Except when they don’t.
Only 1 counter example needed. Alpair 7.3 (metal) vrs Alpair 7p (paper). The A7 broke the stereotyped metal sound, especially when i finished with them. And the paper cone driver sounded like you “expect” a metal cone to sound.
To take that further the (paper) A12pw has voicing that matches better with the metal cone than the paper.
There is a tendency for drivers to take on a sound related to the material they are made of but their execution is very important.
I have heard lots of metal cones with a “signature” metal sound, the MCMs i played with in the late 1990s, The small Founteks (FE87), and Jordans (JX92 and Eikona), the JX150 didn’t really suffer. And the poly TB W5-1611 has a plasticy top end.
dave
My personal SUBJECTIVE ranking in order:
1) Bitches Brew
2) Live Edge Dipoles
3) Bohlender-Graebener Ribbons <-> Shaded Array 12x3.5" - these are roughly in the same league, although still very different from each other
The B-G ribbons (which I just sold) served me very well for 15 years. They sounded great. I put them in sealed enclosures. If I was going to try them again I'd do dipole instead. (Dipole would make a significant difference.)
They had a lot of wonderful qualities, my main frustration with them was that the top one to two octaves had narrow dispersion and it did not have the level of clarity in detail that you would get from a ribbon tweeter, or other kind of very well designed tweeter. They sounded a little "midrangey." That "signature sound" I mentioned earlier.
I get what you're saying about ribbons having a reputation for being "the best" but they don't do bass well; integrating them with subs is hard; and the problem with the BGs as I said was they had a bit of a "plastic film" sound. I thought the Fountek 3.5" drivers had a better top octave.
If I had a bank of ribbon tweeters to cover the top octave, MAYBE matching them with the B-Gs would be even better. But that adds complexity and there are always things that could go wrong. You would lose some of the simplicity and elegance.
The Live Edge Dipoles have a huge soundstage, extremely good imaging, the kind of razor sharp detail and holographic transparency one expects from a Beryllium tweeter, and an incredibly consistent radiation pattern. If you're standing out in the hallway or the next room, the Live Edge Dipoles sound more like real music whereas the BG's sound more like speakers.
(Yes, I know that's not what you were expecting me to say.)
Keep in mind that each of these 4 systems were built around a MiniDSP 2x4HD DSP. Later I upgraded the Bitches Brew to Flex Eight which is better than the 2x4HD.
The Bitches Brews are similar to the Live Edge Dipoles overall with a lot of the same goals and advantages, but they are more powerful, more coherent, more seamless, kind of like the difference between a medium priced bourbon and a really expensive bourbon that is extremely smooth and has no aftertaste. They do everything just a little bit better.
So yes, I am saying that a compression horn coaxial pro 8" or 15" driver mounted on a slab of live edge wood, with the right DSP etc, can sound significantly better than a ribbon, at least in my personal opinion. I think there is a LOT to be said for a consistent well controlled radiation pattern and unfortunately that is not something that ribbons excel at.
Last edited:
To be fair, the handling and treatment of a low or a high DI arrangement is going to be quite different.
No off axis information, each drivers needs full space simulation on baffle.
If you dropped in 12 copys of the same FRD
then no the model assumes they are all mounted on top of each other.
Center to center gets wider and wider for each driver.
So xyz coordinates must be set.
speakers are all same drivers so phase is same of course.
but center to center is farther apart so actual phase of
each driver be different. relative to the listening position you establish.
And phase needs to be correct for model to even be accurate.
otherwise you have factory FRD file.
Model thinks your on a Half space IEC test baffle.
And they are all magical mounted in same place.
Once you model it in full space on a skinny tiny baffle.
Youll find out how unmagical they are.
Then deal with the baffle step that results.
And then the magical cancelation from drivers mounted
millions of miles apart.
So your rolling off the top end to get rid of the cancellation.
From mounted so far apart.
Then find out that 1st crossovers wont fix it.
specially vertical
Maybe 2 or 3 drivers first order maybe would get.
For 12 speakers, center to center of your bottom to top driver.
probably stepper filter, The distance will justify how low
it needs to be in frequency.
Last edited:
Thank you very much for adding more ingredients, well let's see how all this fits into the final cake, as you suggested I have given coordinates to each unit, it is the first time I simulate this in VitixCAD and although the help gives a few touches on each topic I will surely I'm leaving something out again, I appreciate the notes and observations.
but let's see what we have advanced today, I have done the simulation with both units (Fountek and Dynaudio) although to avoid confusion I will only publish those from the main thread (Fountek) the results apart from the frequency response are not very different.
I don't know where the unit width measurements are defined, now with the default width, the curve and position of the drivers in the lateral projection correspond to the model (I think so)
The box needs to be simulated for the port study but that will be at another time.
Un saludo!!!
but let's see what we have advanced today, I have done the simulation with both units (Fountek and Dynaudio) although to avoid confusion I will only publish those from the main thread (Fountek) the results apart from the frequency response are not very different.
I don't know where the unit width measurements are defined, now with the default width, the curve and position of the drivers in the lateral projection correspond to the model (I think so)
The box needs to be simulated for the port study but that will be at another time.
Un saludo!!!
Please note that this position has you listening to the bottom driver. XYZ is relevant to your listening axis. In most cases I put the tweeter center of a two-way at xyz=0. VCAD can be confusing. All of your drivers may have negative values, depending on the height.
Take a look at the "room" tab. It's extremely essential to helping you with this build in particular.
Take a look at the "room" tab. It's extremely essential to helping you with this build in particular.
Very interesting & clever design and build.
When you have multiple drivers in a series array, it is very important to have very high quality drivers, and never over-drive them.
If just one driver becomes faulty (not open circuit), like 'distorted voice coil', the problem affects the sound of the whole series array.
This can create quite a 'perplexing quagmire' in investigative fault finding. Basically a headache to repair.
Last edited:
Good, moved along quickly with the learning curve of simulation.
It is a difficult model to start with and should be great learning experience.
You have learned quickly, very good.
looking at the scribble mess of expected phase.
It appears to be more realistic, and we can see off axis info
In normal speakers, if center to center spacing is tight.
And arranged vertical. Horizontal is easy to make beautiful
and flat. That is the easy part.
Already we can see with many drivers and crossover components
Horizontal response is junk. Not your fault, the concept is nonsense.
For 95 dB of bass and 85 dB of combing high end we could use 2 drivers
and 2 crossover components.
For final deletion switch over to vertical directivity so we can view in sim.
1st orders cant fix the horizontal, and will never ever fix the vertical.
Essentially have added 3x cone area for 9 dB of boost in one area
And 9 dB of cancellation in other areas.
So it needs 9 dB of EQ to fix it.
and combs regardless.
Hence one fullrange + one sub
would do better.
Since the additional 9 dB gained from 12 drivers.
Needs 9 dB of EQ to fix it regardless.
You're a carpenter, right? You should be able to come up with methods to vary the shape of this box.
If you're up for an improvement make some adjustments to the box shape. It will work better that way.
Remove the diffraction signature of a sharp edged box shape.
See it in some ABEC simulation here: https://www.vandermill-audio.nl/smooth-curves-to-avoid-diffraction/
P.S. good to finally see a CBT with frequency dependent shading.
It does work wonders for straight arrays too.
On paper, that might be true. In real life, not the case. If you look very closely at the graph below, you see that in order to get very smooth in room response, I did not need a bunch of high-Q peaks or dips built into my DSP curve.
It has a lot of dB’s of EQ, but none of that EQ was needed for fixing comb filtering problems. The behavior overall is smooth and well behaved.
When you average out the response across multiple angles throughout the room, this is 1/12 octave resolution of actual real room response across 18 positions:
It’s analogous to open baffle designs: they are messier on paper, but they sound better than boxes in real life.
The other story that “on paper”, simulations and measurements don’t tell you is dynamic range. These twelve 3.5s have enough displacement to play full range. But many people will want to match them with a subwoofer and cross them over at 60-100Hz and one of the real advantages that CBT arrays have is a tremendous amount of dynamic range and slam. If you cross these over at 100 Hz to a subwoofer, you have more surface area in the midrange and treble than almost any speaker reviewed in Stereophile. The dynamic capabilities are closer to horns than direct radiators.
I mentioned the guy who brought a CBT array to the Parts Express 2017 competition and won first place in Open Unlimited. Those speakers had dynamic range to burn. I believe they had 8 peerless 6” woofers and maybe 16 or 20 tweeters.
He played “tricycle“ by Flim and the BB’s and I’ve never heard the drums on that song sound as realistic on any speaker.
There is no combination of a single full range driver and sub that could ever do what those CBT arrays did.
So for those who have never built one of these or heard a CBT array in person, I would be cautious about saying you could do a better job with a whole lot less effort making a simple traditional two-way.
Last edited:
That is a contentestable statement. Not something i found true.
Lived with a few of them, heard a bunch of others, all had a common flaw, went bck to boxes… but good ones, mot aren’t.
dave
The path to configuring a closed, open or multi-source speaker has different challenges. That may have a bearing on what we hear about them, but not on their worth as a speaker.
I've been wanting to build a line array for years now with some 3.5" driver like the Peerless TC9 or something similar, was thinking at first with maybe 16 drivers for a half way line to save some cash, or floor to celling if I managed to save a bit more of it. I've always know about the comb filtering issues with them but since some people say that it ain't a problem in practice I've been quite positive that those might be the "ultimate" speakers.
But seeing this with what you describe kind of the same sound as a line but with less comb filtering I can say that I'm really intrigued by this! I mean I've seen the original CBT but I remember them having other downsides so I kind of forgot about those, but these with that filter that makes all the drivers play all the bass looks really interesting!
Just a few questions, if you'd build it with 16 or 18 drivers would you use the same radius of the curve? And how would you do the rest of the filter? Would you tune the port lower with more drivers?
Is there reason why you built them with asymmetrical placement of the drivers btw?
Haven't heard any line or curve irl thought, would be interesting to hear the actual difference between them. I mean in theory the curve seems better, but it's harder to build and not as good looking as a line imo.
If I built it using 16 or 18 drivers, I would use a larger radius. Really, you get to decide for yourself, but I think if the top edge of the speaker leans back more than about 30° it looks a little funny.
Comb filtering is always an issue with line arrays and this is why I like the shaded array approach better.
There’s no law that says that you HAVE to use a curved array. You can use CBT shading in a straight line array, and it will work similarly.
I am guessing the compromise you may need to accept is that a curved array converges on a single virtual point source behind the speaker, and a straight array with shading will distort that to some degree.
Earlier in this thread, https://www.diyaudio.com/community/...th-35hz-25khz-high-output.412154/post-7672809 @Rxdroid was modeling the exact tilt of the array in Vituixcad and you could use his model and compare it to a straight array and see how much difference you think it makes.
If I were using more drivers, I might tune the port a little lower, but not a lot. Every octave of bass extension costs you 12 DB of dynamic range, so with small full range drivers I really wouldn’t try to push the low end of the system below 40 Hz.
Yes, there is a reason why I placed the drivers asymmetrically. It’s so the defraction patterns on one edge do not match the defraction patterns on the other edge. It spreads the peaks around a little bit. The front panel is 7 inches wide and the drivers are 3 inches from one side.
How would I do the rest of the filter with more drivers? For 16 I would add a fourth circuit with four series drivers in the circuit, and I would make sure it still follows the rule that every single driver’s response curve tilts just a little bit further downward than the driver below it.
For 18 drivers, you would need three branches of six drivers wired in series. I think wiring six drivers in series and getting all of the filters to cascade properly would take a LOT of trial and error.
(Or 6 branches of 3 in series, which I suspect might make it even harder to achieve the desired result. It seems like it would be hard to get the bottom six drivers to have the desired relationship with each other. But I haven’t tried to model it.)
I think 16 is easier to pull off.
Either way there’s a fair bit of modeling you need to do to get it the way you want it.
Comb filtering is always an issue with line arrays and this is why I like the shaded array approach better.
There’s no law that says that you HAVE to use a curved array. You can use CBT shading in a straight line array, and it will work similarly.
I am guessing the compromise you may need to accept is that a curved array converges on a single virtual point source behind the speaker, and a straight array with shading will distort that to some degree.
Earlier in this thread, https://www.diyaudio.com/community/...th-35hz-25khz-high-output.412154/post-7672809 @Rxdroid was modeling the exact tilt of the array in Vituixcad and you could use his model and compare it to a straight array and see how much difference you think it makes.
If I were using more drivers, I might tune the port a little lower, but not a lot. Every octave of bass extension costs you 12 DB of dynamic range, so with small full range drivers I really wouldn’t try to push the low end of the system below 40 Hz.
Yes, there is a reason why I placed the drivers asymmetrically. It’s so the defraction patterns on one edge do not match the defraction patterns on the other edge. It spreads the peaks around a little bit. The front panel is 7 inches wide and the drivers are 3 inches from one side.
How would I do the rest of the filter with more drivers? For 16 I would add a fourth circuit with four series drivers in the circuit, and I would make sure it still follows the rule that every single driver’s response curve tilts just a little bit further downward than the driver below it.
For 18 drivers, you would need three branches of six drivers wired in series. I think wiring six drivers in series and getting all of the filters to cascade properly would take a LOT of trial and error.
(Or 6 branches of 3 in series, which I suspect might make it even harder to achieve the desired result. It seems like it would be hard to get the bottom six drivers to have the desired relationship with each other. But I haven’t tried to model it.)
I think 16 is easier to pull off.
Either way there’s a fair bit of modeling you need to do to get it the way you want it.
Last edited:
A couple more things: when I built these in 2017, I modeled them on a spreadsheet and a calculator. I didn’t use Vituixcad and the model was not very precise.
If I were to redo them, I would change the Zobel. The old design has 13 µF and 6 ohms.
The new Zobel for the FE85 would be 4.7 µF and 16 ohms. That would result in a higher impedance to the amplifier and otherwise similar filter behavior.
Your mileage will vary depending on the drivers you are using, and it is a really good idea to use the actual impedance and response curves of any given driver to dial this in. I think with some care it should be possible to get a nearly resistive load above 100 Hz and still hit all of the other design objectives.
Finally, it’s absolutely necessary to use at minimum a 3 inch diameter port. Most people make their ports too small. A large port makes a very big difference at high SPL‘s.
If I were to redo them, I would change the Zobel. The old design has 13 µF and 6 ohms.
The new Zobel for the FE85 would be 4.7 µF and 16 ohms. That would result in a higher impedance to the amplifier and otherwise similar filter behavior.
Your mileage will vary depending on the drivers you are using, and it is a really good idea to use the actual impedance and response curves of any given driver to dial this in. I think with some care it should be possible to get a nearly resistive load above 100 Hz and still hit all of the other design objectives.
Finally, it’s absolutely necessary to use at minimum a 3 inch diameter port. Most people make their ports too small. A large port makes a very big difference at high SPL‘s.
Last edited:
Doesn't this physical resolution/spacing also limit the ability for shading to work (at these higher frequencies in question)?
Perhaps you have a reference to some documentation which shows this?
Shading works across the line, with the driver spacing setting the upper limit since the crossover can't split a response gradient across half a driver, and can't work in the gaps.
I'm not questioning your favourable results, just the role of shading in this case.
Shading works across the line, with the driver spacing setting the upper limit since the crossover can't split a response gradient across half a driver, and can't work in the gaps.
I'm not questioning your favourable results, just the role of shading in this case.
Shading always gives you less comb filtering than equal signal to each driver. Keele's work clearly shows this.
https://www.xlrtechs.com/dbkeele.com/CBT.php
See this graphic in particular
https://www.xlrtechs.com/dbkeele.com/images/Card Back Large.png
4" spacing between drivers is obviously less than optimum, but it's the best one can do with a full range driver. I got very little comb filtering effects in my own measurements.
https://www.xlrtechs.com/dbkeele.com/CBT.php
See this graphic in particular
https://www.xlrtechs.com/dbkeele.com/images/Card Back Large.png
4" spacing between drivers is obviously less than optimum, but it's the best one can do with a full range driver. I got very little comb filtering effects in my own measurements.
Last edited: