VituixCAD v2

[witwald]

Thank you for that. It's much appreciated.

[witwald]

I'm at a loss to understand why someone would want to use HFD as a measurement tool, it's quite bad.

Please note that I am NOT using HFD as a measurement tool. I inadvertently did one rapid frequency sweep without any data capture, so that's the total of my "experience" using HFD as a measurement tool. It sure startled me!

Some software purpose built for measuring like REW or ARTA would be a better solution,...

Not to forget some other measurement systems, such as CLIO, IMP/M, Audiosuite, MLSSA, B&K Pulse, etc.

...where active filter design can be completed by VituixCAD, and filter blocks copied over to HFD simply to push them to the DSP.

In a roundabout way, that is something that I am trying to do, as VituixCAD has much better overall simulation capabilities (of course) than HFD. Funnily enough, I have encountered some major issues using HFD, as it routinely produces "Access violation errors" when I try to read in the text files containing impulse response data. This error seems to corrupt things within HFD's memory space, sometimes in a way that the simulated curves in HFD are modified. The biggest problem that I have encountered is that when I load a tweeter impulse response file into HFD, the frequency response is shifted down in frequency for some unknown reason.

That's where VituixCAD comes to my rescue, allowing me to perform the simulations that I need to be able to work with.

[kimmosto]

My wild guess is that the files are response data for some project originally designed for Hypex DSP, and witwald wants to return it to simulator for possible tuning or conversion for some other DSP gear. The first option wouldn't add much value because VituixCAD is not able to show or do much more than HFD so it must be conversion...🤔

[Reet]

I'm at a loss to understand why someone would want to use HFD as a measurement tool, it's quite bad. Some software purpose built for measuring like REW or ARTA would be a better solution, where active filter design can be completed by VituixCAD, and filter blocks copied over to HFD simply to push them to the DSP.

[kimmosto]

...it would greatly promote interoperability.

VituixCAD does not officially support Hypex due to proprietary biquad coefficient calculation of HFD i.e. simulated responses won't be exactly the same as produced by Hypex DSP. I would expect some co-operation and interoperability promotion also from Hypex.
Anyway, any sample rate between 44100...768000 Hz will be allowed in the next revision, though VituixCAD's main tasks with IR files are:
a) Read raw measurement data, typical sample rates divisible by 22050 or 24000 Hz.
b) Produce/export processed frequency response files or IR files with any - also Hypex's odd sample rates.

[witwald]

I have been trying to load impulse response files written in the MLSSA format into the "Convert IR to FR" tool. When I use a file whose time step between samples is 0.010666666666667 milliseconds, corresponding to a Hypex DSP-compatible sampling rate of 93750Hz, I get an error message: "Samplerate 93750 Hz is not supported".

The first few lines of my file are as shown below:

0
0.010666666666667
65536
-2.082129E+02
3.694307E+03
1.558122E+04
1.405175E+04
-6.690195E+02
:

Is this a bug, or is there a way that I can use my impulse response file?

[witwald]

I am trying to use VituixCAD's "Convert IR to FR" capability. I have an impulse response text file containing a sequential list of 65536 sampled floating-point values, where the sampling rate used was 93750Hz. This sampling rate is used by the Hypex DSP-enabled plate amplifiers. A snippet of the data at the beginning and end of the file is shown below:

-208.21291788177
3694.30732619709
15581.2232067774
14051.7458438195
-669.019517933816
-4681.39384516699
-1202.73093352386
   :
   :
1.65295396072906E-05
-2.75267613005781E-05
2.66568804363487E-06
2.76369109997177E-05
0

When I try to read in the file, I use the "Impulse response files (*.pir,*.mls,*.crp,*.txt,*.wav)" option:



After selecting the file and clicking Open to load it, I get the following error message:



Then I try to select the sample rate of 93750Hz, which is available from the drop-down list:



When I click OK, I then get the following error message:



Would it be possible to support the 93750Hz sample rate, as well as impulse response files that don't have a header? These are in the format used by the Hypex Filter Design program, and if VituixCAD was able to support that format it would greatly promote interoperability.

[witwald]

Thank you for the clarification. With 48 points per octave, starting at 5Hz, that makes the last point in the 5Hz–40000Hz range set to be equal to 5*2^(12+46/48) = 39793.948 Hz, which matches the exported data. I now much better understand how those frequency settings work.

[kimmosto]

All of the above seems to be at odds with having the ability to set the frequency limits using the Options dialog box.

User manual says:
"Frequency axis
Internal frequency range is fixed 5...39794 Hz with density of 48 points/octave, but you can define initial visible scale with minimum 5...5000 Hz and maximum 40...40000 Hz."
So graph scale settings in Options don't limit anything. Just set initial values user can adjust temporarily if needed.

The reason is that limited frequency range is one the most poisonous features in input and output data. Range should be as wide as possible at both ends that VCAD and other applications should not have to extrapolate. Extrapolation is possible if target system has wider range, but it requires conversion to time domain, resampling to sample rate of max freq. x 2 e.g. with some interpolation for HF and assumptions or window function for LF and finally conversion back to freq. domain.
It's not rare that the same data is used for different projects e.g. analog XO and 48k DSP XO. All input data is better to be 5-40k to be compatible with any project.

Of course there is one exception: projects with DSP XO sample rate < 80 kHz . Data is displayed and exported up to Nyquist.

[witwald]

I've been doing some simulations, and I've noticed that I can't use the cursor to read off a point at 20000Hz, even though the frequency limits are set to 5Hz and 20000Hz. Is this by design? See the screen snapshot below for an example.

Interestingly, when I use the cursor to investigate the value at 5Hz, I can do so. Also, if I change the high-frequency limit to <=10000Hz, then I can get the final value.

Also, I have noted that, when results are exported to a text file, the data always covers the range from 5Hz to 39793.9Hz, irrespective of the frequency axis limits that have been set. An example of this behaviour is shown in the following snippet of exported data shown below.

Freq[Hz] dBSPL Phase[Deg]
5.00000 -54.079 -7.161
5.07273 -54.081 -7.264
5.14651 -54.084 -7.369
5.22137 -54.087 -7.474
:
:
38106.8 -71.176 -81.846
38661.1 -71.299 -81.956
39223.4 -71.422 -82.064
39793.9 -71.546 -82.171

All of the above seems to be at odds with having the ability to set the frequency limits using the Options dialog box.

[kimmosto]

2.0.115.0 (2024-07-07)
Main

* Added 'Gray palette' option for directivity Polar map. Activated from context menu of directivity chart -> Contour lines -> Gray palette.
* Default palette' command in context menu restores standard colour circle palette.

Preference rating

* Removed radio buttons for Olive's equations 9 and 10.
* Added Sound pressure slope target SL_SP.
* Added 'Show target' checkbox. Selected target is shown in Power & DI chart.
* Slope targets displayed also in dB/dec (in addition to dB/oct).
* Added 'Full space' button. Initializes slope targets for conventional (boxed) design.
* Added 'Half space' button. Initializes slope targets for flush-mounted design.
* Added 'Constant DI' button. Initializes slope targets for constant directivity design.
* SM and NBD are calculated as slope-independent with all parameters.

[kimmosto]

I am a bit confused why the "with sub" check box has turned into a "ignore LF" button.

I thought button would work but it did not because preference max. of 10.0 requires that LFX x it's weight and LFQ x it's weight should be constants > 0 in the equation to produce certain effect/gain to different variables. So 'with sub' is now restored as it was.

Another change is that Olive buttons are removed. Olive's factors can be set by pressing Ctrl or Shift while clicking 'Full space' but that could stay hidden/secret feature.

Default initial factors may require some tuning to prefer minimum NBD_PIR over maximum SM_PIR i.e. slightly curved PIR is allowed but without narrow resonances visible to many directions. Both can be good at the same time, but real life produces some errors and differences.

[Reet]

I like the idea of displaying the slope target in the main window, it makes it easier to visualize and set the target appropriately. I am a bit confused why the "with sub" check box has turned into a "ignore LF" button. I like Ignore LF as a better description, but why the change to a button? In any case, I still primarily rely on the axial and in-room response optimization over the preference rating tool. Once I find a good result using axial and in-room response at 50/50 weight, if I try to re-optimize using preference rating, any change in the result has been very minor.

[kimmosto]

Sketching PR window...
Slope target is possible to show in Power & DI. Cursor in SL_PIR field or SL_PIR checked alone shows PIR target etc.
Three initial weighting factors for Equation group; default, Olive 9 and Olive 10. Olive presets uncheck slope targets. Factors for Full space with default weights are visible below.
Three initial slope targets; Full space (conventional box), Half space (flush-mounted) and Constant directivity.
Ignore LF resets LFX and LFQ which are >0 with Olive preset.

This may look complex, but should be quite easy to use. Normally just click Full space, select Slope target e.g. check SL_LW and run Optimizer with 'Preference rating'.
Any opinions/comments?

[kimmosto]

Polar map with gray palette is done too, but not much value imo so it could stay hidden reserve.

P.S. Resolution of contour lines is lower due to 8-bits only.

[kimmosto]

I'm cleaning Olive's equations and adding sound power slope target (SL_SP) to PR window because this is design tool. No more comparison to spinorama.org, ASR etc.

Olive's weighting factors would be easy to copy to the fields, but calculated rating is much higher because SM and NBD are insensitive to slope with this custom equation.

[kimmosto]

2.0.114.2 (2024-06-30)

Main

• Unit of group delay traces in GD & Phase chart changed from milliseconds (ms) to cycles (cyc) when 'Normalized' is checked in context menu.

[Reet]

I like it. Limiting the normalization to max 10kHz also allows one to extend the frequency range to 40kHz to show ultrasonic tweeter breakup without negatively impacting the overall chart scaling.

[kimmosto]

2.0.114.1 (2024-06-28)

Main

• Peak level within 20-10000 Hz is normalized to red (dark red with Klippel palette) in directivity Polar map also while 'Normalized' is unchecked.

Enclosure

• Fixed calculation with multiple passive radiators. Bug since version 0.1.4.13 (2016-06-20).

[Reet]

Yes, I would prefer that. As it is, the example you show above will appear very orange/yellow, with a red peak at 15kHz and with no ability to change it.


Another idea would be to provide an option for user adjustment of a single frequency for palette normalization could work well enough for most people, complete speaker or individual driver evaluation. For a complete speaker, something like 1kHz could be used, then user adjusts to perhaps 5kHz for a tweeter only chart, or 200Hz for a woofer chart. The software could provide automatic peak detection to set the normalization frequency for most use cases, with user ability to adjust to a non-peak frequency if desired.

[kimmosto]

For myself, these cases are usually at the frequency extremes, so perhaps peak normalization from 200-8kHz would work well, and peak response outside of that frequency range can be shown towards white spectrum.

Peaking of top octave was probably the reason why normalisation was not originally done, but negative side effect is that we need to adjust generator's voltage to get mid-range close red / reference level (any major grid line).

The cheapest trick to ignore top octave is to look graphs within 20-10000 Hz. Peak is scanned within visible frequency range so HF peak alone does not drop mid-range. But that is also extra temporary adjustment such as generator's voltage.
Fixed range for normalisation is possible and very easy to do. One possible limit is that one could want to evaluate single HF or LF driver such as super-tweeter or sub-woofer without normalisation to reference angle.

For example peak scanning 40...10000 Hz with (equalized) huge peak at the top octave. I could probably live with that...

[Reet]

I like the peak amplitude normalization for the polar map. This is beneficial for the contours, as it ensures that 3dB contour steps are in 3dB increments from peak 0deg amplitude, where previous it appeared arbitrary. It also allows for adjustment on y axis of SPL and power & DI chart without making the polar map look out of place. Thanks!

The edge case would be some instance where some resonant mode or rising response should be shown as "above" the normalized value, which would normally appear white. White colour at the top of the palette however will never occur in this instance, instead the remaining response will be shifted down the scale. I'm not sure how that would be best addressed. For myself, these cases are usually at the frequency extremes, so perhaps peak normalization from 200-8kHz would work well, and peak response outside of that frequency range can be shown towards white spectrum.

[kimmosto]

2.0.114.0 (2024-06-25)

Main

• Added 'Klippel palette' option for directivity Polar map. Activated from context menu of directivity chart -> Contour lines -> Klippel palette.
• Peak level is normalized to red (dark red with Klippel palette) also while 'Normalized' is unchecked in context menu. Colors stay constant regardless of sensitivity and SPL max setting of SPL and Power & DI charts.
• Aspect ratio of directivity Polar map follows W x H in px setting in Options. Height is not forced with dB/dec setting though selected.

Main, Options

• Added 'CTA-2034-A power weights' checkbox to 'Power response & DI calculation' group. Power response is calculated using angle weighting factors with constant angle step of 10 degrees. Factors for angles between 10 deg steps specified in the table are calculated with spline interpolation.
  Note 1: Weighting factors are calculated with traditional integral sine if 'CTA-2034-A power weights' is unchecked or angle step of simulated off-axis angles is not constant.
  Note 2: CTA-2034-A table with 5 deg angle step is ignored (due to bugs) though simulated angle step is 5 deg or less.

Impulse response

• Initial filename of exported impulse response files does not contain crossover variant (var1, var2, ...) anymore.

Normal polar map palette.

Klippel palette. Contour lines are less transparent than earlier to be more visible with darker red.


CTA-2034-A power weights vs. default power calculation using integral sine. Difference is almost invisible.

[mv8]

Poll. Are you happy with XO variant S and R buttons - how they work? E.g. do you prefer that program would not save variants automatically when jumping to some other variant with R button i.e. clicking correct or any S button is required before jumping to some other variant? (project with one variant only would not require S button to save the project).

Single company is asking that kind of functionality, but I have quite high threshold to do changes to existing logic. I think variants are like pages in document editor. S button creates a copy of current page, and R button just changes current page without requirement to click or confirm anything else. Only problem at the moment is that Undo buffer is cleared when variant is changed (to save memory and simplify code).
For example LspCAD has different functionality, but it's difficult and illogical imo.

Hi Kimmo.
The ability to save slots with different filter variations is very useful, but one thing I find really annoying.

When designing filters, I use EQ APO at the same time. When exporting IR files for each driver, VituixCAD writes the individual Variant for that Var cell in its name. This is a big inconvenience for me, because every time I export a different filter scheme from a different Variant cell, I have to manually rename the files in the configuration for EQ APO.

If it were possible to make it so that the IR export saves the same file name, without the Var ending, then it would be possible to switch and export between crossover filter variants without editing the configuration in EQ APO, and this gives the possibility to listen to the saved filter cells in almost real time.

[kimmosto]

2.0.113.3 (2024-05-18)

Main, Crossover tab

• Component/parameter value stepping modified.
  LCR values using 5%/E12/E24/E48 snap are detected from unit instead of component category=passive.
  Added more stepping options with Ctrl and Shift key.
  Snap group not disabled while modifying other than passive components/blocks.

Main, Tune block window

• Added 'LCR snap' group: No, E12, E24 or E48.
  Selection saved to user settings.
• Block attribute value stepping modified.
  LCR values using selected snap are detected from unit.
  Added more stepping options with Ctrl and Shift key.
• Calculated LCR values of inserted blocks are using snap selected in Tune block window.
  Note! Optimized LCR values are still using snap selected in Optimizer window.

Enclosure

• 'Copy as xml', 'Delete rows', 'Duplicate rows' and 'Hide rows' commands disabled if whole rows are not selected.

[Nil L]

Poll. Are you happy with XO variant S and R buttons - how they work?

Yes.

[HTDT89]

Poll. Are you happy with XO variant S and R buttons - how they work?

Generally, yes.

re: auto-save, an optional warning about losing changes (like opening a different project without saving) probably would be easiest way to address the customer request. And after seeing it 21 times, user auto-saves before switching, right?

A visual hint that unsaved changes exist might help, e.g. R2*.

Not important, but I wish XO variants behaved more like browser tabs. Ctrl+PgUp/PgDn and Ctrl+1..8 to switch would be great.

[draki]

No issue with the current form. Having simply accepted as it is, after some practical use it becomes a natural action.

[Reet]

File menu, save project as ...

[JMB13]

I have found these buttons a little confusing at times. I will occasionally bring up one option and work with it and then click on another option and lose the original option of the first. I wonder if there is a way to have it in a cache so that one can work with an option and return to the original after doing so? Maybe a manual save option to temporary back-up cache that can be recalled after "playing" with alternatives? Just thinking out loud here. Thanks so much, Jay

[Reet]

I've no real issue either way. Many times when creating a new variant I want to start fresh, so the only extra tedious task is to select all and delete if I didn't want a copy of the existing crossover variant. Very minor inconvenience not worth talking about IMO.

The improvement I would have in mind would be to leave the functionality of S and R buttons as-is, but pre-populate all 8 variants with blank crossover pages, so you can start fresh if desired by simply recalling an unused variant, or S button can still be used for copy function. Alternative would be remove S button completely in this case, and user can copy/paste if desired.

If the intention would be for the user to hit the S button to save a particular variant before recalling another, otherwise changes made are lost, I am not in favour of that. It would require additional prompt "would you like to save changes?" on every variant change, which I would mostly find annoying.

[kimmosto]

Poll. Are you happy with XO variant S and R buttons - how they work? E.g. do you prefer that program would not save variants automatically when jumping to some other variant with R button i.e. clicking correct or any S button is required before jumping to some other variant? (project with one variant only would not require S button to save the project).

Single company is asking that kind of functionality, but I have quite high threshold to do changes to existing logic. I think variants are like pages in document editor. S button creates a copy of current page, and R button just changes current page without requirement to click or confirm anything else. Only problem at the moment is that Undo buffer is cleared when variant is changed (to save memory and simplify code).
For example LspCAD has different functionality, but it's difficult and illogical imo.

[kimmosto]

2.0.113.0 (2024-04-13)

Power dissipation

• Added Ptot in watts i.e. total power dissipation to tooltip of driver, resistor, coil and capacitor traces. Calculated as power average within simulated frequency range of 5-40k; sum of power values divided by number of simulated frequency points. Power should be selected in View group to show Ptot.
• M-Noise peak response updated.
• M-Noise RMS response updated to produce crest factor and power dissipation more compatible with current Music-Noise.wav files.

Main, Options

• Added 'Generic -a12' to DSP list. Same as 'Generic', but biquad coefficients a1 and a2 are multiplied by -1.
• Added sample rates: 88200, 93750, 352800, 384000, 705600, 768000 Hz.

Auxiliary

• Added sample rates: 88200, 93750, 352800, 384000, 705600, 768000 Hz.

Convert IR to FR

• Added sample rates: 93750, 352800, 384000, 705600, 768000 Hz.
• ETC smooth maximum increased from 8 to 32 samples.

Enclosure

• 'DSP system' combo box removed from Filter tab. Transfer functions are calculated using DSP settings in the main program.

Impulse response

• Added sample rates: 352800, 384000, 705600, 768000 Hz.
• ETC smooth maximum increased from 8 to 32 samples.

[kimmosto]

As a feature request, I think it would be nice to have an option for log scaling of Y axis for impedance charts.

Chart class has nothing ready for log Y scale so that wouldn't be too easy fast.

[Reet]

As a feature request, I think it would be nice to have an option for log scaling of Y axis for impedance charts.

[kimmosto]

Here is my comparison to LspCAD 6.65 Pro. Seas W21EX002 in 35.1 liters, BR, port 7x26.6 cm. About the same Q-factors etc.

Original


Re changed 6.4 -> 3.2 Ohms, BL changed 9 -> 6.505, without autocompute or crosscalc.


Added external resistance of 3.2 Ohms


Inductance simulation should be on in LspCAD 6 to get correct result


Both programs behave as expected with external network.

[Snickers]

I don't understand why you play with parameters and mix them with external circuit.

Not sure I would go into too much detail on that, but it does indicate that something strange is going on. I consider doing an identical simulation in LspCAD for comparison.

[kimmosto]

I don't understand why you play with parameters and mix them with external circuit. The best method to verify is calculation okay with external passive filter components (Cs, Ls and Rs) is to export SPL and Z responses without filter components, load responses to Drivers tab in the main, build test network to Crossover tab, export Driver's frequency response, load it as overlay to SPL chart in Enclosure and finally set up the same filter components and values to Filter tab. Result is exact match at the moment so calculation is correct at least with single driver. Two examples.
Rs=10 Ohms:

Cs=470uF, Ls=2.2mH (0.1 Ohms), Rs=5 Ohms:

Some old versions handled Rs as addition to Re in some calculations, but not anymore. Rs is pure external component which drops voltage at driver terminals Ze/(Rs+Ze), calculated with complex numbers. Very simple and correct.
Crosscalc prioritizes fundamental parameters. Locking makes possible to create impossible combination of overlapped parameters so it's highly unlikely that I would add it.
Row duplication works though it does not inform if whole rows was not selected. Rows should be selected using row header (gray cells on the left).

[Snickers]

I may have found another bug.



So I put a driver in a BR enclosure. Then I modofied the driver to have 2 ohms less Re, and adjusted the Bl back to its original value to see the values for Qts and Qes be updated. Then I put a 2 ohm resistor in series with the driver, so the total serial resistance is back to its original value. This is like taking just 2 ohms of Re and move it just outside the coil. It should give the same response. We should not see a dip like that at 160Hz, neither a raise in pass band sensitivity like this as the circuit is the same.

Either, I have done a huge mistake, or there is a pretty significant bug somewhere in here. I do believe we got a different result in the previos version (where Rg was just next to the input voltage), but my memory may also be playing with me a bit.

While testing this, I also found it hard to clone a driver. There are some menu choices for this, but they did not seem to work. I also find it hard to alter parameters using cross calc as they can not be locked, so I need to figure out in which order they are altered so I can change them back manually.

[JonMarsh]

The Earthworks M30 is a great mic in my experience, highly recommended- of course, it should be, considering the price.

For enclosure interior measurements, I strongly recommend checking detailed specs, and proceeding cautiously with drive levels, as SPL in cabinet is high. Mic's like the ECM8000 and Dayton EMM-6 have maximum working levels around 120-125 dB.

[kimmosto]

Did you receive my email?

Yes, sorry about delay. I have CLIO MIC-01, Earthworks M30, Sonarworks Xref 20 and probably still one very old but unreliable ECM8000. I don't usually do own distortion measurements so no truths or recommendations to share. I would probably try Earthworks M23. My M30 individual does not require calibration. It should tolerate quite high pressure, but this is just guessing.

[tktran]

Hello Kimmo,

Did you receive my email?

[kimmosto]

how are your adventures with Java going?

Not well. No progress since 2023-08-09.

[Reet]

Thanks Kimmo!

By the way, how are your adventures with Java going?

[kimmosto]

2.0.112.0 (2024-03-24)

Merger

• Added 'Delay mm' column to LF response list for travel difference to main/cone radiator on front baffle at typical listening distance (2-4 m) on-axis.
• Directivity to 90 deg is added to LF radiators located on other than front panel i.e. BS (baffle step) checkbox is unchecked. Directivity is simulated for ideal circular piston with 'Diam mm'. Support for rectangular surfaces might be added later.