Imagine you are looking at a butterfly with its wings spread open: an even signal is like a perfect butterfly – its right side is an exact mirror image of its left side. An odd signal, on the other hand, is like a seesaw in a playground: when the right side goes up, the left side goes down by exactly the same amount (and at the midpoint the height is always zero). The beauty is that every crooked, asymmetric signal in the world can be decomposed into two "transparent layers" – one layer symmetric like a butterfly and a second, inverted layer like a seesaw – which, when you add them together, reproduce the original signal exactly.
A signal is called even if it is symmetric about the vertical axis, that is, for every \(t\) we have \(x(t) = x(-t)\) (like \(t^2\), \(\cos(t)\) or \(|t|\)). A signal is called odd if it is antisymmetric, that is, for every \(t\) we have \(x(-t) = -x(t)\) (like \(t\) or \(\sin(t)\)).
The real power of the theory is that every general signal – continuous or discrete, even if it is neither even nor odd – can be decomposed uniquely into the sum of an even component and an odd component: \[x(t) = x_{\text{even}}(t) + x_{\text{odd}}(t)\] where the components are computed as follows: \[x_{\text{even}}(t) = \dfrac{x(t) + x(-t)}{2} \quad , \quad x_{\text{odd}}(t) = \dfrac{x(t) - x(-t)}{2}\]
Geometrically, the superposition (addition) of the two graphs at every point gives back the original signal. For example, in the decomposition of a discrete step signal (which equals \(0\) for \(n<0\) and \(1\) for \(n\ge 0\)), the even component equals \(0.5\) at every \(n\neq 0\) (and \(1\) at \(n=0\)), and the odd component equals \(-0.5\) at the negatives and \(0.5\) at the positives (and \(0\) at \(n=0\)) – so that on the negative side they cancel each other to \(0\), and on the positive side they add up to \(1\).
The wide main graph at the top, in glowing white-gold. In discrete mode it is a Stem Plot.
Bottom left, in neon turquoise, with a soft vertical mirror line on the axis \(t=0\).
Bottom right, in neon magenta, with a glowing anchor point at the origin \((0,0)\).
Switches between continuous curves and stem plots over \(n\) from \(-8\) to \(8\). The same examples exist in both modes.
Pure even signals, pure odd signals and decomposition examples (signals that are neither even nor odd). The selected example is marked in gold.
In continuous mode: a shift slider \(t_0\) that moves the signal to \(x(t-t_0)\), with magnetic stops at \(0\) and at \(\pm\frac{\pi}{2}\). In discrete mode: the "Drag the stems" button, and then you grab any stem in the main graph and drag it up or down.
Moving the mouse over a point (or touching it) places two vertical laser lines: one at \(+t_0\) and one at its twin at \(-t_0\), in all three graphs. A click pins the scanner in place.
In the corner of the main graph: the real numbers inside the formulas of \(x_{\text{even}}(t_0)\) and \(x_{\text{odd}}(t_0)\), the sum check, and the values of both components at the twin \(-t_0\).
From 0% to 100%: it lifts the two lower graphs and lays them on top of each other inside the main graph, over the original signal.
Above 50%, soft bands appear: red where the two components have opposite signs ("Mutual cancellation"), and green where they have the same sign ("Constructive addition"), with a numeric label such as \(0.5 + (-0.5) = 0\).
The icon on the \(x_{\text{even}}\) graph: the right side folds over the left side about the \(y\) axis, and a tag shows the percentage of overlap.
The icon on the \(x_{\text{odd}}\) graph: the graph rotates by \(180^\circ\) about the origin, and a tag shows how well it lands on itself.
The mode and the signal, the shift \(t_0\), the scanner point and its twin with all their values, \(\max|x_{\text{even}}|\) and \(\max|x_{\text{odd}}|\) (zero = the component vanishes), the result of the symmetry test, and the superposition on each side.