Good luck than lol 😀 😀
Personally, I think it's a bit blown out of proportions.
Sure we can try to almost pixel peep every last bit of distortion number.
As long as this is all manageable I personally would extremely doubt you would be able to notice the difference.
So there is quite a bit of choice that way, even a RS150 will do fine a midrange for example.
If I don't do a project for a client, I personally would just try to find something that fits the looks and vibe of what I am going for, sort drivers accordingly that fit this vibe and just pick one.
Just make a spreadsheet incl prices and start crossing things out.
You often end up with just like a handful of candidates.
For a client I do the same thing, but in that case they make the choice for me.
You will be surprised how often aesthetics and looks are a major factor in these things.
One of the reasons why I am having trouble convincing clients about Purifi for example.
1) Someone correct me if I'm wrong, but I don't think the 1.0 -1.4 wl is just a formula, I thought Kimmo has verified this with real experiments
2) I'd assume that several people at SB Acoustics' would actually listen to a driver with their ears before releasing it. The idea that it's just a lemon isn't realistic to me. However, the idea that it is not as good as the CAC is certainly possible.
2) I'd assume that several people at SB Acoustics' would actually listen to a driver with their ears before releasing it. The idea that it's just a lemon isn't realistic to me. However, the idea that it is not as good as the CAC is certainly possible.
Sorry, maybe I didn't say this clearly.
The formula is the underlying principle behind it.
If you really want to know, I have to dive into the acoustics books again, don't know the thing on top of my head.
But it's the same one for any (two) difference sources, incl diffraction problems.
Because that is the same underlying principle with destructive interference.
edit: A = 2Acos(φ/2) or I = 4Icos²(φ/2), but there are many other ways to write this down.
The factor 1.0-1.4 was done with simulations.
Although you could also literally work this out in a big spreadsheet, matlab etc etc etc
If you do the same experiments but with added crossover filters to it, good luck.
Last edited:
Member
Joined 2003
I like lemons...Limoncello on the other hand...(maybe there's good Limoncello in the world, but the stuff we get is Luxardo which is the colour of nuclear waste and tastes of cleaning supplies)
Sorry, back to speaker talking...CTC wavelength is a "rule of thumb" based on previous "normal speaker" criteria I mentioned previously. I wouldn't interpret it as a hard set requirement for success, but more of a "lessons learned" through real world experience. The main point of this rule to be presented at all, is to convince people to move away from the old conception that placing driver as close together as possible is the best thing to do. It usually isn't.
Sorry, back to speaker talking...CTC wavelength is a "rule of thumb" based on previous "normal speaker" criteria I mentioned previously. I wouldn't interpret it as a hard set requirement for success, but more of a "lessons learned" through real world experience. The main point of this rule to be presented at all, is to convince people to move away from the old conception that placing driver as close together as possible is the best thing to do. It usually isn't.
hifijim,
Again thanks for the thoughtful reply.
My impression is that the 1.2 X WL := the C to C dimension is empirical, as in many speakers were built and 1.2 X WL := the C to C dimension is the sweet spot.
Kippel can document it.
Vituixcad or something similar can hopefully model it if you can account for all the variables including phase through the crossover.
Thanks DT
This is why everyone should use VituixCAD, one can do so much with proper data. This c-c thing is very easy to come up by yourself, you can literally tune the c-c with mouse wheel and see response change. It's very easy to optimize system for what ever as it's all real time adjustable. Of course the underlying (measurement) data is perhaps simulated or measured, and when tuning c-c in program it doesn't really modify the source that resulted the measurements, and you'd have to reflect back to sim or to prototype and get new data and verify. But it's just so much joy to tweak almost anything in real time and see all the graphs change, very good tool to build knowledge and intuition. One just needs to be aware how the sim corresponds to reality, how good the data is the sim is based on, and whether the data is setup as captured, because the program lets you do any sims, ideal and unrealistic ones as well.
By the way, It's much fun to play into unrealistic territory. Here is a task for everyone: try to make textbook response graphs with what ever means, and then figure out what it would actually take to be in your livingroom as real physical item 😉 Even the simplified diffraction sim with ideal flat disk transducers remind that wavelength varies dramatically, and any physical objects introduce "error", even the ideal ones, and reality is worse of course.
Last edited:
From Kimmo at htguide
"c-c studies are ridiculously easy with VituixCAD. Just load measurement data of the radiators, create ideal flat on axis response (with Optimizer and G(f) blocks) with estimated XO and tune driver's Y mm until combination of DI and ERDI is the best."
The optimizer and Gf block is a shortcut by basically inverting the response to create a totally flat on axis to eliminate that as a variable and just look at the directivity change from moving the vertical distance between the drivers.
The main point I was trying to explain is that it proves that only under one very specific condition.
Meaning it doesn't prove a lot I think personally, mainly because it's far to simplified from the practical world.
If you just simply take a step back and think what practical influences would already change the phase difference between two drivers, you quickly realize why.
But yeah sure, I do totally agree that it's just another tool that can be used.
Btw, when VituixCAD is being used you MUST turn off between angle interpolation as well as look at much smaller angle differences as well.
If you look at just a default 10 degrees or 15 degrees steps, you will or can miss out on it.
Anyway, anyone is free to believe, my experience is just not the same and I can assure that I have tested and simulated also an awful lot.
But I am 100% open if somebody can show me otherwise with real practical examples incl low enough angle differences.
On top of that show me that vertical directivity is that important to begin with.
Yeah it's the total system DI that can be adjusted with c-c. Increasing it the mainlobe gets narrower but additional sidelobes get introduced, which flattens the DI, more sound toward off-axis angles where would be huge lobe nulls if say ~0.8wl apart. If one likes to listen further away the smoother DI ought to make difference, as it's the early reflections / room one hears mostly, the power response.
I have tried to listen the lobes, and while it feels there is difference, there also isn't 😀 This one is hard to AB test, one would have to have two speakers side-by-side with differetn c-c and also DSP tuned to same axial response in order to try and figure out the difference. Without such test rig one can just listen try and figure out if lobes are audible or not. Here is what happens in my place with single speaker: if I take awkward position either close to speakers or further away where I can move my head at design axis, or above, or below, play sine tone at the xo/lobe frequency, the amplitude seems to vary quite randomly no matter what height or distance head is from speaker. The amplitude modulates if I move to any direction, it could be low on design axis at one listening distance, 20cm further it could be loud again, move up 20cm and it dips again, it just varies spatially no matter where the lobes are. I cannot stay in the lobe perceptually, or avoid it, SPL just seems to vary "randomly".
With music? Not so obvious, for example concentrating listening snare drum, on design axis it sounds fine while bit above it thins out, but I could just check with microphone why. I suspect the whole frequency bandwidth between say 100Hz - 5kHz or so would look quite different on axis and off-axis, also due to room reflections, no hint of lobe amongst all the interference captured. I suspect the lobes are obvious only with very close listening, or in situation where early reflections are late / attenuated but I'm not sure. They just don't seem to be very obvious in situation I have here.
This is great relief so to speak, designer of speaker doesn't have to worry how to get tweeter as close to woofer as possible and forget other things because brain is using resource on this one thing, maybe for nothing. If one just doesn't mind about it too much, just tweak the system to good DI what ever means. For example using a freestanding axisymmetric waveguide tweeter makes the c-c something like 1wl at crossover, and system sound seems to be fine, although I do not have anything to compare to currently. And it would be very difficult to compare in a way that the DI would change, no sufficient roundover would fit, responses need to match within 0,5db and so on 🙂 Perhaps one day try, but it's also fine not to worry too much about it, relaxed c-c yields nice graphs and also the sound seems nice, at least it is not obviously flawed in practice, so I'm not worrying about it. Perhaps I'm after hearing something better, but until then I've got more important things in mind.
It would be interesting how everyone hear this stuff, does it matter? Many say it does, but context is usually missing, like acoustics and listening distance for example (as per does one listen close enough for mostly direct sound or mostly roomy sound) and so on.
I have tried to listen the lobes, and while it feels there is difference, there also isn't 😀 This one is hard to AB test, one would have to have two speakers side-by-side with differetn c-c and also DSP tuned to same axial response in order to try and figure out the difference. Without such test rig one can just listen try and figure out if lobes are audible or not. Here is what happens in my place with single speaker: if I take awkward position either close to speakers or further away where I can move my head at design axis, or above, or below, play sine tone at the xo/lobe frequency, the amplitude seems to vary quite randomly no matter what height or distance head is from speaker. The amplitude modulates if I move to any direction, it could be low on design axis at one listening distance, 20cm further it could be loud again, move up 20cm and it dips again, it just varies spatially no matter where the lobes are. I cannot stay in the lobe perceptually, or avoid it, SPL just seems to vary "randomly".
With music? Not so obvious, for example concentrating listening snare drum, on design axis it sounds fine while bit above it thins out, but I could just check with microphone why. I suspect the whole frequency bandwidth between say 100Hz - 5kHz or so would look quite different on axis and off-axis, also due to room reflections, no hint of lobe amongst all the interference captured. I suspect the lobes are obvious only with very close listening, or in situation where early reflections are late / attenuated but I'm not sure. They just don't seem to be very obvious in situation I have here.
This is great relief so to speak, designer of speaker doesn't have to worry how to get tweeter as close to woofer as possible and forget other things because brain is using resource on this one thing, maybe for nothing. If one just doesn't mind about it too much, just tweak the system to good DI what ever means. For example using a freestanding axisymmetric waveguide tweeter makes the c-c something like 1wl at crossover, and system sound seems to be fine, although I do not have anything to compare to currently. And it would be very difficult to compare in a way that the DI would change, no sufficient roundover would fit, responses need to match within 0,5db and so on 🙂 Perhaps one day try, but it's also fine not to worry too much about it, relaxed c-c yields nice graphs and also the sound seems nice, at least it is not obviously flawed in practice, so I'm not worrying about it. Perhaps I'm after hearing something better, but until then I've got more important things in mind.
It would be interesting how everyone hear this stuff, does it matter? Many say it does, but context is usually missing, like acoustics and listening distance for example (as per does one listen close enough for mostly direct sound or mostly roomy sound) and so on.
Last edited:
Plots like that look worse, but we should think more about what is happening in room. This discussion is not talking about the founding principles at play. First, this method stems from using the CEA2034A standard, which in turn is based on (and validated by) the work at Harman by Toole, et al. CEA2034A includes the response of the entire 360 degree sphere around the speaker. BUT, it weights certain angles higher. For example, listening window has the highest weighting, for obvious reasons. Then a group of metrics called early reflections are given high weighting. The early reflections are comprised of horizontal (30-60 degree IIRC), ceiling reflection (pos 40-60), and floor reflection (neg 20-40). There are other reflections but those are lower weight. The point is the listening window and early reflections are the most important under this standard and that seems a reasonable position based on the work done by Toole. Based on that, you want the smoothest responses in these specific reflections, so similar sound is arriving at the listening position where all of these reflections combine. To illustrate, a reflection directly off a side wall to the listening position is going to be worse than an angle 130 degrees from on-axis, which bounces off four walls before reaching the listening position. Horizontal relflection is easy enough to smooth, but the vertical axis is where the problem is for non-coincident drivers. Recognizing there is going to be lobing/nulling somewhere in the vertical axis, the question becomes where exactly and can we optimize that? If you simulate this, you will see that the typical .7-1 wl ctc causes extremely large nulls around the crossover directed at the ceiling and, to some extent floor reflections. Increasing ctc to 1.2 wl will create much smoother reflections. In fact, at these angles, the response will look much more like a 1/4 wl ctc spacing.
So if you accept the founding premises, consider using 1.2 wl in your designs. If you don't believe certain early reflections are more important than others, don't use it. If you don't believe Toole et al. have established the importance of smooth reflections producing the smoothest possible combined response at the listening position, don't use it.
There are nits to pick if you are considering a specific listening space and designing around that. Different spaces may make certain reflections more problematic than others. But the overall idea is good perspective to start thinking from.
Last edited:
Thanks @augerpro for your answer above.
This is the most complete and 'to the point' answer (especially explaining the context based on which this rule has been derived) I have seen anywhere about this topic.. 🙂
This is the most complete and 'to the point' answer (especially explaining the context based on which this rule has been derived) I have seen anywhere about this topic.. 🙂
Although your idea here is true, it's a bit pulled out of context.
Toole et al only showed a very generic idea of this principle.
But like a couple of things in his book, quite some details and nuances are being skimmed over rather quickly.
Or in other words, yes this idea is definitely true in the bigger picture, but if you start to zoom into the details, definitely not as simple and straightforward.
For example, you can't just dump ALL early reflections, vertical, horizontal, diagonal etc into one big basket and call it the same thing.
The difference between horizontal vs vertical audibility for example isn't the same and very clear.
Unfortunately there isn't that much literature on this and in most cases people (researchers) assume they are equal.
Which is rather doubtful based on how our ears physically work.
Vertical directionality is close to zero for example.
And even if you do, you consider that as a problem in a (theoretical) vacuum.
I believe it's more important to zoom-out from the question and ask what practically is feasible.
Not to mention the total soup of stuff you get in actual rooms and listening positions.
Sure, in a controlled environment these things can become more clear.
But I know very few people listening in controlled environments.
Local absorption and reflections above 6-7kHz is very significant.
Meaning one coffee table or foot rest nearby a speaker and there is absolutely nothing left of that beautiful (vertical) directivity anymore!
I appreciate the comments, @b_force and @augerpro both make good points.
Regarding the importance of directivity index: I have struggled with this a bit in my own listening room. In my main system I have two different Mid-Tweeter units which I switch between every few months. One uses Satori MW16TX + Satori TW29TX. The other uses a Purifi PTT6.5M04-NFA + Satori TW29BNWG waveguide tweeter. A Hypex FA253 provides the 3-channel amps and dsp, and the woofer module handles 200 Hz down. In both cases, there is digital delay applied to the Mid-Tweeter unit to bring it “in-time” with the woofer.
The textreme system has a non-optimal directivity index. The mid and tweeter are closely spaced. The baffle was well designed to minimize high frequency diffraction, but it has some flat surface around the tweeter which creates some diffraction baffle gain around 2k. Both of these aspects combine to create a 2.5 dB hump in the DI curve.
The Purifi-waveguide system was designed from the outset for low diffraction and optimized directivity index.
Comparing the plots for the two, the conventional thinking is that the Purifi-waveguide system would be superior. It has a superior DI curve, ER curve, PIR curve, and sound power curve. But the reality is more complicated.
The two systems actually sound quite similar. From the perspective of tonal balance they are virtually the same. In terms of clarity, detail, and dynamic contrast, they are equal. The differences are mostly in the spatial presentation, and those are subtle. The textreme system has a wider presentation, but localization is slightly less specific. The Purifi-waveguide presents a slightly better illusion of depth.
So what does this mean? Does DI matter or not? Well, perhaps yes and no. In my normal listening position, I am in a large room but I sit fairly close to the speaker, so I am immersed in the direct sound. The influence of reflections is reduced, so most of what I hear is the on-axis response of both speakers, and they are very similar on-axis. @tmuikku has written a lot about this Griesinger effect. If I sat further away, I think I would experience more differences between the two speakers. I am very fortunate to have this listening space. If my two speakers were auditioned in a variety of rooms, I would expect that the Purifi-waveguide system would be easier to work with and more forgiving about placement. The textreme system would be more room sensitive and fussy about location.
My experience with the tower system of post #110 is that it is very easy to place in a room, and I credit the well managed directivity performance.
In a dedicated listening environment where the speakers and listeners can be optimally located, and acoustic treatments can be utilized, I imagine that directivity matters less, since reflections can be minimized. In a more typical domestic listening room, there are limited options for speaker/listener positioning, and dedicated acoustic treatment is rarely used. In this situation, managing the directivity index becomes more important. The listener may be exposed to a lot of reflected sound, so it is important that the reflected sound have a reasonably nice spectral balance. This is the essence of the Toole/Olive research.
Which begs the next question, why am I concerned about the DI performance of this speaker project? Because I don’t know what room it will be used in. This speaker will probably be a gift for a friend. I want this speaker to be suitable for a variety of environments, so I want the DI curve to be well behaved. Another reason is I enjoy the challenge of making speakers which sound nice, look nice, and have nice looking performance plots.
Regarding the importance of directivity index: I have struggled with this a bit in my own listening room. In my main system I have two different Mid-Tweeter units which I switch between every few months. One uses Satori MW16TX + Satori TW29TX. The other uses a Purifi PTT6.5M04-NFA + Satori TW29BNWG waveguide tweeter. A Hypex FA253 provides the 3-channel amps and dsp, and the woofer module handles 200 Hz down. In both cases, there is digital delay applied to the Mid-Tweeter unit to bring it “in-time” with the woofer.
The textreme system has a non-optimal directivity index. The mid and tweeter are closely spaced. The baffle was well designed to minimize high frequency diffraction, but it has some flat surface around the tweeter which creates some diffraction baffle gain around 2k. Both of these aspects combine to create a 2.5 dB hump in the DI curve.
The Purifi-waveguide system was designed from the outset for low diffraction and optimized directivity index.
Comparing the plots for the two, the conventional thinking is that the Purifi-waveguide system would be superior. It has a superior DI curve, ER curve, PIR curve, and sound power curve. But the reality is more complicated.
The two systems actually sound quite similar. From the perspective of tonal balance they are virtually the same. In terms of clarity, detail, and dynamic contrast, they are equal. The differences are mostly in the spatial presentation, and those are subtle. The textreme system has a wider presentation, but localization is slightly less specific. The Purifi-waveguide presents a slightly better illusion of depth.
So what does this mean? Does DI matter or not? Well, perhaps yes and no. In my normal listening position, I am in a large room but I sit fairly close to the speaker, so I am immersed in the direct sound. The influence of reflections is reduced, so most of what I hear is the on-axis response of both speakers, and they are very similar on-axis. @tmuikku has written a lot about this Griesinger effect. If I sat further away, I think I would experience more differences between the two speakers. I am very fortunate to have this listening space. If my two speakers were auditioned in a variety of rooms, I would expect that the Purifi-waveguide system would be easier to work with and more forgiving about placement. The textreme system would be more room sensitive and fussy about location.
My experience with the tower system of post #110 is that it is very easy to place in a room, and I credit the well managed directivity performance.
In a dedicated listening environment where the speakers and listeners can be optimally located, and acoustic treatments can be utilized, I imagine that directivity matters less, since reflections can be minimized. In a more typical domestic listening room, there are limited options for speaker/listener positioning, and dedicated acoustic treatment is rarely used. In this situation, managing the directivity index becomes more important. The listener may be exposed to a lot of reflected sound, so it is important that the reflected sound have a reasonably nice spectral balance. This is the essence of the Toole/Olive research.
Which begs the next question, why am I concerned about the DI performance of this speaker project? Because I don’t know what room it will be used in. This speaker will probably be a gift for a friend. I want this speaker to be suitable for a variety of environments, so I want the DI curve to be well behaved. Another reason is I enjoy the challenge of making speakers which sound nice, look nice, and have nice looking performance plots.
In general this is not a black and white question and answer.
The reality is obviously a grey scale without a discrete jump.
Where is the limit and are boundaries?
Nobody knows, I have never seen any research on this.
Toole et al only proved that there is a certain overal preference.
But never showed any specifics.
Practically, a lot depends on the room as well.
Like i said before, anything above 6-8kHz or so, forget it in a normal living room.
I personally find the 2-4kHz dip in the power response horrible with a regular tweeter, especially with a 6 or 8 inch woofer this is very evident.
It either feels like you have to much energy, or not enough high-end
And personally I think that's the best approach we can do.
We simply don't know beforehand what the best match will be.
Room acoustics and reflections are far to complex for that.
Even more so from a case to case basis.
But certain designs simply have a better change of creating less problems.
Which I personally think is also the answer to your last question.
The reason why we bother, is because we'll have a higher change of having nice sound reproduction.
And a lower change of running into potential issues, which can't be changed after the fact.
A good directivity never performs bad, but has very little gain at worst.
Or in other words, a good directivity is always beneficial in the sense that if we go for the opposite approach we can't go back anymore!!
No I can't hear a difference, nor measure a difference in the usable bandwidth. There is a slight difference in size/shape of the resonance peak, but not enough difference to change an analog or DSP filter. As far as I can determine, they are interchangeable.
I have the same experience.
The CAC is a smidge better, but mostly just on paper.
But both have such incredible low distortion numbers, especially when being used as a mid-range.