My comment was to "In horn a/b tests, a shorter horn always sounds better. More open." - openness being most probable an FR aspect if speaker in same position and same electronics used.If a loudspeaker is designed with a specific FR for a reason, you just can't change that arbitrarily. Doing so, you can easily make it sound worse, just for a sake of comparisonThat makes no sense.
What's typically done is to "equal" an overall level as an average of some frequency band in midrange (and I'm not sure now if based on on-axis or in-room curves) - to subjectively have approximately the same overall loudness. But certainly not "EQ" the on-axis response down to 0,3 dB.
My take was that if you introduce a changed version of a horn (just longer), thus with stronger levels at lower freq., that will change the FR (yes - even if some/most in beginning of the passband) - so my EQ comment was to change it back to the original - as to NOT change the FR between A and B.. when A/B:ing... but how about: adjust the HP part of the x-over? The 0,3 dB figure was to secure that it is not FR you are listening to but some other aspect of the horn - be it dist, dir, HOMs, ... etc. For the case of comparing with a longer throat - what EQ could be needed? a PEQ of max a dB and Q of 2-3? Its not like EQing the whole speaker... just to compensate for the changed FR from the longer version horn.
But probably comparisons are not really possible - right?
//
Yes they are, you can even compare apples to pears. One must decide upon the criteria and as we know that direct sound dominates our perception, I think that you are perfectly right to suggest a comparison based on ON, or maybe better, LW.
This is also why tinkering with slopes for the direct sound is problematic, when the goal is to achieve less HF energy. The best solution is to start with an asymmetric device. With such, the horizontal pattern can be optimized for a constant dispersion across the passband of a tweeter, that is most important for our perception, and the vertical dispersion helps to limit the total energy dissipated.
My point was that different horns (different radiation patterns) may sound best with different direct-sound frequency response. If that's true, and I believe it is, then it's a wrong idea to EQ them to the same direct sound. The comparison simply should be made in the settings in which they sound best individually (at some subjectively equal loudness, for which the room must be included).
Sorry for the wrong wording. I didn't mean the overall depth of the horn, but the throat. I have all "https://horns-diy.pl/" horns. I have tested all 1" with the same speaker. Seos10 was the best.
By changing just the depth of the throat (as with the LF extension discussed here), the radiation pattern doesn't have to change. Then I see no reason why a more efficient device should sound worse than the less efficient one. (Here, an EQ for the same FR would be more than appropriate, of course.)
But that's a special case. In general, many "deeper" horns tend to beam, which indeed doesn't sound very good. I just wanted to say that "deep" doesn't necessarily mean that it will beam. So it's not so simple as short/deep.
But that's a special case. In general, many "deeper" horns tend to beam, which indeed doesn't sound very good. I just wanted to say that "deep" doesn't necessarily mean that it will beam. So it's not so simple as short/deep.
Last edited:
Interesting. What's 'correct' in your opinion then, if ANSI/CTA-2034-A is not?
How can you ask such question? Correct DI calculation uses the correct, i.e. actually radiated, total sound power. ANSI/CTA-2034-A is only a crude approximation using just the H/V polars (for understandable reasons), and as I said, for substantially asymmetric radiation patterns, it's simply not accurate. It's even intuitive, IMHO.
I understand that you always want to push things further, and that I do appreciate this. However, I really do not like your attitude that is constantly creating a high ground for yourself from where you find yourself enabled to ignore the input of others.
I even think you are factually wrong. A waveguide which dissipates too much high frequency energy into the room cannot be corrected properly by sloping the ON. The signature stays. And the sloping of the direct sound will be heard. There might be a certain window to do so, but it is very limited, as we dominantly perceive direct sound and reflections determine only some of our perception of timbre, which is, however, subordinated to direct sound, again.
I might still be wrong, and would not mind if proven otherwise, but this would require evidence. So if you think the abstraction that is ANSI/CTA-2034-A is incorrect, provide proof. After all, you seem to be a knowledgeable software engineer that could present evidence what the shortcomings of a two dimensional representation of a radiation pattern versus a full hemisphere actually means. Until you do not provide this, I cannot take your words at face value.
Please go on. I am very interested to learn about your – actual – findings. Until then, I am skeptical, as for me, 'intuitively', an asymmetrical device appears to be the most logical solution.
Last edited:
Accurately measuring and depicting true 3D dimensional loudspeaker directivity is not easy. Most of the time dual plane data is more than sufficient. When it is not, for example when a pro manufacturer wants to have their speaker used in an array, they will go to the trouble of having the full 3D balloon data captured.
The difference in representation from this
to this
The angular resolution of measurements can hide a lot of features in the same way that smoothing or windowing does. Whether this matters or not depends entirely on what the data is going to be used for.
Because the vertical data occupies the same portion of a sphere as the horizontal does, it is given the same weighting in CTA2034. In perception terms the horizontal response dominates but that is not taken into account.
Perhaps these are the sort of things mabat is referring to.
Yes, I do not question that the two-dimensional representation might have its shortcomings, but it is, as of today, our best means to evaluate what we are doing within a practical scope in speaker design. If Marcel provided insights how this model is insufficient in a way that it occludes actual properties of a waveguide, I am all in.
The situation was different: I had articulated my concerns that sloping the ON would be detrimental to the presentation, and Marcel, ignoring the rather detailed conditions that I had established for my observation, well knowing that it might be a specific case, simply went over the facts this model could provide, to express his rather cloudy, unsupported believes.
I am always happy to learn more, but this would require dialogue and finally, proof. I will be waiting.
Best,
Marinus
The situation was different: I had articulated my concerns that sloping the ON would be detrimental to the presentation, and Marcel, ignoring the rather detailed conditions that I had established for my observation, well knowing that it might be a specific case, simply went over the facts this model could provide, to express his rather cloudy, unsupported believes.
I am always happy to learn more, but this would require dialogue and finally, proof. I will be waiting.
Best,
Marinus
Last edited:
I think we are in all agreement, but may be misunderstanding each other.
According to CTA2034…
“Sound Power
The sound power is the weighted rms average of all 70 measurements…with individual measurements weighted according to the portion of the spherical surface that they represent.
Calculation of the sound power curve begins with a conversion from SPL to pressure, a scalar magnitude. The individual measures of sound pressure are then weighted according to the values shown in Appendix C and an energy average (rms) is calculated using the weighted values. The final average is converted to SPL...
…
Appendix C. Sound Pressure Weighting Values
In order to approximate the total sound power radiated from a source by using the measurements at equal angular increments, the measurement at each angular increment must be weighted by the appropriate value. The weighting value corresponds to the area of the spherical quadrangle at the microphone position for a particular angular position."
The true sound power is a theoretical limit towards which measurements are taken continuously in all 3 dimensions, not 2 orthogonal circles at discrete intervals eg. 10 or 5 degrees.
The more irregular the device under test, the further away the 70 measurements captured by these 70 measurements would be to its true theoretical power response.
Green- shaded region by this author.
“In theory, with complete 360-degree anechoic data on a loudspeaker and sufficient acoustical and geometrical data on the listening room and its layout it would be possible to estimate with good precision what would be measured by an omnidirectional microphone located in the
listening area of that room. By making some simplifying assumptions about the listening space, the data set described above permits a usefully accurate preview of how a given loudspeaker might perform in a typical domestic listening room. Obviously, there are no guarantees, because individual rooms can be acoustically aberrant. Sometimes rooms are excessively reflective (“live”) as happens in certain hot, humid climates, with certain styles of interior décor and in under-furnished rooms. Sometimes rooms are excessively “dead” as in other styles of décor and in some custom home theaters where acoustical treatment has been used excessively. This form of post processing is offered only as an estimate of what might happen in a domestic living space with carpet on the floor and a “normal” amount of seating, drapes and cabinetry. Reverberation time would typically be in the vicinity of 0.4 s and relatively constant with frequency over most of the frequency range.
For these limited circumstances it has been found that a usefully accurate Predicted In-Room (PIR) amplitude response, also known as a “room curve” is obtained by a weighted average consisting of 12 % listening window, 44 % early reflections and 44 % sound power. At very high frequencies errors can creep in because of excessive absorption, microphone directivity, and room geometry. These discrepancies are not considered to be of great importance. Free-field data of the kind shown in Figure 4 are generally trustworthy indicators of performance at very high frequencies.
Figure 11 shows a comparison in which the agreement is impressive above the
transition/Schroeder frequency for the room (300 Hz to 400 Hz). At lower frequencies it is clear that the PIR cannot anticipate the effects of room modes and standing waves, although an ANSI/CEA-2034-A underlying trend seems to be evident … Individual users may experiment with other proportions of direct and spatially-averaged curves to find a good PIR fit for rooms they find themselves working in.
It is important to keep in mind that scientific research has shown that human listeners are farmore analytical than a microphone in a room. While it is tempting to think that a good looking room curve or PIR is a guarantee of good sound, it is not. The first requirement for good sound in a room is a good loudspeaker, something that will be evident from inspection of a data set. A room curve is not a diagnostic tool, in that it reflects a combination of the loudspeaker and the room, however by inspecting the comparison of a PIR and a measured room curve, it may be possible to learn something about the effects of the room, and from that what acoustical modifications might be required.”
Words Italicized by author.
Of course; the relevance of capturing the true power response, the power response, the predicted in-room response and its practical relevance in arbitrary rooms remains to be fully elucidated, as least to this author. For sure it is at least 86% better than a single on-axis curve.
Reference:
https://shop.cta.tech/products/standard-method-of-measurement-for-in-home-loudspeakers
According to CTA2034…
“Sound Power
The sound power is the weighted rms average of all 70 measurements…with individual measurements weighted according to the portion of the spherical surface that they represent.
Calculation of the sound power curve begins with a conversion from SPL to pressure, a scalar magnitude. The individual measures of sound pressure are then weighted according to the values shown in Appendix C and an energy average (rms) is calculated using the weighted values. The final average is converted to SPL...
…
Appendix C. Sound Pressure Weighting Values
In order to approximate the total sound power radiated from a source by using the measurements at equal angular increments, the measurement at each angular increment must be weighted by the appropriate value. The weighting value corresponds to the area of the spherical quadrangle at the microphone position for a particular angular position."
The true sound power is a theoretical limit towards which measurements are taken continuously in all 3 dimensions, not 2 orthogonal circles at discrete intervals eg. 10 or 5 degrees.
The more irregular the device under test, the further away the 70 measurements captured by these 70 measurements would be to its true theoretical power response.
Green- shaded region by this author.
“In theory, with complete 360-degree anechoic data on a loudspeaker and sufficient acoustical and geometrical data on the listening room and its layout it would be possible to estimate with good precision what would be measured by an omnidirectional microphone located in the
listening area of that room. By making some simplifying assumptions about the listening space, the data set described above permits a usefully accurate preview of how a given loudspeaker might perform in a typical domestic listening room. Obviously, there are no guarantees, because individual rooms can be acoustically aberrant. Sometimes rooms are excessively reflective (“live”) as happens in certain hot, humid climates, with certain styles of interior décor and in under-furnished rooms. Sometimes rooms are excessively “dead” as in other styles of décor and in some custom home theaters where acoustical treatment has been used excessively. This form of post processing is offered only as an estimate of what might happen in a domestic living space with carpet on the floor and a “normal” amount of seating, drapes and cabinetry. Reverberation time would typically be in the vicinity of 0.4 s and relatively constant with frequency over most of the frequency range.
For these limited circumstances it has been found that a usefully accurate Predicted In-Room (PIR) amplitude response, also known as a “room curve” is obtained by a weighted average consisting of 12 % listening window, 44 % early reflections and 44 % sound power. At very high frequencies errors can creep in because of excessive absorption, microphone directivity, and room geometry. These discrepancies are not considered to be of great importance. Free-field data of the kind shown in Figure 4 are generally trustworthy indicators of performance at very high frequencies.
Figure 11 shows a comparison in which the agreement is impressive above the
transition/Schroeder frequency for the room (300 Hz to 400 Hz). At lower frequencies it is clear that the PIR cannot anticipate the effects of room modes and standing waves, although an ANSI/CEA-2034-A underlying trend seems to be evident … Individual users may experiment with other proportions of direct and spatially-averaged curves to find a good PIR fit for rooms they find themselves working in.
It is important to keep in mind that scientific research has shown that human listeners are farmore analytical than a microphone in a room. While it is tempting to think that a good looking room curve or PIR is a guarantee of good sound, it is not. The first requirement for good sound in a room is a good loudspeaker, something that will be evident from inspection of a data set. A room curve is not a diagnostic tool, in that it reflects a combination of the loudspeaker and the room, however by inspecting the comparison of a PIR and a measured room curve, it may be possible to learn something about the effects of the room, and from that what acoustical modifications might be required.”
Words Italicized by author.
Of course; the relevance of capturing the true power response, the power response, the predicted in-room response and its practical relevance in arbitrary rooms remains to be fully elucidated, as least to this author. For sure it is at least 86% better than a single on-axis curve.
Reference:
https://shop.cta.tech/products/standard-method-of-measurement-for-in-home-loudspeakers
Last edited:
Had to start googling to learn what this was. Posting this here to save time for anyone else in the same boat as me. (Dayton H6512 is a copy of the JBL PT-F95HF)
Progressive Transition
Waveguide Families:
PT waveguides are grouped into two families. The first is “compact”, and second is “optimized coverage/rotatable”. Compact PT waveguides balance performance in favor of small overall package size. Frequency response is optimal, distortion is superbly low, depth is minimized for use where a shallow enclosure is required. Beamwidth and directivity are optimal in the horizontal plane. Vertical beamwidth and directivity are optimized to provide a good match with JBL low frequency and midrange transducers; however, vertical pattern control does not extend as low as optimized coverage PT waveguides. Figure 5 shows a compact PT waveguide.
Systems with rotatable PT waveguides optimize pattern control both horizontally and vertically. Pattern control is extended to a lower frequency. The installer can easily configure the loudspeaker for horizontal or vertical use. In systems using an optimized coverage PT waveguide, smooth frequency response, and the uniformity of off-axis coverage, and arrayability are all superior. A rotatable PT waveguide is shown in figure 6.
Compact vs. Optimized Coverage PT Waveguides:
Each PT design is appropriate for a wide variety of
applications:
Compact PT waveguides offer these features:
• Minimized enclosure size.
• Optimized low distortion for maximum output.
• Maximum output and superior intelligibility.
Stage monitors, distributed systems, and small
arrays are excellent applications.
Optimized Coverage PT waveguides allow for:
• Rotatable systems: Horizontal or Vertical
orientation.
• Extremely smooth frequency response at all
playback levels.
• Predictable arrayability in engineered loudspeaker systems.
• Superior uniform coverage in difficult acoustical
environments.
• Improved intelligibility.
Source: https://jblpro.com/en/site_elements/tech-note-progressive-transitiontm-pt-waveguides (PDF download from JBL Pro)
Let's keep it technical, this is not about someone's beliefs or attitutes.
When all you have are the H and V (0° and 90° inclination) orbits, the best what you can do is a mere average, giving them both the same weight. There's simply no information what is between. That's what is used in CEA2034, but it doesn't mean it's correct.
I won't go through a whole proof, it's just maths, but here's a hint -
From the following picture it should be obvious that if you measured the orbit at 45° inclination for such an asymmetric source, the data would be closer to 90° (V) than to 0° (H). In other words, on a larger portion of a surface around the source the radiation would be closer to the vertical performance than to the horizontal. This means that for the total power calculation (and the DI), vertical orbit should be given a higher weight than the horizontal.
Now imagine even a simpler example - a square waveguide. The fact that it's actually square, and not perfectly round (!) is completely missing in the data.
The CEA2034 is simply not well suited for horns in general (for round it's of course perfectly accurate). Only after I tried to use it, I found how much it can be forgiving in many cases. If someone shows nothing but the CEA2034 curves, be cautious.
When all you have are the H and V (0° and 90° inclination) orbits, the best what you can do is a mere average, giving them both the same weight. There's simply no information what is between. That's what is used in CEA2034, but it doesn't mean it's correct.
I won't go through a whole proof, it's just maths, but here's a hint -
From the following picture it should be obvious that if you measured the orbit at 45° inclination for such an asymmetric source, the data would be closer to 90° (V) than to 0° (H). In other words, on a larger portion of a surface around the source the radiation would be closer to the vertical performance than to the horizontal. This means that for the total power calculation (and the DI), vertical orbit should be given a higher weight than the horizontal.
Now imagine even a simpler example - a square waveguide. The fact that it's actually square, and not perfectly round (!) is completely missing in the data.
The CEA2034 is simply not well suited for horns in general (for round it's of course perfectly accurate). Only after I tried to use it, I found how much it can be forgiving in many cases. If someone shows nothing but the CEA2034 curves, be cautious.
Last edited:
I haven't written a word about or against this, I actually don't have a strong opinion - I see some disadvantages. I only wrote that I found your data regarding the DI a bit too optimistic. And I already explaind why.