Hi,
My wish is to fully simulate speaker with VituixCAD before buying any of the components and building physical cabinets.
Project in details:
-Bookshelf form
-Passive crossover
-2-way with only 5,5 C 1,5 CP coax driver
-Bass reflex 5-10L of internal volume. Manufacturer gives 8L or 9L as recommended in its documentation, quick sim with WinISD also gives "ok" results in 5+ liters range
-Narrow baffle with massive edge radius, 50mm or similar. Only to fit the driver and then some more for rounding/beveling
I already invested some generous two-digit hours of time in understanding some of the Vituix tools, but... Unfortunately it is a tremendously complex software comparable to the collection of the scientific calculators but still with no clear workflow how to actually simulate the speaker in digital form before starting the physical process.
What I did:
1. Took the freely available measurement files in picture format from dibirama.it. I assume they are made on large baffle, 2pi space, 1000mm at 0 degrees H and V.
2. Made FRD and ZMA files with "SPL Trace" tool in Vituix.
3. Simulated Baffle Step Loss with "Diffraction" tool, opened FRD file from my trace, then exported FRD files with 10 degree steps. All the steps above for both drivers - tweeter and woofer.
4. In "Drivers" tab created 2 drivers and opened bunch of FRD files and ZMA file. Everything seems to look ok.
5. Started to play with crossover, manufacturer suggested as a starting point.
This is what I got with dibirama.it SPL and impedance data AND manufacturer suggested crossover:
dibirama.it data with tuned values of manufacturer crossover:
The same with manufacturer data AND manufacturer crossover values:
I do not worry much about the 100Hz range simulation, as the vented box behavior is pretty much clear to me. Also, I have made all the simulations in the virtual 1m distance, because I think the manufacturer and dibirama.it measured at the same distance.
Below are manufacturer polar diagram of separate tweeter and woofer:
It seems dibirama.it + manufacturer crossover is even better than the manufacturer data alone.
I do believe, that manufacturers make their data by cherry-picking their drivers for measurement or/and simply by painting them to look more beautiful, but my questions are:
-Does the driver look listenable from my simulations?
-If not - what could be reasonable crossover for this driver?
and most importantly:
-Are my simulation steps right?
Here is manufacturer data of the completed speaker with their suggested crossover. All the ripples are in 2-3dB range, but my simulation shows closer to 5-6dB, a pretty big difference:
It is a pretty expensive experiment to do in physical form, as drivers and crossover components will be at least 250-300EUR, I want to make sure it will not disappoint me. There are not many projects with this driver, and also I am scared of crossovers with ridiculous amounts of components (15+...).
Thank you for your thoughts
My wish is to fully simulate speaker with VituixCAD before buying any of the components and building physical cabinets.
Project in details:
-Bookshelf form
-Passive crossover
-2-way with only 5,5 C 1,5 CP coax driver
-Bass reflex 5-10L of internal volume. Manufacturer gives 8L or 9L as recommended in its documentation, quick sim with WinISD also gives "ok" results in 5+ liters range
-Narrow baffle with massive edge radius, 50mm or similar. Only to fit the driver and then some more for rounding/beveling
I already invested some generous two-digit hours of time in understanding some of the Vituix tools, but... Unfortunately it is a tremendously complex software comparable to the collection of the scientific calculators but still with no clear workflow how to actually simulate the speaker in digital form before starting the physical process.
What I did:
1. Took the freely available measurement files in picture format from dibirama.it. I assume they are made on large baffle, 2pi space, 1000mm at 0 degrees H and V.
2. Made FRD and ZMA files with "SPL Trace" tool in Vituix.
3. Simulated Baffle Step Loss with "Diffraction" tool, opened FRD file from my trace, then exported FRD files with 10 degree steps. All the steps above for both drivers - tweeter and woofer.
4. In "Drivers" tab created 2 drivers and opened bunch of FRD files and ZMA file. Everything seems to look ok.
5. Started to play with crossover, manufacturer suggested as a starting point.
This is what I got with dibirama.it SPL and impedance data AND manufacturer suggested crossover:
dibirama.it data with tuned values of manufacturer crossover:
The same with manufacturer data AND manufacturer crossover values:
I do not worry much about the 100Hz range simulation, as the vented box behavior is pretty much clear to me. Also, I have made all the simulations in the virtual 1m distance, because I think the manufacturer and dibirama.it measured at the same distance.
Below are manufacturer polar diagram of separate tweeter and woofer:
It seems dibirama.it + manufacturer crossover is even better than the manufacturer data alone.
I do believe, that manufacturers make their data by cherry-picking their drivers for measurement or/and simply by painting them to look more beautiful, but my questions are:
-Does the driver look listenable from my simulations?
-If not - what could be reasonable crossover for this driver?
and most importantly:
-Are my simulation steps right?
Here is manufacturer data of the completed speaker with their suggested crossover. All the ripples are in 2-3dB range, but my simulation shows closer to 5-6dB, a pretty big difference:
It is a pretty expensive experiment to do in physical form, as drivers and crossover components will be at least 250-300EUR, I want to make sure it will not disappoint me. There are not many projects with this driver, and also I am scared of crossovers with ridiculous amounts of components (15+...).
Thank you for your thoughts
Last edited:
Member
Joined 2003
For diffraction simulation, main problem is that this simulation assumes flat piston, so it becomes quite inaccurate for upper frequency range, let's say >1500Hz, and especially so for any tweeter other than flat faceplate like standard dome. A coax tweeter is effectively horn-loaded by the woofer cone, so don't be surprised when your overall power & DI of the real speaker is quite different from your simulation.
For final speaker design presentation, normalization of polar map should not be used.
Despite the many posts and sticky on this forum suggesting to design using manufacturer data, the correct design workflow would be:
Purchase driver(s)
design and build cabinet
install driver(s) and measure fully, follow measurement guide for VituixCAD to measure and process data
Simulate crossover
optionally simulate crossover via DSP or APO-EQ
build crossover
validate
For final speaker design presentation, normalization of polar map should not be used.
Despite the many posts and sticky on this forum suggesting to design using manufacturer data, the correct design workflow would be:
Purchase driver(s)
design and build cabinet
install driver(s) and measure fully, follow measurement guide for VituixCAD to measure and process data
Simulate crossover
optionally simulate crossover via DSP or APO-EQ
build crossover
validate
Manufacturer data + dibirama just isn't enough data for a good simulation. You might wanna try @vineethkumar01 's full set of measurements for their teardrop shaped speaker with the 5.5C1.5CP. https://www.diyaudio.com/community/threads/a-3-way-design-study.376620/page-31#post-7113704
Dcibel, I have read your post few times... More questions than answers:
Should I learn AKABAK?
Like this one?
Thank you one more time, Dcibel
Did not think about that at all. We need another tool for waveguide simulation at other than 0 degrees. Is that right?
Should I learn AKABAK?
What should I use? The same polar map, but without normalizing fuction? Why? What will I see more from non-normalized chart?
Like this one?
Those are the exact steps I want to do as late as possible. I already have entire collection of 2,5-4 inch fullrange speaker drivers, some in cabinets, some just collecting dust. Most of them just sound plain bad (to my ears) no matter how I EQ them. It was my personal journey to understanding that I just do not like sound of metal cones. Now when crossovers are involved I want to avoid buying capricious driver and then use additional 15+ components to make it ok-ish. I really want to do proper simulations first.
I had installed APO-EQ on my computer. It is indeed a powerful tool, but the question is: If I build "killer" crossover and like the sound - how I translate EQ curve or the filter values into mH and uF values of coils and caps and the topologies of slopes? Is there a tool for this or I still need to use external calculators to build this?
Thank you one more time, Dcibel
Last edited:
On my way reading this, thank you!
Now it is late in my place, will read in couple days.
Member
Joined 2003
Your choice. VituixCAD diffraction tool won't cut it.
Normalized chart simply applies on-axis EQ to pound the response completely flat, so all directivity is shown relative to 0deg axis. Good use for things like evaluating waveguides and horn shapes, or cabinet shape, but a complete speaker is not flat on 0 deg 20Hz-20kHz, so its true directivity result is without normalization. I usually turn on contours at 3dB steps to make the directivity shape clear.
Proper simulation starts with good data. If you want to design speakers, start with a good mic and a USB Audio interface. I use a Line Audio Omni1, which is a great mic for both frequency response and distortion. any 2ch audio interface should do, but I use Motu M4, the M2/M4 are great high performance units that won't break the bank.
Build a simple jig for measuring both acoustic response and impedance:
https://www.htguide.com/forum/articles/do-it-yourself-diy/927384-dual-channel-measurement-jig
Go the other way around, design a passive filter, and run the transfer function through APO-EQ to "simulate" it using impulse response convolution. I made a simple instruction for it here:
https://www.htguide.com/forum/artic...o-filter-simulation-with-vituixcad-and-eq-apo
Adjustments to the passive filter can then be immediately tested and verified through a few clicks at the PC.
Why rely fully on it to begin with?
Even having some odd shaped chamfer and things can change drastically.
These kind of tools are all for estimation purposes.
Even FEM/BEM based solutions.
That's absolutely not what normalization is.
Normalization just takes a given measurement as reference point.
This can be done with any axis (or any measurement for that matter).
All other values are being taken relative from there, instead of showing the absolute value instead.
This IS what directivity is!
Since directivity is purely relative to begin with.
So it's actually the only proper way to look at the directivity of something, since directivity itself is relative.
But I agree that it sometimes can mess up when things are all over the place.
Which on itself clearly shows there is a big issue.
Unless you have super special above human eyes, looking at the non-normalized absolute off-axis response won't show you all the data.
Well technically the data is there, but it's close to impossible to see how things move relatively speaking.
Normalized graphs show things like diffraction problems very easily.