[Commit-gnuradio] [gnuradio] 02/13: fix wrong laplacian random numbers and add testcase

[git]

This is an automated email from the git hooks/post-receive script.

jcorgan pushed a commit to branch master
in repository gnuradio.

commit ebad2162b40b2b144449ff2927dbe89c683c4972
Author: Stefan <address@hidden>
Date:   Tue Sep 1 13:21:51 2015 +0200

    fix wrong laplacian random numbers and add testcase
---
 gnuradio-runtime/include/gnuradio/random.h       | 21 ++++++++--
 gnuradio-runtime/lib/math/random.cc              | 23 ++---------
 gnuradio-runtime/python/gnuradio/gr/qa_random.py | 51 +++++++++++++++---------
 3 files changed, 54 insertions(+), 41 deletions(-)

diff --git a/gnuradio-runtime/include/gnuradio/random.h 
b/gnuradio-runtime/include/gnuradio/random.h
index e01fcb7..454c88e 100644
--- a/gnuradio-runtime/include/gnuradio/random.h
+++ b/gnuradio-runtime/include/gnuradio/random.h
@@ -60,18 +60,33 @@ namespace gr {
     void reseed(long seed);
 
     /*!
-     * \brief uniform random deviate in the range [0.0, 1.0)
+     * \brief Uniform random numbers in the range [0.0, 1.0)
      */
     float ran1();
 
     /*!
-     * \brief normally distributed deviate with zero mean and variance 1
+     * \brief Normally distributed random numbers (Gaussian distribution with 
zero mean and variance 1)
      */
     float gasdev();
 
+    /*!
+     * \brief Laplacian distributed random numbers with zero mean and variance 
1
+     */
     float laplacian();
-    float impulse(float factor);
+
+    /*!
+     * \brief Rayleigh distributed random numbers (zero mean and variance 1 
for the underlying Gaussian distributions)
+     */
     float rayleigh();
+
+    /*!
+     * \brief FIXME: add description
+     */
+    float impulse(float factor);
+
+    /*!
+     * \brief Normally distributed random numbers with zero mean and variance 
1 on real and imaginary part. This results in a Rayleigh distribution for the 
amplitude and an uniform distribution for the phase.
+     */
     gr_complex rayleigh_complex();
   };
 
diff --git a/gnuradio-runtime/lib/math/random.cc 
b/gnuradio-runtime/lib/math/random.cc
index 7170f27..ecf70e0 100644
--- a/gnuradio-runtime/lib/math/random.cc
+++ b/gnuradio-runtime/lib/math/random.cc
@@ -130,18 +130,12 @@ namespace gr {
     return d_gset;
   }
 
-  /*
-   * Copied from The KC7WW / OH2BNS Channel Simulator
-   * FIXME Need to check how good this is at some point
-   */
   float
   random::laplacian()
   {
-    float z = ran1();
-    if(z < 0.5)
-      return log(2.0 * z) / M_SQRT2;
-    else
-      return -log(2.0 * (1.0 - z)) / M_SQRT2;
+    float z = ran1()-0.5;
+    if(z>0) return -log(1-2*z);
+    else return log(1+2*z);
   }
 
   /*
@@ -161,10 +155,6 @@ namespace gr {
 
   /*
    * Complex rayleigh is really gaussian I and gaussian Q
-   * It can also be generated by real rayleigh magnitude and
-   * uniform random angle
-   * Adapted from The KC7WW / OH2BNS Channel Simulator
-   * FIXME Need to check how good this is at some point
    */
   gr_complex
   random::rayleigh_complex()
@@ -172,13 +162,6 @@ namespace gr {
     return gr_complex(gasdev(),gasdev());
   }
 
-  /*   Other option
-       mag = rayleigh();
-       ang = 2.0 * M_PI * RNG();
-       *Rx = rxx * cos(z);
-       *Iy = rxx * sin(z);
-       */
-
   float
   random::rayleigh()
   {
diff --git a/gnuradio-runtime/python/gnuradio/gr/qa_random.py 
b/gnuradio-runtime/python/gnuradio/gr/qa_random.py
index 39d75f3..ee40183 100644
--- a/gnuradio-runtime/python/gnuradio/gr/qa_random.py
+++ b/gnuradio-runtime/python/gnuradio/gr/qa_random.py
@@ -22,78 +22,93 @@
 
 from gnuradio import gr, gr_unittest
 import numpy as np
-from scipy.stats import norm, laplace
+from scipy.stats import norm, laplace, rayleigh
 
 class test_random(gr_unittest.TestCase):
 
+    num_tests = 10000
+
     # Disclaimer
     def test_0(self):
-        print 'NOTE: Following tests are not statistically significant! Check 
out fulltest_random.py for full testing.'
+        print 'NOTE: Following tests are not statistically significant!'
+        print 'Realisations per test:',self.num_tests
         self.assertEqual(1,1)
 
     # Check for range [0,1) of uniform distributed random numbers and print 
minimal and maximal value
     def test_1(self):
         print '# TEST 1'
         print 'Uniform distributed numbers: Range'
-        num_tests = 10000
-        values = np.zeros(num_tests)
+        values = np.zeros(self.num_tests)
         rndm = gr.random()
-        for k in range(num_tests):
+        for k in range(self.num_tests):
             values[k] = rndm.ran1()
         for value in values:
             self.assertLess(value, 1)
             self.assertGreaterEqual(value, 0)
-        print 'Uniform random numbers (num/min/max):', num_tests, min(values), 
max(values)
+        print 'Uniform random numbers (num/min/max):', self.num_tests, 
min(values), max(values)
 
     # Check uniformly distributed random numbers on uniformity (without 
assert, only printing)
     def test_2(self):
         print '# TEST 2'
         print 'Uniform random numbers: Distribution'
-        num_tests = 10000
         num_bins = 11
-        values = np.zeros(num_tests)
+        values = np.zeros(self.num_tests)
         rndm = gr.random()
-        for k in range(num_tests):
+        for k in range(self.num_tests):
             values[k] = rndm.ran1()
         bins = np.linspace(0,1,num_bins) # These are the bin edges!
         hist = np.histogram(values,bins)
         print 'Lower edge bin / upper edge bin / count / expected'
         for k in range(len(hist[0])):
-                print hist[1][k], hist[1][k+1], hist[0][k], 
float(num_tests)/(num_bins-1)
+                print hist[1][k], hist[1][k+1], hist[0][k], 
float(self.num_tests)/(num_bins-1)
 
     # Check distribution of normally (gaussian, mean=0, variance=1) 
distributed random numbers (no assert)
     def test_3(self):
         print '# TEST 3'
         print 'Normal random numbers: Distribution'
-        num_tests = 10000
         num_bins = 11
         hist_range = [-5,5]
-        values = np.zeros(num_tests)
+        values = np.zeros(self.num_tests)
         rndm = gr.random()
-        for k in range(num_tests):
+        for k in range(self.num_tests):
             values[k] = rndm.gasdev()
         bins = np.linspace(hist_range[0],hist_range[1],num_bins)
         hist = np.histogram(values,bins)
         print 'Lower edge bin / upper edge bin / count / expected'
         for k in range(len(hist[0])):
-            print hist[1][k], hist[1][k+1], hist[0][k], 
float(norm.cdf(hist[1][k+1])-norm.cdf(hist[1][k]))*num_tests
+            print hist[1][k], hist[1][k+1], hist[0][k], 
float(norm.cdf(hist[1][k+1])-norm.cdf(hist[1][k]))*self.num_tests
 
     # Check distribution of laplacian (mean=0, variance=1) distributed random 
numbers (no assert)
     def test_4(self):
         print '# TEST 4'
         print 'Laplacian random numbers: Distribution'
-        num_tests = 100000
         num_bins = 11
         hist_range = [-5,5]
-        values = np.zeros(num_tests)
+        values = np.zeros(self.num_tests)
         rndm = gr.random()
-        for k in range(num_tests):
+        for k in range(self.num_tests):
             values[k] = rndm.laplacian()
         bins = np.linspace(hist_range[0],hist_range[1],num_bins)
         hist = np.histogram(values,bins)
         print 'Lower edge bin / upper edge bin / count / expected'
         for k in range(len(hist[0])):
-            print hist[1][k], hist[1][k+1], hist[0][k], 
float(laplace.cdf(hist[1][k+1])-laplace.cdf(hist[1][k]))*num_tests
+            print hist[1][k], hist[1][k+1], hist[0][k], 
float(laplace.cdf(hist[1][k+1])-laplace.cdf(hist[1][k]))*self.num_tests
+
+    # Check distribution of laplacian (mean=0, variance=1) distributed random 
numbers (no assert)
+    def test_5(self):
+        print '# TEST 5'
+        print 'Rayleigh random numbers: Distribution'
+        num_bins = 11
+        hist_range = [0,10]
+        values = np.zeros(self.num_tests)
+        rndm = gr.random()
+        for k in range(self.num_tests):
+            values[k] = rndm.rayleigh()
+        bins = np.linspace(hist_range[0],hist_range[1],num_bins)
+        hist = np.histogram(values,bins)
+        print 'Lower edge bin / upper edge bin / count / expected'
+        for k in range(len(hist[0])):
+            print hist[1][k], hist[1][k+1], hist[0][k], 
float(rayleigh.cdf(hist[1][k+1])-rayleigh.cdf(hist[1][k]))*self.num_tests
 
 if __name__ == '__main__':
     gr_unittest.run(test_random, "test_random.xml")