VituixCAD v2

[kimmosto]

tweeter is old XT25TG30 ring radiator in Visaton WG148 waveguide.

That combination is not very flexible maintaining sound balance in different acoustical conditions and listening directions. You seem to have fitted carpeting which reduces reflected power at HF. Have you measured acoustical parameters at typical listening area? I'm interested in EDT and T20 spectrum. ARTA is good in that task. Probably REW too.

Usually plastic and paper cones 4.5"-5.25" don't need wave guide, but ring tweeter may need some boosting at HF to compensate "excessive" directivity at top octave, especially if listening area is very large.

[kimmosto]

Some statistics with subjective classification: green=good, orange=maybe, red=bad. Calculations done with "custom" preference rating due to better math.

[Reet]

Thank you for your insight. If I understand, end result is that there is no one size fits all criteria for balancing on-axis and power response, especially at the extents of the hearing range. Make it sound good to you in your space and move on. I do want to make note, that the top-end rise I do not find offensive, bright, or "hot", in fact I find the speaker quite balanced. I will make some adjustments with external EQ, so I can A/B compare the two responses above and see if I have a preference, or can even tell the difference between the two.

FWIW, this is the speaker used for the example, tweeter is old XT25TG30 ring radiator in Visaton WG148 waveguide.

https://www.htguide.com/forum/forum/mission-possible-diy/947504-disco-on-the-floor-p

[kimmosto]

...which will be preferred? ... What do you think?

Fortunately top octave can be hot so wave guide or some diffuser is not must with domes and ring radiators. For example Scan-Speak's rings could sound quite dry and dull to off-axis, but they also tolerate longer listening distance.
Generally all this is really complex equation. Much more complex than Olive's Preference rating and reviewers can ever handle. Result depends on room acoustics; length and spectrum of EDT and RT60, location of absorbing panels and carpets, location of diffusers, location and materials of surfaces without acoustic treatment, listening distance, speaker locations, need to listen radically off-axis (in different room), strength of directivity, tilt (dB/oct) of PIR/SP slope, shape of PIR/SP slope, location of woofers and ports in cabinet, SPEAKER CONCEPT and much more.

[Reet]

Another forum had posed a question regarding on-axis lift in the top octave when tuning the overall response. I have personally found that using listening window average is a more useful metric for tuning than single axial response, so ending up with a bit of axial lift provides some better balance in a normal listening situation. I am just wondering if anyone else has any insight/opinions on the subject here.

Here's a simple example, This speaker includes a lift in the top octave, in order to achieve a better slope to listening window, in-room, power response.

Alternatively, this speaker could be tuned for a flatter axial response, which results in a drooping power response. Typical behaviour for many speakers.


But, which will be preferred? If "preference rating" is a good metric, first equation provides very much the same result, with the first option receiving a better rating by 0.1. The second equation gives preference to the second option by about 0.6. Generally speaking, I don't put much importance in the top octave anyway, so this is small details, but one solution must be "better". What do you think?

[kimmosto]

2.0.111.0 (2024-01-05)

Calculator

• Removed 'Product of A responses'.
• Removed 'Directivity of A responses'. Use main program instead due to features and accuracy.
• Removed 'Power of A responses'. Use main program instead due to features and accuracy.
• Added 'Average by off-axis angle A'. Calculates average of the same plane and angle.
• Added Plane column to A responses
• Added possibility to sort A responses by Filename, Plane or Angle by clicking header.

User manual:
Average by off-axis angle A
Calculates average of the same plane and angle. This is handy when crossover is designed with several driver samples having some real life tolerances i.e. differences in frequency responses. All measurements of drives samples are loaded to A responses. Program sorts response list by Plane, Angle and Filename, and calculates and exports average of each off-axis group to response files which can be loaded to Drivers tab in the main program.
Result's magnitude is arithmetic mean of magnitudes to avoid magnitude drop due to possible delay differences between acoustical measurements to the same off-axis direction. Result's phase angle is vector sum to avoid phase unwrapping (which would be required for arithmetic mean). 'Complex Sum/Avg/RMS' is unchecked and disabled to force previous settings.
Note! 'Angle parsing from filename' settings in Options must match with loaded frequency response filenames to get correct values to Plane and Angle columns.

Default result filename prefix is AvgByDir, followed by plane and off-axis angle of the group. Angle coding uses current 'Angle parsing from filename' settings in Options (so you can also change angle coding e.g. from 3D balloon to generic 2D).

[kimmosto]

Happy Holidays!

The latest build of 2.0.110.3 (2023-12-25) finally includes short instructions and few screenshots for Create project sub-folders -feature. That was promised already few months ago with 2.0.110.0 (2023-10-28), but repeatedly forgotten.

Some small but probably invisible logic tweaks has also done to trace widths of Directivity charts. Width difference between "normal/thin", selected and on-axis traces is not yet changed. Selected trace is +1px and on-axis is +2px so they might be difficult to distinguish with 4K display. I also tested multipliers such as 1.5x and 1.7x base width. Multiplier is probably better with 4K, but worse with e.g. 1920x1080px (which I always use though screen of my 17" laptop is native 4K).

[Reet]

Thank you, and Happy Holidays!

[kimmosto]

2.0.110.3 (2023-12-24)

Main

• Line width of line, waterfall and surface directivity charts saved to user settings for the next session.

[Reet]

Is there a possibility to make 2px line thickness for the directivity line chart a default setting? Ever time VituixCAD is closed, or if just the chart type is changed, the line width is set back to 1px.

[kimmosto]

2.0.110.2 (2023-12-13)

Diffraction

• Eliminated program crash (related to automatic Y axis scaling) if magnitude traces are hidden.

Calculator

• Automatic Y axis scaling with A button skipped if magnitude traces are hidden.

[kimmosto]

Extrapolation from 5 Hz down to 0 Hz has always been problematic, and I can almost promise that changes in 2.0.110.1 are not the last.
Previous versions had small bug for several months; real and imaginary part of 0 Hz point was zeroed with all LF slopes, but zeroing was not (accidentally) done to 0 Hz point of "mirrored" frequency response which fills the second half of the FFT buffer. Now both ends are zeroed, but possible problem is that filter of woofer could extend down to 0 Hz. 0 Hz point is not necessary to zero if there's no HP slope at LF.
Much older versions detected LF slope and zeroed or copied values from the next frequency point to 0 Hz if slope is < 3 dB/oct. That was expected to be better approach, but somehow I wasn't happy with it. Maybe it should be evaluated again.
Full range flat frequency response produces natural DC offset to impulse response. That's not a problem, but some DC variation (easily visible in step response) could exist with almost any real life IR depending on combination of sample rate, FFT length, taps and window function because limited part of IR could include DC offset also with high passing slopes.
So there is no perfect solution for FIR with truncated frequency range and limited taps. Some errors will exist at LF.

[Reet]

Excellent update today, thank you.

2.0.110.1 (2023-12-05)

Impulse

• Magnitude and phase extrapolated as zero at 0 Hz to reduce DC offset in impulse response.

[kimmosto]

I think I have a bug- the main screen, moving drivers up or down in the list to reorder them.

in the crossover screen, this order is not reflected there.

Schematic must not change a bit when driver list is re-ordered. There is no such AI which could redesign all possible circuit topologies and connections so that drivers are in certain graphical order from top to bottom without messing up whole wiring and visual clarity. All crossovers are not simple parallel ladders with single driver for each band where that kind of feature would be possible.

[tktran]

I think I have a bug- the main screen, moving drivers up or down in the list to reorder them.

in the crossover screen, this order is not reflected there.

[HTDT89]

Ah, yes, I see. Obvious! Sorry I didn't think to check traces.

[kimmosto]

^Value of Rs and other filter components control visibility of green traces called 'Impedance Magnitude/Phase with passive filter'. Raw impedance without passive components is dark trace (WindowFrame by default). It is normally visible, but possible to hide for good. Check Show cell in Traces window, and it should stay visible also in the next session. I usually click View->Default colors in the main program to restore 'factory defaults'.

[HTDT89]

kimmosto, I believe there's a small bug in the enclosure tool. The impedance curve does not plot when filter tab's Rs = 0. Because Rs is not remembered when closing the tool, there is no impedance graph when you first open it or manually reset Rs to 0.

I am running latest, v2.0.110.0.

Other than this, the embedded filters are great. Much easier workflow than jumping between crossover & enclosure for this basic analysis.

I am a slow thinker, and it took some time to digest your input.
I have gained significant insights from our conversation, thank you very much.
I am sorry I made you upset.

I absolutely agree. You cannot understand my questions and I cannot understand your answers.
It must be me who is too stupid.

[Reet]

Sorry, I do not understand your answer. What is over complicated? I am trying to follow the guidelines, but the guidelines are not applicable to my low frequency config.
Unless there are any guidelines that I have missed. Exactly what guidelines are you referring to?

What is unique about your speaker that makes the measurement guide not apply? Everything I've been leading you towards in past few posts is simply to follow instructions of the measurement guide.

Why is near field v2 useless?

What is so difficult to place it properly? Perhaps review details in aux - Near Field section.

You have not yet answered may question about the merger setup for v2.

No use talking about how to do things poorly so you'll need help from someone else to continue down that route.

[kimmosto]

...guidelines are not applicable to my low frequency config.

Guidelines are applicable with closed and vented LF designs which can be measured at near field and simulated with Diffraction tool. So you could follow instructions.

Don't try to over-complicate things, just follow the measurement guide for VituixCAD. For low freq, "v1" image is correct for 0 deg measurement, continue with far field for all other angles, leave the pipe in front of the driver as you intend for the far field measurements. V2 shows 90 deg far field, near field is useless measurement. Left image is correct for midrange far field, continue far field for all other angles, then move mic up to tweeter level and repeat for tweeter measurements.

Sorry, I do not understand your answer. What is over complicated? I am trying to follow the guidelines, but the guidelines are not applicable to my low frequency config.
Unless there are any guidelines that I have missed. Exactly what guidelines are you referring to?
Why is near field v2 useless? Far field is unreliable under 500Hz even out doors due to ground reflexes.
You have not yet answered may question about the merger setup for v2.

[Reet]

Don't try to over-complicate things, just follow the measurement guide for VituixCAD. For low freq, "v1" image is correct for 0 deg measurement, continue with far field for all other angles, leave the pipe in front of the driver as you intend for the far field measurements. V2 shows 90 deg far field, near field is useless measurement. Left image is correct for midrange far field, continue far field for all other angles, then move mic up to tweeter level and repeat for tweeter measurements.

I don't fully understand what you've done to measure the woofer, but there isn't a need to do anything different than any other driver. It may be better / easier to tip the speaker over on its side to measure the woofer in a forward facing position, and it looks like there is room between the pipe and speaker to place a mic for nearfield measurement. Diffraction should not be 90 deg, in the case of the woofer the egg shaped board is the baffle and it's measured straight on like any other driver.

In the crossover design, enter 90 deg tilt for the woofer, and other coordinates accordingly.

I need to describe what I have done so far. Please see the attached image. Version 1 is my initial measurements.
The high frequency/midrange (HF) measurements are the same in version1 and version 2.
In version1, I measured near field at the baffle surface and center of drivers on-axis, far field at 1m.
For the low frequency (LF), I did not tilt the speaker, I put the microphone 1m above at the driver on-axis.
All measurements were done outdoors, my rooms are too small to enable the recommended 4ms time windows.
I merged the near field and far field for both HF and LF to the best of my understanding.
I divided the merged frequency responses by the HF and LF on-axis diffraction patterns in the diffraction tool to create halfplane SPL versions.
Then I generated both vertical and horisontal directivity files based on the half plane HF and LF SPL files.
In the VituixCAD crossover I set the LF driver as tilted 90 degrees.
This scheme seemed to work rather well and I was able to adjust everything to get a straight SPL frequency response for the system.
However, the listening experience is such that I do not trust my measurements and design to 100%.
It could be room effects, but it makes me want to go through everything once more.

I got the impression that you recommend measurements rather than simulations.
That made me think of a low frequency measurement according to version 2 in the attached drawing.
In version 2 I measure the LF near field 90 degrees off axis near the baffle edge.
In this case the merger might need to be set up differently, hence my questions.

I actually tried to do this v2 measurement outdoors a few days ago but it failed, probably due to the temperature, 1 degree below freezing.
The driver suspensions and cone materials could behave differently enough at this temperature vs. room temperature to make the difference. Or even the microphone.
I made this judgement by comparing the HF measurement to the previous attempt in the summer, and they were quite different.
I need to postpone this v2 measurement project to the spring.

[Reet]

I don't fully understand what you've done to measure the woofer, but there isn't a need to do anything different than any other driver. It may be better / easier to tip the speaker over on its side to measure the woofer in a forward facing position, and it looks like there is room between the pipe and speaker to place a mic for nearfield measurement. Diffraction should not be 90 deg, in the case of the woofer the egg shaped board is the baffle and it's measured straight on like any other driver.

In the crossover design, enter 90 deg tilt for the woofer, and other coordinates accordingly.

I need some advice on how to setup the merger tool for my project.
My low frequency enclosure is a CB, vertically oriented cylinder, 700mm long, 400mm diameter, and the 10-inch woofer is mounted pointing upwards in the top lid baffle.
My high frequency enclosure is a pipe crossing the face of the woofer, facing front, driver center distance to the woofer plane is 150mm.
My crossover frequency is 240Hz.
The front baffle/plane of the high frequency driver is placed at the perimeter of the low frequency enclosure, front side. See the PNG image included.
In my far field measurement I use a distance of 2 meters, the mic height is placed at the midpoint between the enclosure top height and the high frequency driver axis.
My near field measurements for the high frequency driver is conventional, on-axis center about 10mm from the dustcap. No problem to set up the merger tool for this.

My near field measurements for the low frequency driver is different. I measure at the edge of the top woofer baffle, i.e. 90 degrees driver off axis, just below and next to the high frequency driver baffle.

The merger tool manual says that the BS should not be checked, as I am not measuring on the front of the driver. I understand this.
I have generated diffraction responses from the low frequency baffle, distances 2m and 5m, 90 degrees horisontal tilt, see image below.
Should I include any diffraction response at all in the near field section, or in the far field section?

[Reet]

Let's say >1kHz, general use of diffraction applied to a neafield measurement would usually be <500Hz. . As well, any tweeter with anything other than a flat faceplate, off-axis not possible with diffraction tool. Main problem is that calculated diffraction uses a "piston model" for high frequency driver behaviour, reality is a fair bit different factoring cone shape, breakup effects, waveguides on tweeters, etc.

Sure, keeping in mind that calculated diffraction response lacks accuracy at high frequency

Please define high frequency.

[Reet]

Sure, keeping in mind that calculated diffraction response lacks accuracy at high frequency, and its a bit of an annoying tedious process if you want to do that for all off-axis data, since each off-axis diffraction response will need to be processed individually through the calculator tool, it doesn't support multiple files as "B response". Unless the cabinet width varies wildly, it may be easier to do nothing and just load in measurement data for slightly wrong cabinet for ballpark figure, with complete re-measure once the future cabinet is built. The result of a simulation with slightly wrong cabinet measurement may have just as much error as a bunch of math on simulated diffraction data.

For the above inaccuracies mentioned, VituixCAD measurement and design recommendations is to utilize calculated diffraction only for low frequencies, not to be applied to complete half space response.

Follow up question would be why would you want to? Regardless...

Half space data for a driver could be used in the diffraction tool for any future cabinet/baffle design.
Full space data is for the specific driver and enclosure combination where it was measured.
Right?

[Reet]

A real question, not so complicated after all..."How do I generate half space data?".

Follow up question would be why would you want to? Regardless...

1. Measure on infinite baffle. Perhaps "in-wall" installation would fall under this category. Would you need the merge tool in this scenario? Does your room still have a floor and a ceiling and walls? If so, answer is yes.
2. Manufacturer data is often half space equivalent, traced data could be one use case.
3. Approximation of half space can be obtained by dividing far field measured response in-cabinet by diffraction response, use the calculator too. Realize that the result of simulated diffraction is not incredibly accurate at high frequencies, and especially so off-axis.

I don't understand your question. Are you wanting to produce a half space result, or are you wanting to produce a full space result in your cabinet for crossover design? Merge tool simply spices two responses together, while retaining the amplitude and phase of the high frequency part. Whether the spliced result is half space or full space depends on the frequency response data you provide, and selection of baffle loss on the low frequency part.

I do not understand the above. How can I get real half space data under normal circumstances? It is impossible unless I have an infinite baffle, which is very difficult to find, at least in the country where I live. And if I did, why would I need the merge tool?

Generally speaking, the intended use of the merge tool would be to produce a full space response of a speaker measured in the intended cabinet for crossover design. Near field measurement data would be provided for the low frequency part, check diffraction response and load in the diffraction simulation for the low frequency part. Far field windowed data would be loaded to high frequency part. When done correctly, result is reflection free full space response, ie anechoic equivalent.

So there was the answer. It was not so complicated after all.
Thank you.

A follow-up question.
Is it possible to create half space data from a merged full space?
Does it work to use the calculator tool to divide the merged data by the cabinet diffraction response?

[Reet]

I don't understand your question. Are you wanting to produce a half space result, or are you wanting to produce a full space result in your cabinet for crossover design? Merge tool simply spices two responses together, while retaining the amplitude and phase of the high frequency part. Whether the spliced result is half space or full space depends on the frequency response data you provide, and selection of baffle loss on the low frequency part.

Generally speaking, the intended use of the merge tool would be to produce a full space response of a speaker measured in the intended cabinet for crossover design. Near field measurement data would be provided for the low frequency part, check diffraction response and load in the diffraction simulation for the low frequency part. Far field windowed data would be loaded to high frequency part. When done correctly, result is reflection free full space response, ie anechoic equivalent.

The choice is yours. For speaker design, purpose would be the latter, to create an anechoic equivalent response of drivers in their cabinet.

I am sorry, I do not understand your answer. How can you get both?
Please clarify.

[Reet]

The choice is yours. For speaker design, purpose would be the latter, to create an anechoic equivalent response of drivers in their cabinet.

I have a fundamental question about the merger tool.
I have read the manual, but reading and understanding are not the same, unfortunately.

Is the merger tool producing a half-plane SPL trace for the driver without diffraction, or a full space SPL trace for the driver in the enclosure taken at the mic position?

[Nil L]

Thanks for the advice. Perhaps I didn't use all the possibilities. I will try to learn how to use existing opportunities.

[kimmosto]

It’s not a good idea to mentally calculate the denomination that will limit the optimizer’s search.

First you ask more constraints (for limiting XO frequency) and then you don't like or want them. What is the logic?

I have seen program(s) having many times more constraints (including XO frequency) and targets than VCAD. For example LspCAD 6. Problem is that entering multiple constraints and targets could take much time - more than manual initialization of component values, and some small detail in settings could prevent optimizing totally. That is very frustrating.
Initialize network and component values with brain and mouse wheel or optimize response of each individual driver with target, and then fine tune on-axis, LW, ER and SP with optimizer so that bad decisions e.g. with XO and impedance are prevented with constraints which are available during that phase. This method leaves you full control and you won't change to powerless passenger in that vehicle.

[Nil L]

Thanks for the advice. I am aware of this possibility. It’s not a good idea to mentally calculate the denomination that will limit the optimizer’s search. To teach a program that can do this faster than any brain seems to me to be the best idea. But this is not a problem at the moment. In case of arbitrariness of the optimizer, I deprive him of the laurels of the winner and manually do everything normally.

[kimmosto]

...can be limited to adjustment of a single driver circuit by limiting components to optimize.

This is primary method to keep crossover frequency within proper range. Another very good option is to limit component values of passive network. Parameters i.e. component values have min and max constraints. For example series coil of the woofer >= 1.2 mH and/or series cap of tweeter <= 10 uF could maintain some sense in XO freq while total on-axis is optimized.

All this is very natural with active OPA and DSP projects because frequency is main parameter of LP and HP filters. Just limit freq with min and max constraint. FIR projects use transfer functions which are created with target responses having XO freq selected by user.

[Reet]

Optimizer doesn't design the crossover for you, only to adjust part values to achieve the target goal you've provided. Options already exist to limit frequency range of optimization, and define high pass / low pass target to optimize individual driver response. Optimization to overall power and axial response doesn't care about individual driver responses, but can be limited to adjustment of a single driver circuit by limiting components to optimize. Best results if driver response is already in the ball-park of target to begin with and appropriate circuit topology is in place. It's an excellent time saving tool.

[Nil L]

kimmosto,
The optimizer may suggest cutoff frequencies as a result of its operation that should never be used. For example, 600 hertz for two drivers. The tweeter should never play down to 600 hertz. The play of percentages between axial response and power response doesn't help. Offer. Add a parameter that limits the frequency range in which the optimizer can search for the best option. Each driver has its own range. Or a common range for all drivers.
The same thing happens with the level of individual drivers. The optimizer decides that the tweeter level can be 5 dB higher than the midbass level. It would be good to set a range of levels at which the optimizer can play. Each driver has its own range. Or a common range for all drivers.

[Reet]

Great changes, the change in resonant frequency with Fs should make for a bit better agreement when comparing actual measurement to simulation.

FWIW most of my reflex designs use slot ports, when comparing actual measurement and impedance to the simulation, I find Qp around 20-30 is proving better agreement with real world results. Possibly the values I had used for Qa and Ql were bad, or perhaps Qp suggested by Kristian accounts for port ends but still assumes an otherwise perfect port, not accounting for bends, laminar flow vs turbulent flow, back pressure or other factors restricting port efficiency. I'm no expert on the subject though, just trying to determine best values for accurate simulation.

[JonMarsh]

Very useful updates!