Here is a part of a paper by Bob H. Smith from 1953 regarding his HOMs. I have always thought that the real work that could be done re horns will be there! Not in the horn. Let's be honest there are a plethera of great sounding horns. The problems start at the driver, this is where we could real time visual models. Bob Smith talking of High modal resonance (HOMs)in 1953 but his concern is phase plug and throat..Fascinating.
Title: An Investigation of the Air Chamber of Horn Type Loudspeakers
Author: Bob H. Smith
Publication: ASA-J, Vol. 25, No. 2, pp. 305-312, Mar-1953
Affiliation: Division of Electrical Engineering, University of California
URL: none
Abstract: The front air chamber design is treated as a boundary value problem which yields a solution of the wave equation for the general case in which the horn throat enters the air chamber in a circumferentially symmetrical manner.
The following specific cases are analyzed: (1) the case in which the horn throat enters the air chamber by means of a single orifice, (2) the horn throat enters the air chamber by means of a single annulus of radius [r] and width [w], and (3) the horn throat enters the air chamber in [m] annuli of radii [r1],[r2],...[rm] and widths [w1],[w2],...[wm].
The analysis reveals that the radial perturbations caused by the horn throat excites higher order modes. At the resonant frequencies of these modes the horn throat pressure becomes zero and the loudspeaker does not radiate. By suitable choice of annulus radii and widths the first [m] modes may be suppressed and the corresponding nulls in the output pressure eliminated.
Title: An Investigation of the Air Chamber of Horn Type Loudspeakers
Author: Bob H. Smith
Publication: ASA-J, Vol. 25, No. 2, pp. 305-312, Mar-1953
Affiliation: Division of Electrical Engineering, University of California
URL: none
Abstract: The front air chamber design is treated as a boundary value problem which yields a solution of the wave equation for the general case in which the horn throat enters the air chamber in a circumferentially symmetrical manner.
The following specific cases are analyzed: (1) the case in which the horn throat enters the air chamber by means of a single orifice, (2) the horn throat enters the air chamber by means of a single annulus of radius [r] and width [w], and (3) the horn throat enters the air chamber in [m] annuli of radii [r1],[r2],...[rm] and widths [w1],[w2],...[wm].
The analysis reveals that the radial perturbations caused by the horn throat excites higher order modes. At the resonant frequencies of these modes the horn throat pressure becomes zero and the loudspeaker does not radiate. By suitable choice of annulus radii and widths the first [m] modes may be suppressed and the corresponding nulls in the output pressure eliminated.
Last edited:
Earl now, using the same audiophile subjective claim that you would reject from anyone else!
Lets say I accept that your product sounds good and that your waveguides have merit. I have no reason to doubt it. What I am asking for is for some evidence, any evidence, that they are better due to optimization of your unique attributes, reduced HOMs or reduced diffraction related reflections.
Isn't it just as likely they sound good because their basic performance is good? Frequency response and polar curves? How do we know it is one attribute over another?
Here is the story I hear:
"For decades horns were designed to optimize the classical parameters. Tnen one man came along and found heretofore unknown flaws of HOMs and diffraction. These unknown flaws were elusive and hard to measure. Recent horn designs that improved some classical attributes had, in fact, made these new parameters worse. Their traits could be buried under the classical aberations, yet their subjective effect was far beyond what their measurements would suggest."
Sounds like a wondrous myth, doesn't it?
If you want to continue to claim that diffraction is the key (or HOMs) I think you need to answer two issues.
1) show that these are distinctly measurable flaws that can be seperated from all the other flaws of horns, such as mouth reflections and polar aberations.
2) at least suggest why these, as a linear effect, are more audibly significant than any other time or frequency domain aberration of similar delay and frequency.
Again, I don't doubt that these aberrations exist, I am just looking for evidence of their significance.
David S
Dave,
I think what Earl thinks of what HOMs are are seen in the high-resolution polar map. A 'high diffraction/reflection' design shows periodic ripples in the response and those change with angle, so no amount of EQ can correct it. This is in the frequency domain.
In the time domain those same ripples are basically a delayed reflection from the main signal. Earl's study has shown that these become more audible as the playback level increases. Also informal studies done in SynAudCon in the 70s-80s suggest that a clean impulse and a rapid decay with no reflections at or near the speakers (think of protruding stands etc.) gives the most stable image.
From this one could conclude that a clean response from the listening window, especially in the first ms, up to about 10ms, is a good thing to have. I can't think of another way to have this without taking care of diffraction and reflection from the speakers and nearby objects.
I think what Earl thinks of what HOMs are are seen in the high-resolution polar map. A 'high diffraction/reflection' design shows periodic ripples in the response and those change with angle, so no amount of EQ can correct it. This is in the frequency domain.
In the time domain those same ripples are basically a delayed reflection from the main signal. Earl's study has shown that these become more audible as the playback level increases. Also informal studies done in SynAudCon in the 70s-80s suggest that a clean impulse and a rapid decay with no reflections at or near the speakers (think of protruding stands etc.) gives the most stable image.
From this one could conclude that a clean response from the listening window, especially in the first ms, up to about 10ms, is a good thing to have. I can't think of another way to have this without taking care of diffraction and reflection from the speakers and nearby objects.
Hi Art,
I can't really comment on that, primarily because I haven't done a comparison to search for it. My primary experience is with the 80's generation CD devices, both at that time and more recently with installing and tuning cinema products such as the 4675.
But I have to ask, how would you come to such a conclusion based on simple listening tests? Horns are such imperfect devices and all posessed of distinct personalities, but to sit and listen and conclude, "ah, the low/high SPL character difference between the smoother throat JBL horns compared to their diffraction horn counterparts really stands out, doesn't it?"
Again, I'm not disputing that there aren't numerous flaws in available horns, just how we differentiate the effects of one phenomenon from the next.
David
Hi Norman,
That is an interesting collection of measurements and goes some way towards an overview of the subject.
What I see is that each horn has a pretty distinct reflection in the 2 to 3 msec range. These are primarily lower frequency mouth reflections with a path length, per JMLC of twice the horn depth. The 3D plot shows them along with the impulse response. They show up in the frequency response as LF ripple, typically with a bump at 300, 600, and 1k, or therabouts.
The 3D plots show shorter delayed and higher frequency effects, and these vary from horn to horn. JMLC defines these as "diffraction" or possibly HOMs. The path length/arrival time is shorter presumably because they occur on the way out of the horn and no return trip to the throat is required. They show up in the impulse as early hash and ringing. It isn't clear that they can be picked out in the frequency response plots.
Note that no off axis curves are given and we don't know which of these horns maintain their frequency response off axis and which don't.
The question is still whether these are more significant than than the clearly visible mouth reflections? Also, do we need new measurements to evaluate these or is it simple enough to say that a smooth response with a clean impulse is, as always, desireable?
David
Seems that measuring the wavelet function isn't very tough; Zilch did it for the Altec 511 with a DE250 driver:
AudioKarma.org Home Audio Stereo Discussion Forums - View Single Post - Z19 Crossover
Besides the 511 ringing like crazy at several frequencies (which can be fixed), it doesn't look that bad.
AudioKarma.org Home Audio Stereo Discussion Forums - View Single Post - Z19 Crossover
Besides the 511 ringing like crazy at several frequencies (which can be fixed), it doesn't look that bad.
I did not understand this:
In specific, where are two diffraction signals generated where there was only one before?
I am assuming the "correction" is in the form of amplitude correction wherein the frequency variations due to diffraction are corrected for one point? Don't see how this generates any more diffractions...
...probably misunderstanding.
_-_-
In specific, where are two diffraction signals generated where there was only one before?
I am assuming the "correction" is in the form of amplitude correction wherein the frequency variations due to diffraction are corrected for one point? Don't see how this generates any more diffractions...
...probably misunderstanding.
_-_-
Bear
Consider a diffraction signal as a tall impulse followed by a shorter diffracted impulse at some delay, which will be different at every point in space. This is idealized, but will make the point. At some point I cancel the diffraction (in DSP let's say) with a negative pulse. Works just fine at that point. But at every other point the delays are not that same so the DSP cancelation adds a second negative pulse that is no longer lined up with the original diffraction pulse. We no have two delayed pulses rather than one.
In the case of the standing wave, the waves all travel the same paths so the delayed pulse(s) are all at the same delay in the space and if I cancel the pulse(s) at one point they will cancel equally well at all the others.
Dave
I wish that I had the time, finances and resources to provide you with the proof that you desire. Unfortunately I do not. I have provided the supporting evidence, which, I agree, is not conclusive, but it is all there is. To me, having some scientific supporting evidence is far better than most audio claims that have virtually no supporting evidence other than "It sounds good to me."
Consider a diffraction signal as a tall impulse followed by a shorter diffracted impulse at some delay, which will be different at every point in space. This is idealized, but will make the point. At some point I cancel the diffraction (in DSP let's say) with a negative pulse. Works just fine at that point. But at every other point the delays are not that same so the DSP cancelation adds a second negative pulse that is no longer lined up with the original diffraction pulse. We no have two delayed pulses rather than one.
In the case of the standing wave, the waves all travel the same paths so the delayed pulse(s) are all at the same delay in the space and if I cancel the pulse(s) at one point they will cancel equally well at all the others.
Dave
I wish that I had the time, finances and resources to provide you with the proof that you desire. Unfortunately I do not. I have provided the supporting evidence, which, I agree, is not conclusive, but it is all there is. To me, having some scientific supporting evidence is far better than most audio claims that have virtually no supporting evidence other than "It sounds good to me."
Won't the refracted signal that reflects from the diffraction slot (or wherever) travel back to the driver, reflect from there and return as an echo some time later?
Dr. Geddes,
Your example WRT an impulse works perfectly, a single static case. However, in real world situations with an actual music/voice/noise signal there is no secondary impulse to counter. While I understand that your example is meant to explain, it seems not quite right as far as being practical. It seems to me that to obtain a flat response EQ'd to a specific position, the correction is inevitably merely amplitude (although I suppose delay and phase are possible too). While that does mean that for all other locations the EQ is rather wrong, it's not any different than merely varying the source's frequency response (pre-distortion) or having a serendipitous signal that happens to result in a flat response at the target location. That does not generate any other new signals.
...again I am likely misunderstanding.
Your example WRT an impulse works perfectly, a single static case. However, in real world situations with an actual music/voice/noise signal there is no secondary impulse to counter. While I understand that your example is meant to explain, it seems not quite right as far as being practical. It seems to me that to obtain a flat response EQ'd to a specific position, the correction is inevitably merely amplitude (although I suppose delay and phase are possible too). While that does mean that for all other locations the EQ is rather wrong, it's not any different than merely varying the source's frequency response (pre-distortion) or having a serendipitous signal that happens to result in a flat response at the target location. That does not generate any other new signals.
...again I am likely misunderstanding.
Bear If something works or not with an impulse then it has to work the same for any signal. That's why an impulse is so powerful.
Art - in our test of compression drivers you fail to recognized that the drivers were not compared to each other, but to an unmodified signal - no driver, no distortion. Hence the drivers being different or the distortion level being high would have no impact on the results. The listeners could not detect the difference of the distorted driver from no distortion at all. Your doubts can not be correct.
Art - in our test of compression drivers you fail to recognized that the drivers were not compared to each other, but to an unmodified signal - no driver, no distortion. Hence the drivers being different or the distortion level being high would have no impact on the results. The listeners could not detect the difference of the distorted driver from no distortion at all. Your doubts can not be correct.
SPL at driver exit, and in horn throat are well beyond linear properties of air for sound propagation. Speed of sound increases with temperature. Standing waves and traveling standing waves in throat, and further up towards mouth when SPL is raised form complex pattern behaving as diffraction grating.
Signal and intensity allow for continually changing radiation pattern.
Signal and intensity allow for continually changing radiation pattern.
According to your test report you compared nine test stimuli, three drivers X three levels, all of which were highly distorted, being run at 14, 20, 28 Vrms (25, 50 and 100 watts in to 8 ohms) only a 6 dB range of differing degrees of THD ranging from 5-10% at 1000 Hz to 10-35% in the 4-9 kHz range.
The drivers were not compared to an unmodified signal, or an undistorted driver, that would have required more than nine test stimuli.
Had you done a comparison of the original recording (unmodified signal) and the drivers at two orders of magnitude difference, say 1, 10, and 100 watts, the results and your conclusions would likely have been different.
My test allows anybody to compare recordings of drivers tested to the original, and those drivers at a range greater than two orders of magnitude, as they may be used ranging from conversational level listening to pro sound levels.
Since my test has the drivers individually equalized flat, it is difficult to tell the drivers apart (or from the original recording) until the drive level is high.
At higher drive levels, the differences in the drivers sound (distortion) character becomes progressively more apparent.
You can hear the HF drivers with the LF mixed in here:
http://www.diyaudio.com/forums/multi-way/212240-high-frequency-compression-driver-evaluation.html
Or the HF drivers alone here:
High Frequency Compression Driver Evaluation
Give the recordings a listen and subjectively decide for yourself what level of distortion is acceptable to you in the various drivers.
The college students in your test 8 years ago did not have that opportunity with the way the test was conducted.
Art
Attachments
Art - this is where you are wrong. The test protocol was to compare the nine recordings to an unmodified signal and to rate the difference. I always use a reference source to base the results to something.
Your reference only suggests that even the frequency responses may have been undetectable for two of the three drivers, not that real differences could not be detected. That's completely different than what you are suggesting.
Last edited:
Earl,
According to your white paper, "always" "most unfortunately" did not include using a reference source in your GedLee LLC Compression Driver Subjective Test:
"4 General Observations
Subjects will seldom use the extremes of the allowed number scale. Thus a value near zero or unity should not be expected. This means that Sources 1 and 3 may well have been “imperceptible”. There is no doubt that Source 2 has perceptible linear distortion and significantly more than Sources 1 and 3. In hindsight it is apparent that the addition of a dummy source, one that was in fact the reference, would have added value to the interpretation of the results. It is most unfortunate that this possibility was not seen beforehand."
If there was a reference source, why did you write you should have added it?
How do you do the math that you had 9 sources when a reference would have made 10?
I just got a speeding ticket. 8 miles over the limit. I am not happy at all with the metrics used, I think we need longer miles, that way the signs can stay the same, as well as my speedometer. The officer felt that over a period of time the speed limit was adjusted to what a plurality of folks considered safely drivable. I am going to do a small study and attempt to get the scale changed. Bringing this thread as evidence, so please go easy on the new metric point of view until December 3 at 6:30 pm. Thanks.(-:
Last edited:
Art - that's simple. Had there been a "dummy" test where the sample and the reference were identical then the reliability of the listener could have been tested directly instead of indirectly as we did. That is the only reason that I would have added a tenth source.
I used this technique years later at Ford to show that the subjective impressions of the "golden ears" were unreliable.
I always contest tickets - 50% of the time they get thrown out, mostly because the officer does not show up. It's always worth a try. Never have I not gotten a reduced ticket when I showed up. But never argue, I've seen that go very badly.
My last two (con)test results agree with yours 100%, even though we are in different states (of the union).
Ok, the example was a single event.
But we are talking about two things, the "second" diffraction source and correction for flat response on some axis/listening point.
I'm lost on the issue of the "second source" of diffraction, other than changing the frequency response will alter the *amount* of diffraction WRT some frequency at any given instant.
The use of an impulse response w/FFT to determine the response of the system, doesn't change the fact that only one signal is output at any given instance, not two. So, I can't wrap my head around why there should be a "second source" of diffraction in this physical/mechanical system that is being excited by a single excitation. How can this occur?
- Status
- Not open for further replies.