2 - Fourier Transform Sampling Pairs
Understanding Pulse Trains and Fourier Transform: A Simple Guide for Signal Processing
What Is a Pulse Train in Signal Processing?
Let’s talk about something that shows up a lot in signal processing and computer vision: pulse trains.
A pulse train is basically a series of evenly spaced impulses (spikes). Think of it like this:
- One spike → okay, simple
- Many spikes spaced out → that’s your pulse train
These impulses are often used to model sampling, repetitive signals, or structured patterns in data.
The Key Idea: Pulse Train ↔ Pulse Train
Here’s the important relationship:
The Fourier transform of a pulse train is another pulse train.
That might sound odd at first, but it’s actually a core concept in frequency analysis.
The Spacing Rule (This Is What Matters)
There’s one idea you really need to remember:
- If pulses in space (or time) move farther apart → pulses in frequency move closer together
- If pulses in space get closer → pulses in frequency spread out
This comes from the scaling property of the Fourier transform, but you don’t need the full derivation to understand the intuition.
A Simple Way to Think About It
Let’s make it intuitive.
Imagine this:
- You keep increasing the distance between pulses
- Eventually, they’re so far apart that it feels like you only have one single impulse
Now recall this:
- The Fourier transform of a single impulse is a constant (flat line across all frequencies)
Why?
Because an impulse contains all frequencies equally.
So if your pulse train spreads out more and more:
- Its frequency representation gets tighter and tighter
- Until it looks like a dense block (almost continuous)
Visual Intuition
Think of it like stretching and compressing:
- Stretch in space → squeeze in frequency
- Squeeze in space → stretch in frequency
This trade-off shows up everywhere in signal processing and computer vision.
Why This Matters in Computer Vision
This concept is useful when working with:
- Image sampling
- Texture analysis
- Periodic patterns in images
- Frequency filtering
For example, repeating patterns in images (like grids or textures) often behave like pulse trains in the frequency domain.
But there's one other pair that I, I need you to know about. And again we're not going to derive this but it's important. And that is the notion of a pulse train or a set of impulses, okay. And so this is refered to as a pulse train, you can see that here.

And the Fourier transform of pulse train is another impulse train. And the thing that you need to realize is that, and remember the scaling theorem give us this and a bunch of things give us this, right? As the pulses in space get further apart, the pulses in frequency get closer together. And this is a hand-wavy way of thinking about this. Imagine that our pulses got very, very far apart, okay? So far apart that I have an impulse in the middle and the next one is infinitely far away. Well, that would be just like having sort of a single impulse. That's the hand-wavy part. Well, what is the Fourier transform of a single impulse? Well, we said it was a 1 because an impulse gives you back the, when you can involve it, it gives you back the entire image. So therefore the frequency has to keep all the frequencies. So the hand wavy way of thinking about this is, as I stretch out my impulses in space, I have to shove all these impulses and frequency in time infinitely close together to make a solid bar. Okay, and that's one way of thinking about how the Fourier transform of an impulse train is another impulse train, and as they get bigger in space they get closer in frequency.