CODEWITHSUNDEEP
pythonmatlabnumpy
marriam nayyer
marriam nayyer

Reputation: 697

numpy:doesnt give correct for negative powers

i am trying to convert a matlab code in numpy for calculating bit error rate a piece of code is making problem for me this is the matlab code i wanted to convert

SNR=6:22;
display(SNR)
display(length(SNR))
BER=zeros(1,length(SNR));

display(BER)
display(length(BER))
Es=10;


for ii=1:length(SNR)
   variance=Es*10^(-SNR(ii)/10);
   std_dev=sqrt(variance/2);
   noise=(randn(1,length(S))+sqrt(-1)*randn(1,length(S)))*std_dev;
   S_noisy=S+noise;


end
display(variance)

python code SNR=arange(6,23,1)

BER=zeros(len(SNR))  
print(BER)
Es=10


for ii in arange(0,len(SNR)):

   variance=Es*10**(-SNR[ii]/10)
   std_dev=cmath.sqrt(variance/2)
   noise=(np.random.randn(len(S))+cmath.sqrt(-1)*np.random.randn(len(S))) *std_dev
   S_noisy=S+noise

  print(variance)

answer of variance should be 0.063 bt in python it gives 0.01 plzz help

Upvotes: 2

Views: 564

Answers (1)

unutbu
unutbu

Reputation: 879471

SNR is of dtype int32 be default. Dividing an int by an int gives you an int (or raises a ZeroDivisionError) in Python2. So

SNR[ii]/10

gives you the wrong result:

In [15]: SNR
Out[15]: array([ 6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22])

In [16]: SNR/10
Out[16]: array([0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2])

To fix, either put

from __future__ import division

at the beginning of the python code (before the import statements), or else use

variance = Es*10**(-SNR[ii]/10.0)

With this change, the end result is 0.063095734448.

Note: In Python3, int divided by int will return a float by default.

For better performance when using NumPy, you will want to replace Python loops with operations on whole NumPy arrays when possible. Your code would be written like this:

import numpy as np
SNR = np.arange(6, 23)
BER = np.zeros(len(SNR))
print(BER)
Es = 10

variance = Es * 10 ** (-SNR / 10.0)
std_dev = np.sqrt(variance / 2)
noise = (np.random.randn(len(SNR)) + 1j * np.random.randn(len(SNR))) * std_dev
S_noisy = SNR + noise
print(variance[-1])

Upvotes: 3

Related Questions