home.md

@@ -220,8 +220,35 @@ Next I will try to add it.
 ![vitamin_corner_timeseries-2](uploads/7d6dc76a073174931a285db9c8f5048b/vitamin_corner_timeseries-2.png)   
 ![vitamin_corner_timeseries-3](uploads/be766406a9fa79a8fbb0a21ec8a6dd35/vitamin_corner_timeseries-3.png)    
 
-## (9)Whitening comparision   
-   
+## (9)SNR    
+We set the value of each injection parameter as the median of its prior's range which is given to train VItamin. Bilby caculates optimal SNR in each detector:     
+SNR_L1 = 8.25, SNR_H1 = 5.16, SNR_V1 = 0.45    
+However, SNR would be good if it is greater than 10. This can be fixed by adjusting not only distance, but also Antenna response function. In practice we adjust distance from 2000(Mpc) to 1800, theta_jn from pi/2 to pi*5/6. The new SNRs are:   
+SNR_L1 = 21.86, SNR_H1 = 18.43, SNR_V1 = 10.87   
+Now they well meet the requirement.   
+## (10)t_zero   
+In a waveform, t_zero is defined as a time coordinate where the absolute maximum amplitude is. It's also an injection parameter called geocent_time in geocentric reference system.       
+## (11)Whitening comparision   
+Problem: It looks like the amplitude of bilby whitened signal has a greater order of magnitude(10^1, which supposed to be 10^0). So I make a comparision between Hunter's whitening method and mine.    
+First, Hunter gets frequency domain noise-free signals at detector directly, instead of getting time domain signals and make Fourier transform. 
+To get a frequency domain noise-free signal directly:   
+```
+waveform_generator.parameters = parameters
+freq_signal = waveform_generator.frequency_domain_strain()   
+signal_freq_domain = ifos[k].get_detector_response(freq_signal, parameters)   
+# k is the ordinal number of detectors   
+```
+To get a time domain signal, and operate it with Fourier transformation :   
+```
+signal_time_domain= ifos[k].time_domain_strain   
+# k is the ordinal number of detectors   
+signal_freq_domain = np.fft.rfft(signal_time_domain)
+```
+If both of 2 methods above are correct, 2 frequency domain signals they get should be the same, at least very similar.
+However, They are very different.   
+to be continued...    
+## (12)VItamin: Bilby BBH waveform new test  
+In this test we get frequency domain noise-free signal at detector directly, instead of getting time domain signal and make Fourier transform.    
 
 # 2.Data   
 ## (1)Korean merger wavform    
@@ -238,8 +265,8 @@ Next I will try to add it.
 > snrs == A list containing the optimal SNR of each waveform at each observatory (e.g. H1, L1 and V1 ... in that order).  
 > y_data_noisy == The noisy waveform at each observatory with dimension (number of detectors, sampling frequency * duration).   
 > y_data_noisefree == The noisefree waveform at each observatory with dimension (number of detectors, sampling frequency * duration).  
-The parameter values are contained in the key 'x_data', where the values are in the following order:
-['mass_1','mass_2','luminosity_distance','geocent_time','phase','theta_jn','psi','ra','dec']  
+The parameter values are contained in the key 'x_data', where the values are in the following order:    
+['mass_1','mass_2','luminosity_distance','geocent_time','phase','theta_jn','psi','ra','dec']    
 
 # 3.Code  
 ## (1)Adjust BBH waveform's amplitude the same to Korean waveform's