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
302
303
304
305
306
307
308
309
310
311
312
313
314
//! Fixed-point Fast Fourier Transform
//!
//! This crate is intended for use with cores without an FPU and thus
//! can perform a fixed point FFT more quickly. The FFT code uses
//! a signed 16 bit number format, which is interpreted as a Q15
//! format (i.e. one sign bit, 15 fractional bits).
//!
//! The code was written under the assumption that a Count Leading Zeros (CLZ)
//! instruction is available on the target architecture.
//!
//! # Examples
//!
//! ## FFT with complex input data
//!
//! ```
//! use fixed_fft::{Direction, fft_radix2_q15};
//! use num_complex::Complex;
//!
//! let mut samples = [Complex::new(1000, 0); 4];
//! fft_radix2_q15(&mut samples, Direction::Forward).unwrap();
//! assert_eq!(samples[0], Complex::new(4000, 0));
//! assert_eq!(samples[1], Complex::new(0, 0));
//! assert_eq!(samples[2], Complex::new(0, 0));
//! assert_eq!(samples[3], Complex::new(0, 0));
//! ```
//!
//! ## FFT with real input data
//!
//! Note that the length of the output data is N / 2 + 1, where N is the FFT length.
//!
//! ```
//! use fixed_fft::fft_radix2_real_q15;
//! use num_complex::Complex;
//!
//! let mut samples = [1000; 4];
//! let mut result = [Complex::new(0, 0); 3];
//! fft_radix2_real_q15(&mut samples, &mut result, false).unwrap();
//! assert_eq!(result[0], Complex::new(4000, 0));
//! assert_eq!(result[1], Complex::new(0, 0));
//! assert_eq!(result[2], Complex::new(0, 0));
//! ```
//!

#![no_std]

use core::slice;
use num_complex::Complex;

type Q15 = i16;

pub enum Direction {
    /// Normal forward FFT
    Forward,

    /// Forward FFT with output values scaled by 1/N
    ///
    /// This variant can be used to avoid numerical overflows.
    ForwardScaled,

    /// Inverse FFT
    Inverse,
}

#[derive(Debug, Clone, Copy, PartialEq)]
pub enum FFTError {
    InvalidFFTSize,
    InvalidDataSize,
}

const SINETAB_LD_N: usize = 10;
const SINETAB_N: usize = 1 << SINETAB_LD_N;

/// Sine table used to calculate twiddle factors
static SINETAB: [i16; SINETAB_N - SINETAB_N / 4] = [
    0, 201, 402, 603, 804, 1005, 1206, 1407, 1608, 1809, 2009, 2210, 2410, 2611, 2811, 3012, 3212,
    3412, 3612, 3811, 4011, 4210, 4410, 4609, 4808, 5007, 5205, 5404, 5602, 5800, 5998, 6195, 6393,
    6590, 6786, 6983, 7179, 7375, 7571, 7767, 7962, 8157, 8351, 8545, 8739, 8933, 9126, 9319, 9512,
    9704, 9896, 10087, 10278, 10469, 10659, 10849, 11039, 11228, 11417, 11605, 11793, 11980, 12167,
    12353, 12539, 12725, 12910, 13094, 13279, 13462, 13645, 13828, 14010, 14191, 14372, 14553,
    14732, 14912, 15090, 15269, 15446, 15623, 15800, 15976, 16151, 16325, 16499, 16673, 16846,
    17018, 17189, 17360, 17530, 17700, 17869, 18037, 18204, 18371, 18537, 18703, 18868, 19032,
    19195, 19357, 19519, 19680, 19841, 20000, 20159, 20317, 20475, 20631, 20787, 20942, 21096,
    21250, 21403, 21554, 21705, 21856, 22005, 22154, 22301, 22448, 22594, 22739, 22884, 23027,
    23170, 23311, 23452, 23592, 23731, 23870, 24007, 24143, 24279, 24413, 24547, 24680, 24811,
    24942, 25072, 25201, 25329, 25456, 25582, 25708, 25832, 25955, 26077, 26198, 26319, 26438,
    26556, 26674, 26790, 26905, 27019, 27133, 27245, 27356, 27466, 27575, 27683, 27790, 27896,
    28001, 28105, 28208, 28310, 28411, 28510, 28609, 28706, 28803, 28898, 28992, 29085, 29177,
    29268, 29358, 29447, 29534, 29621, 29706, 29791, 29874, 29956, 30037, 30117, 30195, 30273,
    30349, 30424, 30498, 30571, 30643, 30714, 30783, 30852, 30919, 30985, 31050, 31113, 31176,
    31237, 31297, 31356, 31414, 31470, 31526, 31580, 31633, 31685, 31736, 31785, 31833, 31880,
    31926, 31971, 32014, 32057, 32098, 32137, 32176, 32213, 32250, 32285, 32318, 32351, 32382,
    32412, 32441, 32469, 32495, 32521, 32545, 32567, 32589, 32609, 32628, 32646, 32663, 32678,
    32692, 32705, 32717, 32728, 32737, 32745, 32752, 32757, 32761, 32765, 32766, 32767, 32766,
    32765, 32761, 32757, 32752, 32745, 32737, 32728, 32717, 32705, 32692, 32678, 32663, 32646,
    32628, 32609, 32589, 32567, 32545, 32521, 32495, 32469, 32441, 32412, 32382, 32351, 32318,
    32285, 32250, 32213, 32176, 32137, 32098, 32057, 32014, 31971, 31926, 31880, 31833, 31785,
    31736, 31685, 31633, 31580, 31526, 31470, 31414, 31356, 31297, 31237, 31176, 31113, 31050,
    30985, 30919, 30852, 30783, 30714, 30643, 30571, 30498, 30424, 30349, 30273, 30195, 30117,
    30037, 29956, 29874, 29791, 29706, 29621, 29534, 29447, 29358, 29268, 29177, 29085, 28992,
    28898, 28803, 28706, 28609, 28510, 28411, 28310, 28208, 28105, 28001, 27896, 27790, 27683,
    27575, 27466, 27356, 27245, 27133, 27019, 26905, 26790, 26674, 26556, 26438, 26319, 26198,
    26077, 25955, 25832, 25708, 25582, 25456, 25329, 25201, 25072, 24942, 24811, 24680, 24547,
    24413, 24279, 24143, 24007, 23870, 23731, 23592, 23452, 23311, 23170, 23027, 22884, 22739,
    22594, 22448, 22301, 22154, 22005, 21856, 21705, 21554, 21403, 21250, 21096, 20942, 20787,
    20631, 20475, 20317, 20159, 20000, 19841, 19680, 19519, 19357, 19195, 19032, 18868, 18703,
    18537, 18371, 18204, 18037, 17869, 17700, 17530, 17360, 17189, 17018, 16846, 16673, 16499,
    16325, 16151, 15976, 15800, 15623, 15446, 15269, 15090, 14912, 14732, 14553, 14372, 14191,
    14010, 13828, 13645, 13462, 13279, 13094, 12910, 12725, 12539, 12353, 12167, 11980, 11793,
    11605, 11417, 11228, 11039, 10849, 10659, 10469, 10278, 10087, 9896, 9704, 9512, 9319, 9126,
    8933, 8739, 8545, 8351, 8157, 7962, 7767, 7571, 7375, 7179, 6983, 6786, 6590, 6393, 6195, 5998,
    5800, 5602, 5404, 5205, 5007, 4808, 4609, 4410, 4210, 4011, 3811, 3612, 3412, 3212, 3012, 2811,
    2611, 2410, 2210, 2009, 1809, 1608, 1407, 1206, 1005, 804, 603, 402, 201, 0, -201, -402, -603,
    -804, -1005, -1206, -1407, -1608, -1809, -2009, -2210, -2410, -2611, -2811, -3012, -3212,
    -3412, -3612, -3811, -4011, -4210, -4410, -4609, -4808, -5007, -5205, -5404, -5602, -5800,
    -5998, -6195, -6393, -6590, -6786, -6983, -7179, -7375, -7571, -7767, -7962, -8157, -8351,
    -8545, -8739, -8933, -9126, -9319, -9512, -9704, -9896, -10087, -10278, -10469, -10659, -10849,
    -11039, -11228, -11417, -11605, -11793, -11980, -12167, -12353, -12539, -12725, -12910, -13094,
    -13279, -13462, -13645, -13828, -14010, -14191, -14372, -14553, -14732, -14912, -15090, -15269,
    -15446, -15623, -15800, -15976, -16151, -16325, -16499, -16673, -16846, -17018, -17189, -17360,
    -17530, -17700, -17869, -18037, -18204, -18371, -18537, -18703, -18868, -19032, -19195, -19357,
    -19519, -19680, -19841, -20000, -20159, -20317, -20475, -20631, -20787, -20942, -21096, -21250,
    -21403, -21554, -21705, -21856, -22005, -22154, -22301, -22448, -22594, -22739, -22884, -23027,
    -23170, -23311, -23452, -23592, -23731, -23870, -24007, -24143, -24279, -24413, -24547, -24680,
    -24811, -24942, -25072, -25201, -25329, -25456, -25582, -25708, -25832, -25955, -26077, -26198,
    -26319, -26438, -26556, -26674, -26790, -26905, -27019, -27133, -27245, -27356, -27466, -27575,
    -27683, -27790, -27896, -28001, -28105, -28208, -28310, -28411, -28510, -28609, -28706, -28803,
    -28898, -28992, -29085, -29177, -29268, -29358, -29447, -29534, -29621, -29706, -29791, -29874,
    -29956, -30037, -30117, -30195, -30273, -30349, -30424, -30498, -30571, -30643, -30714, -30783,
    -30852, -30919, -30985, -31050, -31113, -31176, -31237, -31297, -31356, -31414, -31470, -31526,
    -31580, -31633, -31685, -31736, -31785, -31833, -31880, -31926, -31971, -32014, -32057, -32098,
    -32137, -32176, -32213, -32250, -32285, -32318, -32351, -32382, -32412, -32441, -32469, -32495,
    -32521, -32545, -32567, -32589, -32609, -32628, -32646, -32663, -32678, -32692, -32705, -32717,
    -32728, -32737, -32745, -32752, -32757, -32761, -32765, -32766,
];

/// complex multiplication in Q15 format
fn q15_cmul(a: Complex<Q15>, b: Complex<Q15>) -> Complex<Q15> {
    let rnd: i32 = 1 << 14; // constant for rounding
    Complex::new(
        ((a.re as i32 * b.re as i32 - a.im as i32 * b.im as i32 + rnd) >> 15) as i16,
        ((a.re as i32 * b.im as i32 + a.im as i32 * b.re as i32 + rnd) >> 15) as i16,
    )
}

fn check_size(size: usize) -> Result<(), FFTError> {
    if !(2..=SINETAB_N).contains(&size) {
        return Err(FFTError::InvalidFFTSize);
    }
    const MSB: usize = (usize::MAX >> 1) + 1;
    let allowed_size = MSB >> (size.leading_zeros());
    if allowed_size != size {
        return Err(FFTError::InvalidFFTSize);
    }
    Ok(())
}

/// Radix-2 in-place FFT
///
/// Perform forward/inverse fast Fourier transform using the radix-2 algorithm.
/// The the length of slice```data``` corresponds to the FFT size and must be a
/// power of 2. The maximum FFT size is 1024.
///
/// Returns `FFTError::InvalidFFTSize` if the requested FFT size is not supported.
pub fn fft_radix2_q15(data: &mut [Complex<Q15>], dir: Direction) -> Result<(), FFTError> {
    let fft_size = data.len();
    check_size(fft_size)?;
    let (inverse, shift) = match dir {
        Direction::Forward => (false, 0),
        Direction::ForwardScaled => (false, 1),
        Direction::Inverse => (true, 1),
    };

    // re-order data
    let mut mr = 0usize; // bit reverse counter
    for m in 1..fft_size {
        const MSB: usize = (usize::MAX >> 1) + 1;
        let l = MSB >> (fft_size - 1 - mr).leading_zeros();
        mr = (mr & (l - 1)) + l;
        if mr > m {
            data.swap(m, mr);
        }
    }

    let mut stage = 1;
    let mut sine_step = (SINETAB_LD_N - 1) as isize;
    // loop over stages
    while stage < fft_size {
        // loop over groups with the same twiddle factor
        let twiddle_step = stage << 1;
        for group in 0..stage {
            // calculate twiddle factor
            let sin_ndx = group << sine_step;
            let wr = SINETAB[sin_ndx + SINETAB_N / 4] >> shift;
            let wi = match inverse {
                false => -SINETAB[sin_ndx],
                true => SINETAB[sin_ndx],
            } >> shift;
            let w = Complex::new(wr, wi);
            // butterfly operations
            let mut i = group;
            while i < fft_size {
                let j = i + stage;
                let temp = q15_cmul(w, data[j]);

                let q = Complex::new(data[i].re >> shift, data[i].im >> shift);
                data[i] = q + temp;
                data[j] = q - temp;
                i += twiddle_step;
            }
        }
        sine_step -= 1;
        stage = twiddle_step;
    }
    Ok(())
}

/// Real value FFT using a complex valued FFT
///
/// The `input` data will be used as a buffer for an in-place CFFT and will thus be modified by
/// this function. The `output` slice will hold the result without redundant data.
/// The length of the slices must be as follows: `output.len() >= input.len() / 2 + 1`
/// The data will be scaled by 1/input.len() if `scale == true`.
pub fn fft_radix2_real_q15(
    input: &mut [i16],
    output: &mut [Complex<Q15>],
    scale: bool,
) -> Result<(), FFTError> {
    let fft_len = input.len();
    check_size(fft_len)?;
    let half_fft_len = fft_len / 2;
    if output.len() < half_fft_len + 1 {
        return Err(FFTError::InvalidDataSize);
    }

    // Assume that Complex<T> is a exactly a sequence of two values { re: T, im: T }
    // This is checked for T == Q15 by the test module below
    let ptr = input.as_mut_ptr() as *mut Complex<Q15>;
    let fft_out = unsafe { slice::from_raw_parts_mut(ptr, half_fft_len) };

    // Do half size CFFT
    fft_radix2_q15(
        &mut fft_out[0..half_fft_len],
        if scale {
            Direction::ForwardScaled
        } else {
            Direction::Forward
        },
    )?;

    let shift = scale as usize;
    // Post processing that yields fft_len / 2 + 1 complex values
    output[0] = Complex::new((fft_out[0].im + fft_out[0].re) >> shift, 0);
    for k in 1..half_fft_len {
        // calculate the twiddle factor w
        let sin_ndx = k * SINETAB_N / fft_len;
        let w = Complex::new(SINETAB[sin_ndx + SINETAB_N / 4], -SINETAB[sin_ndx]);

        // do one post processing calculation
        let zo = Complex::new(
            (fft_out[half_fft_len - k].im + fft_out[k].im) >> (shift + 1),
            (fft_out[half_fft_len - k].re - fft_out[k].re) >> (shift + 1),
        );
        let ze = Complex::new(
            (fft_out[k].re + fft_out[half_fft_len - k].re) >> (shift + 1),
            (fft_out[k].im - fft_out[half_fft_len - k].im) >> (shift + 1),
        );
        let r = ze + q15_cmul(zo, w);
        output[k] = r;
    }
    output[half_fft_len] = Complex::new((fft_out[0].re - fft_out[0].im) >> shift, 0);

    Ok(())
}

#[cfg(test)]
mod tests {

    extern crate alloc;
    use super::{check_size, FFTError, Q15, SINETAB, SINETAB_LD_N, SINETAB_N};
    use alloc::vec::Vec;
    use core::f64;
    use core::mem::size_of;
    use num_complex::Complex;

    /// check sine table
    #[test]
    fn verify_sinewave() {
        for (i, v) in SINETAB.iter().enumerate() {
            let fval = 32767.0 * f64::sin(2.0 * f64::consts::PI * (i as f64) / (SINETAB_N as f64));
            let rval = fval.round() as i16;
            assert_eq!(*v, rval);
        }
    }

    /// Check size of Complex<Q15> to make sure that the unsafe type cast in the real FFT function
    /// works correctly
    #[test]
    fn check_complex_lenth() {
        assert_eq!(size_of::<Complex<Q15>>(), 2 * size_of::<Q15>());
    }

    #[test]
    fn check_size_test() {
        let valid = (1..=SINETAB_LD_N).map(|k| 1 << k).collect::<Vec<usize>>();
        for s in 0..=SINETAB_N * 2 {
            let result = check_size(s);
            if valid.contains(&s) {
                assert_eq!(result, Ok(()));
            } else {
                assert_eq!(result, Err(FFTError::InvalidFFTSize))
            }
        }
    }
}