In a book on optical solitons, the author says that a Gaussian signal pulse has the form-

$u(t) = \exp((\eta + i\beta)t^2)$

where $\beta$ is the chirp parameter.

My question is: what does the $i$ signify here? As a chirped signal is still real, what does the complex nature of the signal represent?

  • $\begingroup$ Dunno if it useful to you, but here is a post i have done on the generalized chirp. $\endgroup$ Commented Aug 12, 2020 at 14:16
  • 1
    $\begingroup$ BTW, your "chirp parameter" $\beta$ is $\frac12$ of the sweep rate. $2\beta$ is the sweep rate. $\endgroup$ Commented Aug 15, 2020 at 15:25
  • $\begingroup$ OP and @robertbristow-johnson, you may find my newest article pertinent and helpful. (I've been trying hard to replicate "Mohonk" exactly in the discrete domain for years. This may be the key.) $\endgroup$ Commented Aug 16, 2020 at 18:54
  • 1
    $\begingroup$ Dang, I thought I had added the link: dsprelated.com/showarticle/1365.php "The Zeroing Sine Family of Window Functions" $\endgroup$ Commented Aug 16, 2020 at 19:17

2 Answers 2


Depending on the context, the use of the complex form could be for mathematical convenience or for a no-kidding need for both real and imaginary parts.

When you factor the expression, you get

$$u(t) = e^{{\eta}t^2}e^{j{\beta}t^2}$$

Where the first exponential is a generic magnitude envelope, in this case Gaussian. The second exponential is the chirp itself and is where all the action is.

So to simplify things, lets assume that the envelope is ideal so that the signal is just the chirp

$$u(t) = e^{j{\beta}t^2}$$

To view the chirp, you can take either the real or imaginary part, and it looks something like this

enter image description here

In a system that does not use I/Q, the real part is what you would expect to transmit as your waveform. The type of system will determine whether using a real signal or some type of I/Q is best. I'm going to use a radar example here.

In frequency-modulated continuous wave (FMCW) radars, a real chirp like the one above is transmitted and has the form

$$x(t) = cos({{\beta}t^2})$$

Which is just the real part of the complex form. It is received after a delay and mixed with itself, and without going into the mixing process, produces a single frequency sinusoid that can be used to determine range. Here, using a real part only is practical. Using I/Q in FMCW is also beneficial (SNR improvement), but is not usually necessary and many systems do not use it.

Another type of radar, pulse-Doppler, benefits greatly from using the complex form. The same chirp is considered, except now the imaginary version is used. This is important because pulsed-Doppler radars usually operate on performing pulse compression, which is just correlating the transmitted waveform with the received one.

The autocorrelation of a complex chirp looks like

enter image description here

Using a complex waveform allows us to mix our signals to baseband, which give the classic autocorrelation responses we expect without additional mixing and filtering.


$i$ is the symbol for $\sqrt{-1}$

There is a very important formula called Euler's Equation.

$$ e^{i\theta}=\cos(\theta) + i \sin(\theta) = (e^i)^\theta$$

"$ e^i $" is a point on the unit circle one radian along the circumference. Any point on the unit circle raised to a power will stay on the unit circle and its distance along the circumference will be multiplied by the power.

$$ (e^{i\theta})^p=e^{ip \theta } $$

Simply factor it.

$$ u(t) = e^{\left(\eta t^2\right)} \cdot \left(e^i\right)^{\beta t^2} $$

The first factor is your real Gaussian (bell curve) acting as an envelope.

The second factor is a point spinning around the complex unit circle. At a steady pace you would get a steady tone. This one's pace isn't steady, but one of linearly increasing frequency (in absolute terms away from the center).

Your signal/function is complex.

$$ \begin{aligned} u(t) &= e^{\left(\eta t^2\right)} \cdot \left[ \cos\left(\beta t^2\right) + i \sin\left(\beta t^2\right)\right]\\ &= e^{\left(\eta t^2\right)} \cdot \left[ \cos\left([\beta t] t\right) + i \sin\left([\beta t]t\right)\right]\\ &= \left[ e^{\left(\eta t^2\right)} \cdot \cos\left([\beta t] t\right) \right] + i \left[ e^{\left(\eta t^2\right)} \sin\left([\beta t]t\right)\right]\\ \end{aligned} $$

  • 2
    $\begingroup$ $\eta$ better the hell be negative. $\endgroup$ Commented Aug 12, 2020 at 14:14
  • $\begingroup$ Thanks, but I get that. I am having trouble understanding how I must "look" at the signal. If the magnitude is in Volts, at any given time, should I just look at the envelope? If not, how does the presence of the imaginary component come into what I actually measure? $\endgroup$
    – Paddy
    Commented Aug 12, 2020 at 14:42
  • $\begingroup$ @Paddy Kind of out of my area. I would venture that your measurement device strips the imaginary part, a graph of what it shows would confirm that. Other systems deal in I/Q signals like this, so the imaginary part is relevant. If your graph is showing the envelope then it is showing $\|u(t)\|$. $\endgroup$ Commented Aug 12, 2020 at 15:02
  • 1
    $\begingroup$ @Paddy Whatever your platform, that is the math of the situation. RB-J is actually quite a chirp expert, especially in the continuous domain. I happen to be studying Discrete Gaussian Eigenvectors of the DFT right now. No hurry on the check, you may get a better answer yet. $\endgroup$ Commented Aug 13, 2020 at 4:24
  • 1
    $\begingroup$ @robertbristow-johnson Where's your sense of adventure? Who doesn't want to see something spiral out of control spectacularly? $\endgroup$ Commented Aug 13, 2020 at 4:25

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.