You have to convert signed short inputs (16bit) into q15 (that is a different encoding scheme).

And vice versa..

## Voice changer

### Re: Voice changer

Pukao Hats Cleaning Services Ltd.

- RogerClark
**Posts:**6726**Joined:**Mon Apr 27, 2015 10:36 am**Location:**Melbourne, Australia-
**Contact:**

### Re: Voice changer

UmmPito wrote:You have to convert signed short inputs (16bit) into q15 (that is a different encoding scheme).

And vice versa..

More processing. Thats a shame. It takes processor cycles

### Re: Voice changer

Not really, I guess. The fft works faster with q15 (??), as it must not mess with saturation (q15 is fractional -1.000 to +0.999 afaik).

There are routines in cmsis from float to q15/31/7 and vice versa, but not from integer - it must not be difficult to convert, though..

There is a type q15_t (q31_t, q7_t) available, I think.

Ie. for 12bit ADC:
or something like that.. Not sure whether to subtract an offset before the shift, in order to create a signed value.. Like
Needs to be double-checked whether that works.. No warranties of any kind

There are routines in cmsis from float to q15/31/7 and vice versa, but not from integer - it must not be difficult to convert, though..

There is a type q15_t (q31_t, q7_t) available, I think.

Ie. for 12bit ADC:

Code: Select all

`Q15_data = (q15_t) ((12bit_ADC_data) << 4);`

Code: Select all

`Q15_data = (q15_t) ((12bit_ADC_data - offset) << 4);`

Pukao Hats Cleaning Services Ltd.

- RogerClark
**Posts:**6726**Joined:**Mon Apr 27, 2015 10:36 am**Location:**Melbourne, Australia-
**Contact:**

### Re: Voice changer

OK

I think I have other problems with the ARM code, not just the data format, unless the FFT can't handle blank data (all zeros)

But, I've had to put this project idea on hold, as I have got some other things I need to get working.

I think I have other problems with the ARM code, not just the data format, unless the FFT can't handle blank data (all zeros)

But, I've had to put this project idea on hold, as I have got some other things I need to get working.

### Re: Voice changer

For sound processing on stm32 i`m using Arduino FFT library.

fft.h
fft.cpp
This is very fast implementation.

fft.h

Code: Select all

```
#ifndef FIXFFT_H
#define FIXFFT_H
#include <WProgram.h>
/*
fix_fft() - perform forward/inverse fast Fourier transform.
fr[n],fi[n] are real and imaginary arrays, both INPUT AND
RESULT (in-place FFT), with 0 <= n < 2**m; set inverse to
0 for forward transform (FFT), or 1 for iFFT.
*/
int fix_fft(int fr[], int fi[], int m, int inverse);
/*
fix_fftr() - forward/inverse FFT on array of real numbers.
Real FFT/iFFT using half-size complex FFT by distributing
even/odd samples into real/imaginary arrays respectively.
In order to save data space (i.e. to avoid two arrays, one
for real, one for imaginary samples), we proceed in the
following two steps: a) samples are rearranged in the real
array so that all even samples are in places 0-(N/2-1) and
all imaginary samples in places (N/2)-(N-1), and b) fix_fft
is called with fr and fi pointing to index 0 and index N/2
respectively in the original array. The above guarantees
that fix_fft "sees" consecutive real samples as alternating
real and imaginary samples in the complex array.
*/
int fix_fftr(int f[], int m, int inverse);
#endif
```

Code: Select all

```
#include <avr/pgmspace.h>
#include "fft.h"
#include <WProgram.h>
/* fix_fft.c - Fixed-point in-place Fast Fourier Transform */
/*
All data are fixed-point short integers, in which -32768
to +32768 represent -1.0 to +1.0 respectively. Integer
arithmetic is used for speed, instead of the more natural
floating-point.
For the forward FFT (time -> freq), fixed scaling is
performed to prevent arithmetic overflow, and to map a 0dB
sine/cosine wave (i.e. amplitude = 32767) to two -6dB freq
coefficients. The return value is always 0.
For the inverse FFT (freq -> time), fixed scaling cannot be
done, as two 0dB coefficients would sum to a peak amplitude
of 64K, overflowing the 32k range of the fixed-point integers.
Thus, the fix_fft() routine performs variable scaling, and
returns a value which is the number of bits LEFT by which
the output must be shifted to get the actual amplitude
(i.e. if fix_fft() returns 3, each value of fr[] and fi[]
must be multiplied by 8 (2**3) for proper scaling.
Clearly, this cannot be done within fixed-point short
integers. In practice, if the result is to be used as a
filter, the scale_shift can usually be ignored, as the
result will be approximately correctly normalized as is.
Written by: Tom Roberts 11/8/89
Made portable: Malcolm Slaney 12/15/94 malcolm@interval.com
Enhanced: Dimitrios P. Bouras 14 Jun 2006 dbouras@ieee.org
Modified for 8bit values David Keller 10.10.2010
*/
#define N_WAVE 256 /* full length of Sinewave[] */
#define LOG2_N_WAVE 8 /* log2(N_WAVE) */
/*
Since we only use 3/4 of N_WAVE, we define only
this many samples, in order to conserve data space.
*/
const int Sinewave[N_WAVE-N_WAVE/4] = {
0, 3, 6, 9, 12, 15, 18, 21,
24, 28, 31, 34, 37, 40, 43, 46,
48, 51, 54, 57, 60, 63, 65, 68,
71, 73, 76, 78, 81, 83, 85, 88,
90, 92, 94, 96, 98, 100, 102, 104,
106, 108, 109, 111, 112, 114, 115, 117,
118, 119, 120, 121, 122, 123, 124, 124,
125, 126, 126, 127, 127, 127, 127, 127,
127, 127, 127, 127, 127, 127, 126, 126,
125, 124, 124, 123, 122, 121, 120, 119,
118, 117, 115, 114, 112, 111, 109, 108,
106, 104, 102, 100, 98, 96, 94, 92,
90, 88, 85, 83, 81, 78, 76, 73,
71, 68, 65, 63, 60, 57, 54, 51,
48, 46, 43, 40, 37, 34, 31, 28,
24, 21, 18, 15, 12, 9, 6, 3,
0, -3, -6, -9, -12, -15, -18, -21,
-24, -28, -31, -34, -37, -40, -43, -46,
-48, -51, -54, -57, -60, -63, -65, -68,
-71, -73, -76, -78, -81, -83, -85, -88,
-90, -92, -94, -96, -98, -100, -102, -104,
-106, -108, -109, -111, -112, -114, -115, -117,
-118, -119, -120, -121, -122, -123, -124, -124,
-125, -126, -126, -127, -127, -127, -127, -127,
/*-127, -127, -127, -127, -127, -127, -126, -126,
-125, -124, -124, -123, -122, -121, -120, -119,
-118, -117, -115, -114, -112, -111, -109, -108,
-106, -104, -102, -100, -98, -96, -94, -92,
-90, -88, -85, -83, -81, -78, -76, -73,
-71, -68, -65, -63, -60, -57, -54, -51,
-48, -46, -43, -40, -37, -34, -31, -28,
-24, -21, -18, -15, -12, -9, -6, -3, */
};
/*
FIX_MPY() - fixed-point multiplication & scaling.
Substitute inline assembly for hardware-specific
optimization suited to a particluar DSP processor.
Scaling ensures that result remains 16-bit.
*/
inline int FIX_MPY(int a, int b)
{
//Serial.println(a);
//Serial.println(b);
/* shift right one less bit (i.e. 15-1) */
int c = ((int)a * (int)b) >> 6;
/* last bit shifted out = rounding-bit */
b = c & 0x01;
/* last shift + rounding bit */
a = (c >> 1) + b;
/*
Serial.println(Sinewave[3]);
Serial.println(c);
Serial.println(a);
while(1);*/
return a;
}
/*
fix_fft() - perform forward/inverse fast Fourier transform.
fr[n],fi[n] are real and imaginary arrays, both INPUT AND
RESULT (in-place FFT), with 0 <= n < 2**m; set inverse to
0 for forward transform (FFT), or 1 for iFFT.
*/
int fix_fft(int fr[], int fi[], int m, int inverse)
{
int mr, nn, i, j, l, k, istep, n, scale, shift;
int qr, qi, tr, ti, wr, wi;
n = 1 << m;
/* max FFT size = N_WAVE */
if (n > N_WAVE)
return -1;
mr = 0;
nn = n - 1;
scale = 0;
/* decimation in time - re-order data */
for (m=1; m<=nn; ++m) {
l = n;
do {
l >>= 1;
} while (mr+l > nn);
mr = (mr & (l-1)) + l;
if (mr <= m)
continue;
tr = fr[m];
fr[m] = fr[mr];
fr[mr] = tr;
ti = fi[m];
fi[m] = fi[mr];
fi[mr] = ti;
}
l = 1;
k = LOG2_N_WAVE-1;
while (l < n) {
if (inverse) {
/* variable scaling, depending upon data */
shift = 0;
for (i=0; i<n; ++i) {
j = fr[i];
if (j < 0)
j = -j;
m = fi[i];
if (m < 0)
m = -m;
if (j > 16383 || m > 16383) {
shift = 1;
break;
}
}
if (shift)
++scale;
} else {
/*
fixed scaling, for proper normalization --
there will be log2(n) passes, so this results
in an overall factor of 1/n, distributed to
maximize arithmetic accuracy.
*/
shift = 1;
}
/*
it may not be obvious, but the shift will be
performed on each data point exactly once,
during this pass.
*/
istep = l << 1;
for (m=0; m<l; ++m) {
j = m << k;
/* 0 <= j < N_WAVE/2 */
wr = Sinewave [j+N_WAVE/4];
/*Serial.println("asdfasdf");
Serial.println(wr);
Serial.println(j+N_WAVE/4);
Serial.println(Sinewave[256]);
Serial.println("");*/
wi = -Sinewave[j];
if (inverse)
wi = -wi;
if (shift) {
wr >>= 1;
wi >>= 1;
}
for (i=m; i<n; i+=istep) {
j = i + l;
tr = FIX_MPY(wr,fr[j]) - FIX_MPY(wi,fi[j]);
ti = FIX_MPY(wr,fi[j]) + FIX_MPY(wi,fr[j]);
qr = fr[i];
qi = fi[i];
if (shift) {
qr >>= 1;
qi >>= 1;
}
fr[j] = qr - tr;
fi[j] = qi - ti;
fr[i] = qr + tr;
fi[i] = qi + ti;
}
}
--k;
l = istep;
}
return scale;
}
/*
fix_fftr() - forward/inverse FFT on array of real numbers.
Real FFT/iFFT using half-size complex FFT by distributing
even/odd samples into real/imaginary arrays respectively.
In order to save data space (i.e. to avoid two arrays, one
for real, one for imaginary samples), we proceed in the
following two steps: a) samples are rearranged in the real
array so that all even samples are in places 0-(N/2-1) and
all imaginary samples in places (N/2)-(N-1), and b) fix_fft
is called with fr and fi pointing to index 0 and index N/2
respectively in the original array. The above guarantees
that fix_fft "sees" consecutive real samples as alternating
real and imaginary samples in the complex array.
*/
int fix_fftr(int f[], int m, int inverse)
{
int i, N = 1<<(m-1), scale = 0;
int tt, *fr=f, *fi=&f[N];
if (inverse)
scale = fix_fft(fi, fr, m-1, inverse);
for (i=1; i<N; i+=2) {
tt = f[N+i-1];
f[N+i-1] = f[i];
f[i] = tt;
}
if (! inverse)
scale = fix_fft(fi, fr, m-1, inverse);
return scale;
}
```