mirror of
https://github.com/ArduPilot/ardupilot
synced 2025-01-30 20:48:33 -04:00
Tools: Web: FilterReview: move to array functions to remove loops where posible
This commit is contained in:
Iampete1
Andrew Tridgell
parent
20d4274e24
commit
d54503346f
@ -4,7 +4,7 @@
|
||||
function hanning(len) {
|
||||
w = []
|
||||
for (let i=0;i<len;i++) {
|
||||
w[i] = 0.5 - 0.5 * Math.cos( (2*Math.PI*i) / (len - 1) )
|
||||
w[i] = 0.5 - 0.5 * math.cos( (2*math.PI*i) / (len - 1) )
|
||||
}
|
||||
return w
|
||||
}
|
||||
@ -13,15 +13,10 @@ function hanning(len) {
|
||||
// linear: 1 / mean(w)
|
||||
// energy: 1 / sqrt(mean(w.^2))
|
||||
function window_correction_factors(w) {
|
||||
let window_sum = 0
|
||||
let window_sum_squared = 0
|
||||
for (let i=0;i<w.length;i++) {
|
||||
window_sum += w[i]
|
||||
window_sum_squared += w[i] * w[i]
|
||||
return {
|
||||
linear: 1/math.mean(w),
|
||||
energy: 1/math.sqrt(math.mean(math.dotMultiply(w,w)))
|
||||
}
|
||||
const mean = window_sum / w.length
|
||||
const rms = Math.sqrt(window_sum_squared / w.length)
|
||||
return { linear: 1/mean, energy: 1/rms }
|
||||
}
|
||||
|
||||
// Frequency bins for given fft length and sample period
|
||||
@ -46,7 +41,7 @@ function run_fft(data, window_size, window_spacing, windowing_function, positive
|
||||
|
||||
var center_sample = []
|
||||
|
||||
const num_windows = Math.floor((num_points-window_size)/window_spacing) + 1
|
||||
const num_windows = math.floor((num_points-window_size)/window_spacing) + 1
|
||||
for (var i=0;i<num_windows;i++) {
|
||||
// Calculate the start of each window
|
||||
const window_start = i * window_spacing
|
||||
@ -90,14 +85,9 @@ function run_fft(data, window_size, window_spacing, windowing_function, positive
|
||||
}
|
||||
|
||||
// Scale all points by the window size
|
||||
fft_x[i] = [];
|
||||
fft_y[i] = [];
|
||||
fft_z[i] = [];
|
||||
for (var j=0;j<positive_index;j++) {
|
||||
fft_x[i][j] = p1_x[j] / window_size
|
||||
fft_y[i][j] = p1_y[j] / window_size
|
||||
fft_z[i][j] = p1_z[j] / window_size
|
||||
}
|
||||
fft_x[i] = math.dotMultiply(p1_x, 1 / window_size)
|
||||
fft_y[i] = math.dotMultiply(p1_y, 1 / window_size)
|
||||
fft_z[i] = math.dotMultiply(p1_z, 1 / window_size)
|
||||
|
||||
}
|
||||
|
||||
@ -129,7 +119,7 @@ function run_batch_fft(data_set) {
|
||||
// Calculate window size for given number of windows and overlap
|
||||
const window_size = math.floor(num_points / (1 + (window_per_batch - 1)*(1-window_overlap)))
|
||||
|
||||
const window_spacing = Math.round(window_size * (1 - window_overlap))
|
||||
const window_spacing = math.round(window_size * (1 - window_overlap))
|
||||
const windowing_function = hanning(window_size)
|
||||
|
||||
// Get windowing correction factors for use later when plotting
|
||||
@ -139,7 +129,7 @@ function run_batch_fft(data_set) {
|
||||
var bins = fft_freq(window_size, sample_time)
|
||||
|
||||
// discard negative spectrum
|
||||
const positive_index = Math.floor(window_size/2)
|
||||
const positive_index = math.floor(window_size/2)
|
||||
bins = bins.slice(0, positive_index)
|
||||
|
||||
var x = []
|
||||
@ -151,10 +141,7 @@ function run_batch_fft(data_set) {
|
||||
for (let i=0;i<num_batch;i++) {
|
||||
var ret = run_fft(data_set[i], window_size, window_spacing, windowing_function, positive_index)
|
||||
|
||||
for (let j=0;j<ret.center.length;j++) {
|
||||
time.push(data_set[i].sample_time + sample_time * ret.center[j])
|
||||
}
|
||||
|
||||
time.push(...math.add(data_set[i].sample_time, math.dotMultiply(sample_time, ret.center)))
|
||||
x.push(...ret.x)
|
||||
y.push(...ret.y)
|
||||
z.push(...ret.z)
|
||||
@ -297,17 +284,17 @@ function get_amplitude_scale() {
|
||||
var ret = {}
|
||||
if (use_DB) {
|
||||
if (use_PSD) {
|
||||
ret.fun = function (x) { return 10.0 * Math.log10(x*x) }
|
||||
ret.fun = function (x) { return math.dotMultiply(math.log10(math.dotMultiply(x,x)), 10.0) } // 10 * log10(x.^2)
|
||||
ret.label = "PSD (db)"
|
||||
} else {
|
||||
ret.fun = function (x) { return 10.0 * Math.log10(x) }
|
||||
ret.fun = function (x) { return math.dotMultiply(math.log10(x), 10.0) } // 10 * log10(x)
|
||||
ret.label = "Linear amplitude (db)"
|
||||
}
|
||||
ret.hover = function (axis) { return "%{" + axis + ":.2f} db" }
|
||||
|
||||
} else {
|
||||
if (use_PSD) {
|
||||
ret.fun = function (x) { return x*x }
|
||||
ret.fun = function (x) { return math.dotMultiply(x,x) }
|
||||
ret.label = "PSD"
|
||||
} else {
|
||||
ret.fun = function (x) { return x }
|
||||
@ -329,7 +316,7 @@ function get_frequency_scale() {
|
||||
|
||||
var ret = {}
|
||||
if (use_RPM) {
|
||||
ret.fun = function (x) { return 60.0 * x }
|
||||
ret.fun = function (x) { return math.dotMultiply(x, 60.0) }
|
||||
ret.label = "RPM"
|
||||
ret.hover = function (axis) { return "%{" + axis + ":.2f} RPM" }
|
||||
|
||||
@ -344,15 +331,6 @@ function get_frequency_scale() {
|
||||
return ret
|
||||
}
|
||||
|
||||
// Helper to apply scale function to array
|
||||
function apply_scale_fun(data, scale_fun) {
|
||||
var x = []
|
||||
for (let i = 0; i < data.length; i++) {
|
||||
x[i] = scale_fun(data[i])
|
||||
}
|
||||
return x
|
||||
}
|
||||
|
||||
// Look through time array and return first index before start time
|
||||
function find_start_index(time) {
|
||||
const start_time = parseFloat(document.getElementById("TimeStart").value)
|
||||
@ -410,36 +388,23 @@ function redraw() {
|
||||
const window_correction = amplitude_scale.use_PSD ? (Gyro_batch[i].FFT.correction.energy * math.sqrt(1/2)) : Gyro_batch[i].FFT.correction.linear
|
||||
|
||||
// Take mean from start to end
|
||||
var fft_mean_x = []
|
||||
var fft_mean_y = []
|
||||
var fft_mean_z = []
|
||||
var fft_mean_x = 0
|
||||
var fft_mean_y = 0
|
||||
var fft_mean_z = 0
|
||||
for (let j=start_index;j<end_index;j++) {
|
||||
// Add to mean sum
|
||||
for (let k=0;k<Gyro_batch[i].FFT.x[j].length;k++) {
|
||||
if (j == start_index) {
|
||||
fft_mean_x[k] = amplitude_scale.fun(Gyro_batch[i].FFT.x[j][k] * window_correction)
|
||||
fft_mean_y[k] = amplitude_scale.fun(Gyro_batch[i].FFT.y[j][k] * window_correction)
|
||||
fft_mean_z[k] = amplitude_scale.fun(Gyro_batch[i].FFT.z[j][k] * window_correction)
|
||||
} else {
|
||||
fft_mean_x[k] += amplitude_scale.fun(Gyro_batch[i].FFT.x[j][k] * window_correction)
|
||||
fft_mean_y[k] += amplitude_scale.fun(Gyro_batch[i].FFT.y[j][k] * window_correction)
|
||||
fft_mean_z[k] += amplitude_scale.fun(Gyro_batch[i].FFT.z[j][k] * window_correction)
|
||||
}
|
||||
if (j == (end_index - 1)) {
|
||||
fft_mean_x[k] /= plot_length
|
||||
fft_mean_y[k] /= plot_length
|
||||
fft_mean_z[k] /= plot_length
|
||||
}
|
||||
}
|
||||
fft_mean_x = math.add(fft_mean_x, amplitude_scale.fun(math.dotMultiply(Gyro_batch[i].FFT.x[j], window_correction)))
|
||||
fft_mean_y = math.add(fft_mean_y, amplitude_scale.fun(math.dotMultiply(Gyro_batch[i].FFT.y[j], window_correction)))
|
||||
fft_mean_z = math.add(fft_mean_z, amplitude_scale.fun(math.dotMultiply(Gyro_batch[i].FFT.z[j], window_correction)))
|
||||
}
|
||||
|
||||
// Set scaled y data
|
||||
fft_plot.data[i*3 + 0].y = fft_mean_x
|
||||
fft_plot.data[i*3 + 1].y = fft_mean_y
|
||||
fft_plot.data[i*3 + 2].y = fft_mean_z
|
||||
fft_plot.data[i*3 + 0].y = math.dotMultiply(fft_mean_x, 1 / plot_length)
|
||||
fft_plot.data[i*3 + 1].y = math.dotMultiply(fft_mean_y, 1 / plot_length)
|
||||
fft_plot.data[i*3 + 2].y = math.dotMultiply(fft_mean_z, 1 / plot_length)
|
||||
|
||||
// Set scaled x data
|
||||
const scaled_bins = apply_scale_fun(Gyro_batch[i].FFT.bins, frequency_scale.fun)
|
||||
const scaled_bins = frequency_scale.fun(Gyro_batch[i].FFT.bins)
|
||||
fft_plot.data[i*3 + 0].x = scaled_bins
|
||||
fft_plot.data[i*3 + 1].x = scaled_bins
|
||||
fft_plot.data[i*3 + 2].x = scaled_bins
|
||||
@ -516,7 +481,7 @@ function redraw_Spectrogram() {
|
||||
const plot_length = end_index - start_index
|
||||
|
||||
// Setup xy data
|
||||
Spectrogram.data[0].x = apply_scale_fun(Gyro_batch[batch_instance].FFT.bins, frequency_scale.fun)
|
||||
Spectrogram.data[0].x = frequency_scale.fun(Gyro_batch[batch_instance].FFT.bins)
|
||||
Spectrogram.data[0].y = Gyro_batch[batch_instance].FFT.time.slice(start_index, end_index)
|
||||
|
||||
// Windowing amplitude correction depends on spectrum of interest
|
||||
@ -525,11 +490,8 @@ function redraw_Spectrogram() {
|
||||
// Setup z data
|
||||
Spectrogram.data[0].z = []
|
||||
for (j = 0; j<plot_length; j++) {
|
||||
Spectrogram.data[0].z[j] = []
|
||||
const index = start_index + j
|
||||
for (k = 0; k<Gyro_batch[batch_instance].FFT[axis][index].length; k++) {
|
||||
Spectrogram.data[0].z[j][k] = amplitude_scale.fun(Gyro_batch[batch_instance].FFT[axis][index][k] * window_correction)
|
||||
}
|
||||
Spectrogram.data[0].z[j] = amplitude_scale.fun(math.dotMultiply(Gyro_batch[batch_instance].FFT[axis][index], window_correction))
|
||||
}
|
||||
|
||||
Plotly.redraw("Spectrogram")
|
||||
@ -664,11 +626,9 @@ function load(log_file) {
|
||||
|
||||
// Remove logging scale factor
|
||||
const mul = 1/log.messages.ISBH.mul[i]
|
||||
for (let j = 0; j < num_samples; j++) {
|
||||
x[j] *= mul
|
||||
y[j] *= mul
|
||||
z[j] *= mul
|
||||
}
|
||||
x = math.dotMultiply(x, mul)
|
||||
y = math.dotMultiply(y, mul)
|
||||
z = math.dotMultiply(z, mul)
|
||||
|
||||
// Add to batches for this instance
|
||||
Gyro_batch[instance].push({ sample_time: log.messages.ISBH.SampleUS[i] / 1000000,
|
||||
|
||||
Loading…
Reference in New Issue
Block a user