Geddes on Waveguides

Jump to Latest
#2,622
markus76 said:
soongsc

Every data from simulations is questionable as long as it's not verified by real world data. I hope Earl and Jean-Michel post comparable data. Otherwise the whole discussion is academic and of no other use than intensifying beliefs instead of knowledge.

Best, Markus
Click to expand...
Well, whenever I question something, I always have a reason so that the original provider can explain whether my question is of real concern or not. Normally we never compare measurements with all sims we do, but rather ensure that medthods sims bring us into a reliable ball park, and that we understand what causes differences between sims and measurements.
 
#2,623
gedlee said:
Most people would have no idea what this information meant.
Click to expand...

Everybody that is seriously interested in good sound reproduction should know what smoothing means and how it affects the display of a frequency response curve.
Admittedly, the windowing time of minor interest when the resulting diagram only shows frequencies that are within the reflection free window time. Not everybody is cutting diagrams at lower frequencies accordingly...

By the way, how would you rate the relevance of using a smoothing based on Zwicker's Bark Scale for frequency responses?

Best, Markus
 
#2,624
markus76 said:

Not everybody is cutting diagrams at lower frequencies accordingly...

By the way, how would you rate the relevance of using a smoothing based on Zwicker's Bark Scale for frequency responses?

Best, Markus
Click to expand...

I don't follow your first sentence.

Some smoothing has to be used, it only makes sense since even the measurement process is never perfect and smoothing helps that. 1/3 octave is about a Bark wide (which is its justification), but I would agree with Floyd that 1/3 octave is a bit too wide as it can cover some serious resonances. But I don't agree that 1/24 octave is required as this seems too narrow to me, but mostly it's really difficult to get good data to this resolution - I can't do it. I use noise and noise always has some uncertainty in it. I find that I can go down to about 1/6 octave, but below that the uncertainty in the noise creates non-repeatable problems, i.e. random peaks and dips. I use 1/3 octave because it seems to be standard. I may start posting 1/6 octave as that is really the right thing to do, but below that I can't do.
 
#2,625
gedlee said:


I don't follow your first sentence.

Some smoothing has to be used, it only makes sense since even the measurement process is never perfect and smoothing helps that. 1/3 octave is about a Bark wide (which is its justification), but I would agree with Floyd that 1/3 octave is a bit too wide as it can cover some serious resonances. But I don't agree that 1/24 octave is required as this seems too narrow to me, but mostly it's really difficult to get good data to this resolution - I can't do it. I use noise and noise always has some uncertainty in it. I find that I can go down to about 1/6 octave, but below that the uncertainty in the noise creates non-repeatable problems, i.e. random peaks and dips. I use 1/3 octave because it seems to be standard. I may start posting 1/6 octave as that is really the right thing to do, but below that I can't do.
Click to expand...

I´m sorry if I have missed this before, but why not use log-sweeps ? I´ve done that for several years now and will never go back to MLS or such. I do use 1/24 oct. smoothing to get log data out of the measurement system.
 
#2,626
-CGL- said:


I´m sorry if I have missed this before, but why not use log-sweeps ? I´ve done that for several years now and will never go back to MLS or such. I do use 1/24 oct. smoothing to get log data out of the measurement system.
Click to expand...
Quite interestingly, I find that MLS reveals more problem areas than a swepted sine wave. I sometimes look at both.
 
#2,627
Earl

I don't follow your first sentence.
Click to expand...

The shorter the window time, the higher the frequency from which on the frequency response is free of room reflections.

I measure with FuzzMeasure which makes use of a swept sine of user definable length. Having high resolution data of 1/48 or 1/24 helps tremendously in identifying narrow irregularities. You will miss them with 1/3 or 1/6. Just look at one of the full range drivers out there...

Bark scale: it would be more like a mixture of different smoothings in certain frequency bands.

Best, Markus
 
#2,628
soongsc said:

Quite interestingly, I find that MLS reveals more problem areas than a swepted sine wave. I sometimes look at both.
Click to expand...

Something wrong ?

Both methods *should* get you the same impulse response.
MLS has some ugly stuff hidden so maybe it´s just that. And, usually it has white spectrum = blown HF drivers...
 
#2,630
I don't know whether the impulse response is the same or not, the the SPL curves do differ, and I did find out the actual problem areas (which may not be a problem to some people), which was very helpfull. Swept sine wave produce smoother looking curves. But this is probably OT.
 
#2,631
soongsc said:
I don't know whether the impulse response is the same or not, the the SPL curves do differ, and I did find out the actual problem areas (which may not be a problem to some people), which was very helpfull. Swept sine wave produce smoother looking curves. But this is probably OT.
Click to expand...

Same IR same FR. Or Vice versa. FFT or iFFT.
 
#2,632
Hello Soongsc,

I guess that Markus76 refered to pulse response recovered by the convolution of a sine logsweep.

This this also my favorite method and it is far less noisy that MLSA method.

More informations about pulse measurement based on logsweep in the following readings.

http://pcfarina.eng.unipr.it/Public/Papers/134-AES00.PDF

http://pcfarina.eng.unipr.it/Public/Papers/226-AES122.pdf

Best regards from Paris, France

Jean-Michel Le Cléac'h


soongsc said:

Quite interestingly, I find that MLS reveals more problem areas than a swepted sine wave. I sometimes look at both.
Click to expand...
 
#2,633
-CGL- said:


I´m sorry if I have missed this before, but why not use log-sweeps ? I´ve done that for several years now and will never go back to MLS or such. I do use 1/24 oct. smoothing to get log data out of the measurement system.
Click to expand...


Dr. Farina's technique is indeed the state-of-the-art and thanks to Jean-Michel for posting those papers. I had read the first, but had not seen the second.

My only excuse for sticking with noise is that I have developed a multitude of custom software over the years that rely on this technique and I am loath to have to change it all now. It's part lazy and part too busy. If, or when, I update my measurement capability, I will use Prof. Farina's techniques. I just need to figure out how to best impliment them into what I already have without causing too much disruption.

I particularly like the impulse responses of the harmonic orders as this is far more meaningful that THD.
 
#2,634
gedlee said:



Dr. Farina's technique is indeed the state-of-the-art and thanks to Jean-Michel for posting those papers. I had read the first, but had not seen the second.

My only excuse for sticking with noise is that I have developed a multitude of custom software over the years that rely on this technique and I am loath to have to change it all now. It's part lazy and part too busy. If, or when, I update my measurement capability, I will use Prof. Farina's techniques. I just need to figure out how to best impliment them into what I already have without causing too much disruption.

I particularly like the impulse responses of the harmonic orders as this is far more meaningful that THD.
Click to expand...

It is the state of the art indeed. What I like about it is that one can hear what the system does during the sweep. MLS nicely hides (expensive) things...
 
#2,636
Jmmlc said:
Hello Soongsc,

I guess that Markus76 refered to pulse response recovered by the convolution of a sine logsweep.

This this also my favorite method and it is far less noisy that MLSA method.

More informations about pulse measurement based on logsweep in the following readings.

http://pcfarina.eng.unipr.it/Public/Papers/134-AES00.PDF

http://pcfarina.eng.unipr.it/Public/Papers/226-AES122.pdf

Best regards from Paris, France

Jean-Michel Le Cléac'h
Click to expand...
I've browsed through the first before, but it's the first time I've seen the second. Thanks for posting it.
 
#2,637
markus76 said:


Why reinvent the wheel? Take a look at http://www.fesb.hr/~mateljan/arta/index.htm

Best, Markus
Click to expand...

I have ARTA, can't do what I want, or at last try it couldn't. If you don't "reinvent the wheel", i.e. do measurements unlike what everyone else does, then you can't improve upon it.

I don't think that you could understand why I need to have my own measurement system, because what I do is completely unique. I don't talk about it because it is proprietary, but I can't simply integrate other software packages into my capabilities without a great deal of work, and many won't do it at all. I know that the Farina technique will work and when I upgrade it will be to that and nothing else. MLS is no different than what I use now and stepped sign is not as flexible as the Farina approach.
 
You must log in or register to reply here.
Top Bottom