i have the same situation here.
my wife is always trying to get me watch "hoarders" and "buried alive"...
my wife is always trying to get me watch "hoarders" and "buried alive"...
Linear distortion in horns can be reduced by the methods employed by Earl Geddes, smooth throats, OS expansion, large round overs, HOM attenuation foam. The reduction in linear distortion using an OS expansion also comes with a reduction of SPL level compared to other horn expansions, but still is adequate for SPL levels generally used in a home environment.
Non linear distortion inherent in the compression driver can be reduced by using drivers with diaphragms with large enough area to achieve the desired SPL at an acceptable level of distortion. Larger diaphragms (above 2-3" in diameter) exhibit more HF distortion, so must be crossed over in the 5-10 kHz range to reduce distortion at high drive levels.
Unfortunately, using two horns and two drivers results in ragged off axis response in the crossover region, the trade off must be weighed against the distortion.
JBL's D2 double diaphragm drivers are a good compromise, doubling SD without increasing diaphragm diameter. BMS markets several co-axial compression drivers which allow a large SD midrange and smaller SD tweeter to be used on the same horn, another good solution.
If a single driver is not capable of providing the desired SPL at an acceptable distortion level, multiple drivers can be attached to a single horn, but provide only marginal benefit due to the increase of non linear air distortion at the high throat SPL levels encountered.
Multiple drivers can be more effectively utilized with the different styles of combiners utilized in various line arrays and DSL's Paraline, but they all suffer from the non linear air distortion problems caused by initial slow expansion and linear distortion from diffraction in the combiner/horn interface.
Art
Hi Art
We are 90% on the same page and maybe we can get to 100% after I give you my latest thinking in the issues.
Picture a graph of level versus perception: the lower bounds of perception is "imperceptible" and somewhere at the top is "unacceptable". There are two types of distortion that concern us here. First is linear distortion (LD) and the second (NLD) nonlinear distortion. Each of these will yield a line of positive slope on the graph - both will become "imperceptible" at some point and both will certainly cross the "unacceptable" line at some point. But we don't know where these lines are located on the graph for any particular example. I think that we can assume that they are straight, at least to first order in this argument. I would argue that the perception of NLD will rise faster than that of LD, because it gets worse and worse with level, while linear distortion only gets unmasked at higher levels, its actual effect is linear.
This hypothesis would explain the high level of variability in opinions that we have because there are an almost infinite number of ways that these lines can fall - nothing audible at low levels, then LD audible for awhile, then NLD at higher levels; or only nothing audible then only NLD, LD being too low to ever be a factor. No matter what, in the end, it will be NLD that prevails because its slope is greater. At mid levels combinations of the two distortion effects might make things very complicated.
In this scenario, the horn dominantly affects the LD while the NLD is dominated by the driver. One may or may not ever see either or both of these effects depending on what range of levels are being considered. Low enough and NLD does not always dominate, and may not even be perceptible, but at higher levels it always will. Below some level neither is a factor - this level would be highly complex to determine. This is especially true since we have no reliable scales (metrics) for either NLD or LD. If we did, then we could plot those line based on these metrics. Until then, it will likely remain a mystery.
We are 90% on the same page and maybe we can get to 100% after I give you my latest thinking in the issues.
Picture a graph of level versus perception: the lower bounds of perception is "imperceptible" and somewhere at the top is "unacceptable". There are two types of distortion that concern us here. First is linear distortion (LD) and the second (NLD) nonlinear distortion. Each of these will yield a line of positive slope on the graph - both will become "imperceptible" at some point and both will certainly cross the "unacceptable" line at some point. But we don't know where these lines are located on the graph for any particular example. I think that we can assume that they are straight, at least to first order in this argument. I would argue that the perception of NLD will rise faster than that of LD, because it gets worse and worse with level, while linear distortion only gets unmasked at higher levels, its actual effect is linear.
This hypothesis would explain the high level of variability in opinions that we have because there are an almost infinite number of ways that these lines can fall - nothing audible at low levels, then LD audible for awhile, then NLD at higher levels; or only nothing audible then only NLD, LD being too low to ever be a factor. No matter what, in the end, it will be NLD that prevails because its slope is greater. At mid levels combinations of the two distortion effects might make things very complicated.
In this scenario, the horn dominantly affects the LD while the NLD is dominated by the driver. One may or may not ever see either or both of these effects depending on what range of levels are being considered. Low enough and NLD does not always dominate, and may not even be perceptible, but at higher levels it always will. Below some level neither is a factor - this level would be highly complex to determine. This is especially true since we have no reliable scales (metrics) for either NLD or LD. If we did, then we could plot those line based on these metrics. Until then, it will likely remain a mystery.
10% as there's no current "metric" to define the level at which disagreements go from being mere disagreements to being contentious i'd say 10% should a non issue.
anything to offer in the way of what can potentially reduce audible distortion in horns?
anything to offer in the way of what can potentially reduce audible distortion in horns?
@gedlee
Am I wrong if I think that in general, you are describing diffraction from edges and discontinuities and the associated combfiltering from those effects, when talking about linear distortion?
Or are you using that term in a broader sence, covering "regular" frequency response and phase response as well?
I know you have claimed earlier that group delay is most audible at high frequencies and at higher leves, however that may be a bit confusing when more specifically it's combfiltering that is taking place (and of course a discrete delay causing this, double spikes when measuring impulse response).
Groupdelay as in typical well implemented crossovers and box tunings does not introduce any combfiltering and as such, should not be bunched together with what happens with various reflections. As a matter of fact, I never heard anyone else using the term "group delay" when dealing with discrete reflections and comb filtering effects.
Am I wrong if I think that in general, you are describing diffraction from edges and discontinuities and the associated combfiltering from those effects, when talking about linear distortion?
Or are you using that term in a broader sence, covering "regular" frequency response and phase response as well?
I know you have claimed earlier that group delay is most audible at high frequencies and at higher leves, however that may be a bit confusing when more specifically it's combfiltering that is taking place (and of course a discrete delay causing this, double spikes when measuring impulse response).
Groupdelay as in typical well implemented crossovers and box tunings does not introduce any combfiltering and as such, should not be bunched together with what happens with various reflections. As a matter of fact, I never heard anyone else using the term "group delay" when dealing with discrete reflections and comb filtering effects.
Pan
I am excluding simple frequency response issues.
To me LD can take a few forms, one is diffraction, the other would be internal reflections. These will both exhibit group delay and most likely a form of comb filtering as well. They are really just two different ways of describing the same thing - comb filter in the frequency domain and group delay in the time domain. One can also think of these LDs as "multi-path".
So I don't exclude the comb-filtering could occur, but I don't know the audibility of that. I do know (from Brian Moore, Cambridge University Prof of Hearing) that group delay does become more and more audible at higher levels and we have a dominate sensitivity to it at about 2-3 kHz. So I tend to think of the group delay as being the audible part rather than the comb filtering.
Remember that the internal reflections are more than just a single reflection, they are like a flutter echo but much much more rapid. Diffraction occurs in horns for a number of reasons - it can be minimized, but never eliminated.
I am excluding simple frequency response issues.
To me LD can take a few forms, one is diffraction, the other would be internal reflections. These will both exhibit group delay and most likely a form of comb filtering as well. They are really just two different ways of describing the same thing - comb filter in the frequency domain and group delay in the time domain. One can also think of these LDs as "multi-path".
So I don't exclude the comb-filtering could occur, but I don't know the audibility of that. I do know (from Brian Moore, Cambridge University Prof of Hearing) that group delay does become more and more audible at higher levels and we have a dominate sensitivity to it at about 2-3 kHz. So I tend to think of the group delay as being the audible part rather than the comb filtering.
Remember that the internal reflections are more than just a single reflection, they are like a flutter echo but much much more rapid. Diffraction occurs in horns for a number of reasons - it can be minimized, but never eliminated.
Ok, thanks for response.
I tend to agree (2nd time this week ;-) that any delayed energy that is significant (in time) to the period of the signal is a problem. IOW all kinds of early reflections/diffraction that also will lead to combing. A delay can be big in absolute time if the frequency is low without much of a negative effect. A 1ms delayed reflection will obviously be nothing looking at 20Hz but a whole lot in the top decade of the audible range.
I tend to agree (2nd time this week ;-) that any delayed energy that is significant (in time) to the period of the signal is a problem. IOW all kinds of early reflections/diffraction that also will lead to combing. A delay can be big in absolute time if the frequency is low without much of a negative effect. A 1ms delayed reflection will obviously be nothing looking at 20Hz but a whole lot in the top decade of the audible range.
Pan - yes, absolutely. I don't think that it is critical to differentiate between comb filter or group delay, or reflection, diffraction, whatever. They are all linear and all audible with an audibility that grows with level. In the loudspeaker itself the time scales are very short when compared to room effects. However, in my system I went to great lengths to minimize reflections and diffractions from objects nearby the speakers and this too was an audible improvement.
We seem not to like very short delayed signals (< 1 ms or so) as our hearing system does not seem to be able to sort them out. At longer time scales (> 20 ms.) they even become pleasant and desirable. Medium time scales (2 -> 10 ms.) is debatable (and a significant source of current debate.)
We seem not to like very short delayed signals (< 1 ms or so) as our hearing system does not seem to be able to sort them out. At longer time scales (> 20 ms.) they even become pleasant and desirable. Medium time scales (2 -> 10 ms.) is debatable (and a significant source of current debate.)
Earl, I think this application and discussion of auditory roughness is relevant to the discussion:
"SRA: A Web-based Research Tool for Spectral and Roughness Analysis of Sound Signals"
http://musicalgorithms.ewu.edu/learnmoresra/files/vassilakis2007smc.pdf
See also
SRA 2.0 © 2008 - Spectral and Roughness Analysis of Sound Signals
Yes, Roughness was a concept that we used quite a bit in noise control. "Roughness" as a property of sound, was first introduced by a colleague of mine Prof. E. Zwicker.
I once started a study on the effect of nonlinearities on music sources with different characteristics (kind of like roughness) - basically statistical representations of the signal samples. I was convinced that this would be a big consideration - I still think that it is likely to be the case - but I was never able to complete it (then lost interest.)
I once started a study on the effect of nonlinearities on music sources with different characteristics (kind of like roughness) - basically statistical representations of the signal samples. I was convinced that this would be a big consideration - I still think that it is likely to be the case - but I was never able to complete it (then lost interest.)
Last edited:
You can look into the two primary actions in horns.. loading and wave propagation, how and when each may be the primary focus, and then read this thread.
i am familiar with the thread. are you suggesting it contains information on the reduction of audible distortion in horns?
sorry correction i should say "useful" or "usable" information on the reduction of audible distortion in horns.
Turk,
Did you read post #263, and not find it useful?
The one thing I forgot in #263 is resonant horn walls lead to another non linear form of audible distortion in horns, (too) thin materials start to flex and "ring" progressively more at higher drive levels. Thicker horn walls, or addition of damping materials can take care of those problems.
As far as compression drivers, other than the few relatively recent additions noted in #263, there have been no advances in reduction of distortion in decades, the improvement in power handling has actually resulted in higher distortion levels being available before coil burnout.
Art
What do you think about this concept. The author was my lecturing professor of electro acoustics at the Technical University, Prague.
Patent US3590169 - Electrostatic horn loudspeaker - Google Patents
Patent US3590169 - Electrostatic horn loudspeaker - Google Patents
Attachments
Pavel,
Interesting device.
Looks like it would be well suited for line arrays, but I don't see it reducing distortion, and likely would be frightfully expensive to produce.
Electrostatics have very limited excursion, so the crossover frequency would have to be high for high output.
Any measured distortion vs. output level of an actual device?
Art
- Status
- Not open for further replies.