Real-world impact of via stubs

Design and Layout
1

Has anyone measured the impact of via stubs on insertion loss or eye diagrams for multi-gigabit signals?
In simulation the effect is obvious, but I’m curious at what stub lengths people have actually seen measurable degradation during lab validation.

1 Like
2

Haven’t done this measurement myself, but from simulation work, two designs at the same data rate can behave very differently depending on transmitter rise time and the overall loss budget, so the stub threshold isn’t always consistent. Most fabs target ≤10 mil residual stub after backdrill for >10 Gbps designs.

1 Like
3

Industry data suggests FR-4 stubs >40-50 mils often show measurable loss at 5-10 GHz, while ≤10 mil residuals (post-backdrill) are the standard target to ensure signal integrity for >10 Gbps. The exact tolerance varies heavily with rise time and modulation.

1 Like
4

One thing that’s often more useful than the physical stub length alone is the stub’s resonant frequency. A relatively short stub may have little impact if its quarter-wave resonance is well above the signal’s spectral content, while a longer stub can create a noticeable notch in the insertion-loss curve if the resonance falls within the frequency range of interest. That’s why designers usually evaluate stubs against the channel frequency content rather than using a single length limit for every interface. The same stub can be insignificant for one link and problematic for another.

1 Like
5

One practical challenge in lab validation is isolating the effect of the via stub from the rest of the channel. Connector launches, package parasitics, and layer transitions can easily dominate the measurement and mask the stub’s contribution.

For that reason, many engineers compare structures that are identical except for the stub length (for example, backdrilled vs. non-backdrilled test coupons) and look for differences in TDR or S-parameter measurements. In many cases, the impact shows up as a localized impedance discontinuity or a notch in the frequency response before it becomes obvious in the eye diagram itself.

1 Like
6

Modern equalizers can mask stub-related degradation entirely: a channel may still pass BER testing even when the via stub is measurably affecting the signal.

Insertion loss or eye height alone doesn’t always tell the full story. In some cases, the first indication of an excessive stub is increased equalizer effort or reduced margin for future channel degradation, rather than an immediate link failure.

1 Like