Reflection Coefficient: Magnitude, Sign and What It Predicts

Saying that a load is not 50 ohms and will therefore reflect is correct and almost useless. The engineering question has two parts: how much comes back, and does the returning wave add to the incident voltage at that node or subtract from it? The reflection coefficient compresses both answers into one quantity, and it is the first calculation worth doing on any interconnect that is being questioned.

The Load Is Not Visible Until the Wave Arrives

When a driver launches a step, the amplitude of the initial wave is set by the source impedance and the characteristic impedance of the line, not by the load. The wave then travels along the line, and only when it reaches the far end does the load determine what happens next.

That is the essential difference between a transmission line and a lumped connection. When the propagation delay is significant compared with the edge, the source cannot know what the far end looks like at the instant it drives. It will discover the answer one round trip later, and the intervening behaviour is governed by the geometry of the line rather than by the component at its end.

What the Coefficient Measures

The reflection coefficient is defined as the ratio of the reflected voltage to the incident voltage at a given interface. For a resistive load it follows directly from the relationship between the load impedance and the characteristic impedance of the line. A positive value means the reflection returns with the same polarity as the incident wave; a negative value means it returns inverted.

The magnitude tells you how large the reflected portion is relative to what arrived. The sign tells you whether the voltage at that node will be pushed up or pulled down. When the load includes reactance, the coefficient becomes a complex quantity, and its phase describes how the reflected wave is shifted. All three pieces of information are available from a single number, which is why it is worth computing before touching the layout.

Three Reference Points Worth Memorising

A perfectly matched load produces a coefficient of zero and no reflected wave at all. An ideal open circuit, where the load impedance becomes very large, drives the coefficient towards positive one, so the reflected wave returns in phase and the voltage at the load roughly doubles. An ideal short, where the load impedance approaches zero, drives the coefficient towards negative one, so the reflection returns inverted and the voltage at the load collapses.

Open and short both produce full reflection in magnitude, and they differ only in sign. The consequence is visible in the waveform at the far end: the same magnitude of returning energy produces opposite effects depending on the terminating condition, which is why the sign matters as much as the size.

The Source Reflects Too

If the source impedance is not equal to the line impedance, the returning wave is not fully absorbed when it arrives. Part of it is reflected again, and the process repeats between the two ends until losses and terminations bring the system to a steady state.

This is why source termination and load termination are not interchangeable solutions to the same problem. The load termination controls what happens at the far end, while the source termination controls how much of the returning wave is re-launched. Analysing a waveform without recording the source impedance, the line impedance, the load impedance and the propagation delay leaves out the four quantities that determine the shape.

Using the Coefficient to Locate a Problem

Start by listing the candidates: connectors, vias, stubs, package transitions and the load itself are all potential impedance discontinuities. Every one of those elements is an impedance discontinuity of some size, and they add to the one the load creates. If the voltage overshoots at the far end, the reflection is in phase; if it sags, the reflection is inverted. That polarity points towards a load that is too high or too low.

The timing of the reflection gives its distance. Measuring the delay between the launch and the arrival of the reflection, or reading the position from a time domain reflectometer trace, converts time into a physical location along the line. In the frequency domain, the return loss from S parameter data shows how much energy is being reflected across the band, while the time domain view shows where. The two views are complementary, and a case that looks ambiguous in one is usually clear in the other.

After any change to the termination or the structure, repeat the measurement and confirm that the reflection changed in the direction and by the amount predicted. That confirmation is what distinguishes a modelled understanding from a sequence of experiments.

Not Every Mismatch Is a Missing 50 Ohms

Fifty ohms is a common system impedance, but reflections come from a mismatch between the line and whatever it actually meets. A trace can be calculated at exactly fifty ohms and still encounter a discontinuity at a connector, a pad, a via, a stub or a device input, and the reflection coefficient at that point is what determines the waveform.

It also follows that some systems do not target fifty ohms at all. The useful comparison is always between the characteristic impedance of the line and the impedance of the structure the wave actually encounters, and the four conditions that matter are the same: magnitude, sign, delay and the possibility of a second reflection.

Turning the Number Into a Design Decision

Once the coefficient is known, the choice of what to do about it becomes narrower. A modest reflection that returns well after the receiver has sampled is usually harmless and needs no action. A reflection that arrives during the sampling window, or that has a magnitude large enough to push the node outside the receiver thresholds, has to be reduced, and the options are to change the termination, to shorten the section that caused it, or to change the geometry so that the discontinuity is smaller.

The geometry options belong in the layout review: keeping the trace width constant through a transition, avoiding stubs, and using a controlled impedance stack-up so that the characteristic impedance of the line is a number the design actually holds rather than one that is assumed, which is a question for fabrication capability as much as for the schematic. Where a connector or a package transition cannot be removed, its contribution can still be characterised and compared against the budget, which is a better position than not knowing whether it matters at all.

The final step is verification on the built board. A test plan that includes a time domain reflectometry measurement or a return loss sweep on the critical nets will show whether the impedance the design intended is the impedance the board has, and the difference between the two is the best available predictor of an intermittent problem in the field.

FAQ

Is a reflection coefficient of zero achievable in practice? Closely enough for most designs, with a matched termination and controlled geometry, but a perfect zero is an idealisation.

What does a positive coefficient look like on a scope? An overshoot at the load end as the reflected wave adds to the incident one.

Can the coefficient be measured? It can be inferred from time domain reflectometry and from return loss measurements, both of which give the magnitude and, in the time domain, the location.

Does terminating only at the source work? It reduces re-reflection at the source, but it does not remove the reflection at a mismatched load.

Summary

The reflection coefficient answers two questions at once: how much of the incident wave returns and how it affects the node voltage. Compute the magnitude and the sign first, use the open, matched and short cases as reference points, and remember that the source contributes a second reflection when it is not matched either. From the formula to the TDR trace to the structural change, it is the step that turns an intuition about impedance into a prediction that can be checked.