AFMG Network Forum
AFMG Network Forum Index -> EASE 4 -> AURA-the suitable particle and length
Post new topic  Reply to topic View previous topic :: View next topic 
AURA-the suitable particle and length
PostPosted: Wed Nov 22, 2017 4:09 pm Reply with quote
ah82318
Member
 
Joined: 22 Nov 2017
Posts: 9




Hi:I'am a graduate student from arch acoustic institute. I have carry a project about improving sound quality of a small live house(173m^3) firstly I measure in the field and I use EASE AURA to make sure the simulation result was similar to the field result.
When I run the AURA ,I set up 100k particle and 1020ms ,but the result wasn't fit about the field measurement. but I use the 20k particle and 1020ms, the result was much better.
So my question is: Is it the more particle ,the more accurate? or it is rely on my room volume or the time length? Shocked Shocked Shocked
View user's profile Send private message
PostPosted: Thu Nov 23, 2017 2:19 pm Reply with quote
thomas
Member
 
Joined: 16 Mar 2011
Posts: 73




Hi,
note that you have to adjust the number of particles to get a
stable simulation result.
I've good experience with length(IR) approx. 2/3 T.

If you want simulation results that fit the measurement results,
you have to adjust the material-/construction parameters
(absorption/scattering coef.). But first you need a stable simulation!

Best wishes
T.
View user's profile Send private message
PostPosted: Thu Nov 23, 2017 5:05 pm Reply with quote
ah82318
Member
 
Joined: 22 Nov 2017
Posts: 9




[quote="thomas"]Hi,
note that you have to adjust the number of particles to get a
stable simulation result.
I've good experience with length(IR) approx. 2/3 T.

If you want simulation results that fit the measurement results,
you have to adjust the material-/construction parameters
(absorption/scattering coef.). But first you need a stable simulation!

Best wishes
T.[/quote]Hi Thomas: thanks a lot for your answer! I''try again to change the length. But I still unsure the number of the particle. When I use the length approx 2/3T, which number I should use?
Is it still the more particle ,the more accurate? or I just go to try until the result was fit the field result?


Best wishes
A. Cool
View user's profile Send private message
PostPosted: Fri Nov 24, 2017 8:59 am Reply with quote
thomas
Member
 
Joined: 16 Mar 2011
Posts: 73




Quote:
or I just go to try until the result was fit the field result?

No, first You have to look for a stable result!
Stable means here, that parameters calculated with the IR don't change
with more particels. Don't change means here that the change is lower
then the just notable difference (JND) of the parameter. For the most
important parameter reverberation time, use 5% (if the EDT, RT10,20, 30)
change up to 5% with more particles, the result is stable.
For this optimization you don't have to simulate the whole audience. In
most rooms 1 or 2 source positions and 2 or 3 listener positions are
enough (less for rooms with a simple shape).
Start with a low particle number, approx. the value proposed by aura for
low resolution is a good starting point. Rise the number by a factor of
5..10, as long as there's a change > 5%.

A second way to optimize:
A rule of thumb you can use: the number of particles in the very first reflection
bins of the reflectogram should be at least 10..20 but not lower than 8.
Simulating the first 100 ms is here sufficent.

good luck!
View user's profile Send private message
PostPosted: Sun Nov 26, 2017 9:01 am Reply with quote
ah82318
Member
 
Joined: 22 Nov 2017
Posts: 9




[quote="thomas"]
[quote]I've good experience with length(IR) approx. 2/3 T.
Thx for your reply,In my field measurement, the T30 was about 0.7-0.8 sec in the 500 Hz, 0.6-0.7 sec in the 1000 Hz and 1.3-2.4 sec in 125 Hz,
Which frequency of T30 would you use to setting the 2/3 T30 ?
I run the two 2/3 T30 length(500 ms and 1500 ms). The 1500 ms obviously fit the field measurement in the low and mid frequency.

But I still have one problem: when I change the particle, the more particle I use , the difference of T30 getting smaller.
(the room has a significant difference T30 in low freq, the front audience part was about 2.4 sec and the back part was 1.3-1.7 sec)
When I use the particle about 20000-25000. the result can show the difference in the front and back part, Is the result regular(correct) or something I didn't consider? Shocked
Thx for your patiently reply !!
View user's profile Send private message
PostPosted: Sun Nov 26, 2017 7:16 pm Reply with quote
thomas
Member
 
Joined: 16 Mar 2011
Posts: 73




Don't compare simulation with the measurement result before your simulation is stable!
Ok, I don't know your room. But 25k particles aren't enough for sure!
I just tested it with a 175 m^3 shoebox (T_mid = 0,7 s) and ended up with 500k particles
for a stable simulation. But it's possible, that the appropriate number for your room is different.
So you have to optimize it by yourself.
While optimizing particle number, I use an IR length, appropriate for 1kHz. When this is done,
I look at the decay curves and the T30-line (for frequencies with longer RT) and chose the
necessary minimum length of the IR. The fast decay at the end of the decay curve has to be
outside of the T30-line (it's here the same as in measuring technology - you need a decay of
approx. 40 dB to determine a 30 dB decay from -5 to -30 dB).

If all is done, you can compare. And start thinking about possible reasons for differences
and wether they're significant.
Possible reasons are:
- your measurement is not correct (tons of possible reasons behind that)
- temperature/humidity differs between meas. and sim. (only important for large rooms)
- deviations of your model (geometry, absorption, scattering) and the real room
- effects due to the wave character of sound are perhaps important, but of course not
considered in simulation


Quote:
In my field measurement, the T30 was about 0.7-0.8 sec in the 500 Hz, 0.6-0.7 sec
in the 1000 Hz and 1.3-2.4 sec in 125 Hz ...
(the room has a significant difference T30 in low freq, the front audience part was about 2.4 sec
and the back part was 1.3-1.7 sec)

There are a lot of possible reasons for that again. Perhaps
- your measurement is not correct or perhaps the calculation of the T30-fit (some older
measurement systems have problems with the automatic calculation, especially for lower
frequencies),
- you have no diffuse soundfield for low frequencies at least (strong seperated single modes
with different damping),
- you have no straight decay curve because of a "second soundfield" with longer RT, but not
evenly distributed throughout the whole room (e.g. a leaky door and a corridor behind with
significant longer RT - you measure sometimes longer RT's for low frequencies near the door)
and so on ...

And keep in mind, that T30 characterizes the whole room. So the means of a well done
measurement and of a well done simulation has to match.

[short summary: don't believe in simulation, don't believe in measurements, at least not blind Wink]
View user's profile Send private message
PostPosted: Tue Dec 05, 2017 5:55 pm Reply with quote
ah82318
Member
 
Joined: 22 Nov 2017
Posts: 9




[quote =“thomas”]
Hi thomas, I try to simulate the different length to represent low and mid frequency(about1.5sec and0.5sec), it was a little embarrassed cause I think I didnt figure out the real meaning of the LENGTH.
I take the length as the interrupt time of the sound, Is it correct? and I didn't realize why the recommend setting of length was 2/3T? not T or exceed T? Sad
View user's profile Send private message
PostPosted: Wed Dec 06, 2017 12:05 am Reply with quote
thomas
Member
 
Joined: 16 Mar 2011
Posts: 73




What is " the interrupt time of the sound"?
I'm talking about the length of the simulated part of the impulse response (IR). To calculate T30, you need a decay of 40 dB (=2/3T).
I'm confused. Perhaps you missed, that 1,5 s = 1500 ms and 0,5s = 500 ms (this are the values aura needs)?
View user's profile Send private message
PostPosted: Mon Dec 18, 2017 10:14 am Reply with quote
ah82318
Member
 
Joined: 22 Nov 2017
Posts: 9




[quote =“thomas”]
Dear Thomas: perhaps I was miss, but I still confused.
the setting "length" I mean was on the AURA->caculation->length(ms), my question was:
Was the length means the decay time or something else?
Should I estimate the expect RT time first and enter over the 2/3T, or just 2/3T ? In EASE tutorial, it recommend the min length should be approx 2/3 expect time.

the other question was:
In different frequency should I setting the different length ? (like I estimate the low frequency RT was 2 sec and mid frequency was 1 sec, so to simulate the accurate T30 in low and mid frequency, I should setting the different length .)

Very appreciate for your patiently reply!
and hope that not bother you so much time~
View user's profile Send private message
PostPosted: Tue Dec 19, 2017 12:52 pm Reply with quote
thomas
Member
 
Joined: 16 Mar 2011
Posts: 73




Quote:
the setting "length" I mean was on the AURA->caculation->length(ms), my question was:
Was the length means the decay time or something else?


->
Quote:
I'm talking about the length of the simulated part of the impulse response (IR).


and, as I wrote, specified in ms, not in s!


Optimize for middle frequencies, after that, calculate a longer IR with respect to low frequencies. Then check, wether IR for middle frequencies is still ok (sometimes it is, sometimes it is not and strange late spikes occur, don't know why).
View user's profile Send private message
AFMG Network Forum Index -> EASE 4
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
All times are GMT  
Page 1 of 1  

  
  
 Post new topic  Reply to topic  


Powered by phpBB © 2001-2003 phpBB Group
Theme created by Vjacheslav Trushkin
Variation by CodeWeavers and AFMG