VituixCAD

[DBMandrake]

Hi Kimmo,

I'm trying out the "Convert IR to FR" feature for the first time using ARTA PIR files, (I usually opened the PIR file in ARTA and exported an FRD file from there to use in Vituxicad) and I seem to have found at least one PIR file that causes Vituixcad to freeze when trying to open it.

I've attached it as a zip file. The file opens normally in ARTA. Strangely I was able to open it once but have not been able to open it again.

Other PIR files I have tried which were created at the same time during the same measurement session in ARTA (using the same FFT/sample rate etc) seem to open OK.

 

[kimmosto]

^It works if you just click Near button after loading pir file (because it's near field measurement) and then adjust Reference time and Left window with cursors or text boxes if necessary.
There is some problem with high sample rates and long time windows forcing to long FFT. I will try to fix that by skipping IR graph updates and FFT calculation while user adjusts something. Too much work within small adjustments with 192k sample rate and 256k FFT. Maybe some recursive/looping problem too.

P.S. I wouldn't use 192k sample rate. 88.2k is recommended in measurement instructions.

 

[DBMandrake]

^It works if you just click Near button after loading pir file (because it's near field measurement) and then adjust Reference time and Left window with cursors or text boxes if necessary.

Unfortunately this doesn't work. As mentioned the program completely hangs and becomes unresponsive when I try to open the file. ("Program not responding" when I try to close it, I have to end task it)

There is some problem with high sample rates and long time windows forcing to long FFT. I will try to fix that by skipping IR graph updates and FFT calculation while user adjusts something. Too much work within small adjustments with 192k sample rate and 256k FFT. Maybe some recursive/looping problem too.

I've tried several other PIR files with the same parameters which do not freeze, so this seems to be a bug triggered by this specific file. Do you see a freeze when opening the file ?

P.S. I wouldn't use 192k sample rate. 88.2k is recommended in measurement instructions.

I have more than 10 years worth of measurements taken at 192Khz sample rate - long before Vituixcad even existed.

For consistency with old measurements I will continue to use 192Khz. (And the same FFT size) I don't really see any reason to change to a lower sample rate. If Vituixcad has difficulty opening 192Khz PIR files I can just continue to export them to FRD in ARTA as I have always done.

 

[kimmosto]

.NET graphics routines didn't handle that much points in line object. Rev. 2.0.51.4 (2020-06-14) has workaround reducing number of drawn data points up to about 4 pts per pixel.

^High sample rates reduce resolution and limit sequence length in practice. I can see from pir file that you're using averaging possibly due to limited sequence length to get decent S/N. Averaging could have some other bad consequences, but you are free to continue with 192k. Still not recommended and period

 

[DBMandrake]

.NET graphics routines didn't handle that much points in line object. Rev. 2.0.51.4 (2020-06-14) has workaround reducing number of drawn data points up to about 4 pts per pixel.

That seems to have solved the problem, thanks.

^High sample rates reduce resolution and limit sequence length in practice. I can see from pir file that you're using averaging possibly due to limited sequence length to get decent S/N. Averaging could have some other bad consequences, but you are free to continue with 192k. Still not recommended and period

I don't recall whether I used periodic noise or sweep exitation in ARTA or whether averaging was used as the measurement was taken 3 years ago. I don't think averaging was used though as I don't normally use additional averaging. (I assume the periodic noise sampling method inherently includes averaging as part of the processing, however I'm not sure whether I even used periodic noise) It's a shame ARTA doesn't record all the sampling modes and parameters in the PIR file.

Getting a good SNR on a port nearfield measurement shouldn't be a problem due to the high SPL at the microphone... I agree the impulse response does look "noisy" to the eye (especially in Vituixcad's impulse display - not so much in ARTA) and I wonder if that is due to "wind noise" on the microphone due to air velocity in the port ?

I've attached the matching cone nearfield measurement - this was taken with the exact same power level and same sampling configuration and mode in ARTA and as you can see it is not noisy, so something about measuring the port is inherently noisy.

 

[kimmosto]

One more fix to rev. 2.0.51.4 available. Date in about box should be 2020-06-15. Previous workaround was too effective filtering also drivers in Driver layout window

 

[kimmosto]

...I wonder if that is due to "wind noise" on the microphone due to air velocity in the port ?

Maybe, though also port measurements have usually been cleaner than your sample. Sometimes measurements just damage without obvious reason. That could be one of those unfortunate incidents.

My opinion about sample rate was general (nothing to do with noise in your sample). Sweep with sample rate having Nyquist freq just above max. simulated frequency is usually the best compromise. That is 88.2k for VCAD2.

 

[DBMandrake]

Maybe, though also port measurements have usually been cleaner than your sample. Sometimes measurements just damage without obvious reason. That could be one of those unfortunate incidents.

So, I set up a quick test to try to reproduce this noisy port measurement to compare the results with periodic noise and sine swept modes with and without a microphone sock, and I was not able to reproduce the problem at all...

With either periodic noise or swept sine modes the impulse response was clear and sharp without any visible noise in the "pre-start" of the impulse as there was before, at normal mangification levels. I could see a small amount of noise if I increased the vertical gain of the impulse response dramatically but that is to be expected.

With the gain increased dramatically it was clear that the swept sine mode has better SNR than the periodic noise mode, but that both had high SNR for this measurement. (Both tested with "number of averages" set to 1, eg disabled)

This is consistent with advice in the ARTA manual regarding the two modes:

"Notes: The swept-sine is an optimal excitation signal for the fast measurement of an acoustical impulse response, without the averaging. It gives a better estimation than other excitation signals in acoustical time-variant environments and slightly nonlinear systems.

The swept-sine is not best excitation signal if the environment generates a large level of the colored or impulsive noise. It also gives a bad estimation in a system that has the frequency sensitive automatic gain control or automatic noise suppression. In those cases the periodic noise excitation gives a better estimation."

So at the end of all this I have no idea why that 3 year old impulse response is so noisy and why only the port measurement is noisy and not the cone measurement when the only difference was moving the microphone. I've obviously made a mistake in the set up for the measurement and should be more careful next time.

I'm about to start a new 3 way project so I'm trying to refine my measurement process a bit including making more use of the more recent features in Vituixcad like the Merger tool and diffraction simulator. (Last time I used ARTA to do merging and estimate baffle step, but I don't think it does these as well especially diffraction)

So many new features have been added in the 3 years since I first used Vituixcad 1.x that I am now reading the manual from cover to cover again to discover some of the new features. (For example I had not noticed the "Variant" feature before today, previously I would save different experimental versions of a crossover to different files and open them in separate copies of Vituixcad to compare them...)

It's amazing that Vituixcad has grown so much since the beginning that it will now do basically everything in between captured impulse response files and finished crossover design.

ARTA + Vituixcad literally is a complete speaker design and measurement system if you're happy to do the actual circuit design and optimisation yourself. (which I am) I can't recommend it highly enough to those who haven't used Vituixcad yet or are only just starting out with it.

 

[b_force]

.NET graphics routines didn't handle that much points in line object. Rev. 2.0.51.4 (2020-06-14) has workaround reducing number of drawn data points up to about 4 pts per pixel.

^High sample rates reduce resolution and limit sequence length in practice. I can see from pir file that you're using averaging possibly due to limited sequence length to get decent S/N. Averaging could have some other bad consequences, but you are free to continue with 192k. Still not recommended and period

Well, if you take a longer FFT length, you technically can get frequency resolution back.
In the end it all depends on context again. For a normal gated in-room measurement, your lowest frequency will be around 150-200Hz
So say around ±5-10Hz or so will be just fine.

For lower frequencies and nearfield measurements, that's not really working anymore. Often I just go to 8kHz, which results is a nice frequency resolution at the lower frequencies.

From an end-user point of view it would also be a lot more intuitive to work that way. People are more interested in what resolution and bandwith they want. Samplerate and FFT length can be calculated accordingly

 

[kimmosto]

^I just try to keep things and instructions simple and clear without requirement of special knowledge or calculations. Recommended settings provide full internal range 5-40000 Hz with 0.336 Hz (or less) frequency step. Not my bad if some individual is smarter and selects something else because "bigger is better"

 

[DBMandrake]

What is the resolution in Hz for a 192Khz sample rate and a 256k sequence length ?

 

[kimmosto]

^It's sample rate divided by FFT length. For example 192k sample / 256k FFT = 0.75 Hz. VCAD has also 512k and 1M FFT, but actual resolution requires also data. For example time window reduces LF resolution if response has dips and peaks.

 

[DBMandrake]

^It's sample rate divided by FFT length. For example 192k sample / 256k FFT = 0.75 Hz. VCAD has also 512k and 1M FFT, but actual resolution requires also data. For example time window reduces LF resolution if response has dips and peaks.

0.75Hz resolution is plenty for me. Only during a nearfield bass measurement would such fine resolution even be useful.

ARTA is limited to a maximum FFT size of 256k as far as I know, I always use this size.

 

[b_force]

^It's sample rate divided by FFT length. For example 192k sample / 256k FFT = 0.75 Hz. VCAD has also 512k and 1M FFT, but actual resolution requires also data. For example time window reduces LF resolution if response has dips and peaks.

Since we use a log scale for the frequency, it would be actually also interesting to use the FFT length on a similar fashion. Since resolution (accuracy) is a relative measure and not absolute, you're basically just wasting resolution because the relationship between samplerate and fft length isn't logarithmic but linear.
As far as I know, that's just the nature of the beast, but I also have never looked into the fact to apply some mathematical tricks to get a similar result done.

 

[DBMandrake]

Quick question - is there an easy way to clear the File->Recent menu ?

I've tidied up my Vituixcad folder structure to come into line with the recommended project folder structure and also moved the base path ("main directory" folder in settings) to a folder in my google drive sync directory where I was actually storing the project files so I don't have to keep manually browsing there.

As a result all the Recent entries are orphaned and I don't see an obvious way of deleting them ? I suppose if I open every file again the invalid ones will be cycled out (assuming there is a limit to the number of items ?) however some of the files don't exist anymore, and most of them are files I don't need to open again any time soon so I'd rather start with a fresh empty recent list.

I quite often create temporary projects to experiment with an abstract idea then delete them later, ending up with orphans in the Recent list.

Is there a way to clear the list without poking around in the registry looking for where the list is stored ? If not perhaps the dialog box that reports an orphaned file can't be found when trying to use the Recent menu could offer an option to remove the file from the recent list ?

 

[DBMandrake]

So at the end of all this I have no idea why that 3 year old impulse response is so noisy and why only the port measurement is noisy and not the cone measurement when the only difference was moving the microphone. I've obviously made a mistake in the set up for the measurement and should be more careful next time.

It turns out I should indeed be more careful in my measurement setup.

Tonight I thought I'd try to go back to original PIR files for the project, and use Vituixcad to do the PIR->FRD conversion, nearfield/farfield merge and baffle diffraction estimation instead of the original method where I used ARTA to do these steps.

I quickly discovered that my nearfield measurements taken for left and right speakers are some 15dB different in test level - as a result one is a little noisy and the other isn't. Mystery solved.

I didn't notice this before because the absolute level of the nearfield measurement is not important since the level is shifted to splice correctly with the farfield measurements anyway - and the left and right farfield measurements were at correct, matching levels for left and right.

(For anyone wondering why I'm measuring left and right speakers separately - left and right drivers due to their age and condition have significantly different responses - different enough that I needed to measure them independently and optimise left and right crossovers separately to get the best final match between left and right speakers...)

 

[kimmosto]

Quick question - is there an easy way to clear the File->Recent menu ?

No. List is stored in user.config xml file, but finding directory of the last version is not very fast. Location for rev 2.0.51.4 is:
\Users\username\AppData\Local\Kimmo_Saunisto\VituixCAD2.exe_Url_...\2.0.51.4\user.config

<setting name="[B]MRUList[/B]" serializeAs="Xml">
  <value>
    <ArrayOfString xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
      <string>C:\Users\kimmo\Documents\VituixCAD\Projects\Epe-3W_demo\Epe-3W_demo_V2.vxp</string>
      ....
    </ArrayOfString>
  </value>
</setting>

Good possibilities to damage the file so it's not recommended

Would be very easy to clean the list from orphans every time when program or menu is opened, but temporary connection lost to cloud/server could cause some harm. There's max. 20 items which is quite fast to clean by clicking the list.

 

[nc535]

Hi Kimmosto:
Still enjoying VituixCAD....

A few observations, wishes and questions:

when I open an old project, Show Normal Phase is always off and Show Target are always on, regardless of state when the project was last saved.

File and directory paths are saved and restored nicely with one exception - the directory for writing calculator results.

It would be really nice if in the room tab I could specify different absorption values for floor and ceiling than for walls.

Once I add an overlay to the power/DI graph, how do I remove it? Seems to be something in the trace setting submenu but having trouble getting to work. Closing project without saving and re-opening seems to work.

When I save all or save selected as overlays, where are they saved and how are they named? Presumably, if I know this I can use open overlay to get them on the screen.

Thanks and regards,
Jack

 

[kimmosto]

Show Normal Phase is always off

Oops, that has always been on by default when program starts, but accidentally switched off while previous overlay modification. I will fix that.

Also target lines in SPL and Power&DI have been on by default, but I will change so that program remembers previous status.

with one exception - the directory for writing calculator results.

This works here without problems. Calculator has separate output directory text box with folder selection button. Just like in Merger tool. Have you noticed that?

It would be really nice if in the room tab I could specify different absorption values for floor and ceiling than for walls.

No problem. Needs just smaller margins between controls to fit everything in Room tab.

Once I add an overlay to the power/DI graph how do I remove it?

Open Traces... window via context menu, select full row by clicking row header on the left and then Delete button. Manual should tell something about this, but selection of full rows is not necessarily logical for all.

When I save all or save selected as overlays, where are they saved and how are they named?

'Save all/selected as overlay' is not saving to file. It saves as overlay curve.
File->Export... saves to files.
'Open overlay...' reads files to curves. Note that drag & drop is recommended because you can select between left and right Y-axis.

 

[nc535]

This works here without problems. Calculator has separate output directory text box with folder selection button. Just like in Merger tool. Have you noticed that?

I am painfully aware of that folder selection button because I have had to use it so many times. I think the issue is when the selected directory gets saved in the project file. I often clone new project files from old ones by creating a new directory and writing the old project file to the new directory, where I will keep all subsequent data - like calculator results and exported files and graphs. It seems that for cloned projects, I have to re-point the calculator results to that new directory every time I open the calculator.

Thanks for the rest. It really helps!
Jack