Jan/WLED_MM_Infinity
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

GEQ: accuracy improvement for Smooth bars

Browse Source 
The previous code was scaling down fftResult[] to matrix height first, then computing smoothed bars.

We now to calculate smooth bars at full fftResult[] resolution first, and afterwards map the result to Matrix height. This improves accuracy and preserves a few more bits of real detail.
This commit is contained in:
Frank
2023-03-18 20:53:31 +01:00
parent d8f0fea1f3
commit eac4fe84e1
1 changed files with 10 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
wled00/FX.cpp
View File

@@ -7132,7 +7132,7 @@ uint16_t mode_2DGEQ(void) { // By Will Tatam. Code reduction by Ewoud Wijma.
int fadeoutDelay = (256 - SEGMENT.speed) / 64;
if ((fadeoutDelay <= 1 ) || ((SEGENV.call % fadeoutDelay) == 0)) SEGMENT.fadeToBlackBy(SEGMENT.speed);
uint16_t lastBarHeight = 0; // WLEDMM: for smoothing out bars
uint16_t lastBandHeight = 0; // WLEDMM: for smoothing out bars
//WLEDMM: evenly ditribut bands
float bandwidth = (float)cols / NUM_BANDS;
@@ -7145,23 +7145,24 @@ uint16_t mode_2DGEQ(void) { // By Will Tatam. Code reduction by Ewoud Wijma.
// Serial.printf("x %d b %d n %d w %f %f\n", x, band, NUM_BANDS, bandwidth, remaining);
uint8_t frBand = ((NUM_BANDS < 16) && (NUM_BANDS > 1)) ? map(band, 0, NUM_BANDS - 1, 0, 15):band; // always use full range. comment out this line to get the previous behaviour.
// band = constrain(band, 0, 15); //WLEDMM can never be out of bounds (I think...)
// frBand = constrain(frBand, 0, 15); //WLEDMM can never be out of bounds (I think...)
uint16_t colorIndex = frBand * 17; //WLEDMM 0.255
uint16_t barHeight = map(fftResult[frBand], 0, 255, 0, rows); // do not subtract -1 from rows here
uint16_t bandHeight = fftResult[frBand]; // WLEDMM we use the original ffResult, to preserve accuracy
// WLEDMM begin - smooth out bars
if ((x > 0) && (x < (cols-1)) && (SEGMENT.check2)) {
// get height of next (right side) bar
uint8_t nextband = (remaining < 1)? band +1: band;
nextband = constrain(nextband, 0, 15); // just to be sure
frBand = ((NUM_BANDS < 16) && (NUM_BANDS > 1)) ? map(nextband, 0, NUM_BANDS - 1, 0, 15):nextband; // always use full range. comment out this line to get the previous behaviour.
// nextband = constrain(nextband, 0, 15);
uint16_t nextbarHeight = map(fftResult[frBand], 0, 255, 0, rows); // do not subtract -1 from rows here
// smooth height
barHeight = (7*barHeight + 3*lastBarHeight + 3*nextbarHeight) / 12; // yeees, its 12 not 13 (10% amplification)
uint16_t nextBandHeight = fftResult[frBand];
// smooth Band height
bandHeight = (7*bandHeight + 3*lastBandHeight + 3*nextBandHeight) / 12; // yeees, its 12 not 13 (10% amplification)
bandHeight = constrain(bandHeight, 0, 255); // remove potential over/underflows
colorIndex = map(x, 0, cols-1, 0, 255); //WLEDMM
}
lastBarHeight = barHeight; // remember barheight (left side) for next iteration
lastBandHeight = bandHeight; // remember BandHeight (left side) for next iteration
uint16_t barHeight = map(bandHeight, 0, 255, 0, rows); // Now we map bandHeight to barHeight. do not subtract -1 from rows here
// WLEDMM end
if (barHeight > rows) barHeight = rows; // WLEDMM map() can "overshoot" due to rounding errors