Where to Place an Unavoidable Impedance Discontinuity

Every high speed link contains at least one place where the impedance is not what the stackup drawing promises. It might be a connector, a package, a via that has to change layers, or a component pad that is wider than the trace. The engineering question is not how to eliminate the discontinuity, since on a real board that is often impossible, but where to put it so that the link has the best chance of working.

The Question Behind the Question

A discontinuity matters because of what it does to the signal at that point and what it does to the channel afterwards. A reflection that returns quickly and is absorbed by the driver is harmless; a reflection that arrives at the receiver during the sampling window is not. A stub that resonates inside the band of interest is worse than a stub that resonates well above it.

Placement is therefore a way of controlling when the reflection arrives and how much of the channel it affects. Two boards with identical component choices can behave differently simply because the discontinuity sat in a different place along the route.

What a Discontinuity Actually Does

Three mechanisms matter. The first is reflection: the impedance change sends part of the incident energy back toward the source, which reduces the amplitude that reaches the receiver and adds a delayed echo. The second is insertion loss: a discontinuity can also absorb or radiate energy, which reduces the total signal that arrives. The third is resonance: an unmatched section behaves as a resonator, and at certain frequencies it can produce a deep notch in the channel response.

The relative importance of the three depends on the data rate, the length of the channel and the equalisation available at both ends. A short channel with a simple driver has little tolerance for reflections, while a long channel with adaptive equalisation can absorb more, provided the discontinuity is not sharp enough to create a notch that no equaliser can invert.

Place It Near the Transmitter, Not the Receiver

The single most useful rule is to keep the discontinuity close to the transmitter. A reflection generated near the source travels a short distance before being partially absorbed by the driver’s own output impedance, and the remaining echo arrives at the receiver with the transmitted energy, where it is less likely to corrupt the decision. Placing the same discontinuity near the receiver inverts the situation: the reflection bounces back toward the source and returns, arriving after the main edge.

The corollary is that where a link has a fixed geometry with a connector in the middle, the connector tends to be acceptable while a badly designed via near the receiver is not. This is why the placement of package vias and of the last layer transition before the receiver deserves particular attention.

Timing: Keep It Away From the Sampling Instant

Whether a reflected echo matters depends on when it arrives relative to the receiver’s sampling instant. Reflections that arrive close to the edge being sampled add directly to the signal and change the crossing point, which is the worst case. Reflections that arrive well after the sampling instant affect the following bit rather than the current one, which is usually less damaging.

This is the practical reason that a discontinuity close to the receiver is dangerous, and it also explains why the exact position matters rather than the raw distance. Reflections that arrive well after the sampling instant affect the following bit rather than the current one, which is usually less damaging.

Avoid Resonant Structures and Long Stubs

A stub that is electrically short compared with the rise time behaves as a small lumped capacitance and is usually tolerable. As the stub approaches a significant fraction of a wavelength, it becomes a resonator, and the resulting notch can be impossible to equalise out. Via stubs in thick backplanes are the classic example, and they are the reason back drilling exists.

Avoid structures that create a narrow resonance inside the band of interest, and keep stubs short by choosing a layer transition that minimises the unused barrel length. Our notes on microstrip and stripline routing describe how the structure of a route affects both impedance and the length of any stub the transition introduces.

<img src="https://www.gopcba.com/wp-content/uploads/2026/09/206-1.jpg" alt="Channel response plot showing a notch caused by a via stub resonance” />

Loss, Equalisation and Channel Margin

Equalisation changes the answer to the placement question. A receiver that can adapt to the channel will compensate for a gradual impedance variation and for a moderate reflection, but it cannot invert a sharp notch or restore energy that has been lost. Where the link includes decision feedback equalisation, the receiver can cancel a discrete echo to some extent, which reduces the penalty of a reflection at the far end.

Understanding what the receiver can and cannot undo is therefore part of the placement decision. A channel with a strong equaliser has more room to place a discontinuity near the receiver than a simple link does, though the safest choice remains the same when there is no reason to do otherwise.

Documenting the Choice

Where a discontinuity is unavoidable, record the decision. Note where it sits, why it is there, what the expected reflection is and how it was considered during design. That record turns a compromise into an engineering decision, and it gives the next person a reason not to move the connector during a mechanical revision.

The record also feeds the review. Our outline of layout verification methods includes checks for exactly this kind of accepted exception, and the check is only meaningful if the accepted exceptions were written down in the first place.

Additional Considerations for This Build

Practical attention to loss budget pays for itself here, because it is one of the items that decides whether the finished board behaves as the drawing intended. Where the requirement is not stated on the fabrication drawing or in the assembly notes, the shop has to assume a default, and that default is rarely the value the design was simulated with. Stating loss budget explicitly, together with the tolerance that applies, removes the assumption and keeps the result predictable from batch to batch.

Process Control and Verification

Reviewing the design before the data is released is far cheaper than correcting it after the panel is in the tank, because every step downstream inherits the decision made at the front end. Where a value sits close to a process limit, the drawing should say so, since the shop can then open the process window rather than working to a nominal figure that carries no tolerance.

The measurements that matter are the repeatable ones: conductor width and spacing, annular ring, finished hole size, plating thickness and surface finish are all verifiable on a coupon that travels with the panel. Documenting the assumption is part of the design work, and a short note on the drawing prevents a question that would otherwise arrive a day later and cost a day of schedule.

Where a value sits close to a process limit, the drawing should say so, since the shop can then open the process window rather than working to a nominal figure that carries no tolerance. Running a first article through the same checks as the production panel confirms that the two agree, and that comparison is the cheapest form of process control available at prototype stage.

Related reading: our fabrication notes, board quality and design release notes cover the same ground.

FAQ

Should I place the discontinuity before or after the connector? On the board side, as close to the transmitter as the layout permits. The connector itself is fixed by mechanics, but the via and pad transitions that surround it are not, and those are usually the parts that can be moved to reduce the effect of the reflection.

How much reflection is acceptable? It depends on the interface specification and on the receiver’s capability. As a practical approach, compare the expected reflection against the amplitude and timing margin available at the receiver, and treat any structure that creates a resonance inside the band of interest as unacceptable regardless of its reflection amplitude.

Can back drilling remove the problem entirely? It removes the stub and therefore the resonance associated with it, which is usually the dominant effect at high frequency. It does not remove the via’s own capacitance and inductance, so the transition still presents a small discontinuity, but a much smaller and better behaved one.