AFMG Network Forum

andrzejhd · Member

This is what I get from the SoundFlow (example structure):

https://www.dropbox.com/s/29y4gid20a4yj3g/Zrzut%20ekranu%202015-08-19%2021.42.59.png?dl=0

Do you have any idea why is that guys? Smile

thomas · Member

Hej andrzej,
are you aware of the frequency range (100 Hz - 10kHz, 1/3 Octave) of
ease? In this range your construction has nearly no absoption!
The absorptionpeak (fr<63 Hz) is below this range. So there will be no
effect of this construction in simulation.
And in reality? The principal construction is 10 mm Plywood on an
150 mm airgap (rough approximation). The maximum of absorption
should be somewhere between 45..55 Hz (estimated) with alpha 0,6..0,8
(with some fibrous insulating material on the back, but not 150 mm) or
0,3..0,5 without fibrous material (pure airgap). One octave below and
above the resonance, you get aprox. half of that absorption (rough).
So, your construction will work mainly below 100 Hz. Is that the absorber,
you are looking for? For using in a large factory hall?

The alpha_r = 2.6 in the shown picture is really strange. I recommend
using absorption-tables of approved constructions as long as you have no
own experiences. And learn about the calculation-program with a tutorial
(if there is one) and with calculating approved constructions you find in
absorption tables. So you get a feeling for what is calculated right, and
where are perhaps inaccuracies.

kind regards
thomas

andrzejhd · Member

Surely I know about the range limitations Thomas.

I designed two different panels for the bass frequencies - one tuned for 50Hz and then the ceiling one affecting 125Hz band. Still - these coefficients doesn't look right.

PS. I'm trying to transform a big factory into a rock music venue.

Regards,
Andrzej

andrzejhd · Member

Okay. Finally I produced some graphs to show you how big is the issue. I've already spent hours looking for the cause and still have no idea how to troubleshoot. I still hope that you could help me guys.

Room data:

https://dl.dropboxusercontent.com/u/1545062/room_data.jpg

Reflectogram:

https://dl.dropboxusercontent.com/u/1545062/reflectogram.jpg

Differences between different RT60 simulation methods:

https://dl.dropboxusercontent.com/u/1545062/Differences.tiff

PS. All the simulations have been produced using 30000 rays and 14th order and 500ms time window. The longer the time - the bigger the differences. Shocked

thomas · Member

Could You send the Sabine Reverberation picture?

thomas

andrzejhd · Member

I can send you Eyring RT if that's okay:

https://dl.dropboxusercontent.com/u/1545062/erring.png

thomas · Member

Ok, eyring is ok, sabine will be a little bit longer.

1. sabine (eyring) ist the best approximation you have at the beginning. (That's ok for the majority of rooms!)
2. your impulse response is really too short! Choose a length according to my first post in this thread, with the longest sabine-value as an good approximation of the RT.
3. Choose 2 listener- and 2 sourcepositions and optimize the number of rays. Increase the number until there is no (significant -> <= (5..10) %) change in simulated RT. That step is maybe very time consuming.
Raytacing is sufficent, you don't need mirror imaging for the calculation of RT. If you really interested in early reflections, optimize the mirror imaging later.
Look at the decay-curves and the quality of fit of the best-fit line. Sometimes ist's no good when the IR is too long (longer than RT).
4. Calculate the RT with the optimized number of rays for all source-listener positions you want. In your room 2 source and 6..12 listener positions are sufficent.

good luck
thomas

thomas · Member

Add on:
Before you do all this simulations I described before, optimize your amount and the positions of absorbing materials with the Sabine RT (With measured or aimed RT-values).
The normal pocedure I use is: measure RT, calbrate your model with Sabine, optimize your simulation parameters, simulate, when simulation ok - put our absorption into the room until the RT-goal is reached (Sabine), optimize your simulation parameters, simulate.

t.

Agustín Arias · Member

I'm almost agree with all the things that thomas wrote.
My recomendation is that you should also add a "Random Tail" to your reflectogram (in Probe module, Edit>Add random Tail) which is calculated according to statistical RT (Sabine or Eyring). This tail is added to your reflectogram until the pulses decay to 0. Then you can calculate the Schroeder RT (calculated from the reflectogram decay, no statistical) and see the differences.
In my opinion, the reflectogram lenght is Ok in this case (maybe you can increase it a little bit, up to 700ms, but the exponential decay can be seen). The reflectogram that you capture is for the 1000 Hz 1/3-octave band, and if you see the Eyring RT for that band, it is 1,88 s. In my practice, I usually use a reflectogram length = half of the Eyring RT at 1000 Hz as a first approximation. In your case it is 1.88/2 = 0,94 = 940 ms. But with 700 ms it should be Ok too.
In the other hand, it will be interesting to increase the amount of rays (i.e. 500.000 rays) to obtain a denser reflectogram.
If you use 700 ms for the reflectogram length, the random tail has to start before, making an overlaping (i.e. the tail starts at 400 ms).

For last, remember that the reflectogram is directly related to your sources characteristics (frecuency response, directivity, etc.). In some cases it can produce very different results in comparison to the measurements if you measured with an omnidirectional source as the ISO 3382 establishes.

If you want, you can send me your project packaged (In the Main EASE window go to file>pack project. It generates a .zip file with all the data). I can check your model and make a simulation to optimize the reflectogram parameters.

Regards,

andrzejhd · Member

Thank you very much for these hints!

I have a problem with good calibration, because the hall will be split in thirds (to make the space smaller) and I simulated only 1/3 of the physical building that I measured.

When simulating with 'Probe' module I use the 'Sphere' speaker in the middle of the room to recreate omni-source.

I forgot to mention that one of the panels is a wooden diffuser. Sabine model doesn't include scattering, right? This is why I started with 'Probe' straight away.

AFMG Frank Siegmann · Member

Please note, that neither Raytracing Impacts nor Local Decay Time can take into account scattering factors. This is only possible when using the AURA module for EASE.

Frank Siegmann
AFMG

andrzejhd · Member

I swear that RT changed after I added scattering coefficients. Well, I guess that explains a lot of my problems...

Ron Sauro · Member

If I might jump in here... Absorption coefficients larger than 1 are correct. The problem lies within the conversion from total absorption to an "absorption Coefficient". The total absorption is divided by the area only and that assumption is incorrect. The other absorption is caused by the edges diffracting sound waves. Diffraction is an absorptive function.
Go to my website at www.nwaalabs.com and look on the front page for three papers on this subject. We are just finishing a 19 lab study confirming this for ASTM
To get the true reverb time you will need to know the total absorption for each surface and the only way is to know the numbers above 1 and multiply the areas by those numbers and still will be not quite correct since the absorption is really based on the ratio of edge to area of each sample