mrrs 0.1.1

Multi rate filtering tools
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
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301

/// Implement a 3-band channelizer with 50% overlap
///
/// Goal is to build a channelizer that takes an input, and returns data that
/// has been tuned by -fs/4, 0, and fs/4, and then ran through a half bandwidth
/// filter.
///
/// Multiple of those channelizers can be cascaded together, though some care
/// needs to be taken to remove duplication. Since the highest channel of one
/// band is the same as the lowest channel of another we only compute the
/// highest channel output on the last leg.
use super::{convolve, shift_down_fs_over_4, shift_up_fs_over_4};
use ndarray::{Array2, ArrayView2, ArrayViewMut2};
use num_complex::Complex;
use num_traits::{Float, NumAssignOps};

/// Perform a single-stage 3-band channelization with 50% overlap
///
/// Uses the naive tuning / convolving implementation to make a clear easily verifiable test case.
pub fn triband_cascade_noalloc<T: Float + NumAssignOps>(
    inp: ArrayView2<Complex<T>>,
    mut out: ArrayViewMut2<Complex<T>>,
    flt: &Vec<T>,
    integer_phase_offset: usize,
) {
    let n_inp_rows = inp.shape()[0];
    let n_inp_cols = inp.shape()[1];

    let n_out_rows = out.shape()[0];
    let n_out_cols = out.shape()[1];

    assert_eq!(n_out_rows, n_inp_rows - flt.len() + 1);
    assert_eq!(n_out_cols, n_inp_cols * 2 + 1);

    for icol in 0..n_inp_cols {
        //Making a copy to use simple naive implementations.
        let col_vec = inp.column(icol).to_vec();
        let col_slice = col_vec.as_slice();

        //While this says no alloc, we're clearly allocating here. The main goal was to get the
        //function signature right for future non-naive implementations.
        let tuned: Vec<Complex<T>> = shift_down_fs_over_4(&col_slice, integer_phase_offset);

        let low_flt = convolve(&tuned, &flt);
        let mid_flt = convolve(col_slice, &flt);

        assert_eq!(n_out_rows, low_flt.len());

        //Copying again to use the simple tuning / convolving methods.
        for irow in 0..n_out_rows {
            out[(irow, icol * 2 + 0)] = low_flt[irow];
            out[(irow, icol * 2 + 1)] = mid_flt[irow];
        }

        if icol == n_inp_cols - 1 {
            let high_tuned = shift_up_fs_over_4(&col_slice, integer_phase_offset);
            let high_flt = convolve(&high_tuned, &flt);
            for irow in 0..n_out_rows {
                out[(irow, icol * 2 + 2)] = high_flt[irow];
            }
        }
    }
}

pub fn triband_cascade_decimate_noalloc<T: Float + NumAssignOps>(
    inp: ArrayView2<Complex<T>>,
    mut out: ArrayViewMut2<Complex<T>>,
    flt: &Vec<T>,
    integer_phase_offset: usize,
    start_row: bool,
) {
    let n_inp_rows = inp.shape()[0];
    let n_inp_cols = inp.shape()[1];

    // Determine the starting index based on the start_row parameter
    let start_index = if start_row { 0 } else { 1 };

    // Calculate the number of decimated rows
    let decimated_rows = (n_inp_rows - start_index + 1) / 2;
    let mut decimated_inp = Array2::<Complex<T>>::zeros((decimated_rows, n_inp_cols));

    for icol in 0..n_inp_cols {
        for irow in 0..decimated_rows {
            decimated_inp[(irow, icol)] = inp[(irow * 2 + start_index, icol)];
        }
    }

    let n_out_rows = decimated_inp.shape()[0] - flt.len() + 1;
    let n_out_cols = decimated_inp.shape()[1] * 2 + 1;

    assert_eq!(n_out_rows, out.shape()[0]);
    assert_eq!(n_out_cols, out.shape()[1]);

    for icol in 0..decimated_inp.shape()[1] {
        let col_vec = decimated_inp.column(icol).to_vec();
        let col_slice = col_vec.as_slice();

        let tuned: Vec<Complex<T>> = shift_down_fs_over_4(&col_slice, integer_phase_offset);

        let low_flt = convolve(&tuned, &flt);
        let mid_flt = convolve(col_slice, &flt);

        assert_eq!(n_out_rows, low_flt.len());

        for irow in 0..n_out_rows {
            out[(irow, icol * 2 + 0)] = low_flt[irow];
            out[(irow, icol * 2 + 1)] = mid_flt[irow];
        }

        if icol == decimated_inp.shape()[1] - 1 {
            let high_tuned = shift_up_fs_over_4(&col_slice, integer_phase_offset);
            let high_flt = convolve(&high_tuned, &flt);
            for irow in 0..n_out_rows {
                out[(irow, icol * 2 + 2)] = high_flt[irow];
            }
        }
    }
}

pub fn triband_cascade<T: Float + NumAssignOps>(
    inp: ArrayView2<Complex<T>>,
    flt: &Vec<T>,
    integer_phase_offset: usize,
) -> Array2<Complex<T>> {
    let n_inp_rows = inp.shape()[0];
    let n_inp_cols = inp.shape()[1];

    let n_out_rows = n_inp_rows - flt.len() + 1;
    let n_out_cols = n_inp_cols * 2 + 1;

    let mut out = Array2::<Complex<T>>::zeros((n_out_rows, n_out_cols));
    triband_cascade_noalloc(inp, out.view_mut(), flt, integer_phase_offset);
    out
}

#[cfg(test)]
mod test {
    use super::*;
    use crate::get_hb_filter;
    use approx::assert_relative_eq;
    use ndarray::{Array1, Array2, Axis};
    use rustfft::{Fft, FftPlanner};
    use std::f64::consts::PI;
    use std::sync::Arc;

    fn find_local_maxima(vv: &Vec<Complex<f32>>) -> Vec<usize> {
        let mut idxs = vec![];
        for ii in 0..vv.len() {
            let lidx = if ii == 0 { vv.len() - 1 } else { ii - 1 };
            let uidx = if ii == vv.len() - 1 { 0 } else { ii + 1 };
            let nrm = vv[ii].norm_sqr();
            if vv[lidx].norm_sqr() <= nrm && vv[uidx].norm_sqr() <= nrm {
                idxs.push(ii);
            }
        }
        idxs
    }
    #[test]
    fn test_run_filter() {
        let flt = get_hb_filter::<f32>(31);
        let sigs = vec![
            Array1::from_iter((0..32768).map(|nn| -> Complex<f32> {
                Complex::<f32>::from_polar(0.5, (-0.17 * 2.0 * PI * nn as f64) as f32)
            })),
            Array1::from_iter((0..32768).map(|nn| -> Complex<f32> {
                Complex::<f32>::from_polar(1.0, (0.35 * 2.0 * PI * nn as f64) as f32)
            })),
            Array1::from_iter((0..32768).map(|nn| -> Complex<f32> {
                Complex::<f32>::from_polar(2.0, (0.21 * 2.0 * PI * nn as f64) as f32)
            })),
        ];
        let mut sum = sigs[0].clone();
        for arr in &sigs[1..] {
            sum += arr;
        }
        let sum_col: Array2<Complex<f32>> = sum.insert_axis(ndarray::Axis(1));

        let flt1 = triband_cascade(sum_col.view(), &flt, 0);

        let mut planner = FftPlanner::new();
        let fft: Arc<dyn Fft<f32>> = planner.plan_fft_forward(flt1.shape()[0]);
        for (icol, col) in flt1.axis_iter(Axis(1)).enumerate() {
            let (mut buffer, _) = col.to_owned().into_raw_vec_and_offset();
            assert_eq!(buffer.len(), 32768-30);
            fft.process(&mut buffer);
            let mut idxs: Vec<usize> = find_local_maxima(&buffer);
            idxs.sort_by(|&i, &j| {
                buffer[j]
                    .norm_sqr()
                    .partial_cmp(&buffer[i].norm_sqr())
                    .unwrap_or(std::cmp::Ordering::Equal)
            });
            let mut max = Complex::<f32>::new(0.0, 0.0);
            let mut max_idx = 0;
            for ii in 0..buffer.len() {
                if buffer[ii].norm_sqr() > max.norm_sqr() {
                    max = buffer[ii];
                    max_idx = ii;
                }
            }

            assert!(idxs.contains(&max_idx));

            for imax in 0..3 {
                let max_idx = idxs[imax];
                let max_freq = if max_idx < buffer.len() / 2 {
                    max_idx as f64 / buffer.len() as f64
                } else {
                    (max_idx as f64 / buffer.len() as f64) - 1.0
                };
                let max_db =
                    (buffer[max_idx].norm_sqr() / (buffer.len() as f32).powi(2)).log10() * 10.0;
                println!("{} {} {} {} {}", icol, imax, max_idx, max_freq, max_db);
                let exp_freq = match icol * 10 + imax {
                    00 => Some(-0.17 + 0.25),
                    10 => Some(0.21),
                    11 => Some(-0.17),
                    12 => Some(0.35),
                    20 => Some(0.21 - 0.25),
                    21 => Some(0.35 - 0.25),
                    _ => None,
                };
                if exp_freq.is_some() {
                    //1 / 30e3 is approximately the FFT bin resolution.
                    assert_relative_eq!(max_freq, exp_freq.unwrap(), epsilon = 1.0 / 30e3);
                }
            }
        }

        let flt2 = triband_cascade(flt1.view(), &flt, 0);
        let mut planner = FftPlanner::new();
        let fft: Arc<dyn Fft<f32>> = planner.plan_fft_forward(flt2.shape()[0]);
        for (icol, col) in flt2.axis_iter(Axis(1)).enumerate() {
            let (mut buffer, _) = col.to_owned().into_raw_vec_and_offset();
            assert_eq!(buffer.len(), 32768-60);
            fft.process(&mut buffer);
            let mut idxs: Vec<usize> = find_local_maxima(&buffer);
            idxs.sort_by(|&i, &j| {
                buffer[j]
                    .norm_sqr()
                    .partial_cmp(&buffer[i].norm_sqr())
                    .unwrap_or(std::cmp::Ordering::Equal)
            });
            let mut max = Complex::<f32>::new(0.0, 0.0);
            let mut max_idx = 0;
            for ii in 0..buffer.len() {
                if buffer[ii].norm_sqr() > max.norm_sqr() {
                    max = buffer[ii];
                    max_idx = ii;
                }
            }

            assert!(idxs.contains(&max_idx));

            for imax in 0..3 {
                let max_idx = idxs[imax];
                let max_freq = if max_idx < buffer.len() / 2 {
                    max_idx as f64 / buffer.len() as f64
                } else {
                    (max_idx as f64 / buffer.len() as f64) - 1.0
                };
                let max_db =
                    (buffer[max_idx].norm_sqr() / (buffer.len() as f32).powi(2)).log10() * 10.0;
                println!("{} {} {} {} {}", icol, imax, max_idx, max_freq, max_db);
                let exp_freq = match icol * 10 + imax {
                    30 => Some(0.21),
                    31 => Some(-0.17),
                    _ => None,
                };
                if exp_freq.is_some() {
                    //1 / 30e3 is approximately the FFT bin resolution.
                    assert_relative_eq!(max_freq, exp_freq.unwrap(), epsilon = 1.0 / 30e3);
                }
            }
        }
    }

    #[test]
    fn test_row_major() {
        let nrows = 30;
        let ncols = 5;
        let mut arr = Array2::<Complex<f32>>::zeros((nrows, ncols));
        for ii in 0..nrows {
            for jj in 0..ncols {
                arr[(ii, jj)] = Complex::<f32>::new((ii * ncols + jj) as f32, 0.0);
            }
        }
        assert_eq!(arr.shape()[0], 30);
        assert_eq!(arr.shape()[1], 5);
        let ptr: *const Complex<f32> = arr.as_ptr();
        unsafe {
            for ii in 0..nrows {
                for jj in 0..ncols {
                    let offs = ii * ncols + jj;
                    let val = *(ptr.wrapping_add(offs));
                    assert_eq!(val.re, (ii * ncols + jj) as f32);
                }
            }
        }
        println!("{}", arr);
    }
}