1

I'm trying to use curve fitting to fit to a function that takes in an array of frequencies as x-values. I keep getting this error and have tried to reshape and make my y-values floats. Not really sure where to go from here since I keep getting the error. 

r_0 = (n_0 - n_1)/(n_0 + n_1)
r_1 = (n_1 - n_0)/(n_1+ n_0)
t_1 = 1 + r_1
t_0 = 1 + r_0
freq_values = np.linspace(108,200,1000)

#function to fit to
def T(freq,ghz=[]):
    Transmittance = []
    for ghz in freq_values:
        X_01 = np.exp((2j*np.pi*0.004724*1.528*ghz*10**9)/(3*10**8))   
        H_1 = X_01 + r_0*r_1*(X_01)**-1
        T = (t_0*t_1)/(H_1)
        hola = list(T.flat)
        Transmittance.append(hola*np.conj(hola))
    return Transmittance

x = freq_values
y = T(freq_values)
yy= np.reshape(y, len(y))
yyy= np.array(yy.real, dtype=float)

plt.plot(x,y, 'r', label = 'calculated')

#fit function, want to match/find optimized n value
def TT(freq,n,ghz=[]):
    Transmittance = []
    for ghz in freq_values:
        X_01 = np.exp((2j*np.pi*0.004724*n*ghz*10**9)/(3*10**8))
        H_1 = X_01 + r_0*r_1*(X_01)**-1
        T = (t_0*t_1)/(H_1)
        hola = list(T.flat)
        Transmittance.append(hola*np.conj(hola))
    return Transmittance

popt, pcov = curve_fit(TT, x, yyy)         #ydata = power (transmission) data
plt.plot(x, TT(x, *popt), 'b', label = 'fit')

plt.legend(loc='upper right')

I expect the code or fit to match the original plot but keep getting an error for yyy in curve_fit(TT,x,yyy)

JellyBelly
  • 13
  • 2
  • what are the values for n_0 and n_1 ? – mathew gunther Jun 03 '19 at 16:16
  • n_0 = 1 and n_1 = 1.528 – JellyBelly Jun 03 '19 at 16:19
  • Search for `[scipy] too deep` produced, https://stackoverflow.com/questions/53284842/scipy-curve-fit-for-two-dimensions-not-working-object-too-deep, https://stackoverflow.com/questions/51522103/valueerror-object-too-deep-for-desired-array-in-optimize-curve-fit – hpaulj Jun 03 '19 at 16:23
  • TT is returning a list of arrays. It should just be a numpy array, not a list of numpy arrays. – mathew gunther Jun 03 '19 at 16:53
  • Debugging helps: a full error traceback, shape and dtype of `x` and `yyy`, and description of what `TT` produces (a list of arrays?). `curve_fit` probably tries to convert the `TT` output into an array, `np.array(TT(...))`. What is that? – hpaulj Jun 03 '19 at 16:55

1 Answers1

0

The code block in question has been modified to work below. Note the difference between what the function returns. Also, I suspect you'll need to provide the fit with a good initial guess.

def TT(freq,n,ghz=[]):
    Transmittance = []
    for ghz in freq_values:
        X_01 = np.exp((2j*np.pi*0.004724*n*ghz*10**9)/(3*10**8))
        H_1 = X_01 + r_0*r_1*(X_01)**-1
        T = (t_0*t_1)/(H_1)
        hola = list(T.flat)
        Transmittance.append(hola*np.conj(hola))
    return np.real(np.array(Transmittance)[:,0])
mathew gunther
  • 180
  • 10