You mean like measuring at high levels or rather dynamic compression?? It's out there, and it's old!!
No, with a signal more like real music that starts and stops. I don't know why people would think distortion is always at the same level. It must go up and and down.
I found this paper by Dan Foley of Audio Precision quite instructive. Of particular interest the capture of higher order harmonics (via FFT) and their audibility even at very low levels as compared to lower order harmonics.
http://www.almainternational.org/ya...tortion_Analysis_Methods_5-2017.171105858.pdf
I have to see how far down this path I can get with what I have.
Thanks for the nudges all.
Barry.
Yea I read it, and thanks for posting it. He does seem to indicate that a musical signal is different, but he doesn't seem to draw any meaningful conclusions. Pretty sure he's measuring a signal anyways, actual speakers have much higher distortion.
Last edited:
When you refer to group delay being more audible (and I presume, more offensive) at higher SPL, I am going to assume that you mean non-flatness of group delay, since equal group delay across the spectrum is just a delay, like sitting further from the speaker...
In that case, what kinds of non-flatness of group delay would you say are the most audible or offensive at high SPL ? Is there a threshold effect where any group delay below X milliseconds is not audible ?
Or does the rate of change of the group delay with frequency also matter ? Are rapid swings in group delay over a small frequency range more audible than the same peak change in group delay spread out over a much wider frequency range ?
I think you missed my point a little though. I don't disagree with your premise, in fact I agree with it, however my remark was to question whether there was anything magical about the "bad" group delay variations caused by diffraction, or whether any similar group delay variations sound bad at high SPL regardless of cause.
In other words, could peaky group delay caused by resonances be just as offensive as peaky group delay caused by diffraction ? I believe the answer to this is yes, but I'm interested to see your position on this.
The difference between the two scenarios of course is that group delay variations caused by driver resonances will tend to show up in the power response in the room and over a wider range of listening angles, while group delay variations caused by diffraction will tend to be averaged out and appear little if at all in the room power response, leaving their effect being primarily in the direct wave and early reflections like sidewall reflections.
Would that be a fair statement ? Diffraction is after all a polar response problem. Does this difference affect which would be more audible in a room ?
One other question I have is that in the work you reference you say that group delay (variations ?) are more audible at higher SPL - which I agree with, however is this not the same thing as saying that a peaky, non-flat frequency response is more audible at higher SPL ?
Since most drivers are minimum phase, a peaky, non flat group delay automatically implies a peaky, non flat frequency response.
What was done in the research to conclude that it was the delay variations that were audible and offensive at high SPL and not that it was the amplitude variations, and perhaps the ringing in the time domain that would accompany it that was the reason for audibility ?
If they are one and the same thing, does it even matter if it is group delay or frequency response variations that we are hearing, if it is the exact same impulse response ?
I'm more objectivist than subjectivist, I'd say I'm at least 80% into the objectivist, measurement based camp, but I don't discount my subjective impressions either, and I am constantly comparing measurements and how it sounds to try to correlate them, and learning what effect a specific measured change to the response has subjectively.
At the end of the day I want to enjoy my listening as well, it's not just an exercise in making a flat line on a frequency response graph, although that's a good place to start, and in this particular design I have only strayed very slightly from my initial target.
And truth be known, the corrections I made in the lower midrange and upper midrange are within my measurement error range anyway, (less than half a dB) so may be a sign that my ears have picked up an error in the response that is a result of designing the crossover based on a measurement in error, (especially the splicing of the woofer nearfield response) rather than me deliberately tuning the response away from flat just because I think it sounds better that way.
I think it's a really interesting topic, and it gets the pure objectivists really wound up - they simply don't believe that the ear can hear something as "small" as half a dB change over an octave.
Over some parts of the spectrum, bass in particular, you can't, but not all frequencies are created equal when it comes to how discerning we are of them.
The ones that are more fussy, where seemingly small changes can be picked out, are those that help form the "image" of the sound, and there are certain frequency ranges whose relative balance to each other is unduly critical.
Such as the balance between lower midrange 200-400ish and the presence region, 2-4Khz, and the balance between the presence region and the top end treble. The balance between these three bands in particular is remarkably fussy to a fraction of a dB.
And when you are making adjustments very close to flat, you're not hearing a difference in tonal balance as such - if you asked me to pick out a half dB shift in 200-400Hz on pink noise in a blind test I doubt that I could.
But play a familiar piece of music that makes use of that frequency range in a certain way to form a cohesive image and I can easily pick out that it's right or wrong without even hearing the reference version of the signal, because the imaging will sound wrong!
So I sort of agree with the pure objectivists who say that we can't pick out a tonal imbalance this small, which is true, certainly on something like pink noise, but when it comes to the effect it has on the imaging, with speech or music, I disagree.
The formation of a cohesive image of a sound source relies partly on the HRTF related changes that our ears cause (as well as other factors like timing etc) and the reality is that some of the frequency response variations caused by the HRTF are really subtle for a significant angular change of sound source.
So our hearing system is extremely sensitive to these very subtle but very specific frequency response "errors" that correspond to localization information. So it doesn't take much of an error at specific frequencies (especially over a whole octave or more) to ruin the cohesiveness of the image.
If you're lucky (or unlucky depending on your viewpoint) your speakers unwanted frequency response variations may correspond to a real HRTF response, which may cause the sound to still image well but seem to come from somewhere it isn't.
A good example is exaggerated response at around 10Khz tends to make the source of the sound seem higher than the speakers, since that frequency gets exaggerated by our HRTF when the elevation of the sound source is above us. I've heard stereo images that seem like they're half way between the speakers and ceiling due to this effect - nice party trick, but not really what we're trying to achieve.
However if the frequency response errors of your speaker don't combine to a "valid" HRTF then all that happens is the imaging becomes unfocused, confused, or is lost altogether - because the brain gets multiple conflicting frequency response cues that don't correspond to any spatial direction that it has learned over its lifetime.
That "confused" image is a clear sign that something is not quite right with the frequency response, assuming the recording is good, and that's what you and I notice when we are tweaking the response slightly.
Over the weekend I did an interesting test - I said earlier I have not touched the baffle step attenuator resistance for the woofer - I've tweaked the rollover point significantly but not the actual amount of midrange attenuation.
So I thought I would lift the midrange slightly to see what effect it had and whether I really had the current level right or not. The midrange is about 3 octave so I know from past experience a +/- 0.5dB change over that 3 octaves is as plain as day (again, mainly on imaging) so I deliberately made a small change that I thought would be just on the borderline of audibility or perhaps below it by increasing it by 0.2dB.
I can see the pure objectivists rolling their eyes already, but I genuinely did not expect to notice much if any difference either.
To my surprise I did notice a difference immediately on familiar music, it sounded pleasant enough on many recordings and was a little bit more "open" sounding on some, but rather than adding something, it sounded more like something had been taken away. And some songs were just slightly pushy sounding in the midrange.
I listened for maybe an hour trying to analyse what it was I didn't like about it, and I guess for want of a better word it just sounded a bit sterile and "flat".
It wasn't until I reversed the change and listened again to some of the same songs I immediately recognised what a change it had made to the imaging.
It wasn't until this point that I realised that the "sterile" increased midrange configuration had stripped quite a bit of the presence and "air" from the image, and simultaneously some of its warmth, and made it a lot drier and less pleasant to listen to.
When you consider the frequency response change that resulted, this makes sense but I was not expecting that much of an obvious problem with the image just by having the midrange 0.2dB too high... kinda gobsmacked really.
0.2dB is well within my measurement error and when I am optimising the filter coefficients for the crossover in the sim I'm trying to make a very bumpy line into a less bumpy line, and it's a bit of a judgement call as to how you shape that bumpy line to the target. You could easily be over or under on average by 0.2dB or realistically a lot more than that when you're doing manual curve fitting with your eyes...
So in hindsight I think a little bit of luck was involved in getting the level spot on the first time!
Last edited:
Why not just use kerfs or some other technique like covering your baffle in fake fur? One of the ways we localize sounds are by diffraction so you can't really calculate how our brains will perceive the sound, so just get rid of as much of it as you possibly can, and with a waveguide you mostly there.
The problem with fur, apart from the aesthetic issues, is that it's only effective at pretty high frequencies, where the depth of the fur is a significant fraction of a wavelength. At the bottom of a tweeters response, say 3Khz, it will have little effect.
I'd rather minimise diffraction using driver directivity. If you get that right it doesn't matter whether your baffle has sharp or rounded edges or is furry or not...
I'm not surprised at all you could hear that over three octaves.
One of the things that got me interested in audio is the mystery of it.
But there's always some doofus who will hear words like "objective" and read how people spend a lot of money on wires and then pick a side because they want be "scientific" when all they are is just completely biased.
The fact remains that there isn't a man made device that can do what an ear can. We can identify amazingly precise details with training and experience (which blind testing generally doesn't take into account).
OTOH we can also turn the tv on and think we hear it only to realize its actually off and we hit the wrong button. Its perception, and at the end of the day no one really knows.
Last edited:
Ive worked for scientists most of my life . Much of it in physics world. In the design seat this is what ive learned.. while I need the "pure objectionists" knowledge, I have learned that in the end They still have belief systems that often prove over time to be missing something. Sometimes that something is small, sometimes, despite what seems obvious from the data, whats missing is HUGE and directly related to what they believe.
This is in no way a dig against anyone specifically here. Its just that IMO there really is no such thing as a "pure objectionist". Every human has finite knowledge and somewhere in our conclusion is a belief. Its simply unavoidable.
This is in no way a dig against anyone specifically here. Its just that IMO there really is no such thing as a "pure objectionist". Every human has finite knowledge and somewhere in our conclusion is a belief. Its simply unavoidable.
+1, yea its not as much on this site, but there is a lot of pretentious objectivism out there, especially with millennials, the tide pod generation.
Nice freudian slip.
I think you meant objectivists, but objectionists might be more apt sometimes. As for my own position, although I consider myself to be primarily in the objective camp when it comes to speaker measurements and design - I still hear what I hear. However I seek to understand in measurements what I hear, instead of waving my hands and saying "measurements can't explain what I hear", as many in the subjectivist camp try to. I'd rather try to figure out what it is I'm hearing and how it can in fact be measured and characterised if you know what to look for. Meanwhile some in the objectivist camp say "I can't see anything statistically significant in the measurements, it must be all in your head". I think the truth is somewhere in between...
Last edited:
Air pressure changes are not sine waves
This is fundamentally wrong.
I believe its also the root of the last 50 years or so of stagnation in loudspeaker development ie modern £million pound plus systems still can not fool the ear brain in the "is it live or is it a recording?"
Assuming we are talking about air borne sound and not water / structural borne sound...
"Ultimately" sound is just air pressure variations.... Increases or decreases above / below ambient pressure.
Nothing more nothing less.
Snapping twig / piano hammer / vocal cord contraction etc,...
They all generate a point source of increased air pressure (compression) which travels out as a spherical pressure wave.
Followed by a decrease in air pressure (rarefaction), then an increase back to ambient air pressure.
There are only two variables:
(1) Level of increase / decrease air pressure. (perceived by our ear brain as volume changes)
(2) TIME. ie the rise time of the impulse and the decay time of impulse.
The impulses may be in rapid succession ( high frequency) or slower ( lower frequency) but the nature of the impulse remains the same ie only compression or rarefaction orf air.
It is more accurate to think of sound as a binary on / off event NOT a continuous sine wave event.
In nature sine waves do not exist...!! only in signal generators!
This is fundamentally wrong.
I believe its also the root of the last 50 years or so of stagnation in loudspeaker development ie modern £million pound plus systems still can not fool the ear brain in the "is it live or is it a recording?"
Assuming we are talking about air borne sound and not water / structural borne sound...
"Ultimately" sound is just air pressure variations.... Increases or decreases above / below ambient pressure.
Nothing more nothing less.
Snapping twig / piano hammer / vocal cord contraction etc,...
They all generate a point source of increased air pressure (compression) which travels out as a spherical pressure wave.
Followed by a decrease in air pressure (rarefaction), then an increase back to ambient air pressure.
There are only two variables:
(1) Level of increase / decrease air pressure. (perceived by our ear brain as volume changes)
(2) TIME. ie the rise time of the impulse and the decay time of impulse.
The impulses may be in rapid succession ( high frequency) or slower ( lower frequency) but the nature of the impulse remains the same ie only compression or rarefaction orf air.
It is more accurate to think of sound as a binary on / off event NOT a continuous sine wave event.
In nature sine waves do not exist...!! only in signal generators!
Nice freudian slip.
As for my own position, although I consider myself to be primarily in the objective camp when it comes to speaker measurements and design - I still hear what I hear. However I seek to understand in measurements what I hear, instead of waving my hands and saying "measurements can't explain what I hear", as many in the subjectivist camp try to. I'd rather try to figure out what it is I'm hearing and how it can in fact be measured and characterised if you know what to look for. Meanwhile some in the objectivist camp say "I can't see anything statistically significant in the measurements, it must be all in your head". I think the truth is somewhere in between...
Ha, good catch on the slip, and I agree btw
Nature abounds with sine waves, in every domain. The Universe may even consist of them entirely, just add enough dimensions.
Yes, that is precisely correct - I was thinking about this myself.
Constant delay is not audible but variations in the delay of different components of a sound would be.
Unknown - never been tested as far as I know.
I have long thought that this would be the case, but it was not what we studied. We looked at HP delayed signals, so they would have group delay and constant delay - they simulated horn diffraction. The factors were; delay - perception got greater with greater delay; level of effect - perception got greater with a greater amount of the HP delayed signal added; level of playback - perception got greater with greater playback level.
This is all I know and you will have to glean out many of the answer to the questions below from the data.
Don't forget that diffraction has excess delay which resonances don't have. Hence, I believe that diffraction is worse.
You would not be wrong is thinking that the evidence implies that, but it is not conclusive.
We did an experimental design which could separate out the three factor using ANOVA.
I hope you don't think that diffraction is minimum phase, because it can't be. It has excess delay which modifies the phase. For instance adding in a peaked response with no excess delay will create a different net result than adding in the same peaked response with excess delay. The two things cannot both be minimum phase.
I simply have to agree to this. Even though I probably fall in the objective camp in general I do realise I still have a large subjective side to me. Preparing to be wrong is something I try to be comfortable with
. However, as this is a hobby for me, and I'm not out to please a crowd, just pleasing me is enough, I do aim, like many here to learn from it and link what I hear/perceive and observe to my measurements.
So far the route of measurements has brought great things. But there's still a fling of mystery too.
I do recognise myself in the stories of small level tweaks over a slightly larger area making rather large perceptive differences at times.
My objectivist side tries to lock down as many variables as I can, but that brain of mine is much harder to control
.
Read the whole post....
You must have missed the opening line....
"Assuming we are talking about air borne sound and not water / structural borne sound..."
There are no sine waves in the natural world of air borne sound as air cannot sustain a compression (or rarefaction) indefinately ie they always decay back to ambient.
Also worth noting is that these Air Pressure Events (APE's) have defined start and stop times ie they are time critical.
In fact our HAS (Human Auditory System) prioritises the critical time domain information over the less important frequency domain information....
This is because we have evolved to locate where a sound is coming from in 3D space before we identify what that sound is.... Our fight or flight response is a survival mechanism.
You must have missed the opening line....
"Assuming we are talking about air borne sound and not water / structural borne sound..."
There are no sine waves in the natural world of air borne sound as air cannot sustain a compression (or rarefaction) indefinately ie they always decay back to ambient.
Also worth noting is that these Air Pressure Events (APE's) have defined start and stop times ie they are time critical.
In fact our HAS (Human Auditory System) prioritises the critical time domain information over the less important frequency domain information....
This is because we have evolved to locate where a sound is coming from in 3D space before we identify what that sound is.... Our fight or flight response is a survival mechanism.
Last edited: