Merge pull request #10 from alvincjk/revert-9-gpu-dev

Revert "sqrt(df) -> sqrt(dt), noise curves added"
alvincjk · Sep 10, 2019 · 1c4d352 · 1c4d352
2 parents 2f20db1 + c04dcfc
commit 1c4d352
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 64 deletions.
diff --git a/examples/pygpuAAKdemo.py b/examples/pygpuAAKdemo.py
@@ -4,73 +4,20 @@
 from AAKwrapper import AAKwrapper as cpu_AAK
 from gpuAAK import GPUAAK
 from pygpuAAK import pygpuAAK
+import tdi
 
 import time
 
-fit_pars = {
-    0.5:{
-        'alpha': 0.133,
-        'beta': 243,
-        'kappa': 482,
-        'gamma': 917,
-        'f_k': 2.58e-3
-    },
-    1:{
-        'alpha': 0.171,
-        'beta': 292,
-        'kappa': 1020,
-        'gamma': 1680,
-        'f_k': 2.15e-3
-    },
-    2:{
-        'alpha': 0.165,
-        'beta': 299,
-        'kappa': 611,
-        'gamma': 1340,
-        'f_k': 1.73e-3
-    },
-    4:{
-        'alpha': 0.138,
-        'beta': -221,
-        'kappa': 521,
-        'gamma': 1680,
-        'f_k': 1.13e-3
-    }
-}
-
-def P_OMS(f):
-    return (1.5e-11)**2*(1 + (2e-3/f)**4)
-
-def P_acc(f):
-    return (3e-15)**2*(1 + (0.4e-3/f)**2)*(1+(f/8e-3)**4)
-
-def S_c(f, dur=4):
-    if dur not in [0.5, 1, 2, 4]:
-        raise ValueError("dur needs to be 0.5, 1, 2, or 4 years.")
-
-    alpha = fit_pars[dur]['alpha']
-    beta = fit_pars[dur]['beta']
-    kappa = fit_pars[dur]['kappa']
-    gamma = fit_pars[dur]['gamma']
-    f_k = fit_pars[dur]['f_k']
-    A = 9e-45
-    return A* f**(-7/3)*np.exp(-f**alpha + beta*f*np.sin(kappa*f))*(1+np.tanh(gamma*(f_k - f)))
-
-
-def LISA_Noise(f, L=2.5e9, f_star=19.09e-3, dur=4):
-    S_n=20./(3.*L**2)*(P_OMS(f) + 4*P_acc(f)/(2*np.pi*f)**4)*(1+6/10*(f/f_star)**2)+S_c(f, dur=4)
-    return S_n;
-
 
 def test():
     iota = 0.2
     s = 0.8
     p = 10.0
     e = 0.7
     T_fit = 1.0
-    init_length = 100000
-    length = 1000000
-    init_dt = 100.0
+    init_length = 14000
+    length = 1400000
+    init_dt = 1000.0
     dt = 10.0
     M = 1e6
     mu = 1e1
@@ -81,7 +28,7 @@ def test():
     phi_S = 0.785
     theta_K = 1.05
     phi_K = 1.05
-    D = 5.0
+    D = 1.0
     LISA = True
     backint = True
 
@@ -119,8 +66,8 @@ def test():
     freqs = np.fft.rfftfreq(len(hIc), d=dt)
     freqs[0] = 1e-8
     deltaF = 1/dt
-    ASD = np.sqrt(LISA_Noise(freqs, dur=4))
-    noise_channels = {'channel1': ASD**2, 'channel2': ASD**2}
+    AE_ASD = tdi.noisepsd_AE(freqs, model='SciRDv1', includewd=3)
+    noise_channels = {'channel1': AE_ASD, 'channel2': AE_ASD}
 
     kwargs = {
         'T_fit': 1.0,
@@ -131,7 +78,7 @@ def test():
     like_class =  pygpuAAK.pyGPUAAK(data_stream, noise_channels, length, dt, init_dt, **kwargs)
     likelihood = like_class.NLL(iota, s, p, e, M, mu, gamma, psi, alph, theta_S,
                             phi_S, theta_K, phi_K, D)
-    num = 10
+    num = 1000
     st = time.perf_counter()
     for i in range(num):
         likelihood = like_class.NLL(iota, s, p, e, M, mu, gamma, psi, alph, theta_S,
@@ -144,7 +91,6 @@ def test():
                             phi_S, theta_K, phi_K, D, return_snr=True)
 
     print('SNR: ', snr, 'Likelihood:', likelihood)
-    import pdb; pdb.set_trace()
 
 if __name__ == "__main__":
     test()
diff --git a/pygpuAAK/pygpuAAK.py b/pygpuAAK/pygpuAAK.py
@@ -41,8 +41,8 @@ def __init__(self, data_stream, noise_channels, length, dt, init_dt, **kwargs):
         # whiten data stream
         self.data_stream = data_stream
 
-        self.noise_channel1_inv = 1./np.sqrt(noise_channels['channel1'])*np.sqrt(dt)  # dt dt df = dt
-        self.noise_channel2_inv = 1./np.sqrt(noise_channels['channel2'])*np.sqrt(dt)
+        self.noise_channel1_inv = 1./np.sqrt(noise_channels['channel1'])*np.sqrt(df)
+        self.noise_channel2_inv = 1./np.sqrt(noise_channels['channel2'])*np.sqrt(df)
 
         self.data_channel1 *= self.noise_channel1_inv
         self.data_channel2 *= self.noise_channel2_inv