Creating a sinewave with logarithmic frequency

56 views (last 30 days)
I'm trying to generate a sinewave with logarithmic increasing frequency.
freq = f_str*((f_end/f_str).^(time/T));
The max frequency in the sinewave is 50 Hz. However, when I construct such sinewave and run a spectrogram on it (STFT), it shows that the max frequency is 150 Hz.
Any thoughts would be appreciated.
% Generate log-frequency sweep then plot it and its spectrogram
NPTS = 225000;
f_str = 7;
f_end = 50;
T = 450;
time = linspace(0,T,NPTS)';
freq = f_str*((f_end/f_str).^(time/T));
Y = sin(2*pi*freq.*time); % Sinewave with logarithmic sweep
fs = (2*NPTS)/900;
subplot(311), plot(time, freq, '-b', 'LineWidth',3), grid on, xlabel('Time [sec]'), ylabel('Frequency [Hz]')
subplot(312), plot(time, Y, '-b', 'LineWidth',3), grid on, xlabel('Time [sec]'), ylabel('SineWave [-]'), axis([0 450 -2 2])
subplot(313), stft(Y, fs, 'FrequencyRange', 'onesided'), set(gca,'Ylim',[0 200]), colorbar off
  4 Comments
dpb
dpb on 25 Oct 2024 at 15:03
Edited: dpb on 25 Oct 2024 at 17:48
% Generate log-frequency sweep then plot it and its spectrogram
NPTS = 225000;
f_str = 7;
f_end = 50;
T = 450;
time = linspace(0,T,NPTS)';
freq = f_str*((f_end/f_str).^(time/T));
Y = sin(2*pi*freq.*time); % Sinewave with logarithmic sweep
fs = (2*NPTS)/900;
tView=[0 1];
y=chirp(time,f_str,T,f_end,"logarithmic",-90);
subplot(321), plot(time, Y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Original'), axis([0 450 -2 2])
xlim(tView)
subplot(323), plot(time, y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Chirp'), axis([0 450 -2 2])
xlim(tView)
tView=[T-0.1 T];
subplot(322), plot(time, Y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Original'), axis([0 450 -2 2])
xlim(tView)
subplot(324), plot(time, y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Chirp'), axis([0 450 -2 2])
xlim(tView)
Compares to using the Signal Processing TB chirp function -- I didn't have time to try to dig into just what the root cause is in the straightforward calculation; perhaps the issue of the large angle is part, but the internals of the chirp function deal with it and do produce the expected number of cycles over the last 100 msec of the time history whereas the original clearly shows that the frequency of the actual time trace itself is right at the 150 Hz, not the expected 50 Hz. OTOMH, I don't have any better explanation, only that one can illustrate the result...
Djamil Boulahbal
Djamil Boulahbal on 25 Oct 2024 at 19:49
Bravo! The chirp function did the trick. I somehow had completely forgotten about it ... Thank you for your help. I am convinced now there is a numerical issue in my approach. Though no time to dwell ... as I have the answer now.

Sign in to comment.

Accepted Answer

Star Strider
Star Strider on 24 Oct 2024 at 23:42
Using the logspace function could work.
Try this ’—
% Generate log-frequency sweep then plot it and its spectrogram
NPTS = 225000;
f_str = 7;
f_end = 50;
T = 450;
time = linspace(0,T,NPTS)';
% freq = f_str*((f_end/f_str).^(time/T));
freq = logspace(-3, log10(f_end/f_str),NPTS)';
Y = sin(2*pi*freq.*time); % Sinewave with logarithmic sweep
fs = (2*NPTS)/900;
subplot(311), plot(time, freq, '-b', 'LineWidth',3), grid on, xlabel('Time [sec]'), ylabel('Frequency [Hz]')
subplot(312), plot(time, Y, '-b', 'LineWidth',3), grid on, xlabel('Time [sec]'), ylabel('SineWave [-]'), axis([0 450 -2 2])
subplot(313), stft(Y, fs, 'FrequencyRange', 'onesided'), set(gca,'Ylim',[0 200]), colorbar off
Make appropriate changes to get the result you want.
.
  5 Comments
Star Strider
Star Strider on 25 Oct 2024 at 21:10
As always, my pleasure!
I am not cerrtain where the problem lies. The code appears to be correct, and iis by my analyses (not shown, since they did not add any useful information).
Djamil Boulahbal
Djamil Boulahbal on 25 Oct 2024 at 22:19
Thanks again, you've been very thorough and helpful.

Sign in to comment.

More Answers (2)

Djamil Boulahbal
Djamil Boulahbal on 25 Oct 2024 at 0:19
Edited: Djamil Boulahbal on 25 Oct 2024 at 0:27
Thinking a little more about it, the errror could be because the angle (2×pi×freq×time) grows very large (reaches 15×10^4 radians), and the precision is lost.
One thing I could try is to define the angle only to within +/- pi ...
Thoughts?
  1 Comment
dpb
dpb on 25 Oct 2024 at 16:56
But the sin function itself deals with long time series ok it seems...
% Generate log-frequency sweep then plot it and its spectrogram
NPTS = 225000;
f_str = 7;
f_end = 50;
T = 450;
time = linspace(0,T,NPTS)';
%freq = f_str*((f_end/f_str).^(time/T));
%Y = sin(2*pi*freq.*time); % Sinewave with logarithmic sweep
Y = sin(2*pi*f_str*time); % Sinewave with starting frequency
tView=[0 1];
y = sin(2*pi*f_end*time); % Sinewave with ending frequency
subplot(321), plot(time, Y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Start F'), axis([0 450 -2 2])
tView=[0 1]; xlim(tView)
subplot(322), plot(time, y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('End F'), axis([0 450 -2 2])
tView=[T-0.1 T];xlim(tView)
subplot(323), plot(time, Y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('Start F'), axis([0 450 -2 2])
tView=[0 1]; xlim(tView)
subplot(324), plot(time, y, '-b', 'LineWidth',1), grid on, xlabel('Time [sec]'), ylabel('End F'), axis([0 450 -2 2])
tView=[T-0.1 T];xlim(tView)
The first 1 sec shows 7 cycles, the last 100 msed is 5, so the two full-length calculations of a fixed-frequency sine wave over the same t vector produce the expected frequencies. Hence one must conclude it isn't internal rounding in the sin() function itself causing the difference in the original code..
As before, I don't have time to really dig, but can show some results that are expected to compare against...

Sign in to comment.


Djamil Boulahbal
Djamil Boulahbal on 25 Oct 2024 at 19:54
Thank you to all who contributed. I truly appreciate that. The chirp function did the trick for me.
I'm now rather curious at the inner workings of the chirp function, ... but that's for another day.
Cheers

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!