|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
192.223.242.21
Hi all -
What all does a speakers phase response effect?
My impression is that phase response impacts a speakers 'truth of timbre' in that the more phase correct the signal is when reproduced, the more accurate / real the reproduced instrument will sound.
I also have an impression that phase response may impact soundstage / imaging, though I am not sure why.
Craig
Follow Ups:
I have heard two demonstration systems, one with DEQX and Magnepan and the other was something related to Music Vault using Harbeth. Switching the correction on/off was a big deal, more so (not surprising) with Harbeth 3-ways. It wasn't just frequency re-balancing with the room correction that helped either. I felt more relaxed with the systems on. And all can articulate now is the sound was more focused on coherent. Not so sure what the AD-DA conversion is going to do with good systems, but the equipment helped in those demos. I have this between horn speakers as well, I think.
I once did some measurement at Guitar Center with OmniMic V2 Precision Measurement System.
In Yamaha HS series, the tweeter showed inverted polarity. In JBL 305, 308, the tweeter showed the same acoustic polarity as the woofer.
I listened to music through these speakers the day before that measurement. I had not felt that Yamaha HS series speakers were particularly bad in 'truth of timbre,' 'timing,' 'rhythm,' 'soundstage,' etc.
Well, HS5 and HS7 had somewhat less bass than HS8 or the JBLs, but I could not detect with my ears that the Yamahas' woofer and tweeter were not in the same acoustic polarity.
And ideally, run in parallel phase at least an octave either side of it. Good phase tracking through the crossover "region" results in flatter summed response and better integration of the drivers. This, of course, is on a fairly narrowly defined design axis. Listen too much above or below this axis and the phase goes wonky and and bumps and dips appear in the frequency response.
Now, this does NOT mean that both drivers must necessarily connected in the same polarity. With some crossover alignments, notably 2nd order Linkwitz Riley, the drivers will be 180 degrees OUT of phase at crossover frequency IF both are connected in the same polarity. The result will be a deep null or "hole" in the response at that frequency. Reverse the polarity of one driver and they will then sum flat, all though deeper into their respective passbands they will be out of phase with each other and this is what shows up in the step response test.
With a 4th order LR, by contrast, the drivers will sum flat through the crossover region when connected in the SAME polarity, and show a null when one is reversed.
In EITHER case, acoustic phase (which is a complex summation of the electrical and acoustical phase of the individual drivers, the LF box loading, and the electrical phase of the crossover filters) does NOT remain flat, or at the same angle. The acoustic phase of even a single "fullrange" driver ROTATES continuously through its operating band. Some higher frequencies will ALWAYS be "out of phase" with lower ones. Add a second driver and a crossover and the acoustic phase of the speaker system may rotate a full 360 degrees multiple times over its operating range. A 4th order crossover will add another 180 degrees of rotation compared to 2nd order.
How audible is this? The consensus seems to be, "not very," as long as the phase tracks well through the crossover region. That is where the ear is most sensitive to errors and anomalies. If some higher harmonics are in incorrect phase relationship to their fundamentals, which is inevitable even with a single driver, we don't really notice that. If we did, all speakers would be intolerable to listen to!
I think that there are good sounding speakers on both sides of the fence. Very few speakers are perfectly time/phase coherent. Different people cue in to different things in music, not everyone is equally sensitive to time/phase coherency issues, etc...
Edits: 01/29/16
nt
The universe is made of electrons, protons, neutrons, and morons.
"You never got me down, Ray!"
. . . is on Parts Express Tech Talk. Thread called "Why does phase matter?" started by mattsk8 on 12/31. I don't think I'm allowed to post links to other discussion boards here, but its about 7 pages back from the newest.
Anyway, a number of advanced builders and designers chime in, with good explanations and multiple graphs showing the phase rotation and tracking of different crossover types. It's one of the best discussions of this topic I've seen yet. Recommended.
http://www.linkwitzlab.com/phs-dist.htm
Phase response means keeping all frequencies of an imput signal in the same time relationship at the speaker output as they were on the recording. Very few speakers do this.
In deed there are arguements both ways how important this is ultimately in reproduction(defining accuracy of signal maintenance is another tale). A good friend who designed speakers thinks it's not too important. he argues changing seats at a live concert changes phase relationships but the sound still sounds like the real thing it is at all these variable seats.
My gut feeling is it may be of secondary importance for listening. But if you're talking about true fidelity it's something to strive for although at the present state of the art it's extremely difficult and costly to do.
Interestingly, if we go to multi amped speakers with digital crossovers, it's a much simpler problem to solve. Digital can be a powerful tool. And I believe you can do it in digital with fast slope crossovers and flat front panels where you need first order crossovers and offset drivers in analog.
> he argues changing seats at a live concert changes phase relationships
I don't necessarily think this is the important aspect of what we think of when we think of linear phase response in speakers. Changing seats at a live concert only changes the phase relationships among the different instruments; however, the phase response of individual instruments will always remain linear regardless of where you sit. The important thing is the phase response of each individual instrument with itself and not the phase relationship from one instrument to another.
I now own speakers that claim to be phase coherent and they sound surprisingly more realistic and transparent compared to all other speakers I've owned in the past.
How do you decide if it's because of linear phase or the other design features of the speaker or even all of the above?
Many great speakers(at least in some peoples' opinions) are not linear phase. In fact since by far most speakers are not linear phase, what does that mean for most highly esteemed speakers?
I don't know the answers! They sure impress me as some of the best sounding speakers I've ever heard, though.
Edits: 02/01/16
Hi Guys,I wish to correct the most common misunderstandings about phase shift and time delays for you. Here are what the mathematics show:
To get any kind of handle on a speaker's phase shift or time-delays, we must first set a proper Phase or Time = Zero point, for when the first 'sound' will arrive at our mic, at our ear, from each driver, BEFORE their signals are altered by any crossover circuits.
According to Dr. Richard Heyser's mathematics, which have been proven correct, 'the first sound to arrive' that we should use as our reference is the beginning of a tone burst (a 'beep') at a driver's highest pistonic frequency, with no crossover used. This would be a burst of a 10-12kHz tone for a good soft dome tweeter and maybe 5kHz on a great mid.
The statements made by Presto that there will be 360, 180, or even 90 degrees (a 1st-order filter) of phase shift at the crossover frequency is the common presentation in EE classes and quickie speaker-design texts. This is not quite right, but I've determined I cannot easily explain here why, because I'd also need to show the derivation of various Phase = Zero reference points along the way... I tried several times, by editing this post, and finally decided to parry it down to the following:
When the crossover-frequency tone, say 3kHz, comes out of both that good mid and good tweeter, through a first-order crossover circuit on each, then their signals will combine at the mic or ear showing NO phase shift, no time delay injected. This is peculiar to the use of first-order circuits. High-order circuits introduce phase shifts (time delays) that change with frequency.
To turn phase shifts into time advances or delays is simple:360 degrees of phase shift, at a certain frequency, means a timing difference of one period OF THAT FREQUENCY. If the frequency is 3000Hz, that has a period of 1/3000th of a second, or 0.33 milliseconds (ms). 360 degrees of phase shift is then 0.33ms time shift at 3kHz (you must specify the frequency). 180 degrees of phase shift is 0.165ms of time shift at 3kHz.
Down at 300Hz, a common crossover point between a woofer and a mid, these numbers would be ten times larger, as the wave period down there is ten times longer, at 3.3ms. For a speaker using high-order crossovers, the time delay imposed on that 300Hz can effectively place the location of 'whatever we think is making that 300Hz tone' nearly three feet (0.8 meters) BEHIND 'wherever the treble tones appear to emanate'.
Phase shifts/time delays in speakers are mostly responsible for what people discuss about imaging differences, transient differences, laid back or forward presentations, of 'apparent detail' (on those speakers, the sound of the 'pick' on a guitar string arrives at your ear well before the string's actual tone develops).
Phase shifts/time delays make speakers sound too bright, even though they measure flat. Why? Because the highs got to you first.
Phase shift/time delay can make the bass sound flabby and the rhythm sound lame. Why? Because the bass got to you late.
'Classic speakers' from the 40's to early 70's have less phase shift/less time delays because they used lower-order crossover circuits than do modern speakers.
One-way speakers have no crossovers, but still have severe phase shifts from cone breakups and whizzer-cone resonances, and from the time-delays caused by any extreme cabinet bass-tuning.
Phase shifts/time delays reduce the apparent efficiency of the speaker, since the peaks of music are not as 'peaky'. The dynamic envelope is reduced since the drivers are not pushing and pulling together from the first dynamic jump.
Phase shifts/time delays psycho-acoustically create image heights (note that we cannot record 'height' for 2-channel). They cause an image to wander at different frequencies.
Phase shifts collapses the depth of any image, and make musicians playing tonally different instruments, such as violins and cellos, sound as if recorded in different halls, having different echoes.
Phase shifts make poor recordings, loud string passages, loud horns, and loud rock music unbearable, especially to women. Phase shifts make some sounds or notes suddenly appear 'far too loud', as when a single note just drills right into the side of one's head. That's why engineers overly-compress and EQ horns, and de-ess singers.
Finally, phase shifts/time delays greatly affect our opinions of everything else- the cables, the gear, recording engineers, the speaker placement, and even the sound of our rooms! Makes us dislike digital more than we need to.
Fix these phase shift/time delay errors and the music is whole again, emotional not mechanical. One's attention remains on the music, as we are never distracted by 'the image' or 'the bass' or 'the airiness', since no part of the sonic spectrum is literally 'sticking out' out in time to attract our attention. But, experience shows one must be a music lover to appreciate that. Those with a limited musical palette seldom 'get this'.
Best regards,
Roy Johnson
Designer
Green Mountain Audio
Edits: 01/29/16 01/29/16 01/29/16 01/30/16
Roy,How have you been?
Yes, it's really hard to try and explain the concept of group delay in laymans terms. It's really a derivative, a rate of change thing. The big deal is that if the group delay is non-constant, the original signal is not preserved. It actually sounds amazingly good despite that that poor waveform looks like!
A true first order *acoustic* design has constant group delay. A capacitor thrown on a tweeter is a 1st order crossover, BUT as you know this does not a 1st order ACOUSTIC design make. It's simply a first order ELECTRIC crossover. Obviously, 4th order LR acoustic targets are going to be far far easier to hit because of the rapid roll-off rate. And, you enjoy quite high power handling even with band-limited drivers and you don't need a tweeter with special power handling considerations and ultra-low Fs.
There are only two ways to go about it really: 1st order acoustic design or some form of trickery, and by trickery I mean passive methods like subtractive delay or "filler drivers" or DSP-based trickery like FIR filters or phase-corrected IIR filters. I've worked with the latter and really it was a lot of fun to be able to compare the pre and post correction impulse response.
A few big myths about group delay.
1. You can "fix" group delay and obtain time/phase coherence by flipping a driver's polarity, as in the case of the 2nd order LR crossover.
2. You can "fix" group delay and obtain time/phase coherence by ONLY time-aligning the drivers using a simple delay.
3. LR4 is time/phase coherent because all drivers are in phase electrically.The last one is where people don't get that although the LR4 is in phase at all frequencies due to the 360 degree phase shift, the group delay is actually worse that 2nd and 3rd order designs. Yep, even with it's inverted mid-range a properly executed BW3 design will have better pulse response than an LR4.
Cheers,
Presto
Edits: 01/30/16
Good heavens. Roy, you've confused nearly everybody with this rambling mess.Phase shift in filter circuits is basic electronics. Let's not make it more complicated than it really is. I believe most users understand the aspects of acoustic crossovers, electrical crossovers, time-coherence, etc.
I suggest to read Presto's post again....accurately this time. You've mischaracterized his statements and also Linkwitz.
Presto's post is absolutely correct in all aspects and much clearer in explanation than yours.Cheers,
Dave.
Edits: 01/29/16 01/29/16
Rambling mess? Really?
Well, he's edited it and now it reads better.IMO, phase distortion is one of those topics that can do without the rhetoric and should be described in stark terms.
It seems most audiophiles have an opinion on the audibility of phase distortion, but very few have evaluated it properly...by removing all other variables in/for an A/B comparison.
As I mentioned in another thread awhile back, "coherence" is a label bandied about and misunderstood by many in the audio industry. I think most audiophiles think of this in the subjective context.Regardless, in my opinion, nearly all conventional speaker systems are not capable of time-coherent reproduction. The physical configuration renders this impossible.
Just my two cents.......worth what you paid for it.
Cheers,
Dave.
Edits: 01/31/16
So sorry to cause any confusion, Dave! Although I still don't believe I was rambling, I did come to the conclusion that I needed to edit my post down to some very basic information. Thanks for poking me about this!Best always,
Roy
Edits: 01/30/16 01/30/16
... overly-compress horns, and de-ess the singers." Etc, etc...Assuming that the above statement accurately reflects the manner in which a majority of available recordings are made...
Do phase/time perfect loudspeakers somehow correct for these types of *mistakes* made by recording engineers? Or, do phase/time perfect speakers effectively exacerbate this dilemna?
In other words, are perfectly made recordings required in order to fully reap the benefits of the phase/time perfect loudspeaker design?
Until the day when perfectly made recordings become the standard, could we be better off using time/phase *imperfect* speakers that would seem to be (or actually are, in a twisted sort of way) a better match for all of the equally imperfect recordings that we might listen to?
Thanks in advance for taking the time to answer questions that you may deem silly. I'm no Einstein.
Edits: 01/29/16 01/29/16 01/29/16
Good question.
When coherence is achieved, such a speaker distorts any and every signal less, including the distortions of a poor recording. So poor recordings sound better, as their distortions are not being further distorted.
The goal of reproduction is accuracy to the RECORDED waveform, not to any REAL waveform. That real waveform is long gone, altered by the microphone, as it does not hear the way we do.
I agree that fidelity to the signal is the goal. And that includes ultimately to include phase distortion. But at the current state of the art linear phase speakers are difficult to do and there is a cost in other fidelity aspects. So given that the audibility of poor phase response seems to be way below the audibility of the more recognized distortions I would opt for solving those problems first. Hopefully in the not too distant future we'll be able to do better.
Allow me to make a correction to the above, thanks.
A 'linear phase speaker' is not a time-coherent speaker.
This is technically-nice-sounding shorthand for "an acoustic phase response that changes constantly and smoothly (i.e., linearly) with frequency." In other words, it has more and more time delay the lower we go down the scale.
When a linear-phase speaker's 'rotates', its phase spins through multiples of 360 degrees (from thinking about phase as if degrees on a compass). Every linear-phase speaker rotates through multiples of 360 degrees as we proceed down the scale. When it is a linear phase design, that rate of rotation is smooth with change in frequency. This linear rate of change produces angled-but-straight lines (i.e., 'linear') for its phase-response, plotted on a lin/log graph.
'Linear Phase' of course also means that this speaker is 'in phase' at any crossover point, i.e., exhibiting no cancellation/no suckout around any crossover frequency. So, a decent speaker has linear phase shift along with a smooth-looking frequency response curve. Its impedance curve is usually not flat.
Linear-phase speakers are the MOST easily designed, because a designer thinks nothing of the time domain, only of the tone balance, a smooth frequency response, with no cancellations. Finito. That's it. Ship it.
FYI, Tannoy, Dali, Dahlquist DQ-10, B&W's, MBL, the new Thiels, M-L, and many more, are linear-phase designs. None is time coherent.
On the other hand, a time-coherent speaker exhibits the SAME TIME DELAY at EVERY frequency. It is not changing the acoustic phase at ANY frequency. These speakers are very difficult to design, but only when a designer lacks the education.
Plenty of good woofers and tweeters now exist for making time-coherent speakers, at no increase in cost, no difficulties with power handling, nor dispersion, nor efficiency.
However, time-coherent speakers take MUCH longer to design, because of the additional time-domain variables. Once a speaker design becomes time-coherent, it turns out that making tweaks to it every six months will not substantially change what is heard. Therefore, marketing is not about offering a 'new and improved' model every year.
This is because, going to your last point above, designing a speaker to be time-coherent is actually MORE important to address BEFORE any other design variable. To get the darn thing to be time-coherent, you must automatically avoid poor drivers with cones that break up or ring (metal), drivers with high distortions (from poor magnets and suspensions), strange cabinet-tunings for bass, and varying-length horns, among others.
Best regards,
Roy
Roy,
I think it might be good to elaborate a bit more on your definition of the term "linear-phase." (I noticed you don't use this term at all on your website.)
Most (I think) audiophiles believe "linear-phase" denotes a system that indeed exhibits a flat phase-response and a constant group-delay through the entire audio range. One comes with the other....so to speak.
I'm not aware of any commercial speakers that are linear-phase but also exhibit multiple phase wraps in the audio range and non-constant group-delay. You might mention a few of those.
Cheers,
Dave.
Hi Dave,
The best I can do, without having you and I working out of a common textbook in a classroom situation, is to reiterate from my previous post that:
"Every linear-phase speaker rotates through multiples of 360 degrees as we proceed down the scale. When it is a linear phase design, that rate of rotation is smooth with change in frequency. This linear rate of change produces angled-but-straight lines (i.e., 'linear') for its phase-response, plotted on a lin/log graph."
What is 'going wrong' is that my use of the word "straight" here is often confused with the word "flat", as in your post just above. What you wrote actually describes a time-coherent speaker, not a "linear phase" one:
"... a system that indeed exhibits a flat phase-response and a constant group-delay through the entire audio range. One comes with the other....so to speak."
Best,
Roy
Thanks for the clarification, Roy.
What about Vandersteen?
To my knowledge, the only speakers that are time-coherent across most of their frequency range are/were Dunlavy, Thiel, Vandersteen and Green Mtn.
The first three have very complicated crossovers, which allows certain driver problems to be 'fixed' for the measurement microphone. However, such complexity reduces dynamic contrasts and blocks the transmission of small, subtle sounds, which together are responsible for music's emotional content.
Best,
Roy
My understanding is the FIR crossovers have NO phase shift thru the passband.
Too much is never enough
''It is easy to get "fooled" by a big image if the phase is off. This usually results in the phantom center being pulled to the sides and creating a "wider" image for the listener. These small delays in phase is causing the insight of the material to be blurry and undefined but "bigger" to the untrained ear. The phase anomalies is usually only measured in a few parts of the spectrum and is therefore only affecting these frequencies. If the phase response shows anomalies in the 200-500hz region you will probably have trouble separating piano , guitars, vocals , snares etc in that area . If one area is affected you can be sure another is affected when moving your head. the suffering in imaging caused by the anomalies in phase response is impossible for me to overlook and is something I would categorise as critical in a speaker.''
Edits: 01/28/16
Well, first of all, there is the electrical phase response of a speaker and the *acoustic* phase response of a speaker.
An LR2 crossover imparts a 180 degree phase shift, which is why the tweeter is connected in reverse polarity. This crossover has non-constant group delay, which means it's not "transient accurate" or what some call "time coherent" or, a bit of a misnomer, phase-coherent.
And LR4 crossover imparts a 360 phase shift, which is how all drivers can be connected with the same relative polarity. So, this crossover design may be phase-coherent but it's not time-coherent aka not transient accurate.
The only crossover that is inherently time-coherent is a proper 1st order acoustic (not electric) design. Putting a single cap on the tweeter does not a 1st order acoustic crossover make. Much equalization needs to be done to get to the correct acoustic "target" response. Often, these speakers are less sensitive that a typical speaker because only "cuts" can be made with passive crossovers, rather than cuts and boosts as is the case with active crossovers.
There are some other crossovers (special ones) that can yield transient accurate (time coherent) response:
- Subtractive delay
- Filler Driver (originally by B&O)
- in DSP-based crossovers with "phase correction" algorithm
I've used the DSP based approach with an emulated 4th order LR crossover c/w phase correction and can obtain the transient response from measurements, comparing the 'before' and 'after' results. This technology does work, but there are some costs. First off, in the DSP realm it requires a time delay since you can't move the delayed part of the spectrum forward in time, but you can move the "relatively forward" parts of the spectrum BACK in time with a special kind of delay. They call it "forward/reverse processing" in that you pre-mangle the group delay so that when the speaker "re-mangles" it, the phase distortion is removed.
Neat stuff.
Cheers,
Presto
to add to what Airtime has said-
if the goal is to generate a series of wave forms that approximate the real live event- they all transducers/drivers should be 'aligned' that is working together- if a driver is Out of Phase (usually expressed in degrees - like a circle) then the contributions of that driver will at best be diminished, at worse cancel out the efforts of other drivers where frequencies overlap-
Happy Listening
Phase is just a measure of time relative to frequency. You can configure a multi-way loudspeaker to have a nice Phase response (Step response in a Stereophile Review) for s place is space relative to the driver. All other places the phase response differs... it's not in phase.
The brain uses timing variation between the ears with similar signals in the mid and low frequencies to assist in location. It is fairly precise tool for live sound, though it can be fooled as well. In stereo reproduction, however, the brain isn't sensing "Real" it is sensing "fake" and has to process what "might be real" and imagine it!
How the individual signals are recorded, and modified in the production process virtually never preserves phase (Binaural and a other specialized methods excluded). Nor the producer and engineer have to create the illusion of direction and do so using many techniques.
In the end, however, playback systems that preserve phase and those that don't can do a fine job of assisting us listeners in creating the illusion.
"The hardest thing of all is to find a black cat in a dark room, especially if there is no cat" - Confucius
An example of phase knowledge is in crossover designs. Mainly to the tweeter and to a lesser extent to the old Midranges - remember them? This is to compensate for the crossover becoming a problem and causing a phase problem. ORR to diminish the added padding/crossover components simply by reversing the wire leads.
Yes, I remember considerable ink spilled over the 'slope' of a crossover, and whether it inverted phase - or merely shifted it (90 degrees versus 180)...
Then there were the questions about gain stages...
and who could forget the "Wood Effect"...
Happy Listening
In a simple nut shell it means which way the drivers are moving at what time. Out of phase generally means while one driver is moving out the other one is moving in.
Real world translation. One driver sounds softer than the other - not a good thing unless that is what you are trying to do.
A single driver will exhibit varying phase at different frequencies. This is all because you're asking spring loaded devices with mass to follow an electrical waveform.
Exactly true, thanks!
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: