What is a Smith Chart?

At Arena Physica, our mission is to build electromagnetic superintelligence. We recently adopted a new logo inspired by the Smith chart, which harkens back to the origins of radio frequency engineering. We wanted to use this opportunity to give a brief technical introduction to what a Smith chart is and why it's so important to us.

What is impedance?

Before we can get to the Smith chart, we first need to understand impedance.

Electrical "resistance" is something that resists the flow of direct current in a circuit (and corresponds to the $R$ in Ohm's law: $V = IR$). Electrical "impedance" generalizes the concept of resistance to include the effects of inductors and capacitors under alternating current (AC), using a complex number $Z = R + jX$, where $R$ is the resistance, $X$ is the "reactance", and $j$ is the imaginary number $\sqrt{-1}$. Writing impedance as $R + jX$ separates two very different behaviors.

The reactance $X$ (the imaginary part) corresponds to energy that a capacitor or inductor exchanges back and forth with an electric or magnetic field. That reactance depends on frequency $\omega$: an inductor with inductance $L$ has a reactance $\omega L$ that grows as the current alternates faster, while a capacitor with capacitance $C$ has a reactance $-1/\omega C$ that shrinks toward zero as the frequency increases.

The resistance $R$ (the real part) corresponds to energy that leaves the source and does not return. We usually picture that as heat, which is right for a resistor, where the energy leaves by warming something up. But some things have a purely real impedance yet send the energy onward instead of dissipating it. The clearest example is the vacuum of free space, whose intrinsic impedance of $\approx 377\,\Omega$ is the ratio of the electric to magnetic field strength of a light wave passing through it. The wave travels forward forever without losing any energy to heat. Inside a circuit, the equivalent concept is the characteristic impedance of a transmission line carrying waves from one part of the circuit to another, and is specified by the ratio of the voltage to the current of the wave traveling through that transmission line. If the two are perfectly in phase with each other, the impedance is purely real, but instead of dissipating power, it transfers it forward through the line.

In practice, what matters most about impedance is what happens when two of them meet. To understand why, we have to talk about reflections.

A good “impedance match” prevents reflections

Consider two ropes tied together, and let's say that the second rope is much heavier than the first. Wiggle one end of the rope and a pulse travels down it. When the pulse hits the knot, part of it travels through — and part of it reflects back at you. This happens if the second rope is lighter than the first, too. The only way to get the pulse to travel through the second rope without any reflections is to make the two ropes exactly the same weight.

Try playing around with the weight of the ropes below and notice what happens to the pulse.

The more different the two rope weights are, the more the pulse reflects back at you. If the second rope is heavier than the first, the pulse slows down when it hits the knot, and the reflected pulse flips upside down. If the second rope is lighter, the pulse speeds up when it hits the knot, and the reflected pulse stays right side up.

It turns out that circuits behave in the same way, but instead of considering the weight of two connected ropes, we have to consider the impedance of two connected components. Picture a transmission line of characteristic impedance $Z_0$ carrying a wave toward a load (the component at the far end that receives the signal) of impedance $Z_L$. If the load matches the line ($Z_L = Z_0$), the wave flows straight in and nothing comes back. If there's a mismatch, part of the wave turns around. And just as with the ropes, the more different the two impedances are, the more of the wave reflects. Being able to predict all of this is critical for an electrical engineer, because it sets the performance of a system: any power that reflects off the load is power you generated but never delivered.

We call the fraction that reflects the "reflection coefficient" $\Gamma$, the ratio of the reflected wave to the incident one:

A matched load gives $\Gamma = 0$ (nothing reflects), a short circuit ($Z_L = 0$) gives $\Gamma = -1$, and an open circuit ($Z_L \to \infty$) gives $\Gamma = +1$. In between, $\Gamma$ is a complex number: its magnitude says how much of the wave returns, and its sign (or more generally, its phase) says whether it comes back upright or flipped.

It's usually cleaner to measure impedance relative to the line, using the normalized impedance $z = Z_L / Z_0$, so that:

A matched load now sits at $z = 1$, and $\Gamma$ depends only on how far the load strays from the line feeding it.

Folding infinity in on itself

Since the impedance $z$ is a complex number, we can view it as a point somewhere in the complex plane, with its real part on the horizontal axis and imaginary part on the vertical axis. Since inductors have positive reactance and capacitors have negative reactance, we can think of the upper half of the plane as being "inductive" and the lower half as being "capacitive", with the horizontal axis being purely resistive. A short circuit lives at the origin, and an open circuit lives infinitely far away to the right of the figure. While this plane is easy to understand, its unbounded extent makes it hard for an engineer to work with.

The reflection coefficient $\Gamma$ comes to the rescue by folding the entire right half of the impedance plane into a single unit disk using a Möbius transformation of the impedance. Our choice of reference impedance sets where the origin $(0, 0)$ is (by convention, we usually let it correspond to $50\, \Omega$). In the reflection coefficient plane, the short circuit moves to $(-1, 0)$ and the open circuit moves to $(1, 0)$. What were once straight vertical and horizontal lines of constant resistance and reactance in the impedance plane are now perfect circles in the reflection coefficient plane.

To get a visceral feel for the mapping between $z$ and $\Gamma$, try grabbing the orange dot on either plot below and watch how its position changes on the other plot as you move it around. Notice how everything on the border of the reflection disk (with maximum reflection) lies somewhere on the vertical axis of the impedance plane with no resistance. Also, the dot shoots off the impedance plane when you drag it close to the right side of the reflection disk:

This disk is now commonly known as a Smith chart, and was independently conceived in 1937 by Tōsaku Mizuhashi, in 1939 by Phillip H. Smith [·], and in 1940 by Amiel R. Volpert, as a graphical calculator for Radio Frequency (RF) engineers. The chart moved every quantity an engineer cared about onto some circle or arc on the reflection coefficient disk. Computations that were once difficult could now be performed easily with a compass and a ruler. Even though we now have powerful computers that can do these calculations for us, as we will soon see, the Smith chart's properties continue to make it a valuable tool for RF engineers today.

Transmission lines rotate phase

A critical component of an RF engineer's toolkit is the transmission line: a cable, trace, or waveguide that carries a signal from one part of a system to another.

So far we've mainly described the reflection coefficient by its real and imaginary parts, but since it lives on a disk, it's often more natural to describe it by a magnitude and a phase. The magnitude $|\Gamma|$ is the distance from the center of the chart, and tells us how much of the wave reflects back. The phase is the angle around the center, and tells us how delayed the reflected wave is relative to the one we sent in.

A device is always driven by a signal oscillating at some frequency. Since a wave takes time to travel down a line, the farther the signal has to go, the more its phase lags by the time it arrives. As the line carries the signal along, that lag keeps growing. On the Smith chart, sliding our measurement point along a lossless line traces a circle of constant $|\Gamma|$ around the center. The distance from the center never changes, only the angle, so the line rotates the phase while leaving the magnitude unchanged.

This picture assumes an ideal, lossless line, one that carries power forward without dissipating any of it, so $|\Gamma|$ stays fixed and the point rides its circle indefinitely. A real transmission line is slightly lossy, bleeding off a little power as heat along the way, much like a resistor. Loss shrinks $|\Gamma|$ as the signal travels, so the circle winds inward into a spiral that slowly closes toward the center.

When a wave reflects off a mismatched load, the forward and backward waves overlap into a standing wave, a pattern of peaks and troughs fixed in place along the line. With no reflection there is only a forward-traveling wave and no standing pattern at all. With a perfect reflection the forward and backward waves are equal, and the standing wave is at its most extreme: the energy gets stuck in place, sloshing back and forth in the fields without ever moving forward. Engineers measure how pronounced this pattern is with the voltage standing wave ratio (VSWR), the ratio of the standing wave's largest voltage to its smallest. A perfectly matched line has a VSWR of 1, and the value climbs as the mismatch worsens. Since it depends only on $|\Gamma|$, a given VSWR is just one of those constant-magnitude circles on the chart.

Try dragging the orange probe along the transmission line to see how the reflection coefficient winds around a circle as you move down the length of the line. You can drag the resistor or reactance to change the load, or drag the cable to add loss and wind the circle into an inward spiral. A voltage maximum along the line lands on the right of the chart, and a minimum on the left. Notice in particular how a "matched" load allows the signal to travel through the line perfectly without any reflections, whereas a "reflecting" load results in a motionless standing wave throughout the line:

Filters selectively pass frequencies

So far, everything we've put on the chart has lived at a single frequency, but real devices respond differently as the frequency changes. A filter is an example of a component that selectively allows certain frequencies to pass through while rejecting others. In 2D, we can plot either the real or imaginary part of the reflection coefficient against frequency, but looking at either component in isolation loses the full story. If we view the real and imaginary parts together as a curve colored by frequency, tracing out a winding path through the Smith chart, we can see the complex frequency response all at once:

Placing a filter between two systems connected to the filter's input and output ports (ports 1 and 2, respectively) allows us to talk about the amount of power reflected or transmitted from one system to the other by referring to the filter's scattering parameters $S_{ij}$ (also known as "S-parameters"). This value $S_{ij}$ is the ratio of the wave leaving port $i$ to the wave incident on port $j$, so when $i = j$ we get exactly the reflection coefficient $\Gamma$ we've been plotting on a Smith chart. When $i \neq j$, we get a transmission coefficient for how much of the wave crosses from port $j$ to port $i$. For a "lossless" two-port filter (one which doesn't dissipate any power), the reflected and transmitted powers sum to the incident power, so $|S_{11}|^2 + |S_{21}|^2 = 1$: anything that isn't reflected must be transmitted.

To get accustomed to reading response curves off a Smith chart, we can look at a few examples of filters below, represented by their circuit diagrams and their Smith chart representations. In addition to the Smith chart, we've included a plot of the magnitude of S-parameters against frequency, where it's clear that anything not reflected is transmitted by the filter:

A bandpass filter transmits only within a certain frequency range, a low-pass filter transmits low frequencies, and a high-pass filter transmits high frequencies. Different types of filters have different properties. A Chebyshev filter can achieve a steeper rolloff at the boundary between the reflected and transmitted frequency range than a Butterworth filter, for example, but at the cost of some "ripple" in the passband. The "order" of the filter counts the number of resonators (inductors or capacitors) it contains.

These examples share a few patterns:

• Higher order filters have a steeper rolloff than lower order filters.
• The Chebyshev filter order controls the number of loops in the response curve.
• At a "resonant frequency", the reflection drops to 0 and intersects the origin of the Smith chart.
• Low-pass filters are mostly "inductive": they're built with inductors in series, and their reflection coefficient curve primarily lives in the upper "inductive" half of the Smith chart.
• High-pass filters are mostly "capacitive": they're built with capacitors in series, and their reflection coefficient curve primarily lives in the lower "capacitive" half of the Smith chart.
• Bandpass filters combine a low-pass and high-pass filter, behaving more like an inductor at high frequencies, and more like a capacitor at low frequencies.

These are all things that an RF engineer has intuition for. When things look off, an engineer can use the Smith chart to quickly diagnose the problem and identify a fix to address it.

Create a matching network yourself

Now that we've set up how to read a Smith chart, we can put ourselves in the shoes of an RF engineer and use it to do calculations for us. Suppose we're handed a component, a "load", whose impedance doesn't match the $50\,\Omega$ that the rest of our system expects. Connected as-is, it reflects some of our signal back up the line and wastes it. The job of impedance matching is to build a network of components in front of the load that transforms its impedance to $50\,\Omega$, which on the Smith chart means driving its reflection coefficient all the way to the center.

The key point to realize is that any component we add moves us along a circle to a new point on the chart. A reactive component placed in series with the load adds to the imaginary part of the impedance. Sliding the point along a constant-resistance circle, a series inductor walks it up toward the inductive top of the chart, whereas a series capacitor walks it down toward the capacitive bottom. A component placed in "shunt" (branching off in parallel toward ground) slides the point along a different family of circles instead, the constant-conductance circles, which sweep across the resistance circles from the other direction. Working outward from the load, we can chain these moves together until the $\Gamma$ point reaches the center.

To be more explicit, take a load $Z_L = 17.5 - j5.0\,\Omega$, for instance. To move it to $50\,\Omega$, we can first add a capacitor in series to bring us to the target conductance $G = 20\,\text{mS}$ while keeping $R$ constant. Then, we can add an inductor in parallel to bring us to the target resistance of $50\,\Omega$ while keeping $G$ constant, completing the match:

But that is not the only way to reach the center. We also have a third move available from earlier: a length of transmission line rotates the point around a constant-$|\Gamma|$ circle, swinging it around to wherever a series or shunt part can take over, all without changing how matched the load is. The same match from above can be achieved with a transmission line followed by a series capacitor:

Notice what's absent from this set of moves: the resistor. We could change the resistance directly by adding one, but a resistor dissipates power as heat, throwing away the signal we're trying to deliver. Matching is meant to be lossless, so we typically build these networks from only inductors, capacitors, and transmission lines.

Reaching the center usually takes two moves, but there are really an infinite number of ways to create a matching network for a given load. To help build some intuition, try it yourself in the mini game below. Given a random load and target frequency, can you build a network that matches it by bringing the reflection coefficient to the center? Drag a part onto the rail to place it in series, or below the rail for shunt. Drag a placed part side to side to tune its value, and click to remove one. Because each component acts across the whole spectrum at once, you can pin the match at only a single frequency. The rest of the range quickly slips away to either side:

Lifting to the third dimension

Everything we've plotted so far has been a passive load, one that can only send back some fraction of what it receives, so its reflection coefficient sits somewhere inside the unit disk (i.e., $|\Gamma| \le 1$). But there's an entire half of the impedance plane we've quietly set aside: the region of negative resistance. An "active" device like an amplifier or oscillator can behave like a negative resistance, reflecting back more power than it receives, so its reflection coefficient has $|\Gamma| > 1$ and lands outside the unit circle.

We can account for this by taking our 2D Smith chart one step further, lifting it into a sphere, with one hemisphere representing our existing Smith chart of passive components, and the other representing active components (as was first proposed by Andrei Müller et al. in 2011 [·]). You rarely need the full sphere in day-to-day work, but it completes the picture nicely:

Where do we go from here?

By now, the Smith chart's original purpose is long gone. It was a calculator. Before computers, working out $(z-1)/(z+1)$ and coming up with a transmission line rotation or a matching network by hand meant an afternoon with tables, but the chart turned it into a few marks with a compass. That arithmetic is instant now, and nobody reaches for a paper chart to compute.

So why make this our logo?

Yes, we think it looks cool, and it's instantly recognizable as an iconic part of radio frequency engineering. But deeper than that, it's a tool that encodes intuition. The field of RF is often regarded as the "black magic of engineering" by outsiders, and typically requires many years of experience to develop the instincts that allow an engineer to "just know" what to do. The Smith chart was a tool that attempted to codify that intuition through a change in perspective, to make the job easier for an expert in the field, and to hand off the knowledge of an expert to the next generation that would be building upon it.

At Arena Physica, we're building and deploying a form of "artificial intuition" with our customers that allows engineers to rapidly design, test, debug, and optimize hardware at scale, using a combination of foundational electromagnetic AI and complex agentic systems that engage with the physical world. We believe that this kind of AI will be essential to the development of the next generation of hardware, just as the Smith chart was for the last.