speaker. If you correct the phase of an LR4 without hitting the LR4 acoustic response, your impulse response will suffer for it no matter how flat the amplitude response is.
Cheers,
Presto
Edits: 03/01/12 03/01/12
What is a 'Behringer or similar' xover?
I use modified two BSS FDS360s.
nt
As I slowly slip into the dark cesspool of audiophalia neurosis. . . .
My speaker building site
Why do you feel I'm a single driver guy? I only offer 1 model with fullrange. Mostly we sell horns. Still I wouldn't offer SEAS x-1 unless I enjoyed its performance.
.....check your facts.....
Oz
Don't worry about avoiding temptation. As you grow older, it will avoid you.
- Winston Churchill
Are you saying you can hear phase distortion?
If so, how did you test yourself to come to this conclusion?
Cheers,
Dave.
There are good speakers and bad speakers and with the number of good speakers that are not touting "phase" or "time" aligned it suggests it is not a key selling point for all good speakers. If course if you screw one of these up it is bad. Yes, many of the good speakers are not flat fronted and so many are. My experiments with it found that it is a lot more important to move your ears from the nipple of the sound field to the boob. That alone makes a large difference, more than moving the tweeter back to get the VCs aligned or to add/subtract components to try for phase adjustments.When you listen to an orchestra with a row of 22 violins, is that phase or time aligned?
I am at work with a meeting in 90 seconds so sorry if some poor wording is making me look like a fool, I know I am not clear on something up there.
If you hear it then more power to you. If you are totally convinced this trait makes it better then more power to you. You have probably had the same gear for years and are happy, more than I can say for me!
Now I have to go discuss why Nigeria is using two brands of CD4 machines and only one is in their national algorithm.
Cheers!
P
As I slowly slip into the dark cesspool of audiophalia neurosis. . . .
My speaker building site
Edits: 02/21/12
This is like the third week in a row that you own the same set of speakers. :)
....Actually it's going on two years. A virtual lifetime for me....
Oz
Don't worry about avoiding temptation. As you grow older, it will avoid you.
- Winston Churchill
A few concentrics exist that are both. Radian makes them as do others. Does depend on crossover used.
Edits: 02/20/12
Why not? I have a pair of Dunlavy speakers and the sound is incredible. You very rarely see these on the used market, and there is a reason why. Look at Lipinski speakers.
(Worshiping at the Universal Music Altar)
Vandersteen is one who claims to but I never liked their sound. I do own one of Richards subs however.
AIUI, this means:
a) filters are 1st order
b) the acoustic centre of each driver is in the same vertical plane?
If I am correct then the latest Magnepans - the 1.7 & the 3.7 - fit the bill. :-))
Regards,
Andy
Have a look at tomservos response and you'll realize that it means no such thing.
None of the drivers acoustic centres are in the same plane and I very much doubt that Tom uses 1st order xovers yet his synergy horns are phase and time aligned as is shown by their ability to reproduce a square wave.
Also it is very difficult (or even impossible) to phase align a 1st order network.
The more honest manufacturers of these claim that the phase shift they introduce (90deg) is benign and possibly inaudible but not that they are inherently in phase.
Many, if not all, larger active monitors are electronically time and phase aligned.
Old (may be the new ones too, I don't know) AlNiCo Tannoys are phase- but not time-aligned.
In the 80s Tannoy introduced a passive all-pass filter to get them time aligned as well but even Tannoys own engineers admitted that this was purely a marketing excersise and that their cure was worse than the problem.
If you've ever lived with one everything else doesn't sound right.
I lived with Quad ESL-63s for a number of years. I like my current monitor speakers better.
-----
"A fool and his money are soon parted." --- Thomas Tusser
Are time aligned.
"Lock up when you're done and don't touch the piano."
-Dr. Greg House
Our full range Synergy horns are, not only that, a speaker like the SH-50 can reproduce a square wave over a decade wide bandwidth, covering both crossover points and not just in one spot.
Unlike normal multi-way speakers, the drivers are also close enough that they couple together coherently into one source, they do not produce an interference pattern like normal multi-way speakers do, evident in a pattern of lobes and nulls in the polar pattern.
These just have one constant directivty lobe, like one single broad band driver. How they work is here
Triangle, Reference 3A....
Dunlavy is gone but Duntech is still alive in Australia. Thiel and Vandersteen as Kava noted.
Thiel?
"Man is the only animal that blushes - or needs to" Mark Twain
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: