Another problem source with common rotation center of speaker is 1/48 pts/oct. frequency scale which could break phase response at HF if radiator drifts a lot while off-axis measurement sequence or virtual speaker rotation while simulation. Common rotation center is not okay especially for large speakers with multiple HF radiators. Therefore only one driver of line array should be measured - rotating around driver's center point, though it doesn't eliminate possible phase wrapping problems while simulation.
Too fancy name for something which does not construct new data. It's just selection of frequency response.
One day I may test response interpolation or simple average if target angle is within 1/3...2/3 of the nearest measured angles. Phase interpolation is not so clear because interpolated response should inherit excess phase/GD features of original responses. Too complex calculation would drop simulation speed.
Some simulators interpolate responses, but I haven't checked how "legal" the results are.
How about using time-domain data instead? Adding impulse response data in the time domain side-steps the issues related to (smoothed / processed) frequency-domain data. I guess this would be a substantial rework of the software, though.
Time domain and FFT would remove the problems with phase unwrapping, but need more time and much more memory to maintain full resolution in frequency domain.
Maybe I'm worrying too much with response interpolation because weighted sum of complex numbers is quite straight forward. Let's wait what happens after I've tested something...
Maybe I'm worrying too much with response interpolation because weighted sum of complex numbers is quite straight forward. Let's wait what happens after I've tested something...
Hi! I tried VituixCAD for the first time yesterday (latest version, running on Windows 7) and have issues with the graphs. I built a three way crossover with driver FR and impedance plots imported from REW measurements.
When I click one of the graphs to see a bigger version, how do I get back to the overview seeing all (SPL, GD&Phase, Filter, ...) of them? So far my only chance is quitting VituixCAD and starting it again.
Bigger problem: In the SPL, GD&Phase and Power&DI windows I only get a single straight line, useful curves only in the Filter and Impedance windows.
Unfortunately searching diyAudio and google didn't bring up answers.
When I click one of the graphs to see a bigger version, how do I get back to the overview seeing all (SPL, GD&Phase, Filter, ...) of them? So far my only chance is quitting VituixCAD and starting it again.
Bigger problem: In the SPL, GD&Phase and Power&DI windows I only get a single straight line, useful curves only in the Filter and Impedance windows.
Unfortunately searching diyAudio and google didn't bring up answers.
- You got the bigger version by double-clicking. So double-click again to get the full six-pack view.
- The measurements must be properly named. From the Preparations for measurement (highly recommend that you read it carefully, also the Manual):
"Saved measurement filenames (.pir) must have valid coding for plane and off-axis angle. ARTA style angle
coding <name-prefix>_deg[+|-]<num>.pir is okay. CLIO style <name-prefix> <num>.mls where angle <num>
is multiplied by 100 is also okay for VituixCAD.
For example M15CH002_hor_deg+110.pir equals M15CH002 to horizontal angle of 110 degrees.
Single axial measurement (without off-axis responses) should also have plane and angle coding with “hor 0”.
schiirrn, see post #1345
How to start
Tips: frequency response file naming and at least single wire from generator to driver.
Few video lessons
These are partly for old version 1.1, but may help figuring out how things work.
How to start
Tips: frequency response file naming and at least single wire from generator to driver.
Few video lessons
These are partly for old version 1.1, but may help figuring out how things work.
Model LTspice
I would like to know, if possible, why in the ltspice model the amplitude value of "Vad" varies according to the loudspeaker and is with a phase of 90 degrees (instead of the usual 2.83v with phase 0).
I would also like to know if in the future Effect of voice coil inductance will be considered.
I would like to know, if possible, why in the ltspice model the amplitude value of "Vad" varies according to the loudspeaker and is with a phase of 90 degrees (instead of the usual 2.83v with phase 0).
I would also like to know if in the future Effect of voice coil inductance will be considered.
Auxiliary LT Spice IV models of Enclosure tool run in acoustical domain. I guess electrical equivalent circuits (instead of acoustical) should be possible. Those are just kinda templates if someone is interested in boxes Enclosure tool does not support.
Le only will not be added because it's unpractical and therefore useless. At least extended impedance model could be added to templates if someone is able to deliver equivalent circuit with Le, Leb, Ke and Rss in acoustical domain.
Last edited:
Word 'approximation' refers to method where power response is calculated from anechoic (or quasi-anechoic) frequency response measurements around the speaker, at least in two planes if speaker is asymmetrical in horizontal and vertical planes. VituixCAD uses this method by default (if 'Intensity on spherical surface' is selected in Options window).
Power response has been measured more directly in diffuse reflecting room with rotating mic. That result could include some half space features at low frequencies.
Main program uses weighting factors calculated from off-axis angle and relative sector measurement covers if angle step was not constant while measurement sequence.
Sector outside measured angle range is not included in calculation so you won't get correct result to full space by measuring half space only.
Average and RMS functions in Calculator tool allow weighting of each frequency response, but that would be very slow process. Not recommended for anything
If I use 0, 5, 10, 15 etc. these are equal steps but they are not equal slices, will Vituixcad weight them for me?
We had a very similar discussion a while ago in this thread. I feel the Vituix documentation does not give the full details required to understand how Vituix calculates power response. That's why I ended up writing my own code that implements the Tylka formulation to calculate power response. Once I compared the results from the Tylka method with the Vituix results, I found that both are similar (but not exactly identical).
No matter which code you use, in principle you need full spherical data to accurately calculate the power response. However, it's certainly possible to get good results by using data that does not cover the full sphere. See here for a comparison of the power response calculated from data covering the full sphere, and the front hemisphere only. The difference was similarly small as the differences observed between the Tylka method and whatever is implemented in Vituix.
No matter which code you use, in principle you need full spherical data to accurately calculate the power response. However, it's certainly possible to get good results by using data that does not cover the full sphere. See here for a comparison of the power response calculated from data covering the full sphere, and the front hemisphere only. The difference was similarly small as the differences observed between the Tylka method and whatever is implemented in Vituix.
Main program weights automatically everything so you can measure with constant angle step e.g. 0, 5, 10, 15, ..., 175, 180 or quite random angle step e.g. 0, 5, 10, 20, 30, 40, 60, 80, 100, 140, 180.
To half space in practice because you would/should check 'Mirror missing' in Options window i.e. hor 0...+90 is mirrored to both hor 0...-90 deg and ver -90...+90 deg.
VituixCAD does not try to guess response of total power to full space if data is to half space only or 'Half space' is checked in Options window. Program just calculates response of total power to half space. Therefore response tilt is usually shallower and power lower to half space than to full space with unidirectional boxed speakers. Symmetrical dipole gives the same result to half space and full space, but symmetrical dipoles are rare in practice so dipoles should be measured to 0-180 deg too to get adequate data for simulation.
..and accepts that this is all the power there is because that is all the responses I have given. Very good
.This way when I build into the corner, say horizontal I only supply -45 to 45 (vertical is different). I will take my own responsibility for how it integrates into the room, but Vituixcad will stop at the angles I give.
Can the SPL trace accept my measured phase data and not insert minimum phase?
I will take my measurements with the speaker installed, but not from the listening position, and I will measure to all extents. I think this will work.
I just found my old source, I though I'd lost it since my last speaker build some years back. Since then I had decided to give Vituixcad a go.
I will take my measurements with the speaker installed, but not from the listening position, and I will measure to all extents. I think this will work.