ProSoundWeb Community
Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => LAB: The Classic Live Audio Board => Topic started by: Merlijn van Veen on October 08, 2015, 05:03:22 AM
-
Hi everybody,
I wrote an article about the audibility of phase shift and how in typical sound systems the harmonics of low frequency sources e.g. electric bass guitar, lead the fundamental by as little as 20 ms (7 meters) or more. You can read it at the link below.
Regards,
Merlijn
https://www.merlijnvanveen.nl/en/study-hall/100-the-harmonics-lead-the-fundamental (https://www.merlijnvanveen.nl/en/study-hall/100-the-harmonics-lead-the-fundamental)
-
Hi everybody,
I wrote an article about the audibility of phase shift and how in typical sound systems the harmonics of low frequency sources e.g. electric bass guitar, lead the fundamental by as little as 20 ms (7 meters) or more. You can read it at the link below.
Regards,
Merlijn
https://www.merlijnvanveen.nl/en/study-hall/100-the-harmonics-lead-the-fundamental (https://www.merlijnvanveen.nl/en/study-hall/100-the-harmonics-lead-the-fundamental)
This reminds me of the contentious debate about audibility of absolute polarity decades ago.
When you speak of harmonics leading the fundamental are you talking about the arrival time for initial transient or the relatively steady tone as the plucked string decays? For the steady state part of the waveform it's relative whether the overtones are leading or lagging. For the initial attack it can make a difference regarding localization.
Phase relationships can affect the shape and symmetry of the compound waveforms. In theory this should not make a difference, but it's easy to imagine asymmetrical signal limits in our mechanical system (ears).
Human hearing is reported to be most sensitive at perceiving such errors in vocals.
JR
PS: I do not claim to hear any of this stuff...
-
Hi everybody,
I wrote an article about the audibility of phase shift and how in typical sound systems the harmonics of low frequency sources e.g. electric bass guitar, lead the fundamental by as little as 20 ms (7 meters) or more. You can read it at the link below.
Regards,
Merlijn
Nice demo video!
In the article you wrote:
"That being said, I know of a few people that I hold in high regards e.g. Mauricio "Magu" Ramírez of Meyer Sound and François "Frankie" Desjardins of Solotech, that have been fortunate enough to actually hear real speakers with linear "flat" phase behavior from virtually DC to light."
You probably have listened to Sony MD 7506 headphones, in which case you also have been fortunate enough to hear real speakers with linear "flat" phase behavior over the audible range.
Listening back, even using the 7506 phones, I couldn't detect the phase shifts...
Art
-
In the sound systems I have aligned for theatrical applications, I have noted a correlation between the systems with the most phase shift, or wraps, through the mid range, and those with the poorest dialog intelligibility. However, yes, correlation is not causation. There are at least a thousand variables not controlled for. What does one do with an observation like that? I dunno, do what one can to avoid purple speakers perhaps.
-
This reminds me of the contentious debate about audibility of absolute polarity decades ago.
When you speak of harmonics leading the fundamental are you talking about the arrival time for initial transient or the relatively steady tone as the plucked string decays? For the steady state part of the waveform it's relative whether the overtones are leading or lagging. For the initial attack it can make a difference regarding localization.
Phase relationships can affect the shape and symmetry of the compound waveforms. In theory this should not make a difference, but it's easy to imagine asymmetrical signal limits in our mechanical system (ears).
Human hearing is reported to be most sensitive at perceiving such errors in vocals.
JR
PS: I do not claim to hear any of this stuff...
Hi John,
AFAICT the sequence between the harmonics' transients and steady states is primarily a property of the instrument or source. The speaker (as one of the links in the signal chain) doesn't differentiate. The group delay of the speaker is technically non-linear behavior and changes the natural order of things. Being most severe in LF range I can imagine a noticeable change without it.
Verzonden vanaf mijn iPad met Tapatalk
-
In the sound systems I have aligned for theatrical applications, I have noted a correlation between the systems with the most phase shift, or wraps, through the mid range, and those with the poorest dialog intelligibility. However, yes, correlation is not causation. There are at least a thousand variables not controlled for. What does one do with an observation like that? I dunno, do what one can to avoid purple speakers perhaps.
As non-linear behavior, less group delay is off course commendable but should be realized within reason. With more than one speaker in installs, relative phase offset between various makes and models is the more likely culprit for impaired intelligibility.
In this corner of the world, most theaters' fill speakers exhibit different phase behavior in comparison to the main speakers which can be treated most of the with careful use of all-pass filters.
Verzonden vanaf mijn iPad met Tapatalk
-
Nice demo video!
In the article you wrote:
"That being said, I know of a few people that I hold in high regards e.g. Mauricio "Magu" Ramírez of Meyer Sound and François "Frankie" Desjardins of Solotech, that have been fortunate enough to actually hear real speakers with linear "flat" phase behavior from virtually DC to light."
You probably have listened to Sony MD 7506 headphones, in which case you also have been fortunate enough to hear real speakers with linear "flat" phase behavior over the audible range.
Listening back, even using the 7506 phones, I couldn't detect the phase shifts...
Art
Hi Art,
Thanks for the compliment and the revealing tip!
I'm anxious to experience those cans with renewed interest :-)
Verzonden vanaf mijn iPad met Tapatalk
-
This reminds me of the contentious debate about audibility of absolute polarity decades ago.
When you speak of harmonics leading the fundamental are you talking about the arrival time for initial transient or the relatively steady tone as the plucked string decays? For the steady state part of the waveform it's relative whether the overtones are leading or lagging. For the initial attack it can make a difference regarding localization.
Phase relationships can affect the shape and symmetry of the compound waveforms. In theory this should not make a difference, but it's easy to imagine asymmetrical signal limits in our mechanical system (ears).
Human hearing is reported to be most sensitive at perceiving such errors in vocals.
JR
PS: I do not claim to hear any of this stuff...
Nice article & video Merlijn (like all your work).
I have also been doing a little work on this subject.
I posted plans for a speaker I built on Soundforums.net https://soundforums.net/threads/12075-60-Degree-DIY-Mid-Hi
It has an almost perfect impulse response. The amplitude and phase response is almost perfectly flat from about 200Hz up. It will produce a reasonable looking square wave from about 100Hz, both on and off axis …. And it goes LOUD
To my ear the flat phase response version doesn’t really sound any better, just more real. It’s like the difference looking out through a window with clean class and opening the window (no glass) if that make sense. You notice that with a minimum amount of phase wrap the drum and bass sounds are very “tight” and have a lot of impact.
My experience even though the phase response was not perfectly flat to DC matched what Mauricio and François described.
-
You notice that with a minimum amount of phase wrap the drum and bass sounds are very “tight” and have a lot of impact.
This is why higher-order bandpass woofer designs fail to pass sonic muster.
-
To my ear the flat phase response version doesn’t really sound any better, just more real. It’s like the difference looking out through a window with clean class and opening the window (no glass) if that make sense. You notice that with a minimum amount of phase wrap the drum and bass sounds are very “tight” and have a lot of impact.
My experience even though the phase response was not perfectly flat to DC matched what Mauricio and François described.
Back decades ago when I was designing hifi gear, I used to joke if you can't hear the improvement through the screen door while standing in the yard, it isn't good enough. :o
Subtle subjective phenomenon are always difficult to get a handle on.
One test I've heard speaker designers use is to loop through a system for multiple passes. Modern digital recorders should be linear enough, so the mic becomes the only other variable and mics are generally better than speakers. After running a recording through a system (better yet while in an anechoic environment) for multiple passes, any errors will accumulate and become easier to identify.
A square wave seems like a difficult gold standard. Do you see Gibbs phenomenon due to LPF scrubbing off upper harmonics?
JR
-
One test I've heard speaker designers use is to loop through a system for multiple passes.
This was common "back in the day" with reel to reel recorders and "generational loss".
We did this awile back.
It is REALLY interesting that after just a few (think 4) passes, you can't even recognize the song anymore with some speakers.
Yes the "errors" pile up on top of each other.
It makes you feel really bad about how "accurate" loudspeakers actually are.
Yes the mic used to make the recordings will also pile on the errors. In this case we used an Earthworks M30-a pretty flat and accurate mic.
-
Back decades ago when I was designing hifi gear, I used to joke if you can't hear the improvement through the screen door while standing in the yard, it isn't good enough. :o
Subtle subjective phenomenon are always difficult to get a handle on.
One test I've heard speaker designers use is to loop through a system for multiple passes. Modern digital recorders should be linear enough, so the mic becomes the only other variable and mics are generally better than speakers. After running a recording through a system (better yet while in an anechoic environment) for multiple passes, any errors will accumulate and become easier to identify.
A square wave seems like a difficult gold standard. Do you see Gibbs phenomenon due to LPF scrubbing off upper harmonics?
JR
My measurment system, environment and speaker is not that good that you can see that :-)
FWIW at high frequencies the FFT is reduced to just the fundamental (f) + 1/3 cos 3f.
-
My measuring system, environment and speaker is not that good that you can see that :-)
FWIW at high frequencies the FFT is reduced to just the fundamental (f) + 1/3 cos 3f.
-
I don't expect to see much Gibbs at 100Hz... Good job...
JR
-
I don't expect to see much Gibbs at 100Hz... Good job...
JR
I don't have any copies of higher frequencies ... but here is the impulse response.
-
I recall in the early days of CD players there was a lot of angst about the apparent pre-ring of Gibbs...
I think (hope) the phools have finally gotten over that and moved on.
JR
-
Yep, limited bandwidth. To few harmonics. Love this video.
https://youtu.be/cIQ9IXSUzuM
-
This is purely a subjective opinion.
We have a couple of products in which if you flip the polarity of the HF driver, the response does not change much (just the "wiggles" are a little different) but the phase goes through a 180° shift. The response is a "little" bit flatter with the phase shift-but not much.
I have set up tests on this in which I simply hook up a DPDT switch so that the polarity can easily/quickly be changed.
I got a group of "ears" together and would play a number of different tracks, and all they knew was that one "sound" was A and the other was B.
They had no idea what I was doing by flipping the switch.
They all picked the position in which the phase as flat through crossover.
I know it was not real "scientific", but when given the choice-the "picky guys" chose the flatter phase response vs amplitude.
-
This is purely a subjective opinion.
We have a couple of products in which if you flip the polarity of the HF driver, the response does not change much (just the "wiggles" are a little different) but the phase goes through a 180° shift. The response is a "little" bit flatter with the phase shift-but not much.
I have set up tests on this in which I simply hook up a DPDT switch so that the polarity can easily/quickly be changed.
I got a group of "ears" together and would play a number of different tracks, and all they knew was that one "sound" was A and the other was B.
They had no idea what I was doing by flipping the switch.
They all picked the position in which the phase as flat through crossover.
I know it was not real "scientific", but when given the choice-the "picky guys" chose the flatter phase response vs amplitude.
In general amplitude errors as small as a fraction of a dB in the sensitive midrange can be audible especially for short term A/B tests.
It's good that they preferred the "as designed" mode.... ;D
JR
-
This is purely a subjective opinion.
We have a couple of products in which if you flip the polarity of the HF driver, the response does not change much (just the "wiggles" are a little different) but the phase goes through a 180° shift. The response is a "little" bit flatter with the phase shift-but not much.
I have set up tests on this in which I simply hook up a DPDT switch so that the polarity can easily/quickly be changed.
I got a group of "ears" together and would play a number of different tracks, and all they knew was that one "sound" was A and the other was B.
They had no idea what I was doing by flipping the switch.
They all picked the position in which the phase as flat through crossover.
I know it was not real "scientific", but when given the choice-the "picky guys" chose the flatter phase response vs amplitude.
Hi Ivan, all
The thing is, listening with headphones to an all pass filter or addition of a simple high pass and low pass filter electronically which acts like an all pass filter, these produce a change in the time or Z domain and nothing else. In the hifi world, this test was used to argue you can’t really hear crossover phase shift or more correctly, it is only detectable / audible under some conditions, with some kinds of music.
Loudspeaker drivers on the other hand radiate in 3 dimensions X, Y and Z and so most often two drivers rarely add coherently like two electrical signals do when added through resistors etc. More often two drivers radiate (if more than about ¼ to 1/3 wavelength apart) as two independent sources. This can be seen if one does a polar plot of the two, where they interact, they usually produce an interference pattern recognized by a series of lobes and nulls (one might think of this as comb filtering in 3d).
In that case, what you measure so far as phase depends on the filters but also the individual path lengths to each source and the latter part makes what you get depend on location. As a result of radiating in 3 dimensions. The phase response may or may not be a good predictor of what you hear, two system which measure identically may sound quite different if they spatial radiation is different for each.
On the other hand, with Ivan's switch test, you ARE dealing with the relationship between individual sources and so it is more complicated than the coherent addition case using headphones and now how these sources also radiate as a sum in X and Y can be altered by that switch change or not altered if they add coherently (less than ¼ wl apart).
Also, one can use FIR filters to electronically alter the Z relationship however the sources still interact in the X and Y domains as well so when there are problems (incoherent addition) in X and Y, the DSP correction only works for the range on axis or where the data was taken to derive the correction and not globally.
The loudspeaker generation loss testing we did starting 11 years or so ago was a good reality check and did several things.
First, it separates you from your automatic hearing process which seeks information and discards problems without your awareness. It more easily lets you hear a loudspeakers warts, even just listening to a loudspeaker with a good measurement mic and headphones will usually let you hear flaws which you would have a harder time hearing straight up. Second, it’s a good indicator of how faithful the speaker is in reproducing the music your using, obviously the more generations you can go, the more faithful it is. Third it is humbling reality check so far as showing how bad loudspeakers are compared to every other part of the signal chain. Make a speaker that sounds good on the third pass and it will sound unusually good listening directly. Also, often when you hear a speaker that sounds bad through a measurement mic, once you become attune to the flaws, you will also be able to hear them much more clearly without the headphones.
Intelligibility; There is an article on the fort page at PSW written by Pat Brown. It mentioned a seminar he conducted which was an eye opener and like every Synaudcon seminar I have attended, I learned something(s)..
Inteligibility unlike musical quality is measureable and thanks to the STIpa measurement, there is a window to what hurts intelligibility. STIpa is measured using a speech bandwidth version of the optical MTF or modulation transfer function where the signal is rapidly gated on and off and the corruption of the off periods is what reduces intelligibility.
The optical version is here;
http://www.edmundoptics.com/technical-resources-center/optics/modulation-transfer-function/
https://en.wikipedia.org/wiki/Optical_transfer_function
One can picture that preservation of time (related to phase) is key to this but also in a room, the directivity is equally important as reflected sound also corrupts the “off” periods. This is partly set by how the multiple sources interact, where the lobes and nulls go and go on to produce delayed arrivals. Interestingly, intelligibility is not always important for music reproduction, for example a choir in a large church there is hardly any information which is time sensitive and a large reverberant room sounds good.
On the other hand, if the music has transient information or a lot of detail, then it matters a lot as that stuff can be lost and is usually lost in large array based sound systems due to the scattering of time information and reflected sound.
Ivan gets to go to large stadium installations pretty often (we have speakers in about half of the 100,000+ stadiums so far) but I don’t get to go that much so for me, it has been a lot of fun hearing the difference preserving time / phase can make vs the live sound style array systems they replaced.
I have made a few recordings with a cheapo canon vixia camcorder which might be fun if you have headphones on your computer. These demonstrate both effect of the coherent driver addition, phase and directivity so far as its impact on intelligibility and musical articulation.
Like any system though, especially one that is signal faithful, stinky sonic garbage in equals stinky sonic garbage out.
Try these with headphones, the loudspeaker are all in the scoreboards
https://www.dropbox.com/s/0u58zsinsz1gd4j/20140805114837.mts?dl=0
https://www.dropbox.com/s/gwlxss1uxioi4mq/20140805115158.mts?dl=0
https://www.dropbox.com/s/tnsw5mb4v5vdlwq/20120726122124.mts?dl=0
https://www.dropbox.com/s/2cl3l4blx6r9r11/20120726114539.mts?dl=0
https://www.dropbox.com/s/va4mihvefqyxk24/20130723140018.mts?dl=0
https://www.dropbox.com/s/oyosfc3adc6j1du/20130723135350.mts?dl=0
Best Regards
Tom
-
In general amplitude errors as small as a fraction of a dB in the sensitive midrange can be audible especially for short term A/B tests.
It's good that they preferred the "as designed" mode.... ;D
JR
I have done other side by side tests with "identical" speakers.
Depending on the track that was played-people would prefer one over the other. Play a different track and get a different result.
At least the people agreed.
Yes, I agree that under controlled conditions, even a few tenths of a dB can make an audible difference.
Everywhere from the pick on the string, to the sound of the rosin on a box, to the spit in the voice.
What is truly accurate is hard to say.
-
Nice article & video Merlijn (like all your work).
I have also been doing a little work on this subject.
I posted plans for a speaker I built on Soundforums.net https://soundforums.net/threads/12075-60-Degree-DIY-Mid-Hi
It has an almost perfect impulse response. The amplitude and phase response is almost perfectly flat from about 200Hz up. It will produce a reasonable looking square wave from about 100Hz, both on and off axis …. And it goes LOUD
Peter,
Your measurements of your speakers look incredible and I'm impressed with what you've come up with. You say that it maintains a good looking square wave on AND off axis. Can you elaborate on how much variation there is when going off axis vertically? I would assume it varies more with frequency when at the extremes of dispersion than it does on axis.
-
Peter,
Your measurements of your speakers look incredible and I'm impressed with what you've come up with. You say that it maintains a good looking square wave on AND off axis. Can you elaborate on how much variation there is when going off axis vertically? I would assume it varies more with frequency when at the extremes of dispersion than it does on axis.
The vertical axis falls apart the quickest. Once you start to come off axis more than about 15 -20 degrees the frequency response becomes too uneven ... but within reasonable limits a square can be maintained.
Here are some examples - I think they are around 400Hz
top picture off axis - horizontal
bottom picture off axis - vertical
-
Hi Ivan, all
The thing is, listening with headphones to an all pass filter or addition of a simple high pass and low pass filter electronically which acts like an all pass filter, these produce a change in the time or Z domain and nothing else. In the hifi world, this test was used to argue you can’t really hear crossover phase shift or more correctly, it is only detectable / audible under some conditions, with some kinds of music.
Best Regards
Tom
Hi Tom,
Forgive my ignorance but I fail to see how physical displacement between drivers introducing phase OFFSET, affects the audibility of phase SHIFT or the D/R ratio of different coverage patterns for the matter?
The HPF/LPF examples in my video were only there to prove that the plugins actually do something for those wondering.
Sincerely and wi all due respect,
Merlijn
Verzonden vanaf mijn iPad met Tapatalk
-
Hi Tom,
Forgive my ignorance but I fail to see how physical displacement between drivers introducing phase OFFSET, affects the audibility of phase SHIFT or the D/R ratio of different coverage patterns for the matter?
The HPF/LPF examples in my video were only there to prove that the plugins actually do something for those wondering.
Sincerely and wi all due respect,
Merlijn
Verzonden vanaf mijn iPad met Tapatalk
The offset drivers will cause a different phase response at different seats.
This is due to the fact that the arrival distance from each driver to the listener will be different at different seats.
With a single source, the distance will remain the same at different seats.
-
The offset drivers will cause a different phase response at different seats.
This is due to the fact that the arrival distance from each driver to the listener will be different at different seats.
With a single source, the distance will remain the same at different seats.
Yes but AFAIK that's the average compound phase shift produced, by 2 or more drivers, other speakers, reflections, the acoustic source, etc...
It sympathizes with the strongest contributing signal, be it direct or indirect.
I chose my demo very carefully and intended to demonstrate how hard it is to perceive the phase shift by inserting e.g. an all-pass filter upstream.
This doesn't affect the inherent interaction by design between multiple drivers in an (a)symmetric n-way speaker, co-entrant or coaxial.
-
Yes but AFAIK that's the average compound phase shift produced, by 2 or more drivers, other speakers, reflections, the acoustic source, etc...
It sympathizes with the strongest contributing signal, be it direct or indirect.
I chose my demo very carefully and intended to demonstrate how hard it is to perceive the phase shift by inserting e.g. an all-pass filter upstream.
This doesn't affect the inherent interaction by design between multiple drivers in an (a)symmetric n-way speaker, co-entrant or coaxial.
Hi Merljin
What I was trying to get at was that the demo using an all pass filter or summing a high pass and low pass crossover electrically which also produces an “all pass” phase response is valid for the case where the sources combine coherently into one new signal, like adding two signals through resistors.
Two acoustic sources can add coherently / unilaterally as well BUT only when they are close together acoustically like close coupled subwoofers (at a spacing less than about ¼ wl apart at the highest frequency of concern) but more often they are farther apart and radiate as independent sources and the result depends on location as they are producing an interference pattern ( as shown by a series of lobes and nulls in the polar pattern).
This condition is often referred to as constructive and destructive interference and is not unilateral or coherent addition. The issue is that the pattern of lobes and nulls depends on the acoustic spacing of the sources and often lobes are produced in directions which are harmful or exacerbate the room acoustic problems.
In that case, it would be possible to have two loudspeakers with essentially the same magnitude and phase response (on axis) , one producing an interference pattern and one not, but when used in a room the two might sound pretty different as what one hears is a combination of direct and reflected or late sound and in a large venue like a stadium, the late sound can destroy or greatly limit the articulation and intelligibility of voices.
Also the complex interference pattern if present is very audible if the wind blows and in that the spectral balance changes with position and distance and usually limits the usable working distance of that system. This is very audible especially in large scale sound as one can hear the difference (absence of these effects) in those links of several large stadiums.
We used to do a lot of generation loss recordings to zero in on what we were doing at work with the idea being the more generations a loudspeaker could go through and still be listenable, the more faithful it was to music or voice. Using headphones to listen and a good measurement mic to sample the sound by passes your ear / brain system that seeks information and rejects problems but does so without ones awareness. With a good mic and headphones, one can hear the loudspeakers warts MUCH more clearly and while it doesn’t tell you what to fix or what’s wrong, it can be an arrow point in that direction.
Even a good loudspeaker will sound bad after just a few generations; they are by far the weakest link in the reproduction chain.
What you might want to try is to take your “all pass” test and see how many generations that takes before it is obvious as I have measured loudspeakers that had a thousand or more degrees of phase shift from high to low.
Hope that helps, i probably could have explained it better.
Best Regards,
Tom Danley