Complementary means summing to flat frequency magnitude response, and both sides having the same phase trace.
Like any order LR .
Important thing, as I figure you know, is that acoustic complementary is the xover goal....not electric complementary.
The nice thing about steep linear-phase xovers, is that the realized acoustic xover becomes damn close to the electric xover without much effort.
The steepness of the electric xover dominates the driver's natural response rolloffs (and out of band peaks etc). All it takes is keeping the passband usage confined to the region of naturally flattish mag and phase. Back to making a case for 4-ways !
Complementary acoustic is very important to using the linear-phase xovers though, without it the odds of pre-echo rise substantially.
Also, keeping drivers c2c close (as with any speaker) is probably more important with lin-phase xovers than IIR, because potential for off-axis pre-echo is added to the normal lobing issues.
That's the theory anyway...someday I'm gonna take a big set of off-axis impulse/step responses and see how it plays out.
I use Q-sys design software and various hardware Cores for processing. The software is open architecture and allows any kind of design. Designs are made in Windows. The hardware is basically a Q-sys dedicated PC running Linux. Lowest model, a Core110f is good for 8ch of FIR, 8k taps per channel. A Core 510i is good for at least 15 ch of 16k taps per ch. (what i'm running now). 48kHz is the only sample rate supported.
Some/many folks seem to care about higher sample rates. Ironically, for FIR usage, I find lower sample rates are more valuable.
YMMV
Once you add a dedicated sub to a well designed 3 way I don't really see the gain except for added complexity and cost. Unless the drivers chosen have very narrow bandwidths. I have a 14" sub used with Revel F-206's so a pair of 6" and 4" and 1" dome. Don't need another driver in the mix.
Also have a bi-amp version of L250 Jubilees 4 ways where you have 14, 8, 5 and 1. I don't see how you could squeeze another driver in there
If you worried about sd just go big 15" sub 15" mid-bass 10" mid and large format CD. There you get higher sensitivity as a bonus only issue is size.
Rob
Hi macka,
Thank you for your comments and advices. I see many of your advices in line with my plan. I happened to use B2500.1 in my current system, and I prefer pulp cones for bass/midbass. For a higher WAF, I'd like to use 2 8" instead of single bigger one for the woofer section. If counting the SVS in, I am more like planning to end up with 5-way, but I am talking about only the part to plan and build in the future, so calling it 4-way.
There seems to be no Bliesma 2" dome, and you didn't mention Satori, so I take it that you don't recommend it.
. I am also looking at Scanspeak 21mm textile dome, so let me know what you think.There are other textile dome midhigh options you didn't mention, so I wonder what you think about HiVi and Tang Band 3" ones.
BTW, Is it JBL 4344? I heard it a few times in 2 places, and the last one sounded pretty charming and pleasing, though it wasn't exactly to my taste. Surely a masterpiece.
I am assuming the listening room is already stuffed with chairs, sofas, desks, cabinets, shelves, lamps, and other furnishings, and the speakers need to be squeezed in somewhere. Of course in a "real listening room", the speakers and listener's chair are positioned first, and everything else is positioned around these so as to not interfere. In that second kind of room, you can have anything you want.
JH> Thank you for bringing up good examples which are more relevant to my project due to the WAF requirement. It looks quite difficult to even just copycat the shapes of baffles/waveguides. Revel's and Vivid's crossover information is very useful. I would say Vivid's 880Hz crossover point is quite daring, and also maybe both encouraging and misleading. I am tempted.
Instead of these fancy, difficult-to-make shapes, I am dreaming of something like below. But this must be still very difficult for me to make. I just want smaller ones in those tubular shape. Due to the size, woofers in that form factor is practically impossible(to even place somewhere), and I'd just go with a mildly fancy-shaped box. Obviously not the frame which must be irrelevant to acoustics.
JH> This makes very good sense, though I am not inclined toward hard diaphragm materials at this time. Maybe in somewhat further future, I'd make modules that I can easily swap with currently planned for different sound. It would like several modules mixed/matched into dozens of different speakers(combinations). I am actually interested in trying out ceramic, DMD and ribbons for tweeter/super tweeter.
JH> Yes. I want to place drivers as close as possible. Woofers will be placed on the top part of the woofer module baffle. Your process requires some measurements I am not used to, but hopefully I can find a way for it. I don't know if I will/can resort that far to "3d simulation".
The first article is very encouraging as I want to do "very small panel" approach to everything except woofers. Maybe it didn't mean to recommend such for anything other than tweeters, but I tend to think, why not?
Attachments
I was thinking of lower ranges/crossover points for midbass and midhigh. Isn't 1600Hz a bit risky? 300~400Hz woofer/midbass could be working well.
I am sorry that 10" is not an option for me, but glad that I used Volt woofer in mine(B2500.1).
You mentioned many interesting things, and Manger's BWT is very very interesting. Though I have no idea what material property the BWT diaphragm takes to emulate a single point sound source, but kept wondering if DSP could emulate BWTs with a very flexible diaphragm attached with multiple bobbins/voice coils and DSP-processed signals driving them.
I am not a big fan of NOS DAC, and actually skeptical for technical reasons. Just saying...
I don't know what I can do for measurement at this time, but will let you and others will to help know.
I am taking a bit of different approach from using class-D amps with built-in crossover, and I think I like the way audio signal processing is configured in CamillaDSP S/W, and the huge processing power built in the Macs/PCs. Then I can keep the downstream simple, clean and proven to work - pro digital gears and high-quality DACs.
I will keep this in mind, though my plan of stacking separate enclosures seems to make it difficult to practice this idea...
I see your point. To be honest, I don't have a technical argument against it. It just sounds fancy and fun to do a 4-way, since I have a successful(IMHO) 3-way one. Maybe, there could be a slight improvement which hobbists can appreciate. Maybe I can use some crazy 8" drivers to get crazy deep bass for the size and compensate upper band loss with a midbass. I just don't know yet.
Yes. I asked one of my professional friend about concerns about 800Hz crossover, and he told me to worry more about 2~4kHz.
Got a question for you. How applicable is the idea of 'smallest possible flat baffle area around the tweeter' to the midrange/midbass?
Sharp edges near the drivers will cause significant diffraction leading to bumps and dips in the response. It is unlikely to be a good idea if you are interested in high fidelity which requires the radiation pattern of adjacent drivers to blend smoothly. On the other hand perhaps it will sound good and make a good audiophile speaker (the production chaps in the videos earlier had 15 dB bumps and dips in the upper midrange of their speaker and it sounded extraordinarily good apparently) or perhaps it will sound poor. It is hard to say.
Large radii will reduce the strength of the diffraction at the higher frequencies and is likely worth considering if you really want to stack cylinders. Unlikely to look as good though. (Back in the 70s I built a small cylindrical speaker with a KEF T27 and B110 with a large radius on the thick end caps. A year later Rogers released their JR 149 which looked much neater with the thin sharp edged end caps.)
3d simulation is likely only needed for curved shapes that are intended to accurately guide the sound waves. If you want to lean toward high fidelity then chamfers are likely the most straightforward approach as illustrated in the last 2 links. Perhaps a stack of boxes with chamfers, wider bottoms than tops and dimensions that match the neighbouring boxes in the stack. This could keep the "good" edge and avoid "bad" edges although the joins might need some thought/attention.
Because of the wavelengths involved it is likely to be important for tweeter and upper mid but less so for lower mid and woofer.
As applicable as with any size transducer. Edge diffraction is about the same no matter what size transducer, except everything is lower in frequency for bigger sized stuff and higher in frequency for smaller transducers, and objects. Hense the minimal baffle means no flat around drivers, and is as effective on all sized transducers and systems. Minimal baffle in absolute dimensions is not very meaningful, only in relation to wavelength.
It relates to sound wavelength, which is basically physical size of sound and, sound at any frequency, any wavelength isbjust sound and reacts the same. This means any frewuency (wavelength) reacts similarly with same sized physical objects. If ideal 1" transducer in 4" baffle shows problems from 4" wavelength up to 1" wavelength, a 10" transducer in 40" baffle would show exactly same problems/behavior, only everything is now on 10x longer wavelength, 10x lower in frequency.
But, which one is more audible? hearing system at 4" wavelength is less sensitive to phase differences between ears, while at 40" its less sensitive to amplitude difference between ears. This is due to head being about 8" in size.
If you want you can think sizes of objects in frequency, or sound frequency as physical size, and this stuff is easy to imagine after a while. My head is roughly 1700Hz in size, I'm about 180Hz tall. 4000Hz fits in palm of my hand, fingernail is about 20kHz. My room dinensions are 55Hz, 42Hz and 140Hz.
If you check your tweeter diffraction ripple on measurements its roughly from baffle width to tweeter diameter. If it was 20cm baffle and 2.5cm tweeter it would be roughly from 1700Hz to 13kHz, the whole bandwidth of the tweeter. Whats the diffraction ripple with 20cm cone on the same 20cm baffle? from 1700Hz to 1700Hz, so about none. Diffraction still happens, but not the problematic variety which wpuld be higher in frequency, but because the 20cm transducer beams, less and less sound gets toward edge and there isn't much.
Only if distance from transducer edge to baffle edge is relatively long (compared to driver size). If the edge is very close to transducer there is no wiggle as explained above, only main diffraction hump would be there which has relatively uniform behavior to all directions and can be EQ:d on axis, but it makes response narrow a bit. Any roundover would of course be even better, would smoothen even the main hump, lower DI. Some times one wants to utilize higher DI here, sometimes lower, to compensate total DI where there is vertical lobing for example.
Frequency response wiggle measured in listening window illustrates there is secondary sound, the edge, wiggle is interference of direct sound and the edge. Audibility of it isn't that obvious though, very hard to compare low diffraction and high diffraction systems, as they would differ in many ways.
The ubiguous 1" dome tweeter on flat 4" flange is just bad in my opinion, legacy tech, as it basically makes the whole tweeter bandwidth have diffraction interference, unless generous roundovers beyond the flange. If it was 1" dome on no flange, one could use very small roundovers, even none, just put it on a stick top of your woofer box. Of course tweeter would have "the bafflestep", which could be a problem with passive xo, perhaps no problem with DSP, as long as system SPL requirement isn't too high and low DI system is on the table.
If higher DI system is desired, there must be bulk around the transducer, so better make it such it doesn't make edge diffraction backwave happen. Properly designed waveguides and roundovers, even grude slants to approximate round shape, whatever, to keep the flat baffle minimal = edges nonexistent or very close to transducer. Another option is to put edges very far, so that there is no difference in early room reflections and edge diffraction. Like soffit mounting.
Last edited:
Yes, you and I have been tossing this concept about for several years now... As you pointed out back then, the best tweeter baffle is a radius that begins at the edge of the diaphragm. I am encouraged by the trend of newer tweeters having face plates in the 60-75mm range... Bliesma T25, Wavecor TW030WA23/24, and ScanSpeak D3004/604000 for instance.
The diffraction of a DXT-type waveguide can be put to constructive use but the proposal is to closely stack cylinders. This introduces edges of the neighbouring cylinders which are not axially symmetric. It introduces bumps and dips that are largely avoidable with better design (e.g. baffles like Revel, Vivid, Heissmann, etc...). Better design that is w.r.t. technical performance but other factors are of course of importance. If the op is still keen and has an interest in technical performance then the question becomes how large are the bumps and dips? Shouldn't be too difficult to quantity but would require some effort.
yes, and looks and ability to manufacture and all kinds of things, cost, so perhaps edge diffraction related ripple isn't most important thing on many projects, but it's still one thing among others which one might consider to address somehow. After all it's easy to notice on measurement graphs and even by looking at the object, and means to affect it are surprisingly many. For some reason eyes have become quite sensitive to edge diffraction, although perhsps it's not so obvious to ear. But same phenomena, acoustics, is related to perceived system performance in many other ways as well. Overall directivity, what the response is to various directions, and what that might be perceotually considering the room and placement. I think it is enough is to be aware of the stuff so there is a chance to address some of it. And the more, the better.
Thinking about four way system from directivity perspective, or is it? not much different than what 2 or 3 way can do. Perhaps harder to do high DI system unless OB. I'm thinking waveguides are harder to fot, ubless MEH. Perhaps bit easier do cleaner horizontal coverage if lower DI system target. Wider baffle could be used to raise DI of course. Just use some tactic with the edge
Thinking about four way system from directivity perspective, or is it? not much different than what 2 or 3 way can do. Perhaps harder to do high DI system unless OB. I'm thinking waveguides are harder to fot, ubless MEH. Perhaps bit easier do cleaner horizontal coverage if lower DI system target. Wider baffle could be used to raise DI of course. Just use some tactic with the edge
I ran a WAF test by showing her both JBL 4344 and Sonus Faber Stradivari. "Too big" for JBL, and "I'd rather take the other" for Sonus Faber.... lol
Good point. Maybe some things can be done on the surface of or around the cylinders to reduce the adverse effect?
Complementary means summing to flat frequency magnitude response, and both sides having the same phase trace.
Like any order LR .
Important thing, as I figure you know, is that acoustic complementary is the xover goal....not electric complementary.
The nice thing about steep linear-phase xovers, is that the realized acoustic xover becomes damn close to the electric xover without much effort.
The steepness of the electric xover dominates the driver's natural response rolloffs (and out of band peaks etc). All it takes is keeping the passband usage confined to the region of naturally flattish mag and phase. Back to making a case for 4-ways !
Thanks for the explanation.
These Q-sys stuffs are a totally different world from usual pro DSP stuffs. Which industry are these in? It got a lot of performance, and looks very promising. I think some latest server ARM cores can be better, or maybe even some RISC-V based stuffs, mainly for heat/power/fan noises.... So you use Intel with Linux, and I want to use Apple ARM cores with macOS, which are strong rivals in the computing world.
Hopefully the Apple M1 chip offers enough performance for crossover processing of my 5.1 systems. Current system has 3x 3-way crossover, and each channel consumes about 1200 MACs/cycle @48kHz Fs. FFT-based filtering seems to keep the processor cycle consumption low enough. Going 4-way, it will be increased significantly if I do the same between 2x 8" woofer and the midbass. Crossing over at as low as 150Hz seems to be a steep increase of filter length, which may still be okay in terms of cycle consumption. However, I use my systems mostly to watch BDs, so further increased latency can possibly break the A/V sync completely. I guess using LR4 for woofer/midbass is a decent compromise.
Since you corrected me on LRx crossover, maybe LR8 or LR12 to get similar to FIR filtering in the aspect of filter slope.
48kHz is the only sample rate supported.
Some/many folks seem to care about higher sample rates. Ironically, for FIR usage, I find lower sample rates are more valuable.
I think I have the option of 96kHz. The professional gear I plan to use support up to 96kHz, and Apple M1 seems to sustain the performance well for the S/W BD player processing and the similar crossover to what I use in the surround processor at 96kHz.