My situation is as follows: i am trying to generate a waveform the hard way, by constructing the samples one by one and then saving the result to a .wav file using Python.

When the frequency is constant, everything is fine: i use $y(t) = \sin(2 \pi \cdot f \cdot t)$. However, if i change the frequency to be a function of time, things go wrong. If the function is a linear one of the form $f(t) = a + bt$, it still works. But if i choose, for example, $f(t) = 40 + 10 \sin(t)$, the max frequency increases over time, reaching a maximum higher than the expect 50Hz.

I have read something about instantaneous frequency, namely, this: Why does a wave continuously decreasing in frequency start increasing its frequency past the half of its length?. But doing the integral evaluation makes the sound even weirder. And the method i currently have works for linear function of time, so i figured there must be something else wrong.

I also tried to generate chunks of sound in the frequency i need, at each time, and then glue them together. I calculated the period of the oscillation, so that a chunk has as many samples as necessary to make a whole period on that frequency, so that there are no "jumps" between different frequencies. But generates a cracking sound in the sample.

Here is an example of the kind of a JavaScript implementation of the kind of sound i need: http://jsfiddle.net/m7US6/4/.

Any help would be appreciated. Thanks.

  • $\begingroup$ Can you post the bit of your Python code that generates the waveform? $\endgroup$
    – Phonon
    Commented May 22, 2014 at 9:10

1 Answer 1


Consider the function $\sin(2\pi f t)$. When $2\pi f t$ goes from $0$ to $2\pi$ you get oscillation of the sine wave, $2\pi$ to $4\pi$, another, and so on. So every time the argument changes by $2\pi$ you get one oscillation.

Now lets plot $2\pi f(t) t$ where $f(t) = 10 + 10 \sin(2 \pi t)$. (I've modified it a bit to show the effect more):

enter image description here

As $t$ increases, this value changes faster and faster, meaning $\sin(2\pi f(t) t)$ will have higher and higher frequency.

Another way of thinking about it - the frequency is the derivative of $f(t) t$. For the example this is $20 \pi (\sin(2 \pi t)+2 \pi \mathbf{t} \cos(2 \pi t)+1)$. Note the $t$ multiplier in there - $t$ increases thus the frequency increases.

So what you really want is $\tfrac{d}{dt} f(t) t = 40 + 10 \sin(t)$ which gives $f(t) t = 40t - 10\cos(t) + c$ and for your example try $ \sin(2\pi(40t - 10 \cos(t)))$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.