To quote Geddes:
Next, he went on about how "some" think that all reflections are good, but clearly this is not the opinion of Geddes. All box reflections would be called early and fall under the <10 ms, right?
Geddes said:
Which makes me think that Geddes wouldn't see it the way you do.
I've asked Geddes about this subject in the past, in this forum. It was a long standing disagreement he had with Toole. But after talking to Toole in more recent years they came to some kind of mutual understanding. Toole even softened up his statement in the most recent pressing of his famous book. Let's see if I can find me questioning Geddes about this...
https://www.diyaudio.com/community/...eo-phantom-center.277519/page-26#post-4715727
Geddes paper said:
Next, he went on about how "some" think that all reflections are good, but clearly this is not the opinion of Geddes. All box reflections would be called early and fall under the <10 ms, right?
Geddes said:
Which makes me think that Geddes wouldn't see it the way you do.
I've asked Geddes about this subject in the past, in this forum. It was a long standing disagreement he had with Toole. But after talking to Toole in more recent years they came to some kind of mutual understanding. Toole even softened up his statement in the most recent pressing of his famous book. Let's see if I can find me questioning Geddes about this...
https://www.diyaudio.com/community/...eo-phantom-center.277519/page-26#post-4715727
And wanting to maximize these results cannot be a bad thing. So a little goes a long way 😉.
(see fluids graph)
(see fluids graph)
Same baffle width, difference is 200mm vs 400m depth of cabinet, from Vineeth's thread (which I think you linked before), no new evidence just the fact that I showed you the difference more clearly in a gif.
Raw images are here and in the post below
https://www.diyaudio.com/community/threads/a-3-way-design-study.376620/post-6831404
Are we moving towards being on the same page?
Don't let colors blind you!
Check the frequency responses from the exact same posts, its db or two difference, nothing to do with polar pattern, its mere ripple riding on what the front front of the box makes it be. Mind you, the scales are different so its easy to get fooled with these as well, trace one line and then compare to the other. Even better, make new comparable graphs if you still have the project easily accessible. Perhaps dB scales differ in your GIF as well?
ps. cycle through the images, similar effect as with GIF
pps. "the forward leaning waves", anomaly that moves higher in frequency with increasing off-axis angle seen in these graphs show diffraction from back of the box, very easy to spot on measurement data. These waves just shift in frequency when back of the box changes, but polar pattern is pretty much the same. Both show diffraction related secondary sound source, interference in frequency response, which one is better? I say it doesn't matter, they are equivalent in this sense. Perhaps the other one is better sounding, but which?
Check the frequency responses from the exact same posts, its db or two difference, nothing to do with polar pattern, its mere ripple riding on what the front front of the box makes it be. Mind you, the scales are different so its easy to get fooled with these as well, trace one line and then compare to the other. Even better, make new comparable graphs if you still have the project easily accessible. Perhaps dB scales differ in your GIF as well?
ps. cycle through the images, similar effect as with GIF
pps. "the forward leaning waves", anomaly that moves higher in frequency with increasing off-axis angle seen in these graphs show diffraction from back of the box, very easy to spot on measurement data. These waves just shift in frequency when back of the box changes, but polar pattern is pretty much the same. Both show diffraction related secondary sound source, interference in frequency response, which one is better? I say it doesn't matter, they are equivalent in this sense. Perhaps the other one is better sounding, but which?
Last edited:
Of course this all is just semantics, what we are interested in if any of it affects perceived sound. If it does, then just round the box back and ripple disappears. Polar pattern doesn't change, ripple changes. Adjust baffle size and shape if you want to change polar pattern. We can all see this, right? signifiganse is for everyone to decide.
Spherical enclosure would not show any of these effects in frequency response indicating no delayed sound, no secondary sound source. Also Mabat thread shows waveguides without signs of secondary sound source.
Spherical enclosure would not show any of these effects in frequency response indicating no delayed sound, no secondary sound source. Also Mabat thread shows waveguides without signs of secondary sound source.
Ugh, there is no more than 1-2db difference, why don't you look the numbers on the graphs instead of colors? Frustration is both ways and it shows.
Alright I get it its not worth it to spend time here arguing for silly things so, its over and lets avoid this alright. There is enough data for anyone to look and draw their conclusions. Mine is not superior to yours, but I'd like to get on the right conclusions in order to make better sounding system.
Alright I get it its not worth it to spend time here arguing for silly things so, its over and lets avoid this alright. There is enough data for anyone to look and draw their conclusions. Mine is not superior to yours, but I'd like to get on the right conclusions in order to make better sounding system.
I am terribly sorry it went like so, not the intent at all. Kind of grateful this happened though as there is lesson to learn
Well, that might be significant. I have little expertise in measuring or designing for directivity below 500 Hz. I don't have the measurement set-up and I don't have the software simulation tools. But allow me to pull on a thought thread here...
A typical monopole box speaker has about 5 to 6 dB of DI at 1000 Hz. This means the on-axis is 5 to 6 dB higher than the sound power. At 100 Hz, the DI is near 0. If a speaker was able to maintain a directivity index of 5 dB down to 100 Hz, it would be very significant. This is the whole point of cardioid and dipole speakers. So the range of potential impact between omnidirectional and highly directional is about 5 dB. A difference of 1 or 2 dB is a difference of 20% to 40%.
I have never auditioned a true cardioid speaker, and I look forward to doing so someday. Most people can hear a difference compared to a monopole speaker.
j.
Hi hifijim,
as you say DI is difference between sound power and on-axis response and 1-2db difference in DI would be significant, sound outside of listening window be reduced.
Like you write when baffle is big enough compared to wavelength DI is up to about 6, cardioidish response as the speaker radiates half space. As frequency drops radiation transitions to full space, bafflestep. On bafflestep the on-axis response drops in relation to sound power and DI drops with it.
Now lets think it through how to maintain high DI to lower frequency:
1. One can lower the bafflestep, basically make the baffle bigger.
2. The other way is to introduce additional sound sources beyond baffle to cancel sound behind the speaker, to reduce sound power. As on-axis response drops with bafflestep the sound power needs to drop as well to maintain their relationship, DI.
Diffraction shows as interference in on-axis response but sound power is not affected as diffraction related secondary sound source, that is responsible for the interference ripple, just redistributes the sound differently to various directions on bandwidth where it happens. Diffraction does not add additional sound to increase sound power. Interference ripple on-axis means DI has ripple as well.
as you say DI is difference between sound power and on-axis response and 1-2db difference in DI would be significant, sound outside of listening window be reduced.
Like you write when baffle is big enough compared to wavelength DI is up to about 6, cardioidish response as the speaker radiates half space. As frequency drops radiation transitions to full space, bafflestep. On bafflestep the on-axis response drops in relation to sound power and DI drops with it.
Now lets think it through how to maintain high DI to lower frequency:
1. One can lower the bafflestep, basically make the baffle bigger.
2. The other way is to introduce additional sound sources beyond baffle to cancel sound behind the speaker, to reduce sound power. As on-axis response drops with bafflestep the sound power needs to drop as well to maintain their relationship, DI.
Diffraction shows as interference in on-axis response but sound power is not affected as diffraction related secondary sound source, that is responsible for the interference ripple, just redistributes the sound differently to various directions on bandwidth where it happens. Diffraction does not add additional sound to increase sound power. Interference ripple on-axis means DI has ripple as well.
Last edited:
I do not wish to re-open debate, but for anyone following who doesn't really understand, this to me is what is seen in the data. The polar heat maps and the polar curves are exactly the same data, this is because the polar curves are generated in vacs from the heat map data. When you have stared at either sort of graph for a long time and been able to manipulate resolution, highlight the individual curves etc. it does make it easier to see patterns and identify where features are occurring and what it was that changed them. Some graphs are easier to see certain features in and in this case I think the heat map is the easier to understand. The polar curves show the same but the near axis dominates vision in these making the far off axis less easy to pick. The gif on the previous page shows that with a smaller depth of cabinet the main diffraction hump is higher in frequency with a defined single peak from a null that forms at 150 degrees. This is a common feature of anything that is tending towards super cardioid response, a null forms there creating that directivity, it pinches in the 130 to 160 degree range.
In the graph for the deeper cabinet the main diffraction hump has been moved lower as the combined path length is now longer. In this specific case because of the ratio between width and depth there is two nulls that form but neither quite as strong, this has the effect of making the whole response have more diffractive ripple higher in frequency and two diffractive humps that extend out to off axis angles.
That is the difference that comes solely from changing the cabinet depth, depending on the design, the frequency range and any number of factors it may or may not be important, or even useful, but it is a difference that exists and it can be used to try and fine tune a design, to reduce the impact of rear diffraction or make it useful in changing the pattern to something slightly more desirable.
In the graph for the deeper cabinet the main diffraction hump has been moved lower as the combined path length is now longer. In this specific case because of the ratio between width and depth there is two nulls that form but neither quite as strong, this has the effect of making the whole response have more diffractive ripple higher in frequency and two diffractive humps that extend out to off axis angles.
That is the difference that comes solely from changing the cabinet depth, depending on the design, the frequency range and any number of factors it may or may not be important, or even useful, but it is a difference that exists and it can be used to try and fine tune a design, to reduce the impact of rear diffraction or make it useful in changing the pattern to something slightly more desirable.
I don't also want to get back to the debate, because its not constructive and nobody wants or needs it. I'm not sure if my communication skill is any better yet but here goes, small nugget.
Now that there is observations on the data one can start asking questions about it and reason answers from the data and turn the graphs into useful information. First question might be which one of the two depths would be better? Or, why the interference ridges that form in the graphs bend forward, higher up in frequency to more off axis but don't extend all the way to 180 deg, or why they dont extend past about baffle width wavelength? Or, why my home measurements dont show these back edge diffraction ripple, but for example klippel scanner measurements do?
After finding answers to first layer of questions there is more: how do I use the information for my advantage, what kind of box shape yields results I'd like?
Now that there is observations on the data one can start asking questions about it and reason answers from the data and turn the graphs into useful information. First question might be which one of the two depths would be better? Or, why the interference ridges that form in the graphs bend forward, higher up in frequency to more off axis but don't extend all the way to 180 deg, or why they dont extend past about baffle width wavelength? Or, why my home measurements dont show these back edge diffraction ripple, but for example klippel scanner measurements do?
After finding answers to first layer of questions there is more: how do I use the information for my advantage, what kind of box shape yields results I'd like?
Lack of frequency resolution in home measurements would be the primary reason for not seeing these effects so clearly, below 1K at 5ms you are looking at general trends only. Simulation is a good way to bridge the gap, much like it is at very low frequencies.
The first part is not so simple. The width and shape of the front baffle will set the first onset of diffraction effects before the combination of width, depth, ratio, rounding etc. will begin to interfere. The smoother you make the transition from driver to surrounding air the smoother the overall diffraction signature will be and it will appear more as a change in directivity than a ripply mess. Observation fields are a good way to see how the wavefront is moving and what obstacles do to it at different frequencies.
Most forward firing boxes will produce less sound to the rear than they do to the front and the back corners will often create enough of a secondary effect to still get the 130 to 160 degree pinch. The sort of box that wesayso created doesn't do that, it seeks to present as little obstacle to the sound as possible, by having the curve change smoothly and continuously it is very much is like a waveguide with rollback.
You have to decide what result you are targeting to know which box shape will help or hinder. Do you want the smoothest diffraction response, do you want to try and constrain directivity to a lower frequency by using baffle area, do you want to have less sound to the rear or some combination? Apart from general trends that can be seen from the few examples that exist, what is needed is to try out different options that can fit within your design constraints and see what the outcome is and whether you like it or want to tweak it. mabat has made this much simpler for the uninitiated to try for themselves in ABEC if they can get the hang of the script. There are still limitations there, to remove them a deeper understanding of the program and how to create the meshes is needed.
exactly. Lets make thought experiment, simplifying a bit, ignoring baffle edge diffraction for second and think about the back edge and how back of the box contributes in graphs.
If box is 40cm deep and we observe ~on-axis, it is roundtrip for sound to travel from front edge to back edge, where opposite polarity secondary sound source happens. When this comes after direct sound to our observation point ts roughly 80cm late. Too late, and too low frequency effect, it doesnt fit to window we have to use.
This kind of thing happens to frontal sector angles but observe past that there is (almost) only front edge and back edge as sound source, as box edge shadows direct sound from driver and things change bit. Another change happens quite fast past that when observation angle is big enough as opposite side back edge is visible to observation point. Here the effect of closer back edge diffraction secondary sound source to response starts to diminishs as it becomes part of the direct sound, and now the secondary sound source we see diffraction is what comes around opposite side back edge, which is also direct sound and much louder than the opposite polarity backwards rafiating diffraction secondary sound source ever is.
This interference also shows in out semi-anechoic measurements as it now fits the window because the extra delay is missing. Wiggle starts to show up on the high degree observation angles as now there is basically no extra delay if rotation axis is on the baffle plane, and we see sound interference again. Also, if the box is smallish sound diffracting around above and below also appears here, coming all around the speaker.
There really is great deal of information in the graphs and can be fed to imagination nicely. It depends how far one wants to take it but after using imagination for this a while it became obvious to me that situation on the back of the box is so complex at least simple box shapes cannot control the sound there, its ripple there ripple here kind of situation. Ripple /interference could be minimized or left as it is, but its about impossible to harness it for effective pattern control. It would be cool if AI / optimization proces could come up with some (complex) shape or something that would make the backside useful. Otherwise we are left pretty much with options that are more or less chaotic response there.
If box is 40cm deep and we observe ~on-axis, it is roundtrip for sound to travel from front edge to back edge, where opposite polarity secondary sound source happens. When this comes after direct sound to our observation point ts roughly 80cm late. Too late, and too low frequency effect, it doesnt fit to window we have to use.
This kind of thing happens to frontal sector angles but observe past that there is (almost) only front edge and back edge as sound source, as box edge shadows direct sound from driver and things change bit. Another change happens quite fast past that when observation angle is big enough as opposite side back edge is visible to observation point. Here the effect of closer back edge diffraction secondary sound source to response starts to diminishs as it becomes part of the direct sound, and now the secondary sound source we see diffraction is what comes around opposite side back edge, which is also direct sound and much louder than the opposite polarity backwards rafiating diffraction secondary sound source ever is.
This interference also shows in out semi-anechoic measurements as it now fits the window because the extra delay is missing. Wiggle starts to show up on the high degree observation angles as now there is basically no extra delay if rotation axis is on the baffle plane, and we see sound interference again. Also, if the box is smallish sound diffracting around above and below also appears here, coming all around the speaker.
There really is great deal of information in the graphs and can be fed to imagination nicely. It depends how far one wants to take it but after using imagination for this a while it became obvious to me that situation on the back of the box is so complex at least simple box shapes cannot control the sound there, its ripple there ripple here kind of situation. Ripple /interference could be minimized or left as it is, but its about impossible to harness it for effective pattern control. It would be cool if AI / optimization proces could come up with some (complex) shape or something that would make the backside useful. Otherwise we are left pretty much with options that are more or less chaotic response there.
Thisnalso contains answers why the "ridges" dont extend all the way to 180deg or why the effects extend only to about bafflestep and below.
I no longer bother with this kind of thinking. Wave propagation is both simple in the sense it obeys it's own rules and quite complex, it can be hard to predict how these rules will interact with each other to produce the end result. The more I learn about how things work the more I realise I really don't understand it nearly as well as I thought I did.
I see the glass as half full not half empty. Use the right tool, test out the options evaluate the results, rinse, repeat until satisfied. For anyone who can afford or has access to COMSOL or equivalent and someone well versed enough to drive it, finding an answer to a question you pose it is simple. Without it there is more work and you have to create the variations yourself.
augerpro used the scripts I gave him to get started and has produced a design he is satisfied with, tried some options to see the effects and picked the one he liked. This is no quantum shift just making the best of what you have within the constraints that are set.
Yeah exactly, same conclusion as I came about, sound propagation is simple on the baffle and at the baffle edge because the source is simple enough but past that all bets are off I totally agree as the sound source becomes much more complex than simple point source. There is still value in thinking as it helps understand the graphs some. For example if one wants to do something about "the ridges" then change depth of the box, round the back edge to get less. If radiation on back angles needs to be manipulated, then manipulate also size of the back and so on, gives pointers for what kind of simulations to run so there is no need to run them all. Although computers are good at running things.
Thought experiment for others, if there is someone trying to catch up. Lets simplify, one woofer on a box. Lets observe the system front side. Its pretty easy to predict and calculate what ideal response would be by the size of the transducer and diffraction that happens at the edge, as we have a point source.
Now rotate the box in your imagination so that you dont see the woofer anymore. Now there is no point sources visible from this perspective, but line sources, front edges of the box! or some other edges you see, probably some front edges are also hidden. And the line sources are not playing in syncrony but have delay to them, as we had point source originating the sound it has varying path length to edge.
Complexity of things went high up immediately past the baffle edge, and if we need to have control on the sound the edge is last frontier, after which we are losing the battle and left with not much options other than try and reduce all kinds of reflections/diffraction, smooth things out. Especially if the edge is long. Of course its all frequency dependent, low enough frequencies are not affected by the edge/box. Its midrange that is affected.
If one wants to make cardioid box for example, having sound power cancelling secondary sound source close the baffle edge seems logical as it has still some chance to approximate sound that came over the particular edge and cancel it out effectively. The further back this cardioid forming secondary source is the more chaotic sound field there is to cancel as sound intensity comes from multiple linesources now more than from a singular one, smeared, and I'd imagine cancellation is not as effective. Again, the midrange performance suffered, low frequencies are not affected the longer the wavelength do box size.
Thought experiment for others, if there is someone trying to catch up. Lets simplify, one woofer on a box. Lets observe the system front side. Its pretty easy to predict and calculate what ideal response would be by the size of the transducer and diffraction that happens at the edge, as we have a point source.
Now rotate the box in your imagination so that you dont see the woofer anymore. Now there is no point sources visible from this perspective, but line sources, front edges of the box! or some other edges you see, probably some front edges are also hidden. And the line sources are not playing in syncrony but have delay to them, as we had point source originating the sound it has varying path length to edge.
Complexity of things went high up immediately past the baffle edge, and if we need to have control on the sound the edge is last frontier, after which we are losing the battle and left with not much options other than try and reduce all kinds of reflections/diffraction, smooth things out. Especially if the edge is long. Of course its all frequency dependent, low enough frequencies are not affected by the edge/box. Its midrange that is affected.
If one wants to make cardioid box for example, having sound power cancelling secondary sound source close the baffle edge seems logical as it has still some chance to approximate sound that came over the particular edge and cancel it out effectively. The further back this cardioid forming secondary source is the more chaotic sound field there is to cancel as sound intensity comes from multiple linesources now more than from a singular one, smeared, and I'd imagine cancellation is not as effective. Again, the midrange performance suffered, low frequencies are not affected the longer the wavelength do box size.
Last edited: