VituixCAD v2

[DaveFred]

At the risk of splitting hairs, I am wondering how to best model enclosure diffraction when the cabinet has beveled facets.

Here is the cabinet drawing, 350mm tall, 180mm wide, driver 190mm from bottom, driver SD 54cm^2,



And here are some diffraction simulations,

Orange - 180mm cabinet width with square edges ignoring the bevel
Red - 180mm cabinet width, treating the bevel as a roundover based on where the bevel lands in the middle of the woofer
Purple - 180mm cabinet width, treating the bevel as a roundover based on where the bevel is at the top of the woofer
Green - 150mm cabinet, treating the cabinet width as being the bevel in the middle of the woofer.

I realize that for the purpose of baffle step, these are all within 1/2 dB and not going to make a world of difference, I was just wondering which simulation would you consider most correct?

My poorly educated guess would be "Red".



Thank you,

David.

[Reet]

For the intended purpose of diffraction simulation, which would be to apply to nearfield response of a woofer, I would probably just take the size of the facet directly horizontal to the centre of the woofer and apply that as a roundover and call it good, so red trace if I understand your descriptions properly.

I know you have some completed speakers with these big facets, why don't you compare with your own real-world measurements? Which one of these responses when applied to the nearfield response and merged, has the best alignment with the final speaker measured response?

[kimmosto]

2.0.132.0 (2025-11-25)

• Linkwitz-Riley HP/LP linear-phase active filter blocks support 10th...16th order (60...96 dB/oct) slopes.
• Linkwitz-Riley HP/LP target in Optimizer window supports 10th...16th order (60...96 dB/oct) linear-phase slopes (Lin.pha checked).
• Development environment updated from Visual Studio 2022 to 2026 (VS 18).

[kimmosto]

2.0.133.0 (2026-02-05)

Main

• Minor modifications to target zones in CTA-2034 chart.
• Added 'Danville 96k' and 'LEA 96k' to DSP system list in Options window.

Preference rating

• Added NBD_LW, NBD_SP and NBD_SPDI values to variable table.
• Minor modifications to default values for 'Full space', 'Half space' and 'Constant DI' slope targets (SL_).
  For example default values of SL_LW, SL_PIR and SL_SP for 'Full space' are equal to 'Multiple Regression Model for Predicting Loudspeaker Preference Using Objective Measurements Part II' by S. Olive.
• One slope target can be activated at a time to clarify monitoring of target line in CTA-2034 chart.

[kimmosto]

Target zones in CTA-2034 chart as a function of standard directivity index slope, since rev. 2.0.133.0:
Red circles are default values of target slopes (SL_) in Preference rating window: Close to Constant DI @DI-slope=0.3 dB/oct, Half space @DI-slope=0.78 dB/oct, Full space @DI-slope=1.18 dB/oct. Ideal 'Constant DI' does not have default values (all -0.8 dB/oct). It's too theoretical.

[kimmosto]

For example default values of SL_LW, SL_PIR and SL_SP for 'Full space' are equal to 'Multiple Regression Model for Predicting Loudspeaker Preference Using Objective Measurements Part II' by S. Olive.

Chapter related to slope targets. Last two rows are calculated from Olive's table, not included in the original.

[Reet]

PIR should probably have slightly higher slope target than ER, but those two and often very close together perhaps it's splitting hairs. I often turn off ER trace on CTA-2034 chart and just look at PIR to reduce clutter.

[kimmosto]

There are few typos in Olive's paper...

There are also a couple of ambiguities or illogicalities. Preferred speakers in the big group have target LW slope = -0.2 and target SP slope = -1.75, but target SPDI slope = 1.4. It should be 1.75 - 0.2 = 1.55 if the same speakers were preferred. Maybe each variable has own group of preferred speakers...

Speakers in the small group were 2-ways having higher directivity slopes due to directivity of smaller box and less crossover bands. Can we assume that lower slopes would be more preferred also with 2-ways, but really good 2-ways with lower slopes were not available and tested? Or was stronger directivity more important in that room acoustics than better (=lower) directivity slopes? Combination of reverberation time and listening distance may not support well-designed small 2-ways with healthier directivity slopes.

Speakers were not listened off-axis or off-room. Speakers were not listened how they are normally listened to. Monophonic only, one speaker at 3 m, directly in front of the listener. In the consumer market, everyone does not sit in the same listening spot whole time. If someone does, they are most likely listening in stereo or multi-channel, meaning the angle between listener's ears and the speaker was not correct either (except for HT center). Music samples were almost the same elevator muzak. Few sharpened drum beats, but no real transients, heavy loads or real acoustic orchestra for sure. Listeners were led to focus on the tones of the singer, guitars and a few other instruments. Why not, because there was no intention to study anything else due to preconditions. The study was so limited that for example directivity slopes, dynamics, non-linear distortion, direction of wavefront, sound stage / diffraction problems dropped to zero or close to. So here we are - too many variables has become irrelevant on some forum(s) because they were not properly studied when it was possible.

[kimmosto]

PIR should probably have slightly higher slope target than ER...

Agree. This case dependent. In some rare cases ER slope is very close to PIR slope, but usually PIR slope is slightly steeper.

[kimmosto]

Chapter related to slope targets...

Quoted message updated completely. Added text by S. Olive, target values in dB/oct and calculated textbook power response and DI slopes.

Few comments to Olive's short text:

The ideal target slope for the on-axis and listening window curves (0 and –0.2) is identical for both test samples, which indicates that the on-axis curve should be flat, while the off-axis curves should tilt gently downwards.

Agree, but other directivity patterns than unidirectional boxed were not (presumably) studied so this this cannot be generalized. I'm quite sure horizontal on-axis cannot be generalized because nothing in practice has indicated that. More constant directivity (flat horizontal DI or widening dispersion at top octave) -> more on-axis tilt down.

The degree of tilt varies among curves for Test One and the larger sample. Test One includes mostly 2-way designs whereas the larger sample includes several 3-way and 4-way designs that tend to have wider dispersion (hence smaller negative target slopes) at mid and high frequencies. This suggests that the ideal target slope may depend on the loudspeaker’s directivity.

3- and 4-ways don't necessarily have wider dispersion at H.F. than small 2-ways. They can have smaller mid so dispersion at upper mid can be wider. More ways and larger front baffle increase directivity at lower mid ... upper bass so large traditional multi-ways usually have more linear directivity index and lower off-axis slopes. Agree that this suggests that the ideal target slope may depend on the loudspeaker’s directivity, but dependence locates probably in on-axis, listening window and power response. Not (so much) in predicted in-room because it represents perceived tonal balance at mid...far field better than on-axis and listening window alone. Power rules more to far field + off-axis. In addition, speakers in Test One were quite bad in average so we can ignore the results and conclusions quite safely, though slope targets in the table are calculated from top-10 %.
Dipole usually produces extra reflected phase matching pressure to mid-bass so on-axis and listening window slopes can be more horizontal than with cardioidish. Dips at L.F. in room response reduce energy and tilt balance up so large boxed or cardioid woofer arrays with solid wavefront towards the listener can have more horizontal on-axis than e.g. small bass cardioid requiring intentional bass boost. And so on, not forgetting the tilt of room reverberation balance.

[Reet]

Hi Kimmo, over at DIY there has been some "crowd-sourced" development of a classic monkey-coffin style 3-way speaker, and a question was raised on the use of "force to gradient" in the merge process. I've always understood this feature to be more of a "fix" for measurements with low SNR where the windowing function has made for incorrect off-axis response at low frequency, however the follow up question is "what should one expect for DI at low frequency for a monopole speaker?".

In my own measurements, I have not found a monopole speaker to be fully omnidirectional at 100Hz and below, however slightly better than this particular speaker. My speakers designs have used a smaller baffle width, and usually with a port on the back side, so DI approaches zero and even negative value at low frequency.

The speaker being discussed in a wide cabinet with 12" woofer in a closed/sealed cabinet, and measurements completed at 1m distance in-room. The measured data is of good quality, so I did not find the need to use "force to gradient" and the resulting DI at low frequency is around 2.4dB. Of course, if force to gradient is used with 100% monopole selected, the result becomes a DI of 0dB.

I am wondering if you have any insight or opinions on what would be considered the more correct result, this:


Or this:

[Reet]

Minor bug report, tooltip on "expand SPL scale" and "compress SPL scale" at the top of the main window shows 40dB regardless of actual SPL span.

[kimmosto]

I've always understood this feature to be more of a "fix" for measurements with low SNR where the windowing function has made for incorrect off-axis response at low frequency, however the follow up question is "what should one expect for DI at low frequency for a monopole speaker?"...

As far as I recall, the most important reason for adding 'Force to gradient' was unreliable measurements; environment and input noise, possible drifting of D.C. in the mic input and "mathematical" effects of cropped impulse response by the time window. For example ARTA had some issues with input channel opening causing variance to L.F. It's also certain that it's difficult...impossible to measure actual far field in a room due to distance dependencies. Relative difference of delays between 0 and 180 deg measurements and front & rear radiators is bigger to short than long measurement distances. That is one of the reasons for DI > 0.x dB at very L.F.

Common sense tells that unidirectional full space concept with sealed cabinet and front driver is omni at frequency lim->0 Hz. DI > 0 dB above 20 Hz. The same story with front vents.

Boxed with front active driver and rear vent/passive is dipole at frequency lim->0 Hz. DI is slightly negative close to tuning frequency because active cone does not move much and rear vent/passive produces most on the pressure. Below tuning freq. DI returns back to 0 dB - and positive while closing to 0 Hz. We could estimate theoretical minimum level of negative dip in DI at tuning freq. by simulating vent's/passive's response to 180 deg by Diffraction tool or The Edge. This assumes that active cone is silent i.e. system is ideal without any damping. For example D=100mm vent in the center of 400x1000mm rear baffle produces ca. 0.25 dB at 27 Hz to distance of 30 m. That would be the minimum level of dip in DI as negative value i.e. -0.25 dB.

I usually use 'Force to gradient' in Merger so that directivity of full space concept with rear vent is forced to omni at tuning frequency. Pressure drops quite fast below that so dipole effect closing to 0 Hz is insignificant for sound balancing.
Larger sealed and vented full space concepts with front vents are forced to omni at 20 Hz if measurements have obvious errors such as clear level variance to different angles, or DI >> 0 dB @ 20 Hz. Very large; wide and tall constructions are closing to half space concept in practice so forcing to omni for zeroing DI does not make much sense anymore. Large and flat box on solid front wall is acoustically half space, DI >= 3 dB.

You can copy-paste this to the other thread/forum if needed.

[Reet]

Thank you for taking the time, this is very helpful.

[kimmosto]

2.0.133.1 (2026-02-17)

Main

• Improved accuracy of contour lines in Directivity Polarmap with Klippel palette. Cost is decreased performance so Default palette or some other chart type is recommended while adjustments.
• Tooltip of Compress and Expand SPL scale buttons updates after clicks.

[kimmosto]

2.0.133.2 (2026-02-18)

Main

• Fixed svg export of Directivity Polarmap. Bug since yesterday.

[kimmosto]

2.0.133.3 (2026-02-28)

• Performance of minimum phase calculation improved.

SPL Trace

• Impedance response extrapolated below the lowest traced frequency using R + L||C||R electrical equivalent circuit fitted to traced curve at 0 Hz with entered Re, f_min and f_peak. Linear impedance responses without adequate resonance peak are extrapolated as linear, limited with entered Re.
• Minumum-phase response is calculated from traced data before resampling to 5-40kHz 1/48 ppo.

[kimmosto]

2.0.133.4 (2026-03-03)

Main, Enclosure

• Fixed crash when chart with logo is zoomed in with double-click.

This bug has not appeared since the logo addition in 2.0.74.0 (2021-06-11), so .NET 4.8 could work differently in hiding chart grid columns and rows than previous versions.

[Reet]

Those guys at Purifi must be having a laugh...

[kimmosto]

2.0.134.0 (2026-03-21)

Optimizer

• Fixed bugs in Minimum impedance and Maximum gain limitation. Frequency range was limited while optimization to accelerate performance, but finding actual minimum and maximum requires full range calculation.
• Calculated range while optimizing Preference rating depends on selected variables (weights). Optimized range is 100-16000 Hz with NBD weight = 0. NBD weight > 0 needs 84-19200 Hz to cover 1/4 oct. below and above normal range. Start frequency with LFX/LFQ weight > 0 is 10 Hz. Frequency limits are set to text fields when Preference rating is selected.
• Weighting factors for Minimum impedance, Maximum gain, and Ref axis linearity have been increased to make the limits much steeper.

Preference rating

• Allowed two slope targets; ON or LW, and PIR or SP. ON target alone without power averages could be a bit unstable in some cases. Dual target system is also more compatible with Optimizer having two targets.
• 'Show target' checkbox removed. ON/LW slope target is shown in SPL chart. PIR/SP slope target is shown in CTA-2034 chart.

Main

• Frequency response, Frequency response of Driver, Listening window, Group delay, Filter response of Driver and Polar frequency responses exported with unwrapped phase.

[kimmosto]

2.0.135.0 (2026-03-22)

Optimizer

• Added Limit minimum EPDR checkbox.

[witwald]

Hello. I would like to make a small suggestion for an added feature. When I am using the Enclosure Tool, I often work with various high-pass and low-pass filters. I use the Group Delay panel to display the filter gain, but I tend to swicth off the display of the Group delay trace, leaving just the Filter gain trace, which is what I am interested in seeing. Below is an example:

To make the display of filter gain make more sense in this particular use case, would it be possible to change the "Group delay" graph heading to "Filter gain"? I imagine that it should not be very difficult to do that. Ideally, it would also be great if the dB axis was now displayed on the left-hand side of the plot, and the group delay axis was completely omitted in this instance.

If both "Group delay" and "Filter gain" are being displayed simultaneously, would it be possible to change the "Group delay" heading to "Group Delay and Filter Gain"?

If only group delay is being displayed, then would it be possible to completely omit the dB axis that is labelled on the right-hand side?

Taken together, those relatively minor changes would considerably enhance the UX.

Thank you for giving due consideration to my request.

[kimmosto]

^I need to stabilize Y-scale in Frequency domain tab of Impulse response window anyway so I can look Enclosure too.

[Reet]

Hi Kimmo,
Just want to mention that the VituixCAD help page appears to be missing today - 404 Not found.

[kimmosto]

Just want to mention that the VituixCAD help page appears to be missing today - 404 Not found.

Few days ago service provider informed that my site has exceeded maximum data transfer so I temporarily removed user manual to maintain download service without extra payments. Site uses <3 % of maximum size, but data transfer hits the limit.
Negotiations to balance maximum size and data transfer have started. Situation is that customer service did not understand Finnish or was too busy read my questions.
It's also possible to change service provider or move setup and help docs to somewhere else (shared cloud). Less revisions per month might not help because all - thousands of users would download within the same month anyway.

[kimmosto]

Online manual restored. Data transfer limit will be removed today.

[kimmosto]

2.0.135.1 (2026-04-10)

Main

• EASERA .etx file format added to Filter response of Driver -export. Sampling rate is special 79587 Hz to cover internal frequency range. Time samples is fixed 32768 so row count is 16385 including 0 Hz. Unit is 'Mag', Re + Im magnitude of transfer function at driver terminals.

Impulse

• Y-axis maximum and span of chart in Frequency domain tab are maintained until Y-axis unit (SPL/gain) changes to maintain temporary scale settings by user.

Enclosure

• Title of Group delay chart changes based on visible curves.

[kimmosto]

Plan is to update source code for .NET 8.0, convert old practices to new syntax recommendations and rules, and simplify the code. .NET (core) is more modern and significantly faster in some tasks than .NET Framework. Basic work including more than 3000 modifications and code refactoring is already done. It's basically ready for beta testing. This was possible within a week while conversion to Java or something else would require a year or two.

.NET 8.0 will drop Windows 7 and 8 users out. .NET 5.0 would support Win 7 and 8, but I've not yet tested is it possible to step back from 8 down to 5. The latest .NET LTS version is .NET 10 supported until November 14, 2028, and .NET 8.0 is supported until November 10, 2026 so I already have small pressure to go forward.

What do you think?

[Reet]

Windows 7/8 support cycle ended 10 years ago. I'm sure there's plenty of these legacy systems still in service, but it's up to you on what level of support you want to offer to these users. At some point we all need to "get with the times" even if a modern system appears as a lateral or backwards movement.

My personal preference would be a move towards Java or something else that's more OS agnostic in order to provide Linux native installation and the max users too. I see many people moving away from Microsoft these days, however it will still be a while until businesses follow suit. At least, it is starting with some companies reconsidering their Office license agreements.

One suggestion on the Java migration, as much as well all love to hate AI, in some cases it can be employed well to streamline operations, this code conversion may be one place where you could save some time by having code rewritten by a machine model, and then sift through it with a fine comb to ensure its clean and error free.

[kimmosto]

...all love to hate AI, in some cases it can be employed well to streamline operations, this code conversion may be one place where you could save some time by having code rewritten by a machine model, and then sift through it with a fine comb to ensure its clean and error free.

AI is our change to get application to some other language and platform. VCAD already has some algorithm which is optimized by copilot. It has been unable to produce anything complex and working so far. For example, it tried to get "2nd opinion" for minimum phase extraction. It tried about ten times without luck. In every step it explained that "sorry it is wrong, but this new version works, and this principle is used by REW and some others... blaa blaa". But it did not calculate right so I uploaded working source. After that it was happy and praised my solid approach and was able to make it ca. 20 % faster.
I guess it should be some other than MS's AI for converting C# to non-Windows platform 😁

[Reet]

I guess it should be some other than MS's AI for converting C# to non-Windows platform 😁

Probably . I hear Claude does quite well for code generation.

[kimmosto]

2.0.135.2 (2026-04-25)

• This is presumably the last revision of VituixCAD 2.0.
• Auto update modified to download and install VituixCAD 3.0

[kimmosto]

3.0.0.0 (2026-04-26)

• First release of version 3.0.
• Requires Microsoft Windows 10 or 11.
• Requires Microsoft .NET Desktop Runtime 10.0

Convert IR to FR

• Charts' background color follows Dark mode setting after session file is opened.

Enclosure

• Enclosure project can be opened with double click also when 'Auto open' is checked, but project list has just one row.

Auto update from 2 to 3 will be activated in a week or two - after some experience how 3 works in different setups.

[Reet]

I found no problem so far running Version 3.0 on a Windows 11 VM.

[kimmosto]

3.0.0.2 (2026-04-27)

• Built with ApplicationHighDpiMode=DpiUnawareGdiScaled to avoid severe problems with display scaling other than 100 %.

[BartL]

V2.0.130.1

On the option Tab you have the possibility to interpolate off axis responses.
When I turn this feature on or off I got very different results.
I made measurements 0-15-30-45-60 degrees.
What is the correct setting for this option?





I'm confused 🤔

Do you need some adjustments on the "Listing window" settings to match the steps off the measurements angles? (step of 15 degrees in my case)
What is the meaning of user hor & ver?

[kimmosto]

User manual says:

Frequency responses are interpolated between off-axis angles and planes loaded to drivers when Interpolate is checked. Program selects the closest off-axis angle loaded to driver when Interpolate is unchecked. This could cause some errors and discontinuities, but enables simple tests and projects with few responses only in Directivity chart.

Angle density of simulated and visualized off-axis directions is selected with Angle step list box. Available options are 0, 5, 10, 15, 20 and 30 deg. Off-axis angles loaded to drivers are simulated when 0 deg is selected. Initial value is 10 deg.

Generic 2D is configurable option allowing user to specify how plane and angle value are coded in frequency response filename. Plane keywords define how to distinguish between horizontal and vertical axis. Horizontal axis is selected if keyword of vertical plane is not found in the filename. Search from beginning/end defines whether VituixCAD should start parsing angle value from the beginning of the filename (hor +150 drivername.txt) or the end (drivername hor +150.txt). Angle multiplied by defines how angle value is formatted. For example if 1500 represents 15 degrees off-axis, use Angle multiplied by 100. This option is compatible with ARTA, CLIO in 2D mode.

[kimmosto]

Listening window (LW) is specified in CTA-2034 to include -30,-20,-10,0,10,20,30 deg off-axis in horizontal plane and -10,+10 in vertical plane. That is one small reason why you could measure and simulate with 10 deg steps. You should also measure 0-180 deg to simulate CTA-2034 parameters for normal full space design.
VituixCAD is very flexible with interpolations enabled. It does not care much about angle step and coverage, but you could follow recommendations (in measurement instructions) and play with default settings to get more accurate results.

[BartL]

User manual says:

Still confusing 🤔 read it 3 times.
Still don't quiet understand what would be the correct setting in my case 😕.

The above picture is on axis. 0 degree measurement.
Still, toggling the interpolation on/off shows different results.

[Reet]

When driver coordinates are entered, woofer may be 5.6 degrees angle to the listening location centered on-axis with tweeter at 2500mm. With measurements at 10 degree intervals, interpolation would be required to provide an estimate of the result at 5.6 degrees. Without interpolation, it will simply use 10 degree measurement as-is since it is closest to the required angle.

Most will leave interpolation enabled as it is by default, unless you have some high resolution of angular measurements I don't really see any reason not to use interpolated data.

[kimmosto]

'No interpolation' is primarily designed for:
1) Coaxial 2-ways having both drivers at X,Y,Z=0,0,0 mm and R,T=0,0 deg because measured and simulated exit angles match to all off-axis angles.
2) Professionals who can measure in large anechoic at 2-3 m without time window, keeping rotation center (of mic or speaker) at the same position for all drivers. This enables setting all drivers to X,Y,Z=0,0,0 mm in the schematic because all exit angles match between measurement and simulation. Distance from each driver to mic will change while (vertical) rotation - especially with large constructions so there will be some dense phase wrapping in measurement data. That does not harm summing at the mic, but GD cannot be calculated to off-axis angles. Fortunately it's not usually needed.

One small thing in BartL's screen captures is testing with inverted polarity of mid. That method is not valuable while simulation. Phase mismatch is one possible tool to control direction of main axis usually in vertical plane, power balance and timing (reduce...minimize excess group delay). Inverted polarity method can be used as sanity check to compare timing of simulation and actual speaker. Sometimes mic or speaker stand bends causing error to timing and phase responses if designer forgets to double check actual distance from rotation center to microphone before measurement sequence. Required accuracy is just few millimeters if the highest XO frequency > 3 kHz.

So common people like us keep Interpolation checked. Mid-woofer and tweeter of tiny designs can be measured using the same mic position for both drivers. Interpolation does not help much with tiny speakers because error between simulated and measured exit angles is small. Interpolation is best to keep on that you don't forget it off.

With 15 deg angle step while measurements it's safest to simulate with 15 deg angle step to minimize effect of interpolation due to exit angle error. 5 of 5 measurements within 0-60 deg are probably quite close to simulated 0, 15, 30, 45, 60, ... deg.
Options:

3 of 5 measurements (0, 30, 60) would be close to measurement data if simulated angle step is 10 deg.
0 deg setting in Options uses measured angles only.

[kimmosto]

3.0.1.0 (2026-05-05)

• Added File -> Export -> Filter response of All Drivers.

[kimmosto]

3.0.1.1 (2026-05-13)

SPL Trace

• Added image conversion to bitmap to 'Paste image Ctrl+V' command. Image formats that can be automatically converted to bitmap are supported.

This was mysterious because pasting e.g. jpeg was working with 2.0 using .NET Framework, but not anymore with 3.0 using .NET 10. Just a small change was needed to get other than direct bitmaps supported without File - Open image command.