VituixCAD v2

[Reet]

That's quite the download count. I didn't think loudspeaker design could be so popular . If I were to dissect the stats a bit, can I assume that the download count of 85775 is not unique downloads, but includes all individual download. That is, downloads for every user for every update as well, so user like myself that updates the software with every change might have 50 downloads throughout the year.

[kimmosto]

What's the grand plan for 2022?

To drop weight few kgs.

Have you been interviewed?

Yes, both employers before hiring, in 1988 and 1995.

Where can we learn more about you?

Nowhere. My person is not important.

[kimmosto]

Added PIR weight LW, ER and SP text boxes to Frequency responses group. Enables deviating from PIR calculation of CTA-2034-A standard.
Bigger weight e.g. 40% is recommended for LW while optimizing with Preference rating Equation 9 to reduce error in axial responses.

Example about preference rating optimized with equation 9 and non-standard PIR weights: LW=40%, ER=30%, SP=30%. Target SL_LW=-0.25. Woofer's LP is fixed so that XO is ca. 2.9 kHz, but otherwise iterated automatically. Slope is eliminated from NBD so rating is not comparable, but that's irrelevant while designing speaker.

[kimmosto]

It's Taipuu Speakers Kero Active with FIR preset

[tktran]

Kimmo,

Some years ago a fellow by the nickname of Pjay had a website www.helarc.com
At some stage he interviewed various gurus eg. Loudspeaker, amplifier designers etc and asked them interesting questions.
The interview questions were something like:

What makes a good loudspeaker?
What don’t we know, about making good loudspeakers?
What advances do you think the next decade would bring?
What is the most important decisions you have to make when you are designing?
Are you a vinyl person, or a disc person?
When you are at a party, and someone finds out that you design speakers; what do often get asked?

I don’t remember the exact questions.
Some answers were very insightful.
Some were fun(ny).
But all were enjoyable to read, and explained that designers go through all kinds of challenges/compromises when designing.

Sadly, I cannot find them via the Way Back Machine.
Someone with a sticky memory like Wolf might recall, or interviewees like John Krevskovsky.

Re: Taipuu speakers- Very nice measurements!

Have these gone to market?
What FIR capable DSP are you using?

[tktran]

Deleted

[kimmosto]

Rev 2.0.81.0 (2022-01-07)

General
* Added crosshair cursor to Line charts (except Merger tool). Cursor is activated by clicking bold curve.
* Value display close to mouse cursor shows Y value of selected/bold curve at pointer's X.

Impulse response
* Added unit to left Y scale, Pa with SPL and V with transfer function. Was unitless -1.0...+1.0.

Options
* Added text boxes for negative hor and ver angles of Listening window to enable asymmetrical LW.
* Crossover variants can have different DSP system and Sample rate.

Preference rating
* Added PIR weights: LW %, ER %, SP % for information.


Quite massive changes so heads up - bug risk alarm.

[kimmosto]

What makes a good loudspeaker?
What don’t we know, about making good loudspeakers?
What advances do you think the next decade would bring?
What is the most important decisions you have to make when you are designing?
Are you a vinyl person, or a disc person?
When you are at a party, and someone finds out that you design speakers; what do often get asked?

20 years on discussion forums has sucked almost everything from me. Philosophies, theories, opinions and experiences wrote too many times already. Just sadness and frustration left. Very difficult to handle all hate, prejudice, trolling, strong beliefs in superficial/slack science, pointless and endless debate about politics, covid, snake oils etc. Not much forums left free from all that sh*t.

Re: Taipuu speakers- Very nice measurements!
Have these gone to market?
What FIR capable DSP are you using?

That's just simulation with measurement data produced according my own instructions so there's some difference to actual IRL.
Kero Active is still in R&D. Few recording/mastering studios are interested to test. Passive version (for home) is in the market.
Taipuu uses FourAudio DSP amps having both IIR and FIR options.

[kimmosto]

2.0.82.0 (2022-01-17)

Preference rating
* Added 'custom' equation including SM_ON, SM_LW, SM_ER and SM_SP with adjustable weight 0-2 for each.
Option is designed for Optimizer.
Initial weighting factors: SM_ON=0.5, SM_LW=1.0, SM_ER=0.75 and SM_SP=1.5.
SM values are calculated within 130-16000 Hz instead of standard 100-16000 Hz.
Maximum rating is 10.
Also this equation requires target for SL_ON or SL_LW (or both) to be stable while optimizing.
* Added SM_ER and slope_ER values to PR table.
* Removed PIR weight information.

Impulse response
* IEEE 32/64-bit wav files are exported without normalization to max 1.0 to include gain differences also when individual IR files are exported with different taps and IR window.

Options
* Removed PIR weight LW, ER and SP text boxes.

[Reet]

Hi Kimmo,
In the instruction for SoundEasy for export of impulse response to VituixCAD, we have a requirement for SoundEasy V24 or later. This was due to a change in SoundEasy for single column impulse response data that follows (roughly) the same format as the standard MLSSA txt file.

I was chatting on another forum with a user of V20 SoundEasy, and he posted an impulse response file. The older file includes 2 columns of data, which causes an error in VituixCAD. It is easily resolved by simply disregarding the first column of data. I wonder if it's worth the effort to allow for the 2 column data file to be imported to VituixCAD in order to support SoundEasy versions prior to v24.

Hopefuilly this link works to the impulse response file so you can see for yourself and use for testing. File is recorded at 96kHz.

https://diy.midwestaudio.club/uploads/editor/ja/c2em4x8fua4s.txt

[kimmosto]

2.0.82.1 (2022-01-18 )

Impulse response
* Left side time window changed to start at tap 0.

Convert IR to FR
* Values are read from column 2 if measurement .txt file contains more than one data column.
* Minimum length of measurement (txt) file changed from 512 to 128 samples.


Luckily reading from column 2 supports also some other programs exporting time in column 1. For example this format: Time[s] Real[Pa] Imag[Pa]

[Reet]

Great stuff, thanks!

[Reet]

I was speaking with a user of Ultimate Equalizer recently, and it was explained that UE is not just a standard DSP software implementation. Rather, it takes a driver response and applies an inverse filter to conform it to a target response within some maximum gain limitations. It allows for precisely accurate acoustic output as far as linear distortion on the single axis that the driver frequency response was generated.

I understand that perfect conformance to a target textbook filter transfer function may not actually be the desired goal for a given driver, however with the recent discussion in DIYAudio of using VituixCAD impulse response features in conjunction with APO EQ, I wonder if there is a possibility for a modern implementation of this "Ultimate Equalizer" concept that can be applied to in-room and power response of a system. What I am thinking of would be something along the lines of an "inverse impulse response" that when run through convolution provides a normalized driver response that can then be filtered/shaped to achieve a desired target response. This process would be an "Auto EQ" of sorts that could replace a string of PEQ bands to shape the response.

Does the concept of a modernized Ultimate Equalizer appeal to anyone?

[kimmosto]

At the moment VituixCAD supports mirror EQ for on-axis and listening window, but ER and SP or PIR is not yet supported so directivity index should be very smooth in order to make very flat ON or LW with FIR. This feature is quite secret inside TF button in Optimizer. Process takes a minute or two so it's again "almost too easy".

User manual is very short: "Target response divided by response to reference angle or listening window can be exported with TF button in top right corner. Response file is loaded into Transfer function file G(f) block currently selected in crossover. This could be helpful while designing FIR filters with exported impulse response files because crossover can include single G(f) block and active buffer for the speaker or each driver/way, and response is shaped automatically to ideal textbook curve with tilt and HP/LP given or detected from total SPL target line."

It's not separately mentioned, but target response should be minimum phase and close to system response at LF if max taps is limited because any EQ at sub range requires huge amount of taps.

[Reet]

That's brilliant! I have run through the process and indeed it was quite easy. At the very least, this process can replicate the functionality of UE completely.

1. Connect driver directly "single line" in the crossover.
2. Using the optimizer window, define the target slope and high pass or low pass function as desired.
3. Press the -> TF button to export the target transfer function.
4. Repeat steps 1-3 for each driver in your system.
5. Connect all drivers with active circuit blocks in the crossover schematic, each driver with it's own G(f) block. For Each G(f) block, load the previously saved transfer function file at the bottom of the window.
6. Add in any additional response shaping features such as active peak/notch as desired to shape the response further.
7. Complete IR export per the previous instruction to export IR and utilize within APO EQ for active processing.

[Reet]

One thing I noticed when playing around with the active peak/notch filters, is that if they are set with a gain of zero, the optimizer will not adjust them from zero. For example, if I place a few PEQ blocks together with gain of 0 set, run the optimizer, nothing will happen with these blocks. I'm not sure this is intended behaviour or not, but I would hope that the gain is adjusted if it is flagged for optimization regardless of its starting value, and the open/short functions can be used to remove the block from the equation if desired.

[kimmosto]

^That feature comes with Nelder-Mead simplex solver algorithm (amoeba) implemented by Rick Sladkey for C#. Algorithm does not include "gain test" and tries new values by multiplying so problem exists with any value close to zero - no matter is it dB or not. dB values are the most annoying because algorithm turns very slow close to 0 dB. It's perfectly normal to go from negative via zero to positive or vice versa at constant speed.
This has been known feature since the beginning, but I can try again to convert at least dB parameters to gain for optimizer. That should be possible with XO because parameters contain unit field. Previous attempts were not good enough.

[kimmosto]

Finally motivated to fix this. Thanks Reet!

2.0.82.2 (2022-01-21)

Optimizer
* Improved optimizing of parameters/attributes which can be smaller or greater than zero such as dB, distance and time values by converting values between min and max to absolute between 0.1 and 10 for the iterator.

[Reet]

Thanks Kimmo! Your description above is a good explanation as well of why I’ve always found the optimization of PEQ blocks to not be nearly as aggressive as passive components. I’ve made a habit of setting PEQ blocks to roughly where I want to apply correction, then running the optimizer to fine tune.

[dwk]

For anyone interested, I tried using VituixCAD on my Macbook Air M1. Used a Windows-ARM VM under Parallels, and it at least started up and allowed me to load and navigate around a reasonably complex project (the Directiva project from ASR). I haven't tried starting a project from the ground up yet to see if there are some pitfalls, but I was pleasantly surprised.

[Reet]

VM is perfectly fine for VituixCAD and many other software. I run VituixCAD in a VM within Linux daily. Wine can work as well, but the VM is a guarantee.

For the M1 you should be able to use REW directly with it, and possibly ARTA within the VM, provided there isn't major issues with the VM passing the hardware through directly to the OS.

[dwk]

Sure, running on an x86 platform in a VM isn't noteworthy. It's specifically that it seems to be OK on a Windows ARM VM on Apple Silicon that might be of interest. I have lots of other x86 hardware to run measurements etc if needed, but I'm trying to make the Macbook Air my daily driver laptop so having Vituix on it would be a big advantage.

[nc535]

I like the TF method of getting G(f) of a mirror filter.

I had been doing it the slightly harder way in the xo schematic with a LPF, G(f), HPF blocks lined up to make the target with G(f)=mirror from calculator. I need the filter IR for my DSP so I export it from the buffer Ax option. To find if I have enough taps, I need to turn the IR back into a G(f). To do that I import it into REW, adjust gain to unity, and then export it. That is the hard part, to the extent that any of this is hard.

Is it possible to convert IR to FR in Vituix? Or can the function be added? Its still needed even when using TF.

Thanks,
Jack

[kimmosto]

convert IR to FR in Vituix?

Convert IR to FR

[JonMarsh]

For anyone interested, I tried using VituixCAD on my Macbook Air M1. Used a Windows-ARM VM under Parallels, and it at least started up and allowed me to load and navigate around a reasonably complex project (the Directiva project from ASR). I haven't tried starting a project from the ground up yet to see if there are some pitfalls, but I was pleasantly surprised.

That's quite interesting! I've been too lazy and busy to get Parallels installed on my late model Mac Pro, but that there are M1 possibilities is very intriguing- have a M1 MBP and a M1 Mini. But for a portable, I think I'd need the screen space the MBP16. I normally run VituixCAD 2 on a 2014 Mac Pro that boots WIN10 for my work CAD programs on a 43" 4K monitor, but I also find it very usable on an LG Gram17- but a smaller screen would be hard to imagine.

[nc535]

I'm shocked and embarrassed. I did RTFM, or try to, but I must have had blinders on. I only asked when I couldn't search the online help and all the while didn't even see the selection at the bottom of the tools menu.

[kimmosto]

Few quite insignificant changes

2.0.82.3 (2022-01-26)

Options, Angle parsing

* 'Generic dual plane' renamed to 'Generic 2D'.
* 'CLIO 3D balloon' renamed to 'CLIO 3D'.
* 'EASE 3D balloon' renamed to 'EASE 3D'.
* 'VACS 3D balloon' renamed to 'VACS 3D'.
* Added 'MF 3D' radio button for V[mmm]H[ppp] file naming. Earlier this was included in 'EASE 3D balloon' option.
* Added 'MF 2D' radio button. Assigns V000H[ppp] to horizontal plane and V[ppp]H000 to vertical plane.
* Swapping of planes enabled also with 'EASE 3D' and 'MF 3D' options.
* Initial filename for parsing test depends on selected file type.

[kimmosto]

2.0.83.0 (2022-01-31)

Main
* Added support for drag&drop phase overlays to Filter chart.
* Driver's target curves are not saved to library blocks.
* Adding comment text to XO uses previous text color.

Preference rating
* Added SM_ERDI, SM_SPDI, slope_ERDI and slope_SPDI to table.
* Added AAD_ON, NBD_ON, NBD_PIR, SM_PIR, LFX and LFQ to custom equation (for optimizer).
* Maximum weighting factors of custom equation increased from 2 to 5 to support the same factors with equation 9 and 10.

Optimizer
* Allowed optimizing of Preference rating without SL_ON and SL_LW. Warning about possible unstability is given, but continuation is possible if optimizing cannot damage slopes.

Convert IR to FR
* Default window function changed from Tukey 0.25 to Tukey 0.50 to reduce overshooting at LF.

[Efalegalo]

Would it be possible to add scaling adjustment to the overlays one pulls into VituixCAD - perhaps under the "Trace" setting on the SPL graph? I constantly go back and forth between modeling xover in VituixCAD and confirming results via ARTA. After taking measurements in ATTA, I then export an FRD file drag it into VituixCAD SPL graph. Next, I adjust the voltage (up or down) in the Crossover tab to scale my simulated response to the level of the imported measured response. Certainly not a big deal if too cumbersome to implement.

[kimmosto]

Would it be possible to add scaling adjustment to the overlays

There is scaling. User manual might help. Move mouse cursor over the overlay (turns bold) and Shift+mouse wheel or Shift+Ctrl+mouse wheel to fine tune.

[Efalegalo]

There is scaling. User manual might help. Move mouse cursor over the overlay (turns bold) and Shift+mouse wheel or Shift+Ctrl+mouse wheel to fine tune.

Brilliant! I've reviewed the manual - but it's been a long time. Maybe it's time to go back and see what other cool features have been added. Apologies for requesting something that's already there - and thank you again for all your effort (and patience).

Donated

[Reet]

Brilliant! I've reviewed the manual - but it's been a long time. Maybe it's time to go back and see what other cool features have been added. Apologies for requesting something that's already there - and thank you again for all your effort (and patience).

Don't feel too bad, I asked just the same question a few weeks back. Personally, from a software design standpoint I'm not a fan of hidden features like this that are only available from keyboard hotkey, and require intimate knowledge of the user manual to know they exist. Keyboard hotkeys should be for "shortcuts" only IMO, providing a quicker method of performing GUI actions for power users. That is, all functions available via keyboard should also be present somewhere in the GUI. For the overlay scaling, I would suggest a column for scaling in the traces window, maybe add a column for delay adjustment in there as well for good measure so both amplitude and phase are user adjustable.

Can't complain too much though, since the UI overall in VituixCAD is very good.

[kimmosto]

Mouse wheel is standard for adjustments in this program, and Alt key is partly reserved by Windows for menus and shortcuts so Ctrl and Shift modifier keys are almost unavoidable but also easy and fast i.e. obvious without need to read user manual imo. Traces window is... clumsy

[kimmosto]

2.0.84.0 (2022-02-10)

Main, Options
* Added 'Save chart overlays to project' checkbox. Check to save and load overlays from project file.
Uncheck to retain existing overlays when project is opened and not save to project file.

Note! Chart overlays are not read from older 'Convert IR to FR' session files due change in format. New format is more packed to avoid huge file size.

2.0.83.2 (2022-02-07)

Main, Optimizer
* Export TF button moved outside Textbook target group to enable export with target file.
* Added Free LF checkbox. Check to export TF with constant gain below frequency range low limit. Uncheck to extrapolate gain below low limit.
* Added Free HF checkbox. Check to export TF with constant gain above frequency range high limit. Uncheck to extrapolate gain above high limit.

2.0.83.1 (2022-02-02)

Convert IR to FR
* ETC window is positioned according to impulse peak.
* ETC length reduced from 32k to 16k samples.

[kimmosto]

Overlay saving/reading is not yet added to Enclosure, Merger and Diffraction projects. Just to Main and Convert IR to FR. Personally I don't give much value for this feature, but project-dependent overlays could be important for some(one).

[Reet]

Just a minor bug report. When "crossover of driver" is selected in the enclosure model align tab, the simulation voltage appears to lock at 2.83V and is not adjustable. I have to select the power button to adjust the driver output in this mode.

[kimmosto]

It's not a bug. Generator and possible amplifier in the crossover feed the signal in that mode.

[kimmosto]

Preliminary version of measurement instructions for CLIO just uploaded to server.
This will be new master document including more generic and post-processing information than earlier versions for ARTA, REW and SoundEasy. So please read and give feedback no matter which program you're using.

[ergo]

Hi Kimmo,

Great work! I've only skimmed through and need to go through with more though, but there are two "footmarks" that might be of use for some that Audiomatica does not properly document. These could cause one to spend a lot of time scratching the head.

[kimmosto]

^Thanks! I will add short comment about length limit though it does not matter if users stay strictly in this context i.e. conversion from IR to FR and export to txt/frd is done with VituixCAD which reads millions of frequency points. Maybe you or someone else has already asked Audiomatica to add frequency points to calibration file, but I can try it too.

It's common and basically good thing that IR files don't include mic calibration. I've never tested is calibration included in .sin files of sinusoidal analysis with mic measurements. Probably not due to same reason.

[kimmosto]

These could cause one to spend a lot of time scratching the head.

I got an answer from Audiomatica. Maximum points in frequency calibration file is documented in user manual. Section 5.4.2 UNITS CONVERSION, GUIDELINES TO REALIZE .CAL FILES. The first line "Note: The maximum number of correcting points allowed is 100." Like this in CLIO 12.55 QC's manual.
That is not changing so longer files need (manual) modification also in the future.

Earthworks has quite short cal file - 55 points or so but some others may have >900 points. Long files are probably exported from measurement program or A/B calculation.

[Reet]

Something has dawned on me from speaking with someone who was completing diffraction investigation. He wanted to observe the diffraction effects of various methods of creating a baffle edge, and at various radius to see the real world effects.

It occurred obvious to me after viewing the data, that as the edge radius is increased, the baffle width is effectively smaller, and diffraction ripples are pushed up in frequency as a result. In VituixCAD diffraction simulation this effect is not apparent, or if it is, it is much smaller than reality. Increasing edge radius in VituixCAD appears to only smooth out the diffraction ripples, which should be a combination of smoothing and decreased effective baffle width. I'm not sure this is something worth changing, as diffraction model is supposed to be used for low frequency behaviour only, so it's just an observation.

[kimmosto]

Effective width is reduced by radius. Edge rounding causes also low passing to diffracted response which changes density of response peaks and dips. Result is that ripples will move towards higher frequencies.

[Reet]

Effective width is reduced by radius. Edge rounding causes also low passing to diffracted response which changes density of response peaks and dips. Result is that ripples will move towards higher frequencies.

[ATTACH=CONFIG]32008[/ATTACH]

I don't recall what I was looking at the other day when I posted, but looking again today I think everything is as expected. The comparison I was making may not have been very good since the real world measurements only had roundover applied to the horizontal edges, and the vertical edge was left as a sharp 90 degree angle. Of course this type of arrangement is not available in VituixCAD, but for all intensive purposes it doesn't need to be either. Sorry, this discussion has been a bit of a waste of time.

Anyway, if anyone is interested in some baffle roundover real-world measurements, David has posted his tests at DIY Audio here:
https://www.diyaudio.com/community/t...8/post-6976150

Davis is continuing his study with the vertical edge now, comparing both roundover and tweeter distance to the vertical edge. I'm sure most of us can predict the result, but he appears to be someone who enjoys learning from real world experience.

[tktran]

By the way Kim, how come your website no longer have all the old projects posted?

It was great fun reading all the different projects you've done in the past.

By the way, any future plans to have VituixCAD interface with any automatic turntables?

Here is one which I feel is reasonable value. I think there's work to scale it up. I understand it would be simple to interface with VituixCad and George is very personable and if you were interested it would make a good upgrade opportunity for serious users. :W

You'll probably to skip the queue that I'm in to purchase one.

https://augene.blogspot.com/2022/03/new-turntable-for-loudspeaker.html