My first project is a 4 way speaker...

[Zvu]

I'd find the tweeter that works ok to me and then i'd have somebody make me new aluminium front plates that mach the opening on the cabinet. Factory tweeter plate would serve as template.

It's 3.25mm aluminium circle plate with couple of holes drilled. Nothing complicated really.

[ergo]

Joaquim,
I'm not super close to you, but still in Europe (Estonia). If interested I can help out with the 3D printing. I have a Prusa i3 mark 2.5. In last few days I've been doing a very similar project for my friend....
You seem upto speed with 3D modelling. If you can come up with a model for pieces you need to couple a new tweeter to the existing hole then I could print these out. 3D printer needs STL files in end but if you'd have a model in Sketchup file or Fusion 360 file then I could export from those as well.
Hope the below pics give some idea what to keep in mind. The 3D printing can not print mid air - so if there are overhangs than the print would need to use a temporary "support structure" that can be broken off later with relative ease. Material I've used for these samples is PETG plastic that is temperature stable upto 70..80deg celsius or so. PLA can have problems with >50 - experienced that with a mobile holder I printed to my car dashboard.

[cochinada]

Wow! That's awesome Ergo! Appreciate your kind offer and if I decide to go with plastic I will certainly ask your help.
Thanks again!

[cochinada]

I'm considering changing the philosophy of my crossover, so even before changing the tweeter I was doing some simulations to see if it was possible to improve what I have right now.

Basically, as we all know MTM has the unavoidable problem of lobbing and my speaker is no exception to the rule of course. So, I decided that it would be a good idea to do as many do, which is to destroy the symmetry of the MTM by simply forcing one of the mediums to roll off before the other. It is what B&W does on the 800 Matrix and more recently Rockport with the Lyra for instance, but I'm sure there are plenty of other examples. What do you think of the idea?

This is what I have achieved so far. First picture shows my current crossover and the second one is this 4.5 way or 3.5 if we ignore my active subwoofers.



Some points to notice:
1. On axis response is flatter.
2. Vertical directivity is a lot better (blue spots are almost completely gone)!
3. The woofers have a much less active role (don't know how good this is, considering I have 4).
4. The mediums start at lower frequency (I think I'm very close to the limit or perhaps below it already).
5. Passive all pass circuit is gone on the tweeter.
6. Zobel has gone on the medium crossover
7. New kid on the block -> L1 does the trick!

I've done only a few iterations so far and I'm sure there is many room to improvement. Perhaps I should try to move the crossover point between the woofers and the mediums a bit higher?
Any comments/suggestions will be highly appreciated!

[ergo]

New design looks good in many ways indeed. But the other variable from current version is the tweeter response. In new design it needs to play lower than before (evident from tweeter xover gain plot also), but if you choose a capable one with enough linear excursion that could be well doable.

For woofer / mid it's harder to say. You'd need to analyze their cone excursion and see from that if this gets them too close to their limits or not.

[cochinada]

Thanks ergo for calling my attention upon that. I was looking elsewhere and totally missed it. Indeed you are right. Perhaps I can tweak the xo and increase that point or perhaps the SEAS T35C002 suggested by roadrune can cope with that. This looks a very interesting tweeter! Troels has nothing but good things to say about it.

How about T35C002? You can change the faceplate on it, so you can get one made that fit...

http://www.seas.no/index.php?option=...=50&Itemid=360

[cochinada]

Does anyone knows how to calculate Qms or Rms that are missing from manufacturer's datasheet for Seas Excel T35C002 E0055?

This is all the data published:

Nominal Impedance - 6Ω
Recommended frequency range - 1500 – 25000Hz
Short term power handling - 250w
Long term power handling - 100w
Sensitivity - 95db
Voice Coil diameter- 35mm
Voice Coil Height - 2.0mm
Air gap height - 3.0mm
Linear Travel (p-p) - 1.0mm
Voice Coils Resistance - 4.6Ω
VCA Inductance - 0.06mH
Force Factor - 4.3 N/A
Free Air Resonance - 620Hz
Moving Mass - 0.47g
Effective piston area - 11.9cm2
Magnetic Gap Flux Density - 1.6T
Magnet Weight - 35g
Total Weight - 0.37kg

[sdl2112]

Sorry to see you have so many issues with your speakers, they are very impressive. Another alternative would be to use the Scan-Speak D2908/714000.

https://www.scan-speak.dk/datasheet/pdf/d2908-714000.pdf

It is 120mm in diameter, 6mm thickness and 90mm cutout. Sensitivity is 92db.

I wonder what your actual hole size is...if there were a little margin you could tape off the hole diameter for protection and deepen the cutout carefully using a router with a template/pattern adapter or how Troels has done. Probably a precision template and test on scrap piece first.

http://www.troelsgravesen.dk/tips.htm#MAKING_A_DRIVER_HOLE_BIGGER

[cochinada]

I wonder what your actual hole size is...

...the hole for the Transducer Lab is very peculiar as it has an outer diameter of 120mm, an inner of 92mm and only 3,25mm deep, so finding a straight fit is mission impossible I dare to say. I could build an adapter provided that the new tweeter is smaller than 92mm plus the thickness of the adapter wall itself.

Thanks but it doesn't fit as it is too thick.

EDIT: I would not like to ask my carpenter to modify the cabinets anymore. :roll:

[Steve Manning]

Sorry to see you have so many issues with your speakers, they are very impressive. Another alternative would be to use the Scan-Speak D2908/714000.

https://www.scan-speak.dk/datasheet/pdf/d2908-714000.pdf

It is 120mm in diameter, 6mm thickness and 90mm cutout. Sensitivity is 92db.

I wonder what your actual hole size is...if there were a little margin you could tape off the hole diameter for protection and deepen the cutout carefully using a router with a template/pattern adapter or how Troels has done. Probably a precision template and test on scrap piece first.

http://www.troelsgravesen.dk/tips.ht...ER_HOLE_BIGGER

Good idea Scott ..... it would be an easy fix to make the original hole deeper and since the drivers are the same diameter you don't really have to worry about messing up the paint. Make a template to match your existing hole attach with double stick tape and use a flush trim bit to deepen the hole.

[cochinada]

No matter how I try, I simply can't get this D2908/714000 to integrate well enough with the other drivers. The pronounced curve doesn't help at all. Moreover I will probably need it to go low as in my previous simulation; at least 2KHz or so and I'm not sure it can handle it.

[ergo]

Does anyone knows how to calculate Qms or Rms that are missing from manufacturer's datasheet for Seas Excel T35C002 E0055?

As a bit of an hassle option I could trace back the impedance, then import it to LspCAD and then allow it to "optimize" the T/S of the respective driver parameters until it gets a fit with the imported curve. I have used that method for finding parameters for a passive radiator before, but it should work for tweeter as well.

[cochinada]

As a bit of an hassle option I could trace back the impedance, then import it to LspCAD and then allow it to "optimize" the T/S of the respective driver parameters until it gets a fit with the imported curve. I have used that method for finding parameters for a passive radiator before, but it should work for tweeter as well.

I never used LspCad but I do have the impedance already traced. Is it possible to download a working version or is it just demo?

[ergo]

http://www.ijdata.com/LspCAD_demo.html

Seems it should do the trick in demo mode... does not allow save or export but you would not need that. But it is sorta tricky process and I had LspCAD open for lookin at TEK-s project so you'r in luck - try these values

[cochinada]

Thanks a lot ergo! :T
The values are bit different from the ones the manufacturer tells but it suits my purpose which was just to evaluate the cone excursion.

[kimmosto]

I just tweaked VCAD's Calculate T/S parameters for tweeters, though I think it's not so valuable... Mms or BL should be entered if secondary ZR cannot be measured. This might be simpler procedure than previous.

[ergo]

Great work Kimmo! You are getting closer and closer to making LspCAD obsolete

[kimmosto]

^Hopefully not, though I believe that also Ingemar has some decent daily work

[cochinada]

Thanks Kimmo! How do you calculate the 'Error' on the Extended Z model? Is it the sum of the square deviations between the simulated and measured curves?

[kimmosto]

^Sum of squared deviations divided by number of frequency points to normalize error value regardless of bandwidth and density of frequency scale.
Frequency high limit is 20 kHz and low limit is smaller of fs*10 and 2 kHz. Density of frequency scale is 1/6 oct.

Basic impedance model is solved first to eliminate lossy voice coil inductance before fs and Q detection from measurement(s). Frequency low limit is smaller of fs*10 and 3 kHz. 3k limit improves accuracy with tweeters, though it's not perfect & universal workaround.

[cochinada]

...Frequency high limit is 20 kHz and low limit is smaller of fs*10 and 2 kHz. Density of frequency scale is 1/6 oct..

Makes sense, and it explains why a better fitting curve in lower frequencies does not necessarily give lower Error but If I'll use a waveguide and extend the response of the tweeter lower than 2KHz what happens then?

Do you have defined limits also for W/M?

[kimmosto]

If I'll use a waveguide and extend the response of the tweeter lower than 2KHz what happens then?

Not much in impedance curve.

Do you have defined limits also for W/M?

Previous low limits are in use for all driver types. For example if driver's fs=40 Hz, range of extended Z model solver is 400 Hz...20 kHz.

[cochinada]

On the picture below, on top we have SEAS-T35C002 with a possible adapter/waveguide with the same acoustic center of the medium. On the middle, a intermediate version and below the standard manufacturer face plate. It can't be a big waveguide. After all the customized diameter is 120mm vs. the original 110m.

What do you think? I have zero experience with waveguides, although I've followed the one in this forum with great interest. There is also this detail of the screws that can't be countersunk as the thickness is not enough.

Image not available

[TEK]

There is also this detail of the screws that can't be countersunk as the thickness is not enough

Using a countersunk screw head instead of the flat one in the figure that should not be a problem as far as I can understand.

[ergo]

The problem with the waveguides is that by just looking at profile it is pretty impossible to guess how it will measure. The two ways to find out are basically to create all 3 and measure (even if on a test baffle) or find someone with access and know how to simulation tool like Comsol. Latter is much harder.
There are some guys over at DIYaudio who are trying to do this with a lower cost simulation software ABEC but by the looks of it it is such a challenge to learn that I've not even contemplated trying.

Have you considered some small flange tweeters like the SB small beryllium.

http://www.sbacoustics.com/index.php/products/tweeters/dome/sb29bnc-c000-4/

3D printed or metal milled adapter for that might fit your existing space too ?

[Juhazi]

I have been struggling with similar adapter plate challenge. I bought very rare active speakers that have box made of stone plates, WWTMM configuration. I replaced it's analog xo with minidsp 2x4HD and changed configuration to WWMT with better MT drivers. Also two of woofers were broken.

The SS D2608 Tweeter fits in the opening of mid, but it is difficult to fit because it's front plate is actually a very shallow waveguide. I have modified the shape of adapter, but measured response is ugly. So far I used plywood for adapters and the project is on hold now, I will continue it later on with my son who is now abroad as exchange student.

Some pictures

[cochinada]

Using a countersunk screw head instead of the flat one in the figure that should not be a problem as far as I can understand.

Now, why didn't I think of that? Maybe because I have the other type of screws and also round head but don't remember seeing nice black wood screws with countersunk head. :roll:

EDIT: they are rare and I just found this M2x4...16 kit on eBay. Very thin...
EDIT #2: found imperial #8 (~M4) on Amazon here.
EDIT #3: Bingo! However, it looks like the head might have more than 3,5mm of height which in my case won't do. Perhaps the M3 have a shorter head...

The problem with the waveguides is that by just looking at profile it is pretty impossible to guess how it will measure. The two ways to find out are basically to create all 3 and measure (even if on a test baffle) or find someone with access and know how to simulation tool like Comsol. Latter is much harder.
There are some guys over at DIYaudio who are trying to do this with a lower cost simulation software ABEC but by the looks of it it is such a challenge to learn that I've not even contemplated trying.

Now, that seems like the kind of challenge Kimmo likes to tackle! If someone can do it, he's the man! :B

Have you considered some small flange tweeters like the SB small beryllium.

http://www.sbacoustics.com/index.php/products/tweeters/dome/sb29bnc-c000-4/

3D printed or metal milled adapter for that might fit your existing space too ?

Yes, indeed I have. For that one in particular I would need an adapter but it would not be a waveguide.
As a matter of fact I have complied a list of the possible contenders so far, also with the suggestions I have received. Here they are:

SB Acoustics
SATORI TW29D-B
SATORI TW29DN-B
SATORI TW29DN-B-8
SATORI TW29B-B
SATORI TW29BN-B
SATORI TW29BN-B-8
SB26ADC-C000-4 -> not sensible enough
SB29BAC-C000-4
SB29BNC-C000-4 -> adaptor needed (not waveguide)
SATORI TW29R-B
SATORI TW29RN-B
SATORI TW29RN-B-8
SB29RDNC-C000-4 -> adaptor needed (not waveguide)

SEAS
T35C002 E0055

Scan Speak
D2908/714000 -> in all my simulations it doesn't integrate well enough with the other drivers... :roll:

Like I said I do fancy the SATORY line...

I have been struggling with similar adater plate challenge. I bought very rare active speakers that have box made of stone plates, WWTMM configuration. I replaced it's analog xo with minidsp 2x4HD and used configuration to WWMT with better MT drivers. Also two of woofers were broken.

The SS D2608 Tweeter fits the opening of mid, but it is difficult to fit because it's front plate is actually a very shallow waveguide. I have modified the shape of adapter, but measured response is ugly. So far I used plywood for adapters and the project is on hold now, I will continue it later on with my son who is now abroad as exchange student.

Some pictures

That's very disappointed but it's good to know I'm not alone in this kind of challenge, although you seem to have more problems than I do. :roll:
This D2608 has 72mm of Magnet diameter vs. 74.5mm of the SEAS T35C002. You have one thing in favor though which is that the SS has a 26mm voice coil whereas the SEAS has 35mm and a more prominent dome which I guess intuitively would make the response with such a waveguide even more ugly! But if the hole you have is similar to mine with 120mm you have a bit more room for a waveguide than me.

[cochinada]

Like I wrote previously some tweeters are virtually impossible to integrate and this has mostly to do with phase.

Whenever the tweeter has this stupid curve below like a saddle, it doesn't integrate with the mediums properly...


But, whenever the phase is a nice almost straight curve/line that follows the ones of the mediums all is fine...


How do you people correct this issue? I've used all possible circuits I could imagine with no success, like Passive All Pass, even of second order! The only way I can correct this is by "cheating" and entering a different delay for the tweeter on Vituix Cad.
What delay should I use in the simulation of new tweeters? The same as in the original crossover or something else? In the original all delays are equal.

[kimmosto]

What delay should I use in the simulation of new tweeters? The same as in the original crossover or something else? In the original all delays are equal.

Start point of time window i.e. reference time in your original measurements is about -138 us. You should keep that value for woofers and mids, because -138 us shifts timing very close to minimum phase without XO components.

BUT (at least I suppose) you are using traced data for tweeter. Traced responses are already minimum phase so you need to enter 0 us to Delay parameter for tweeter in Drivers tab to get close to equal timing with the others. In addition, traced responses do not include effect of driver mechanics to acoustical center, possible non-minimum phase features and baffle effects. Therefore this is not adequate method for evaluating compatibility of responses. Especially designing of XO with mix of simulated and measured data might be just waste of time.

[cochinada]

...Especially designing of XO with mix of simulated and measured data might be just waste of time.

Much better, thank you very much.
Your assumption is correct but then how should I simulate another tweeter? Should I discard the real measured values of my mediums and woofers and also use the respective data from the manufacturer? I won't surely buy all tweeters on my list just to be able to measure them and use the real values on the simulation. Heck! I won't even buy a single one as money doesn't grow on trees on these parts.

[kimmosto]

I guess you did not simulate original set of drivers, but selection was okay because responses were flat enough and USPL of mids and tweeter was adequate for 4x woofers.
Now you have close to same situation. Problems are not expected if USPL of new tweeter is equal or more, response is fairly flat to half space, dome is as deep or deeper compared to baffle surface and driver is mechanically compatible or otherwise installable. But there's no 100% guarantee - with or without pre-simulations.

[cochinada]

But I did, I did! :B
Of course that previously i selected the drivers based on those criteria and some.

... dome is as deep or deeper compared to baffle surface ...

Didn't quite follow this. Taking as example post #581, do you mean only 1st picture is ok? Original tweeter is the last picture and that would pretty much excludes most tweeters I know

[kimmosto]

Taking as example post #581, do you mean only 1st picture is ok? Original tweeter is the last picture and that would pretty much excludes most tweeters I know

It looks that N26MGR-G is convex dome where surround is close to same level with face plate. Or is it concave? Possible difference in acoustic centers will not be a problem because most of modern domes are deeper with some miniature wave guide. Except the Seas you've been playing with. In addition, you already have Lattice all-pass network for tweeter, and also 3rd order HP with the same polarity with mids is possible. So phase match will be guaranteed, and main problems will be USPL, mechanical compatibility and especially sound quality compared personal preference. Just leave initial XO simulation to that Finnish guy who made the original design.

[cochinada]

It looks that N26MGR-G is convex dome where surround is close to same level with face plate. Or is it concave? Possible difference in acoustic centers will not be a problem because most of modern domes are deeper with some miniature wave guide. Except the Seas you've been playing with. In addition, you already have Lattice all-pass network for tweeter, and also 3rd order HP with the same polarity with mids is possible. So phase match will be guaranteed, and main problems will be USPL, mechanical compatibility and especially sound quality compared personal preference. Just leave initial XO simulation to that Finnish guy who made the original design.

N26MGR-G is convex and dome surpasses the plan of the face plate:


There you go again talking about this mysterious "Lattice all-pass network" and UPSL... :roll:
Isn't the sole purpose of the 1st one to add or subtract 90º of phase at the frequency X? How important is that on the initial XO and how will it be with another tweeter, e.g. this SEAS in particular? Can I live without it or something terrible will happen otherwise?
Pardon my ignorance but what is USPL after all?

P.S. I'll certainly leave the initial XO simulation to that "Finnish guy who made the original design" which we all know who he is, since he kindly offers to help. :rofl:
In fact I'll most surely need some guidance whenever I'm ready to measure the new tweeter again as i completely forgot how to use ARTA and how I did at the time but this might take some time.

[kimmosto]

Isn't the sole purpose of the 1st one to add or subtract 90º of phase at the frequency X? How important is that on the initial XO and how will it be with another tweeter, e.g. this SEAS in particular?

It makes phase matching easier and enables decent time alignment between M and T without moving tweeter mechanically for example with wave guide or stepped baffle. Acoustically 3rd-4th order slopes might be needed to get phase match without all-pass. But that would cause extra high jump to GD because mids are also mechanically further than tweeter in your speakers.

Pardon my ignorance but what is USPL after all?

USPL=dB/2.83V/1m.
92.5 dB with N26MGR-G, and for example with Seas T29CF002 Crescendo.

[cochinada]

Does anyone have pictures or even better, technical drawings of Satori TW29BN disassembled? SB acoustics advertises it has "Two part aluminium faceplate" and I would like to understand how to make an adapter for replacing the outer ring.

[ergo]

My guess would be that the two part means the inner circle versus the outer circle where the screw holdes reside. The two color model would have the outer circle in bare aluminum color....

Not aware of a cut drawing but here is a nice high res unboxing video that shows it from various angles at least

https://www.hifisound.de/Do-it-yourself-Products/Speaker-Drivers/Tweeter/SB-Acoustics-Satori-TW29BN.html

[cochinada]

I've seen the same video already.
Yes, my understanding is the same. The outer ring should be possible to replace i guess but i would need to know the exact profile and distance of the 3 screws in order to design my own.
I just sent an email to SB Acoustics. I don't expect they send me a technical drawing but let's see...

[cochinada]

New exercise. From the most extravagant version on top to the original face-plate on the bottom.



NOTE: 1st (topmost) should have the same acoustic center of the mediums. Responses expected should also be more different from the manufacturer's datasheet as we move towards the top.


...and the 2nd option from the bottom on the baffle. Here it is clearly visible the vertical offset between the acoustic centers.

[cochinada]

This is still with my current tweeter.

original


new version


One thing that bothers me is the modest role of my woofers under this simulation...

[cochinada]

And this is a pre-simulation as Kimmo calls it with my best candidate tweeter and traced responses from the manufacturer:



I was wondering why this bump on the power response, mainly on the high frequencies and suddenly I had an epiphany! Could it be since I only have three angle curves (0º, 30º and 60º) and due to the higher DI of the tweeter that the real values are further from the reality on this region? In other words, when I input the real off-axis responses over 160º I'm expecting that the bump will disappear or be much smaller at least. Am I right or am I wrong?

Mainly, the trade offs between the original crossover and the new 'asymmetric' one are the loss of 2-3dB sensitivity vs. the much better Directivity (ver) and also a flatter on-axis response. That is after all the main goal of this redesign and I can sacrifice 2 or 3dB without issues.
As drawbacks or something to think about, we have as I said the lesser role of the woofers and the predominance of the mediums, specially the one before the L1. I'm not too worried about the tweeter although perhaps it would be more prudent to use a 3rd order HP filter just to protect it better from any bursts of pink noise.

[Juhazi]

Bass response (blue) must be from just one driver, but you have four of them. Same applies to other multiple drivers in the simulation. Scale is in dB, but it could be also voltage required to get that total spl.

[cochinada]

Bass response (blue) must be from just one driver, but you have four of them. Same applies to other multiple drivers in the simulation. Scale is in dB, but it could be also voltage required to get that total spl.

That is true. I forgot to upload the power consumption of both XO:

Original


New

[kimmosto]

And this is a pre-simulation as Kimmo calls it..

I think it was called as

designing of XO with mix of simulated and measured data might be just waste of time.

Please proceed but remember that results are simply wrong.

[cochinada]

...Please proceed but remember that results are simply wrong.

For the new tweeter yes but for the current one certainly not.