Huge improvement in display. It was basically a bug that made it so hard to read. Playing with an IF shift feature, and filtering and stuff. I think the overlap-discard is not quite right yet....
This commit is contained in:
Michael Colton
parent
9d045be4ce
commit
6a7b46d523
1 changed files with 45 additions and 27 deletions
|
|
@ -39,14 +39,14 @@
|
|||
// Sample pragmas to cope with warnings. Please note the related line at
|
||||
// the end of this function, used to pop the compiler diagnostics status.
|
||||
#pragma GCC diagnostic push
|
||||
#pragma GCC diagnostic ignored "-Wunused-parameter"
|
||||
#pragma GCC diagnostic ignored "-Wmissing-declarations"
|
||||
#pragma GCC diagnostic ignored "-Wreturn-type"
|
||||
//#pragma GCC diagnostic ignored "-Wunused-parameter"
|
||||
//#pragma GCC diagnostic ignored "-Wmissing-declarations"
|
||||
//#pragma GCC diagnostic ignored "-Wreturn-type"
|
||||
|
||||
void dac1SetValue(uint16_t value);
|
||||
void dac2SetValue(uint16_t value);
|
||||
void ddsPrefix(void);
|
||||
void sendToDds(uint16_t data1, uint16_t data2);
|
||||
//void ddsPrefix(void);
|
||||
//void sendToDds(uint16_t data1, uint16_t data2);
|
||||
#define FFT_SIZE 256 //supported sizes are 16, 64, 256, 1024
|
||||
#define FFT_BUFFER_SIZE 512 //double the FFT_SIZE above.
|
||||
__IO long long millis = 0;
|
||||
|
|
@ -63,7 +63,7 @@ uint16_t menuLastState = 1;
|
|||
uint16_t menuCount = 10;
|
||||
uint32_t frequencyDialMultiplier = 1;
|
||||
|
||||
long vfoAFrequency = 10001460;
|
||||
long vfoAFrequency = 7236400;
|
||||
long vfoALastFreq = 0;
|
||||
int encoderPos, encoderLastPos;
|
||||
|
||||
|
|
@ -74,6 +74,8 @@ uint8_t mode = 0;
|
|||
float agcLevel = 0;
|
||||
float agcScale = 160; //Higher is lower volume.. for now
|
||||
|
||||
int ifShift = 0;
|
||||
|
||||
void polarToRect(float m, float a, float32_t* x, float32_t* y)
|
||||
{
|
||||
*y = m * arm_sin_f32(a);
|
||||
|
|
@ -131,13 +133,13 @@ void populateCoeficients(int bandwidth, int sideband, int offset)
|
|||
return; //Skipping all the later stuff doesn't seem to make a huge difference yet...
|
||||
|
||||
//2:
|
||||
float x, y;
|
||||
for(i = 0; i < FFT_SIZE; i++)
|
||||
{
|
||||
polarToRect(fftFilterCoeficient[i], fftFilterCoeficient[FFT_BUFFER_SIZE - 1 - i], &x, &y);
|
||||
fftFilterCoeficient[i] = x;
|
||||
fftFilterCoeficient[FFT_BUFFER_SIZE - 1 - i] = y;
|
||||
}
|
||||
// float x, y;
|
||||
// for(i = 0; i < FFT_SIZE; i++)
|
||||
// {
|
||||
// polarToRect(fftFilterCoeficient[i], fftFilterCoeficient[FFT_BUFFER_SIZE - 1 - i], &x, &y);
|
||||
// fftFilterCoeficient[i] = x;
|
||||
// fftFilterCoeficient[FFT_BUFFER_SIZE - 1 - i] = y;
|
||||
// }
|
||||
|
||||
//3:
|
||||
arm_cfft_radix4_instance_f32 fft_co;
|
||||
|
|
@ -186,7 +188,7 @@ void populateCoeficients(int bandwidth, int sideband, int offset)
|
|||
}
|
||||
|
||||
|
||||
void applyCoeficient(float *samples)
|
||||
void applyCoeficient(float *samples, int shift)
|
||||
{
|
||||
//See DSP Guide Chapter 9 (Equation 9-1)
|
||||
int i;
|
||||
|
|
@ -196,9 +198,13 @@ void applyCoeficient(float *samples)
|
|||
filterTemp[i * 2 + 1] = samples[i * 2 + 1] * fftFilterCoeficient[i * 2 + 1] + samples[i * 2] * fftFilterCoeficient[i * 2];
|
||||
}
|
||||
|
||||
int shifter;
|
||||
for(i = 0; i < FFT_BUFFER_SIZE; i++)
|
||||
{
|
||||
samples[i] = filterTemp[i];
|
||||
shifter = i + 2 * shift;
|
||||
if(i < 0) samples[i] = 0;
|
||||
if(i > FFT_BUFFER_SIZE - 1) samples[i] = 0;
|
||||
samples[i] = filterTemp[i + 2 * shift];
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -778,7 +784,7 @@ main(int argc, char* argv[])
|
|||
encoderPos = getPos();
|
||||
if(encoderPos != encoderLastPos)
|
||||
{
|
||||
mode = (encoderLastPos - encoderPos) % 3;
|
||||
mode += (encoderLastPos - encoderPos) % 3;
|
||||
Adafruit_GFX_setTextSize(1);
|
||||
Adafruit_GFX_setCursor(150, 190);
|
||||
int i;
|
||||
|
|
@ -886,14 +892,26 @@ main(int argc, char* argv[])
|
|||
// timeMeasurement = millis - timeMeasurement;
|
||||
//
|
||||
|
||||
|
||||
//TODO: Should I shift away from 0Htz? to get away from 1/f noise? It didn's LOOK bad, but maybe it is negatively effecting things.
|
||||
//I could do something where the dial moves around on screen, but if you get too close to the edge, the DDSs start moving the frequency
|
||||
//Hmm, I think that's kind of a cool idea. It would be cool in two ways: it would allow you to shift the IF so you could get away from
|
||||
//birdies, and it would mean that while tuning around locally, the waterfall would stay aligned in a useful way. Eventually, when I have
|
||||
//sufficient display performance, I'd like to move (and scale, if applicable) the waterfall so it is properly aligned.
|
||||
|
||||
//Speaking of 1/f noise. It doesn't seem to be much of an issue on this radio, I wonder why? Did I design something right?
|
||||
//Also, early on, I thought it had an issue with microphonics, but it turned out that it was the connection to the computer.
|
||||
//Also since this is a form of direct conversion receiver (two of them together) I was worried about AM broadcast interference
|
||||
//but I haven't noticed any, again, maybe I did something right? Beginner's luck?
|
||||
|
||||
arm_cmplx_mag_f32(samplesDisplay, magnitudes, FFT_SIZE);
|
||||
|
||||
float fftMax = 0;
|
||||
float fftMin = 100;
|
||||
float fftMaxMax = 0;
|
||||
float logMax;
|
||||
float fftMax = 0; //AH! These are being reset each time! Static makes them persistant right? Does it also ensure they are
|
||||
float fftMin = 100; //only initialized once? Have to try it when I get home. It would certainly be nice if the waterfall
|
||||
static float fftMaxMax = 0; //didn't change in brightness so much. Later, I may want to fix these values, or at least, make them
|
||||
static float logMax; //manually controllable, sorta, you know?
|
||||
uint8_t i;
|
||||
for(i = 0; i < 255; i++)
|
||||
for(i = 1; i < 255; i++) //If bin 0 is the DC offset, should we skip it in this calculation?
|
||||
{
|
||||
float mags = magnitudes[i];
|
||||
if(mags > fftMax) fftMax = mags;
|
||||
|
|
@ -901,9 +919,9 @@ main(int argc, char* argv[])
|
|||
}
|
||||
//logMax = log2(fftMax);
|
||||
|
||||
if(fftMax > fftMaxMax) fftMaxMax = fftMax;
|
||||
if(fftMax > fftMaxMax) fftMaxMax += fftMax * 0.1;
|
||||
logMax = log2(fftMaxMax);
|
||||
|
||||
fftMaxMax *= 0.99;
|
||||
|
||||
// TODOne: SWITCH THESE AND FLIP THEM. So that higher frequencies appear higher on screen.
|
||||
// TODO: Got rid of the first bin because it's just DC offset, right?
|
||||
|
|
@ -912,7 +930,7 @@ main(int argc, char* argv[])
|
|||
//uint16_t i;
|
||||
for(i = 1; i < 120; i++)
|
||||
{
|
||||
mags = (log2(magnitudes[i] + 1)) / fftMaxMax * 32;
|
||||
mags = (log2(magnitudes[i] + 1)) / fftMaxMax * 32; //Log needs to be at least 1 right? We could do a + (1-fftMin) maybe? Worth it?
|
||||
//mags = magnitudes[i] / fftMaxMax * 32;
|
||||
Adafruit_ILI9340_drawPixel(waterfallScanLine, (120 - i), gradient[(uint8_t) mags]);
|
||||
}
|
||||
|
|
@ -1023,7 +1041,7 @@ void processStream()
|
|||
|
||||
//arm_cmplx_mag_f32(samplesA, magnitudes, FFT_SIZE);
|
||||
|
||||
applyCoeficient(samplesA);
|
||||
applyCoeficient(samplesA, ifShift);
|
||||
|
||||
arm_cfft_radix4_init_f32(&fft_inst, FFT_SIZE, 1, 1);
|
||||
arm_cfft_radix4_f32(&fft_inst, samplesA);
|
||||
|
|
@ -1054,7 +1072,7 @@ void processStream()
|
|||
arm_cfft_radix4_f32(&fft_inst, samplesB);
|
||||
// Calculate magnitude of complex numbers output by the FFT.
|
||||
//arm_cmplx_mag_f32(samplesB, magnitudes, FFT_SIZE);
|
||||
applyCoeficient(samplesB);
|
||||
applyCoeficient(samplesB, ifShift);
|
||||
|
||||
arm_cfft_radix4_init_f32(&fft_inst, FFT_SIZE, 1, 1);
|
||||
arm_cfft_radix4_f32(&fft_inst, samplesB);
|
||||
|
|
@ -1088,7 +1106,7 @@ void processStream()
|
|||
//arm_cmplx_mag_f32(samplesA, magnitudes, FFT_SIZE);
|
||||
|
||||
//arm_cmplx_mag_f32(samplesA, magnitudes, FFT_SIZE);
|
||||
applyCoeficient(samplesC);
|
||||
applyCoeficient(samplesC, ifShift);
|
||||
arm_cfft_radix4_init_f32(&fft_inst, FFT_SIZE, 1, 1);
|
||||
arm_cfft_radix4_f32(&fft_inst, samplesC);
|
||||
|
||||
|
|
|
|||
Loading…
Reference in a new issue