I will consider that.
But, let first discuss two things?
1) why you need more resolution in directivity plots.
2) why ARTA uses smoothing
I suppose that you think that higher resolution will more "deeply" reveal side lobes at higher frequencies.
It is true if you analyse theoretically generated response (the Fraunhofer integral will always give deep side lobes when integrating over limited radiating surface). In practice you always have baffled horns and slight off-centering, it changes response - somewhere adds smoothing, somewhere gives new side lobes.
What we hear is sum of direct sound, diffraction and room reflections. All these change response further with new FR peeks and deeps. The body reflections also add some kind of smoothing (Pierce in Acoustics has shown that smoothing in frequency domain is equivalent to spatial response integration on single frequency).
To conclude this:
A mild smoothing gives better representation of response "that we hear". It slightly covers some FR deeps but they will still be recognized as such. A higher resolution usually will not help us, it will just add more information in the graphs that can make the mess
.Best,
Ivo
I'm talking about gated measurements for speaker design, so room and body reflections aren't a concern. For speaker design I want to see diffraction effects as they are.....not smoothed. If this is your position then why offer higher resolution for single gated measurements? Not a criticism, ARTA is my preferred measurement suite, just a question.
I told you that I will probably add 1/24oct. smoothing. Why not?
But, I want to point out that smoothing is not bad thing even if remove reflections by proper gating of loudspeaker response. In my lecture I always treat measurement as process of "estimation" of measured value in the presence of random noise, nonlinearities and disturbances. The best result of estimation is one that give best expetation and lowest variance. Smoothing is one of techniques that lower the variance (the other is averaging of results of several measurements). The price payed for smoothing is little lower frequency resolution, although in practice you can notice all trends (peak and deeps) that exist in unsmoothed curves.
The other reason for smoothing in frequency domain is that is gives better trends to interpolation or curve fitting in in space (polar angle) domain. Actually resolution in space domain is worse than in frequency domain - interpolation can hide some details while curve fitting to series of cylindrical or spherical functions can generate false - aliased response data.
I hope that you are aware of all these facts.
In my work I usually use 1/6 octave smoothing and gets directivity patterns that I can read with confidence.
Best,
Ivo
Could you take the following also into consideration for future releases:
- allow the user to enter values (in ms) for gate start position and gate length (to simplify ARTA automation)
- add a button "uncheck all" in the FR overlay manager
- add a "db offset" function in the polar plot: this way colours can be scaled the same for comparing measurements, even when some sharp resonance is present
- add “Variable-Octave Complex Smoothing” for frequency response smoothing like described here: http://www.aes.org/e-lib/browse.cfm?elib=14322&rndx=212630
- allow the user to enter values (in ms) for gate start position and gate length (to simplify ARTA automation)
- add a button "uncheck all" in the FR overlay manager
- add a "db offset" function in the polar plot: this way colours can be scaled the same for comparing measurements, even when some sharp resonance is present
- add “Variable-Octave Complex Smoothing” for frequency response smoothing like described here: http://www.aes.org/e-lib/browse.cfm?elib=14322&rndx=212630
I will put first three requirements in ToDo list, but for last one I will first need some more information.
There are lot of request and some solution with smoothing on bark (loudness) scale, but none of them was actually approved for dynamic signals. Zwicker definitions are for stationary signals, while Moore filters are removed from last edition of his book.
Best,
Ivo
The real situation with Windows XP is that nobody wants to improve computers especially laptops.
In future versions I will make more enhancement in ARTA only for non-XP OS versions.
A good thing for owner of WinXP is that generally they can install Win8 without any problem and have modern computer if they have 2GB of memory and processor faster than 1.5 GHz.
Contrary to many report on bad Win8 user experience I have excellent experience. It gives fast work and much better audio stability than in Vista or Windows 7.
Ivo
In future versions I will make more enhancement in ARTA only for non-XP OS versions.
A good thing for owner of WinXP is that generally they can install Win8 without any problem and have modern computer if they have 2GB of memory and processor faster than 1.5 GHz.
Contrary to many report on bad Win8 user experience I have excellent experience. It gives fast work and much better audio stability than in Vista or Windows 7.
Ivo
I missed that thanks, and I also appreciate your response. Your reasoning makes sense.
Don't know what the average user here has in a laptop but have had my latest for 4 years now. Purchased for $650, 17" full keyboard + 10 point multi touch pad, i7 quad core, dual drive bays, Win7. Added 4gig for 8. Even has blueray and usb3, hdmi and dual audio out. Never had issue with Arta, HolmImpulse, REW etc. Only wish RTA was closer to real time.
Ivo, your response to include suggested changes and explainations are top notch. Thank you.
Ivo, your response to include suggested changes and explainations are top notch. Thank you.
Dual gate method is an arbitrary method that is not fully theoretically approved. ARTA follows methods that was introduced by MLSSA system, as it was request from many users.
I will not change that method, until I find some method that has better theoretical background.
Ivo
You could allow the use to set the gates to smaller values also. Minimum 70 ms for gate 2 I have always found very large. Minium 5 ms for gate 1 is also a little to big when measuring inside, as the first reflection most of the time arrives within this period (always if you have a ceiling height of 2.5m or less)...
I understand, I just thought that instead of dual gate method, the method was effective and it seems that one in Sound Easy or Omnimic. This method is of course also only quide, but has better results. MLSSA standard support, of course, too, there is still nothing better ...
Hello Ivo and All. Happy New Year
Sorry Ivo for my rather long silence.
For my rig centered around Focusrite Scarlet 2i2, ASIO driver on Windows7 system, you advised the following for calibration:
" The trick in calibration is that you would calibrate front balanced mic. input connected to rear balanced output, and reference line input connected to headphone output. In balanced XLR connector signal is measured between pin 2 and 3"
I get the first part about mic in calibration. For the reference input to H/P output connection I make an unbalanced to balanced connection ? - T of Input to T of Headphones, Ring of input to S of H/P , S of input and R of H/P unconnected.
Another problem:
In ARTA Impulse response, when I change settings of OUTPUT VOLUME in the Impulse response/ Signal recording window, the difference is clearly audible. But after windowing the impulse responses measured at vaious settings of Output Volume, the resulting smoothed FRs are fixed at a particular dB level.( Dual channel measurement mode - unticked)
What am I missing?
Regards
Soundrex
Sorry Ivo for my rather long silence.
For my rig centered around Focusrite Scarlet 2i2, ASIO driver on Windows7 system, you advised the following for calibration:
" The trick in calibration is that you would calibrate front balanced mic. input connected to rear balanced output, and reference line input connected to headphone output. In balanced XLR connector signal is measured between pin 2 and 3"
I get the first part about mic in calibration. For the reference input to H/P output connection I make an unbalanced to balanced connection ? - T of Input to T of Headphones, Ring of input to S of H/P , S of input and R of H/P unconnected.
Another problem:
In ARTA Impulse response, when I change settings of OUTPUT VOLUME in the Impulse response/ Signal recording window, the difference is clearly audible. But after windowing the impulse responses measured at vaious settings of Output Volume, the resulting smoothed FRs are fixed at a particular dB level.( Dual channel measurement mode - unticked)
What am I missing?
Regards
Soundrex
