Acoustic parameters (C50,C80, D50 etc.) calculation bug

[wbinek]

Function GetSumLimit() in projet_calculation.cpp calculates from time=0, whereas all parameters should be calculated starting from time when direct wave approaches the receiver.

I've made a temporary solution at:
wbinek@827530d

[nicolas-f]

Thanks for feedback.

We have to fix this issue by changing the fromTime value. (without copying array)

[nicolas-f]

After reading again the code:

The implementation follow the ISO 3382:1997(F). The reverberation time RT30 is computed using the regression on a Schroeder curve from -5db to -35db.

The variable wxFloat32 fromTime is the time when the value is -5 dB inferior than the higher value of the Schroeder curve.

[nicolas-f]

All this post-processing code will be moved into lib_interface and included into the automated unit test framework.

[wbinek]

The RT calculation is ok. The bug is in energetic parameters calculation (mainly C and D). The integration should start when the impulse arrives to the receiver not from t=0.

[wbinek]

If you look at for example projet_calculation.cpp in C param calculation there is:

wxFloat32 wj_CurrentFreq=GetSumLimit(idFreq,0.f,te,timeTable,tab_wj)/GetSumLimit(idFreq,te,-1.f,timeTable,tab_wj);

The problem is that instead of starting integration from fixed 0 it should start when direct sound approaches the receiver. It can be changed in parameter calculation funcition or GetSumLimit method.
This applies to all energetic parameters.

[nicolas-f]

Yes sorry, I was focused in RT and EDT. I reopen the issue and I will fix it by updating

Compute_ST_Param
Compute_C_Param
Compute_D_Param
Compute_TS_Param
Compute_LEF_Param
Compute_LF_Param
Compute_LFC_Param
Compute_LG_Param
Compute_G_Param

[Picaut]

I try to validate this modification. However, I obtained NaN or °Inf values for all parameters (except reverberation parameters).
Need to verify step by step the calculation, in debug mode.

Below, a simulation (v1.3.4) for a 10m cubic room, source at the center, receiver at 2m from the source:

50Hz to 1250Hz, increasing absorption from 0, step 0.05, d=1 (100% diffuse)
1600 and 2000 Hz, absorption 0.1 and 0.2 respectively, no diffusion d=0

[Picaut]

May be a problem in this function. Is-it for finding the time step of direct field? lastdb should not be updated at each time step?
Time step of the direct field (if exists) could be estimated simply by calculating the distance from source;
If no direct field ?

https://github.com/Ifsttar/I-Simpa/blob/dca3a488d558d3756fb2a90a0f52877e2238c9bd/src/isimpa/data_manager/projet_calculation.cpp#L114-L126

[nicolas-f]

Yes it could be that. GetTimeDecay should return the time before the noise level is decaying .

[Picaut]

Are you sure it finds the time step of the direct field (or a valid change/increasing of energy)?

[nicolas-f]

With I-Simpa in dev branch I have this:

What is the timestep in spps ?
issue_7.zip

[nicolas-f]

ok, there is an issue when sound power received is < 0 dB. I will try to fix that

[Picaut]

May be that results are correct. I have prepared a I-Simpa project with the possibility to compare results with expected values for C, D...

[Picaut]

I have compared SPPS and expected theoretical results for 100% diffusing room, in many room size (verifying also the effects of the Time Step and distance from source). It seems that the procedure for calculating the clarity C80 within I-Simpa seems correct. It probably needs further validation with measurements or other reference data, but it seems quite acceptable for now.