Re: Re: Re: Re: Re: SL: "Much is hypothesized, little is proven and much is overrated when it c
Using theory, your position is that it can’t be defined as min phase.
However, both empirically and analytically, it is observed that at one point in space, it is min phase. The difference can’t be attributed to relative amplitudes: the diffraction effect is easily observable in the composite amplitude, and therefore phase response.
How then does one reconcile the clear difference in observed behaviour, vs. the modeled behavior? Or do you agree that at one point in space, it can be viewed min phase?
This isn’t an academic discussion, for the following very important reason. That the diffraction is delayed doesn’t mean that it can’t be equalized with min phase EQ, at one point in space, if the composite response at the observation point is indeed min phase. By definition, it could be that this diffraction delay is the minimum delay possible to arrive at the composite frequency response, at that point.
I of course understand your position that the delay itself is paramount, and as the delay changes, so does the audibility and the residual subjective degradation. Sean Olive, Toole etc have done a great job defining what to listen for. However, I’m contending we can just eq it out with min phase (ie simple) eq, at one observation point.
I also fully appreciate your sentiment that all this may not matter, since this eq will be wrong over distributed space, being correct only at one point in space. That’s obvious from measuring speakers over angle, and from your analysis of the acoustic formulae. However, to use your standard of proof as the metric: I’d like to see subjectively tested proof of the importance of eq’ing the power response for diffraction, and that its not sufficient to just eq the first arrival at the observation point; thresholds, different conditions etc. Without this, it’s only an opinion with all the pifalls you pointed out earlier
Regarding decency: thanks, for the compliment. For me, it matters far less who is right, than what is right. Thanks for sticking with this. I appreciate that you’re living with a hectic schedule these days, and your responses are generous.
gedlee said:
Using theory, your position is that it can’t be defined as min phase.
However, both empirically and analytically, it is observed that at one point in space, it is min phase. The difference can’t be attributed to relative amplitudes: the diffraction effect is easily observable in the composite amplitude, and therefore phase response.
How then does one reconcile the clear difference in observed behaviour, vs. the modeled behavior? Or do you agree that at one point in space, it can be viewed min phase?
This isn’t an academic discussion, for the following very important reason. That the diffraction is delayed doesn’t mean that it can’t be equalized with min phase EQ, at one point in space, if the composite response at the observation point is indeed min phase. By definition, it could be that this diffraction delay is the minimum delay possible to arrive at the composite frequency response, at that point.
I of course understand your position that the delay itself is paramount, and as the delay changes, so does the audibility and the residual subjective degradation. Sean Olive, Toole etc have done a great job defining what to listen for. However, I’m contending we can just eq it out with min phase (ie simple) eq, at one observation point.
I also fully appreciate your sentiment that all this may not matter, since this eq will be wrong over distributed space, being correct only at one point in space. That’s obvious from measuring speakers over angle, and from your analysis of the acoustic formulae. However, to use your standard of proof as the metric: I’d like to see subjectively tested proof of the importance of eq’ing the power response for diffraction, and that its not sufficient to just eq the first arrival at the observation point; thresholds, different conditions etc. Without this, it’s only an opinion with all the pifalls you pointed out earlier
Regarding decency: thanks, for the compliment. For me, it matters far less who is right, than what is right. Thanks for sticking with this. I appreciate that you’re living with a hectic schedule these days, and your responses are generous.
Hi
Agree 100%.
On the other hand John K's "razor sharp" analysis separate two baffle effects that "have nothing to do" with each other and at least for me it was necessary to sharpen my view on that.
1) different path length from front and rear source due to the baffle beeing an obstacle
2) effects of edge diffraction due to different size / shape of the edge
Agree 100%
John K, thanks for presenting that simulations. It illuminates an aspect that EDGE doesn't tell us. It might be asked for too much but at the speed you set up that simulation couldn't you write a tool that extends EDGE by baffle thickness and speaker directivity over FR?
Greetings
Michael
Agree 100%.
On the other hand John K's "razor sharp" analysis separate two baffle effects that "have nothing to do" with each other and at least for me it was necessary to sharpen my view on that.
1) different path length from front and rear source due to the baffle beeing an obstacle
2) effects of edge diffraction due to different size / shape of the edge
Agree 100%
John K, thanks for presenting that simulations. It illuminates an aspect that EDGE doesn't tell us. It might be asked for too much but at the speed you set up that simulation couldn't you write a tool that extends EDGE by baffle thickness and speaker directivity over FR?
Greetings
Michael
DDR,
It's not something that needs EQ. Flattening overall response is relatively easy using baffle shape and dimensions. The problem is in the time domain, and the result is quite similar to the cause of image distortion with line arrays. With line arrays the effect is different in that the time difference resulting from the sound arriving last from the top and bottom drivers distorts the size of the image. With these closer to point source OB's, I find that the illumination of the edges of the baffle with the delayed sound makes it easier to localize the speakers instead of disappearing in the stereo illusion. I don't believe any type of EQ will change that.
It's not something that needs EQ. Flattening overall response is relatively easy using baffle shape and dimensions. The problem is in the time domain, and the result is quite similar to the cause of image distortion with line arrays. With line arrays the effect is different in that the time difference resulting from the sound arriving last from the top and bottom drivers distorts the size of the image. With these closer to point source OB's, I find that the illumination of the edges of the baffle with the delayed sound makes it easier to localize the speakers instead of disappearing in the stereo illusion. I don't believe any type of EQ will change that.
JohninCR,
Is the delayed sound that causes localization of drivers a standing wave phenomena on an OB?
With the drivers I have dealt with, just by themselves, the standing waves on the emitter surfaces are what cause driver specific localization and their removal places all intelligible sound sources and noise behind the driver, in perceived local. This is also true of the monopole baffle that they are mounted upon and the elimination of standing wave energy across that baffle surface also eliminates the baffle as a localized source of perceived sounds, again moving them back behind the actual physical baffle location.
This driver or baffle specific localization is is what I have always thought edge diffraction was. It does seem to me that if you remove the cause of the localization, that the other effects might go unnoticed, except at 90 degrees off axis.
Bud
Is the delayed sound that causes localization of drivers a standing wave phenomena on an OB?
With the drivers I have dealt with, just by themselves, the standing waves on the emitter surfaces are what cause driver specific localization and their removal places all intelligible sound sources and noise behind the driver, in perceived local. This is also true of the monopole baffle that they are mounted upon and the elimination of standing wave energy across that baffle surface also eliminates the baffle as a localized source of perceived sounds, again moving them back behind the actual physical baffle location.
This driver or baffle specific localization is is what I have always thought edge diffraction was. It does seem to me that if you remove the cause of the localization, that the other effects might go unnoticed, except at 90 degrees off axis.
Bud
BudP said:
Bud,
With the baffles pictured, which are the worst offenders I've built, it's probably a combination of issues, but I don't know if any of them would result in standing waves. First, there's the front and rear assymetry especially where the edge is closest to the driver. Then there's diffraction where the pressure change is greatest. That high velocity flow at the edge also stimulates vibration of the baffle, as others have pointed out, and it likely another source of audible problem as well. About the only thing they do well is smooth overall response because the dipole ripples are all but eliminated, and the diffraction effects are spread across a wide range.
Multipath Transmission
Good point by Dr. Geddes. The length of the this discussion confirms my impression that acoustical diffraction is not well represented in the literature, at least in the AES Journal. Maybe all the money goes into underwater research instead (US Navy funding). And no, I don't mind the brief side-track at all. It's been most interesting, actually, seeing people attack this thing from different angles and come out with results that are 180-degree opposites from each other!
Everyone knows I'm not strong on theory at all, and what I know is mostly rubbed off from the Tektronix Spectrum Analyzer guys. I view the system as responding to an impulse generated by the driver - as the impulse expands into a hemisphere, secondary radiation sources are generated whenever the spherically expanding impulse encounters a change in radiation impedance. This is how Time Domain Reflectometers (TDR's) or cable-radar operates - even quite small kinks in the cable are easily seen on the display.
The microwave/radar guys routinely use foam sprayed with carbon as termination strips on circuit boards - for example, where semi-rigid coax meets a circuit trace. This is a physical 2D to 3D transformation that requires smoothing techniques, otherwise there are big reflections that can severely degrade a spectrum analyzer or radar system.
Back when I was working on the Ariel and using near-realtime MLSSA measurements, the diffraction emission off the edges of the cabinet appeared to be greatest at a 135-degree angle with respect to the front surface of the cabinet. There was still emission at 90 degrees (parallel to the direct wave from the driver) and 180 degrees (parallel to the front surface of the baffle), but this was substantially reduced in magnitude and HF content. My impression was diffraction behaved like an acoustic prism, with HF response strongly correlated to emission angle.
I've been thinking more about acoustic multipath. My limited understanding of a non-minimum phase system is that the identity between frequency response and time response is broken, so minimum-phase equalization cannot simultaneously correct time response. This is significant, because in electronic systems, pre-emphasis and complementary de-emphasis are routinely used to restore both FR and TR on the receiving end - and this is only possibly when the whole chain is minimum phase.
Multipath is a different story. Multipath in television appears as "ghosts" on-screen, and NO conventional minimum-phase equalization can remove the ghosts. The only way to get rid of them in TV is a tapped delay line, with individual delays tuned for every reflection - each delay has to be precisely adjustable in time, magnitude, and of course inverted polarity. The biggest delays are nulled out, and the picture is restored to close to its original quality.
In TV, multipath becomes non-minimum phase once it is spatially summed at the receiving antenna. After the summation, the only way to remove the reflections is through a time-delay system that is very precisely tuned to remove each reflection. The tapped time-delay system, implemented either in analog or digital domains, is NOT a minimum-phase system.
In analog systems, particulary those with a subcarrier, multipath results in degradation. In digital systems, it results in either correctable or uncorrectable data loss, depending on severity. The American 8VSB HDTV over-the-air modulation scheme has suffered from an inability to reject multipath (picture blocking and freezing), while the European DVB-T system uses active multipath detection and digital delay-line compensation. So multipath matters a lot in the real world of mobile phones and broadcasting.
I don't see any difference between edge diffraction and multipath in television. Both have a direct signal that arrives first and is typically largest in magnitude by 6~20 dB, with many discrete reflections that arrive later and are summed at the detector.
The human listener, with two detectors and pinna spatial filtering, uses a form of "diversity antenna" to analyze the ambient acoustic . Unlike FFTs or even cross-correlators, the human listener has multiple anticipatory functions that are expecting the music signal to progress in a certain way - the listener not only has a knowledge of the immediate past (unlike an FFT), but expects the signal to go a certain way in the immediate future (unlike a cross-correlator, which knows nothing about the future).
Any sort of acoustic mis-termination to the spherical wave expanding from the diaphragm generates secondary reflections. The perceptual impact of these delayed signal is a function of their delay, magnitude, and spatial displacement from the main emissive surface. The reason I'm discussing resistive mesh surfaces at the edge of the baffle is to acoustically smooth out the abrupt edge of the baffle - I see the zero-width, knife-edge baffle as the worst-case, in terms of the most violent mis-termination, guaranteed to produce the sharpest, most discrete reflection, just an intentional nick in a microwave cable would produce a sharp reflection, easily visible on a TDR.
The function of the mesh is to act as a lossy, broad-area leak path from front to rear of the baffle. If it were simply loose fabric or thin leather, sound would go right through it, not what I want. I want a lossy, reasonably broadband "curtain" between front and rear - not rigid, not transparent, but halfway between.
gedlee said:
Good point by Dr. Geddes. The length of the this discussion confirms my impression that acoustical diffraction is not well represented in the literature, at least in the AES Journal. Maybe all the money goes into underwater research instead (US Navy funding). And no, I don't mind the brief side-track at all. It's been most interesting, actually, seeing people attack this thing from different angles and come out with results that are 180-degree opposites from each other!
Everyone knows I'm not strong on theory at all, and what I know is mostly rubbed off from the Tektronix Spectrum Analyzer guys. I view the system as responding to an impulse generated by the driver - as the impulse expands into a hemisphere, secondary radiation sources are generated whenever the spherically expanding impulse encounters a change in radiation impedance. This is how Time Domain Reflectometers (TDR's) or cable-radar operates - even quite small kinks in the cable are easily seen on the display.
The microwave/radar guys routinely use foam sprayed with carbon as termination strips on circuit boards - for example, where semi-rigid coax meets a circuit trace. This is a physical 2D to 3D transformation that requires smoothing techniques, otherwise there are big reflections that can severely degrade a spectrum analyzer or radar system.
Back when I was working on the Ariel and using near-realtime MLSSA measurements, the diffraction emission off the edges of the cabinet appeared to be greatest at a 135-degree angle with respect to the front surface of the cabinet. There was still emission at 90 degrees (parallel to the direct wave from the driver) and 180 degrees (parallel to the front surface of the baffle), but this was substantially reduced in magnitude and HF content. My impression was diffraction behaved like an acoustic prism, with HF response strongly correlated to emission angle.
I've been thinking more about acoustic multipath. My limited understanding of a non-minimum phase system is that the identity between frequency response and time response is broken, so minimum-phase equalization cannot simultaneously correct time response. This is significant, because in electronic systems, pre-emphasis and complementary de-emphasis are routinely used to restore both FR and TR on the receiving end - and this is only possibly when the whole chain is minimum phase.
Multipath is a different story. Multipath in television appears as "ghosts" on-screen, and NO conventional minimum-phase equalization can remove the ghosts. The only way to get rid of them in TV is a tapped delay line, with individual delays tuned for every reflection - each delay has to be precisely adjustable in time, magnitude, and of course inverted polarity. The biggest delays are nulled out, and the picture is restored to close to its original quality.
In TV, multipath becomes non-minimum phase once it is spatially summed at the receiving antenna. After the summation, the only way to remove the reflections is through a time-delay system that is very precisely tuned to remove each reflection. The tapped time-delay system, implemented either in analog or digital domains, is NOT a minimum-phase system.
In analog systems, particulary those with a subcarrier, multipath results in degradation. In digital systems, it results in either correctable or uncorrectable data loss, depending on severity. The American 8VSB HDTV over-the-air modulation scheme has suffered from an inability to reject multipath (picture blocking and freezing), while the European DVB-T system uses active multipath detection and digital delay-line compensation. So multipath matters a lot in the real world of mobile phones and broadcasting.
I don't see any difference between edge diffraction and multipath in television. Both have a direct signal that arrives first and is typically largest in magnitude by 6~20 dB, with many discrete reflections that arrive later and are summed at the detector.
The human listener, with two detectors and pinna spatial filtering, uses a form of "diversity antenna" to analyze the ambient acoustic . Unlike FFTs or even cross-correlators, the human listener has multiple anticipatory functions that are expecting the music signal to progress in a certain way - the listener not only has a knowledge of the immediate past (unlike an FFT), but expects the signal to go a certain way in the immediate future (unlike a cross-correlator, which knows nothing about the future).
Any sort of acoustic mis-termination to the spherical wave expanding from the diaphragm generates secondary reflections. The perceptual impact of these delayed signal is a function of their delay, magnitude, and spatial displacement from the main emissive surface. The reason I'm discussing resistive mesh surfaces at the edge of the baffle is to acoustically smooth out the abrupt edge of the baffle - I see the zero-width, knife-edge baffle as the worst-case, in terms of the most violent mis-termination, guaranteed to produce the sharpest, most discrete reflection, just an intentional nick in a microwave cable would produce a sharp reflection, easily visible on a TDR.
The function of the mesh is to act as a lossy, broad-area leak path from front to rear of the baffle. If it were simply loose fabric or thin leather, sound would go right through it, not what I want. I want a lossy, reasonably broadband "curtain" between front and rear - not rigid, not transparent, but halfway between.
Hey guys, keep it up - this is a great discussion! At this stage , is it helpful to point out that (topologically) an open baffle is the "same" as a simple ported cabinet?
Guess this makes the perimeter of the baffle the "same" as the inside of a port. Dunno if it helps to think about what happens to port action as diameter and length change?
Cheers,
Mike
Guess this makes the perimeter of the baffle the "same" as the inside of a port. Dunno if it helps to think about what happens to port action as diameter and length change?
Cheers,
Mike
Bratislav said:
Waves of equal magnitude and opposite phase are travelling in the same direction and have travelled the same distance, so they remain directly out of phase and have the same magnitude, so they net to 0.
At the edge of the baffle the front and rear waves both go from half space expansion to full space expansion. The primary role our baffles play is to make the rear wave travel further to get to our ears at the listening position. This added distance changes the phase relationship away from being directly out of phase, and that phase relationship determines how much bass we hear. Since our delay distance is fixed and wavelengths get longer with lower frequency, the lower you go the closer to directly out of phase the 2 waves are at the listening position and the less bass you get. That is what causes the 6db/oct dipole bass roll-off.
If they cancelled at the edges, dipole bass would be easy and would be the same as monopole bass.
johninCR said:
John, that's the best argument I've heard yet; that eq'ing doesnt rectify the situation at one point, since diffraction is non-coincident with the direct. Great point.
The offshoot is that small changes in head movement would render diffraction localizable in space (not in the classic sense, but as image distortion) due to the non-coincidence, and that the hrtf would make the amplitude "summation" at the pinna differ from what a mic sees.
This gets back to Lynn's comment much further back of philisophically being uneasy with multi speakers in the same frequency range.
This is a good argument for a small baffle in a regular box speaker, and a possible explanation for why they image so much bigger in space. Some (e.g. REG) see small baffle imaging (spacious) as a distortion, but based on this, I'm not convinced. Of course for OBs, a larger baffle is required. I don't see why edge treatment of a large OB would be un-needed.
Edge Diffraction, again
A lot has been written, and as much as I hate to do so, here is my 2 cents worth.
Think of edge diffraction as the result of air movement over the edge. Kinda like blowing over an edge. It creates vibrations that result in an edge diffraction.
In an open baffle, using the example of the driver going forward, there is positive pressure on the front of the baffle, and negative pressure on the rear. When this wave hits the baffle edge, the air moves from the front to the rear, with twice the velocity that it would if there was no rear negative pressure. Thus, twice the amount of edge diffraction in an open baffle than a closed box.
Please tell me where I am wrong.
A lot has been written, and as much as I hate to do so, here is my 2 cents worth.
Think of edge diffraction as the result of air movement over the edge. Kinda like blowing over an edge. It creates vibrations that result in an edge diffraction.
In an open baffle, using the example of the driver going forward, there is positive pressure on the front of the baffle, and negative pressure on the rear. When this wave hits the baffle edge, the air moves from the front to the rear, with twice the velocity that it would if there was no rear negative pressure. Thus, twice the amount of edge diffraction in an open baffle than a closed box.
Please tell me where I am wrong.
johninCR said:
But not when baffle is preventing them from interferring, right ? So at any point along the baffle there is a pressure gradient (traveling outwards). If you measure sound pressure along the baffle, it is not zero. Only after you go into free air (disregard the discontinuity at baffle edge for the moment) waves are allowed to go through destructive interference.
Now, placing a solid obstacle (piece of paper) in continuation of the desctructive plane of interfernce (extending the baffle if you wish) instantly separates the wavefronts, and pressure gradient reappears (and will go positive/negative as waves move outwards).
THAT will cause your piece of paper to vibrate.
My argument is only that having piece of paper vibrate at the baffle edge does not prove diffraction; simply as there is at least one more source for paper vibration.
Note that diffraction can still exist (I'm not negating this premise at all) and may as well ADD to the paper vibration. I couldn't care less. I am only pointing that your experiment is flawed and conclusion drawn from it is wrong.
That is all.
Too bad the scientists can't explain exactly what happens with edge diffraction, at least I haven't seen an explanation. Then we'd be armed better to attack them. Lynn refers to them as reflections, and though I don't think there's an actual reflection of sound at the edge, I think you get a better view of the picture.
Rudolf explained to me, in a way that made me see the light, that the pressure change at the edge causes a new sound source to occur there. In the case of a bipole with a sharp edged baffle, there is no pressure change and no sonic event would occur other than the 2 waves to combine and continue the same expansion. With a monopole there is a pressure change at the edge, because the wave has been constrained to half space expansion by the baffle, and at the edge there is a pressure drop because the wave suddenly has a larger space for expansion. With dipoles the pressure change at the edge is double, because the high pressure portion of the wave is meeting the low pressure portion coming from the other side. Since the pressure change is greater, the resulting sound created must be greater too.
Rudolf explained to me, in a way that made me see the light, that the pressure change at the edge causes a new sound source to occur there. In the case of a bipole with a sharp edged baffle, there is no pressure change and no sonic event would occur other than the 2 waves to combine and continue the same expansion. With a monopole there is a pressure change at the edge, because the wave has been constrained to half space expansion by the baffle, and at the edge there is a pressure drop because the wave suddenly has a larger space for expansion. With dipoles the pressure change at the edge is double, because the high pressure portion of the wave is meeting the low pressure portion coming from the other side. Since the pressure change is greater, the resulting sound created must be greater too.
johninCR said:
Excuse me ?
Try Huygens principle. Discovered in 1600's.
http://www.launc.tased.edu.au/online/sciences/physics/diffrac.html
Re: Edge Diffraction, again
Most of the apparently on-going discussion on this issue was focused on symmetric out-of-phase (i.e. an actual dipole) pressure that does not have an "edge" to react to.
(..and note that with an edge it isn't really a dipole.. depending on the passband of operation, at lower freq.s than the baffle can support it may well be a dipole - but this also depends on the driver's radiation).
In such a case (symmetric dipole with no edge) - it isn't high pressure to low pressure (or conversly low pressure to high pressure), but rather High/or Low to NULL PRESSURE.
Think of the NULL zone like a vacuum. What happens when sound enters a vacuum?
Jon Ver Halen said:
Most of the apparently on-going discussion on this issue was focused on symmetric out-of-phase (i.e. an actual dipole) pressure that does not have an "edge" to react to.
(..and note that with an edge it isn't really a dipole.. depending on the passband of operation, at lower freq.s than the baffle can support it may well be a dipole - but this also depends on the driver's radiation).
In such a case (symmetric dipole with no edge) - it isn't high pressure to low pressure (or conversly low pressure to high pressure), but rather High/or Low to NULL PRESSURE.
Think of the NULL zone like a vacuum. What happens when sound enters a vacuum?
Bratislav said:
Bratislav,
That demonstration simply shows there is significant energy in the form of a pressure change and it is greatest at the edge of a thin baffle. That means diffraction can't possibly be zero. It's either zero or it's double, there is no in between answer once your baffle edge approaches 0 thickness.
Re: Edge Diffraction, again
Jon,
I think you're pretty close, and that's the way I used to visualize what occurs (as a flow). Rudolf convinced me otherwise, and to thing of it in terms of a pressure change. While I have no doubt that it is stimulating vibration in my baffle, even if it were perfectly rigid and didn't vibrate at all, there would still be sound resulting from the pressure change alone.
Jon Ver Halen said:
Jon,
I think you're pretty close, and that's the way I used to visualize what occurs (as a flow). Rudolf convinced me otherwise, and to thing of it in terms of a pressure change. While I have no doubt that it is stimulating vibration in my baffle, even if it were perfectly rigid and didn't vibrate at all, there would still be sound resulting from the pressure change alone.