basic_dsp 0.2.1

Digital signal processing based on 1xN (one times N) or Nx1 vectors in real or complex number space. Vectors come with basic arithmetic, convolution, Fourier transformation and interpolation operations.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237

extern crate basic_dsp;
extern crate rand;
extern crate num;
pub mod tools;

#[cfg(feature = "slow_test")]
mod slow_test {
    use basic_dsp::*;
    use basic_dsp::conv_types::*;
    use basic_dsp::window_functions::*;
    use tools::*;
    use std::os::raw::c_void;
    
    #[test]
    fn compare_conv_freq_multiplication_for_rc() {
        for iteration in 0 .. 3 {
            let a = create_data_even(201602211, iteration, 1001, 2000);
            let delta = create_delta(201602212, iteration);
            let time = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let fun: RaisedCosineFunction<f32> = RaisedCosineFunction::new(0.35);
            let freq = time.clone().fft().unwrap();
            let points = time.points();
            let ratio = create_delta(20160229, iteration).abs() / 10.0; // Should get us a range [0.0 .. 1.0] and hopefully we are not that unlucky to get 0.0
            let time_res = time.convolve(&fun as &RealImpulseResponse<f32>, ratio, points).unwrap();
            let freq_res = freq.multiply_frequency_response(&fun as &RealFrequencyResponse<f32>, 1.0 / ratio).unwrap();
            let ifreq_res = freq_res.ifft().unwrap();
            let left = &ifreq_res.data();
            let right = &time_res.data();
            assert_vector_eq_with_reason_and_tolerance(&left, &right, 0.2, "Results should match independent if done in time or frequency domain");
        }
    }
    
    #[test]
    fn compare_conv_freq_multiplication_for_sinc() {
        for iteration in 0 .. 3 {
            let a = create_data_even(201602214, iteration, 2001, 4000);
            let delta = create_delta(201602215, iteration);
            let time = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let fun: SincFunction<f32> = SincFunction::new();
            let freq = time.clone().fft().unwrap();
            let points = time.points();
            let ratio = create_delta(201602216, iteration).abs() / 20.0 + 0.5; // Should get us a range [0.5 .. 1.0]
            let time_res = time.convolve(&fun as &RealImpulseResponse<f32>, ratio, points).unwrap();
            let freq_res = freq.multiply_frequency_response(&fun as &RealFrequencyResponse<f32>, 1.0 / ratio).unwrap();
            let ifreq_res = freq_res.ifft().unwrap();
            let left = &ifreq_res.data();
            let right = &time_res.data();
            assert_vector_eq_with_reason_and_tolerance(&left, &right, 0.3, "Results should match independent if done in time or frequency domain");
        }
    }
    
    #[test]
    fn compare_optimized_and_non_optimized_conv() {
        for iteration in 0 .. 3 {
            // This offset is small enough to now have a big impact on the results (for the RC function)
            // but the code will use a different non-optimized branch since it won't recognize ratio as an
            // integer
            let offset = 2e-6; 
            let a = create_data_even(201602217, iteration, 2002, 4000);
            let b = create_data_even(201602218, iteration, 50, 202);
            let delta = create_delta(201602219, iteration);
            let time = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let fun: RaisedCosineFunction<f32> = RaisedCosineFunction::new(0.35);
            let ratio = iteration as f32 + 1.0;
            let left = time.clone().convolve(&fun as &RealImpulseResponse<f32>, ratio, b.len()).unwrap();
            let right = time.convolve(&fun as &RealImpulseResponse<f32>, ratio + offset, b.len()).unwrap();
            assert_vector_eq_with_reason_and_tolerance(&left.data(), &right.data(), 0.1, "Results should match independent if done in optimized or non optimized code branch");
        }
    }
    
    #[test]
    fn compare_vector_conv_freq_multiplication() {
        for iteration in 0 .. 3 {
            let a = create_data_even(201601171, iteration, 502, 1000);
            let b = create_data_with_len(201601172, iteration, a.len());
            let delta = create_delta(201601173, iteration);
            let time1 = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let time2 = ComplexTimeVector32::from_interleaved_with_delta(&b, delta);
            let left = time1.clone().convolve_vector(&time2).unwrap();
            let freq1 = time1.fft().unwrap();
            let freq2 = time2.fft().unwrap();
            let freq_res = freq1.multiply_vector(&freq2).unwrap();
            let right = freq_res.ifft().unwrap();
            assert_vector_eq_with_reason_and_tolerance(
                left.data(), 
                &conv_swap(right.data())[0..left.len()], 
                0.2, 
                "Results should match independent if done in time or frequency domain");
        }
    }
    
    #[test]
    fn compare_smaller_vector_conv_with_zero_padded_conv() {
        for iteration in 0 .. 3 {
            let a = create_data_even(201601174, iteration, 1002, 2000);
            let b = create_data_even(201601175, iteration, 50, 202);
            let delta = create_delta(201601176, iteration);
            let time1 = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let time2 = ComplexTimeVector32::from_interleaved_with_delta(&b, delta);
            let left = time1.clone().convolve_vector(&time2).unwrap();
            let time2 = ComplexTimeVector32::from_interleaved_with_delta(&conv_zero_pad(&b, time1.len(), true), delta);
            let right = time1.convolve_vector(&time2).unwrap();
            assert_vector_eq_with_reason_and_tolerance(
                left.data(), 
                right.data(), 
                0.2, 
                "Results should match independent if done with a smaller vector or with a zero padded vector of the same size");
        }
    }
    
    #[test]
    fn compare_smaller_vector_conv_with_zero_padded_conv_real() {
        for iteration in 0 .. 3 {
            let a = create_data_even(201601177, iteration, 1002, 2000);
            let b = create_data_even(201601178, iteration, 50, 202);
            let delta = create_delta(201601179, iteration);
            let time1 = RealTimeVector32::from_array_with_delta(&a, delta);
            let time2 = RealTimeVector32::from_array_with_delta(&b, delta);
            let left = time1.clone().convolve_vector(&time2).unwrap();
            let time2 = RealTimeVector32::from_array_with_delta(&conv_zero_pad(&b, time1.len(), false), delta);
            let right = time1.convolve_vector(&time2).unwrap();
            assert_vector_eq_with_reason_and_tolerance(
                left.data(), 
                right.data(), 
                0.2, 
                "Results should match independent if done with a smaller vector or with a zero padded vector of the same size");
        }
    }
    
    fn conv_zero_pad(data: &[f32], len: usize, is_complex: bool) -> Vec<f32> {
        if is_complex {
            let mut result = vec![0.0; len];
            let points = len / 2;
            let data_points = data.len() / 2;
            let diff = points - data_points;
            let left = diff - diff / 2;
            for i in 0..data.len() {
                result[2 * left + i] = data[i];
            }
            result
        }
        else {
            let mut result = vec![0.0; len];
            let data_len = data.len();
            let diff = len - data_len;
            let left = diff - diff / 2;
            for i in 0..data.len() {
                result[left + i] = data[i];
            }
            result
        }
    }
    
    /// This kind of swapping is necessary since we defined
    /// our conv to have 0s in the center
    fn conv_swap(data: &[f32]) -> Vec<f32> {
        let len = data.len();
        let mut result = vec![0.0; len];
        let points = len / 2;
        let half = 2 *(points / 2 + 1);
        for i in 0..len {
            if i < half {
                result[i] = data[len - half + i];
            }
            else if i >= len - half {
                result[i] = data[i - half];
            }
            else /* Center */ { 
                result[i] = data[i]; 
            }
        }
        result
    }
    
    /// Calls to another window with the only
    /// difference that it doesn't allow to make use of symmetry
    fn unsym_triag_window() -> ForeignWindowFunction<f32> {
        ForeignWindowFunction {
            window_data: 0,
            window_function: call_triag,
            is_symmetric: false
        }
    }
       
    extern fn call_triag(_arg: *const c_void, i: usize, points: usize) -> f32 {
        let triag: &WindowFunction<f32> = &TriangularWindow;
        triag.window(i, points)
    }
    
    /// Calls to another window with the only
    /// difference that it doesn't allow to make use of symmetry
    fn unsym_rc_mul() -> ForeignRealConvolutionFunction<f32> {
        ForeignRealConvolutionFunction {
            conv_data: 0,
            conv_function: call_freq_rc,
            is_symmetric: false
        }
    }
       
    extern fn call_freq_rc(_arg: *const c_void, x: f32) -> f32 {
        let rc: &RealFrequencyResponse<f32> = &RaisedCosineFunction::new(0.35);
        rc.calc(x)
    }
    
    #[test]
    fn compare_sym_optimized_window_with_normal_version() {
        parameterized_vector_test(|iteration, range| {
            let a = create_data_even(20160116, iteration, range.start, range.end);
            let delta = create_delta(201601161, iteration);
            let time = ComplexTimeVector32::from_interleaved_with_delta(&a, delta);
            let triag_sym = TriangularWindow;
            let triag_unsym = unsym_triag_window();
            let result_sym = time.clone().apply_window(&triag_sym).unwrap();
            let result_unsym = time.apply_window(&triag_unsym).unwrap();
            let left = &result_sym.data();
            let right = &result_unsym.data();
            assert_vector_eq_with_reason_and_tolerance(&left, &right, 1e-2, "Results should match with or without symmetry optimization");
        });
    }
    
    #[test]
    fn compare_sym_optimized_freq_mul_with_normal_version() {
        parameterized_vector_test(|iteration, range| {
            let a = create_data_even(201601162, iteration, range.start, range.end);
            let delta = create_delta(201601163, iteration);
            let freq = ComplexFreqVector32::from_interleaved_with_delta(&a, delta);
            let rc_sym: RaisedCosineFunction<f32> = RaisedCosineFunction::new(0.35);
            let rc_unsym = unsym_rc_mul();
            let ratio = create_delta(201601164, iteration).abs() / 20.0 + 0.5; // Should get us a range [0.5 .. 1.0]
            let result_sym = freq.clone().multiply_frequency_response(&rc_sym as &RealFrequencyResponse<f32>, 1.0 / ratio).unwrap();
            let result_unsym = freq.multiply_frequency_response(&rc_unsym as &RealFrequencyResponse<f32>, 1.0 / ratio).unwrap();
            let left = &result_sym.data();
            let right = &result_unsym.data();
            assert_vector_eq_with_reason_and_tolerance(&left, &right, 1e-2, "Results should match with or without symmetry optimization");
        });
    }
}