VituixCAD v2

[kimmosto]

If I connect a first order 1000Hz 6dB HS and a first order 1000Hz 6dB LS in series I expect them to give a straight 6dB gain, no frequency dependence.

There are two different specifications for corner frequency of shelving filters. I would call them "analog" compatible and "RBJ digital" compatible. Corner frequency is kinda "-3 dB point" with "analog", and middle gain (in dB) with "RBJ digital". For example Sigma Studio and miniDSP use "RBJ digital", and most of European brands use "analog" specification. The same problem exists with 2nd order shelf and Bessel HP/LP filters. This is really annoying imo too.

So you need to select suitable DSP system in Options window to be compatible with your gear or app. Generic should be compatible with Sigma Studio. Another gear using Sigma project is Platin SM 48k, but it's for 48k sample only. I use that DSP amp quite often.

[bnilsson]

I have a question about low and high shelf filters in VituixCad.
If I connect a first order 1000Hz 6dB HS and a first order 1000Hz 6dB LS in series I expect them to give a straight 6dB gain, no frequency dependence.
This is how such a filter combination behaves in REW, and also in SigmaStudio DSP design software.
However, in VituixCad I get a bump of +2dB at 1000Hz over the +6dB flat. Why?
This behaviour make a design tuned in VituixCad not transferrable to e.g. SigmaStudio, which is annoying.

[bnilsson]

Dear all,

I have a question about the Diffraction Tool.
I have a cylindrical enclosure standing vertical with the driver mounted on top, pointing upwards.
Consequently the orientation is 90 degrees vertically in reference to the listener.
I have selected a listening/microphone distance of 1000mm, which will be in the same plane as the driver mount, sideways to the cylinder axis.
In the layout graph describing the baffle, driver and microphone, where should the microphone be located to conform to this geometry?
Should it be in the middle of the driver, or at the bottom, outside the baffle?

BN

[tktran]

Kimmo,

You are correct that the 8260A was the first coaxial that was made into a Genelec. It was Ilpo's idea but the design was such they could only make 6 a day.

After this they decided to use 2 oval woofers, above and below the centrally mounted large "baffle" 8351A - Genelec.com but to do this, they needed to hire a transducer designer. (Genelec could not design/simulate driver's prior to this) The smaller Ones came out later. 8341A, 8331A, together marketted as The Ones: gen_theones_brochure_210x279_12pager_lowres_0.pdf

The larger 8361A and updated 8351B came later.

But putting subwoofer against a hard surface might be OK for subwoofer egs. <120Hz, but woofers that cross up as high as 320Hz to meet the coaxial.

Well the designers must be listening with their eyes. Or people are measuring incorrectly and not paying attention to Time Domain.

[tktran]

I can. They are good for beginner level. Only problem is he is using single channel measurement technique and Xsim.

As a previous course user, I reach out to him and said, you should update your course for dual channel and using VituixCAD2, and he vehemently agreed. When he has some time...

[tktran]

I asked this question in 2019...

https://www.audiosciencereview.com/forum/index.php?threads/should-we-i-get-into-speaker-testing-measurement.7904/post-191634

Today it's just a tool for marketing using some old science from the 20th century, from Floyd Toole's book to KEF and Genelec products that generate a clean looking spinorama.

They can't even measure a Soundbar, nor make sense of what makes a good one from a bad one. And whenever serious comment or a question is raised, there's a heap of hand waving going on. eg. my recent exchanges with Lars or Jack

[Reet]

I would say yes, but I also know that writing a book is a lot of work, and I already have a full time job. Also, in the effort of keeping with the times, a video series would be much more popular with "internet education" in mind. Again, a lot of work to do it well and cover many topics in depth.

For sale, Marius from Audio judgement sells some courses here:

https://audiojudgement.com/courses.html

I can't vouch for how good they are or the accuracy of information contained within, however.

[kimmosto]

Where should one go to find a consolidated gospel of truth, laid out in a concise progression path like a textbook for a complete understanding of acoustics as it related to loudspeaker design?

I agree that available textbooks and internet sites are not very inclusive, and data has been scattered to many places. The best books for beginners are DIY speaker books such as Dickason's (in English) and Tuomela's (in Finnish). CTA-2034-A standard fills few holes in those because it's just 2+ decades old. For example ER, PIR and SP. Lack of goals/recommendations and best practices is a big problem. Usefulness of old generic basic/filtered studies (by Toole), deep theories (by Beranek&Mellow) and publications in AES Library is quite light for designing imo.

AI can search and summarize, but it could lie smoothly or just quote forum discussions so everything must be questioned and verified.
Maybe we should write something better

[Reet]

You may be surprised how many are oblivious to much of these "general basic things known for four decades", especially in DIY Audio where most are trying to educate themselves via the internet. Most are struggling with basic concepts and interpretation of SP and PIR, let alone group delay, step response, etc. Where should one go to find a consolidated gospel of truth, laid out in a concise progression path like a textbook for a complete understanding of acoustics as it related to loudspeaker design?

[kimmosto]

Relevant: https://www.aes.org/tmpFiles/elib/20250731/17839.pdf

I'm developing an allergy to Toole's documents and Sound Reproduction book. They seem more useless every day. Close to nothing for speaker designing and different concepts. Just general basic things known for four decades, added with indefinite statements that can also be interpreted as false claims.
For example timing / phase distortion in 3rd edition, chapter 4.6, pages 77-78: "...humans are very sensitive to variations in amplitude vs. frequency, and very insensitive to phase shifts. In other words, we do not "hear" waveforms per se, but we are very sensitive to their spectral content."
Everything is left to interpretation. He himself interprets it as not having to worry about the excess group delay - of any length, so the phase response is either systematically omitted from the documents or downplayed so much that gullible reader becomes convinced of its complete insignificance. However, the audibility is easily detectable even with headphones with phase distortion equivalent to some actual speakers+subwoofer. Tested again a few minutes ago.

[Reet]

Relevant: https://www.aes.org/tmpFiles/elib/20250731/17839.pdf

[augerpro]

I don't remember what authoritative source I heard that from, as you said some form of it has bounced around the forums for a long time. I'm almost sure it was in Sound Reproduction, but I just don't remember anymore.

[kimmosto]

augerpro Your Waveguides page contains this image:

I tried to find the reference for "-6dB 100-10khz drop on axis vs power response found by Harman to be most preferred in listening tests". In the beginning also ChatGPT answered that ca. -1 dB/octave slope is recommended, but in the end it didn't find any references, papers or books and page numbers. Some posts in diy audio forums, and general statements by Toole et al. without exact values or ranges. Do you remember where you got that phrase from?

This question is related to slope targets in VituixCAD. My plan is to clarify the ranges and logic between different slope targets in Preference rating window. So far I've been using statistics of measured speakers and concepts, and referred book "Tee itse hifikaiuttimia" by P. Tuomela's including this text: "A suitable power response reduction is typically 5-6 dB between 200 Hz and 12 kHz. An exact guideline value cannot be given, as it is also affected by the free-field response and the directivity of the speaker". Those sentences leave out acoustics, listening setup, "bad" 80s recordings etc. so more tolerance may be needed.

Current calculation formulas for target ranges can be found here:

Slope target limits as C# code for...

[kimmosto]

2.0.128.0 (2025-07-29)

Diffraction

• Added very simple emulation for directivity of a waveguide. Does not emulate change in efficiency and on-axis SPL. 'Waveguide mimic' group includes Enable checkbox, Cutoff in Hz and Slope in dB/oct text boxes. A circle in green dashed line is drawn outside the driver(s) when emulation is enabled.
• Axis Hor and Axis Ver can be non-zero at the same also with rectangular driver.

Poetry in the user manual:

Waveguide mimic
This is a simple emulation for directivity of a waveguide. Emulation does not include change in efficiency and on-axis SPL. There is no response boost (and anyhing for impedance) so result cannot be used for designing equalization. Actual waveguide/horn simulation would be much more complicate requiring all dimensions, shape of cross-section(s) and driver parameters.
Emulation is activated with Enable checkbox in 'Waveguide mimic' group. A circle in green dashed line is drawn outside the driver(s) when emulation has been enabled.
Frequency where directivity starts is adjusted with Cutoff textbox in Hz.
Change in directivity as a function of frequency is adjusted with Slope text box in dB/oct. Value to be adjusted is slope of off-axis response to 90 degrees in half space (infinite baffle without directivity due to baffle edges). Range is from -0.1 to -6.0 dB/oct.
Root response for emulated off-axis responses is transfer function of the 1st order low-pass filter. That is manipulated with Slope parameter and direction (angle) from the driver to virtual mic/listening point. Waveguide is assumed to be circular. Directivity calculation of the driver (as an ideal piston) is limited between -50 and +50 degrees i.e. opening angle of the throat is assumed to be ca. 100 degrees.

[augerpro]

Looks great to me!

[kimmosto]

Test app simplified to circular due to some mathematical challenges

[kimmosto]

Mockup with simple directivity. That could use ca. transfer function of 1st order low-pass at cut-off manipulated by slopes and off-axis angles (up to 90 deg).

[kimmosto]

I wonder if you could create a driver type for the baffle diffraction simulator that would mimic a waveguide? Define a cutoff frequency and directivity slope above that slope.

Yes I could. The simplest solution would be tiny throat (without directivity of the driver) and constant circular directivity slope down to cut-off frequency, but I'm afraid directivity of the driver to small off-axis angles, closer to CD with steeper cut-off, and elliptical, square or rectangular wave guide with different parameters for horizontal and vertical planes will be asked quite soon. This may open Pandora's box

[augerpro]

I wonder if you could create a driver type for the baffle diffraction simulator that would mimic a waveguide? Define a cutoff frequency and directivity slope above that. Currently when prototyping where I intend to use a waveguide on the tweeter, I just use a driver diameter (say, 4") that is similar to the actual waveguide response around the intended crossover frequency so I can understand how the response will combine with the midrange response. But above some frequency the driver response drops off quickly offaxis, which the actual waveguided tweeter does not do.

[kimmosto]

2.0.127.3 (2025-07-16)

Enclosure

• Added 'Equivalent Peak Dissipation Resistance' (EPDR) trace to impedance chart.

[Reet]

I am reminded of a paper from Bohdan Raczynski from a few years back "Perfecting 'Punch' In Your Loudspeakers", here:

https://bodziosoftware.com.au/Perfecting_Punch.pdf

The paper makes no reference to step response however, but there references at the end to a lot of related research which may be of interest.

Complete list of papers and internet sources
1. M.Zaunschirm, J.Reiss and A.Klapuri, A High Quality Sub-Band Approach to Musical
Transient Modification, Computer Music Journal, Volume 36, Number 2, Summer 2012,
pp. 23-36

2. M.Goodwin and C.Avendano, Enhancement of Audio Signals Using Transient Detection
and Modification, AES Convention Paper 6255, October 2004

3. "Detection and modeling of fast attack transients" by Xavier Rodet and Florent Jaillet

4. S.Fenton., B.Fazenda, and J.Wakefield, 'Objective quality measurement of audio using
multiband dynamic range analysis'. Institute of Acoustics (IOA) Conference- November
2009,

5. S.Fenton and J.Wakefield, Objective profiling of perceived punch and clarity in produced
music.132nd Audio Engineering Society Convention, April 2012,

6. Steven Fenton, Hyunkook Lee, and Jonathan Wakefield in "Elicitation and Objective
Grading of 'Punch' Within Produced Music" Convention Paper 9043, Presented at the
136th Convention 2014 April 26–29 Berlin, Germany,

7. Hybrid Multiresolution Analysis Of 'Punch' In Musical Signals. Steven Fenton,
Hyunkook Lee, and Jonathan Wakefield, Convention Paper 9229 Presented at the 138th
Convention 2015 May 7–10 Warsaw, Poland

8. Towards a Perceptual Model Of 'Punch' In Musical Signals, Steven Fenton and
Hyunkook Lee, Convention Paper 9381Presented at the 139th Convention 2015 October
29-November 1 New York, USA

9. Peter Dowset ―Audio Production Tips: Getting the Sound Right at the Source‖, CRC
Press, 538 pages.

10. URL Source: http://www.stereophile.com/reference/706deep/index.html

11. http://music.tutsplus.com/tutorials/...-mixes--audio-
1083

12. http://subaqueousmusic.com/transient...-to-add-punch/

13. http://www.thesoundkeeper.com/top-5-...esign-plugins/

[kimmosto]

It appears that this is not strictly a magnitude response problem, the time domain is just as important.

Possible differences in magnitude responses, directivity, wavefront, compression/dynamic features of the radiators (including ports) and location of the radiators make difficult to evaluate effect of time domain only. Therefore the only controlled method to distinguish timing is to compare the same speaker with different phase linearisation with FIR gear/app: from minimum-phase version via original timing to added excess group delay. Test speakers should be dynamic, preferably closed to have short group delay at LF, and wavefront towards the listener to not limit perceptibility of timing with something else than timing of the source.

The only study and convention paper including close to proper method is "Audibility of Loudspeaker Phase Distortion" by Greenfield and Hawsford, 1990. Test speaker was Celestion SL700 (2-way with XO at 3 kHz) and they listened two music tracks only. That study is not valid to tell anything about 3...5-ways or using subwoofers crossed at mid...high-bass. So insignificance of time domain is just religious cult in "objective" forums such as Audio Science Review.

[Reet]

The most ideal phase and step response with certain magnitude response is minimum-phase. Step response like with Dunlavys.
That is possible to set as a reference overlay by selecting "Total SPL MP" to 'Signal to show' list box and locking Y scale with IR max checkbox. For example this in theoretical limes for 8260A shown below, assuming that speaker would have adequate phase linearisation with FIR convolution:

Thank you, this is quite helpful. It aligns well with my own perception, from previous examples shown, speaker 2 has the most impactful, "punchy" bass, despite a quite modest mid tier driver being used. It appears that this is not strictly a magnitude response problem, the time domain is just as important.

[JonMarsh]

Thank You! I should be looking closer, but I've been working on cabinets and hardware so much lately!

[kimmosto]

I would request a built in Duelund Reference transfer function generator for VituixCAD- you only need two input variables to define it relative to 0 db, but one could always add a scaling function for extra credit!

VituixCAD has two Duelund active 3-way library blocks using generic active blocks. Two parameters: f and a.

It also has two Duelund target transfer functions for LP/HP and BP, defined in Optimizer window. These are designed for passive 3-way speakers. May require some knowledge or testing how f and a should be set:

[kimmosto]

Is an ideal response to have a flat RMS step, or some sort of slope to this curve?

The most ideal phase and step response with certain magnitude response is minimum-phase. Step response like with Dunlavys.
That is possible to set as a reference overlay by selecting "Total SPL MP" to 'Signal to show' list box and locking Y scale with IR max checkbox. For example this in theoretical limes for 8260A shown below, assuming that speaker would have adequate phase linearisation with FIR convolution:

Reality with early production series and original GLM is quite much lower maximum pressure and RMS compared to minimum-phase:

Perceivability of pressure drop with ears and body is more complex topic. As far as I know, there are no valid studies and convention papers available for that. Perceivability depends on many other things too so it's difficult to define minimum group delays for speakers without FIR gear/app. Designers are forced to use their own experience or design principle without requirement for any psycho-acoustic science. Minimizing excess group delay and avoiding subsonic filter and subwoofers is quite safe way, but for example pro audio sector could have some higher priorities such as total latency and durability which could prevent from aiming for zero excess group delay and short GD at LF.

[JonMarsh]

To be sure everyone understands, just kidding...
But it would be useful, to me.

[JonMarsh]

Now, if Kimmo doesn't have enough to do, and is looking for a new extra credit homework assignment, I would request a built in Duelund Reference transfer function generator for VituixCAD- you only need two input variables to define it relative to 0 db, but one could always add a scaling function for extra credit!

[JonMarsh]

Generating a transfer function set and exporting these to a text file, they can be imported into VituixCAD to use as response target references.

[JonMarsh]

And, for anyone interested, a quick reminder of how Duelund filters work...


the individual LP, MP, HP equations and total transfer function, based on a defined center frequency and coefficient aleph:




MathCAD plots for different coefficients of aleph:





Phase examples for those versions of aleph:





For an aleph of 1.414 or less, the transfer function devolves to a two way.

[JonMarsh]

OK, it's not a slow Saturday morning here, but this is a long time favorite topic of mine, so I thought I'd drop by with a hint or two, or rather, observation and passing explanation for commercial loudspeakers.


Let's consider the measured step response of two commercially available loudspeakers that are fairly well known (well, in some quarters). The Magico M2, and the Vimberg Mino. Both are three way speakers. Vimberg is sort of a subsidiary of Tidal Audio as I understand it, and I love the esthetics and design choices, and they were an influence in the Ardent D project.

So lets look at the step response, courtesy of online reviewers


Magico M2



Scuttlebut says the Magico's use a multi-way LR4 crossover.



Vimberg Mino






Scuttlebut and Stereophile review describes a crossover that sounds an awful lot like a Duelund variation- certainly possible to achieve with the drivers chosen.





A key requirement is having the effective acoustical Z axis origin in the same plane, and the Accuton aluminum sandwich woofers do a good job of meeting that requirement. These appear to be customized versions of the AS168-9-470.

[Reet]

How best to interpret "RMS of Step after pre-delay"? Is an ideal response to have a flat RMS step, or some sort of slope to this curve?
I have 3 examples below of 3-way tower speakers, each with quite different unique features. Which is okay, great, or awful as far as the step response is concerned?
Speaker 1:


Speaker 2:


Speaker 3:

[kimmosto]

2.0.127.2 (2025-07-08)

Main

• Added 'Logo height' text box to Options window. Height is defined as percent of chart area height. Range is 5-25 %.
  So logo file can be HD, and program scales it down to specified height.

Impulse

• RMS checkbox updated if visibility changed in Traces... window.

M-S Defender has started to nag and block/delete VituixCAD2.exe with few users. May be due to new obfuscator version: Eziriz Intellilock 3.5.0.0. VirusTotal reports Trojan:Win32/Wacatac.B!ml with Microsoft's scanner. No alerts in my PCs so this could be false alarm.

[kimmosto]

any details about the specific speaker?

The 1st Ones. Measurement was done in 2012 -11. I don't know if the excess delay was reduced (about a decade) later by updating the firmware and GLM.

In current models such as this Ones, the issue has been fixed by linearizing phase within allowed latency limits to input signal. That doesn't produce perfect minimum-phase response even in "extended linearisation mode", but it's better than the older active 3-way models. Transients also sound stronger and soundstage is more logical than earlier, but subwoofer can/will still damage timing at bass.

[JonMarsh]

Great example... any details about the specific speaker?

[kimmosto]

Once again my favorite timing example with new 'RMS of Step after pre-delay' traces.
Actual loudspeaker, flawless magnitude responses and directivity, but otherwise 😭 Very confusing experience to listen speaker with so much right, but still too much wrong.

[kimmosto]

They should really drop "Science" from the name IMO.

Harman's research (R.I.P) focused on the most significant and easily audible parameters to conquer the low- and mid-priced consumer market. Basic sh*t for everyone except hi-fi enthusiasts. At least that would explain why they filtered out almost everything else than parameters based on magnitude response variations. The result of this work was generalized to "listener preference" which was chosen as the law book and the bible of loudspeaker reviews on ASR. Few other subjectively selected parameters are measured, and the rest have been arbitrarily discarded though errors can be objectively radical. That shouldn't affect to loudspeaker designers, but at least some other manufacturers have made mistakes by trusting narrow and/or invalid old studies and simplifications by Harman. Some may optimize their products to succeed in certain "objective" reviews and measurements such as CTA-2034 by Klippel NFS.

I don't want to support forum acting like that so my collaboration with VituixCAD has ended. Next step would be ending public releases so I hope ASR begins to show development and openness to clearly measurable errors even if they are not within Harman's consumer-level preference.

[Reet]

I don't find there's much DIY effort going on there regardless, the DIY section of the forum is quite barren compared to the rest of the forum. They should really drop "Science" from the name IMO.

[kimmosto]

You may need a legal team to enforce that one, but I commend your efforts to thwart misuse.

I don't try to enforce. Just recommending some other forum such as diyaudio or HTguide especially for loudspeaker DIY. ASR is too toxic, religiously limited, hopeless, authoritarian and nowadays commercially connected too so DIY community could exit for good.

[Reet]

"You may not publish any material produced or processed with VituixCAD on Audio Science Review (ASR) Forum"
You may need a legal team to enforce that one, but I commend your efforts to thwart misuse.

[kimmosto]

dB/octave slope of a curve

Tooltip indicating X, Y and Y2 values and name in a tiny borderless window at mouse cursor is global constant feature of chart class i.e. the same code serves all line charts in the package.
I'm not very keen to make additional custom code for some individual chart.

driver impedance using a logarithmic scale

Chart class does not support logarithmic Y scales. Logarithmic scale has been programmed for frequency X axis only. Of course dB scale is kinda logarithmic, but the values are saved as linear in dB.
Logarithmic Ohm scale has always been very annoying imo. "Certain" inductance behavior can be detected with single series coil (Ls) though unrealistic clean Le is quite recognizablewithout log scale anyway.

[kimmosto]

2.0.127.1 (2025-07-05)

• License updated.

Main

• Crossover schematic dimensions extended from 200x140 to 400x300 coordinate units.

Impulse

• Added RMS checkbox. Shows quadratic mean of step response after pre-delay.
• IR view tab and chart renamed to Time domain.
• FR view tab and chart renamed to Frequency domain.

[witwald]

For your consideration, I would like suggest a new feature for inclusion in VituixCAD's Enclosure simulation tool.

Would it be possible to have an option to display the magnitude of the driver impedance using a logarithmic scale? This would make certain log-linear behaviour characteristics much easier to identify.

[kimmosto]

2.0.127.0 (2025-06-21)

• Added 'Export data' button to Traces window. Exports data points of all or selected traces to txt/csv file and clipboard.

Power dissipation

• 'Export chart data' command removed from context menu of chart.

Enclosure

• 'Export six-pack data' command removed from context menu of charts.

[kimmosto]

2.0.126.1 (2025-06-18)

• In frequency response smoothing, the imaginary and real parts of the phase response are smoothed separately without phase unwrapping.
  Magnitude smoothing has not been modified.

Main

• Frequency axis maximum value updated on XO variant change.