Engineering Summary

Backdrilling becomes necessary when via stub resonance consumes enough channel margin to threaten 112G PAM4 compliance.

At lower data rates, via stubs may have minimal impact. At 112G, the combination of higher frequencies, reduced eye openings, and tighter channel budgets makes stub-induced reflections a significant source of signal degradation.

In many 112G systems, backdrilling is no longer a performance optimization. It is a risk mitigation strategy required to preserve channel integrity and compliance margin.

The engineering question is not: "What is backdrilling?" 
The real question is: "When does a 112G design become unable to meet performance requirements without backdrilling?"

Why 112G Changes the Design Rules

Many PCB structures that function correctly at 10G, 25G, or even 56G become problematic when deployed in 112G PAM4 systems.

The reason is simple:

Available signal margin decreases dramatically as data rates increase.

Every discontinuity consumes part of that margin.

Examples include:

• Connectors
• Via transitions
• Material variation
• Surface roughness
• Manufacturing tolerances

At 112G, via stubs often become one of the largest remaining discontinuities in the channel.

What was previously acceptable becomes a measurable source of compliance risk.

For high‑speed PCB designs, this shift is critical.

The Hidden Failure Mechanism: Via Stub Resonance

A plated through-hole via typically extends beyond the signal layer transition.

The unused section is known as a via stub.

Although electrically inactive from a routing perspective, the stub behaves as a resonant structure at high frequencies.

The failure mechanism follows a predictable sequence:

The longer the stub, the stronger the resonance effect.

As frequencies approach those required by 112G PAM4 signaling, the impact becomes increasingly significant.

This is why many channels that appear acceptable in layout reviews fail during laboratory validation.

What Happens When Backdrilling Is Not Used

Many engineers assume that material upgrades alone can solve high-speed channel issues.

In reality, stub-related reflections often remain even after low-loss materials are introduced.

Common consequences include:

Return Loss Degradation

Reflected energy increases. Receiver equalization becomes less effective. Channel margin decreases.

Eye Diagram Closure

Reflections interfere with the primary waveform. Typical symptoms include reduced eye height, reduced eye width, increased deterministic jitter, and higher receiver sensitivity.

COM Margin Reduction

Channel Operating Margin (COM) is widely used to evaluate 112G performance. Via stubs consume valuable margin that could otherwise compensate for connector losses, material variability, and manufacturing tolerances. Many compliance failures occur because multiple small impairments accumulate. Via stubs are often among the most preventable contributors.

Increased Bit Error Rates

As reflections increase, receiver stress increases, equalization becomes more aggressive, and error rates rise. A design may successfully pass simulation yet fail hardware validation because actual via structures introduce losses that were underestimated during modeling.

When Does Backdrilling Become Mandatory?

This is the most important engineering question.

Backdrilling is not automatically required for every PCB.

Its necessity depends on data rate, via length, PCB thickness, channel loss budget, compliance targets, and receiver architecture.

General industry guidelines are shown below.

Additional conditions that increase the need for backdrilling include thick multilayer PCBs, long backplane channels, AI server systems, data center switches, telecom infrastructure, and high-layer-count networking hardware.

In these environments, eliminating via stubs often provides greater benefit than further reducing dielectric loss.

For multilayer PCB designs, backdrilling is a key enabler.

Backdrilling Versus Alternative Solutions

Engineers frequently evaluate alternatives before selecting backdrilling.

Blind Vias

Advantages: Reduced stub length. Limitations: Higher fabrication cost, increased process complexity.

Buried Vias

Advantages: Improved signal transitions. Limitations: Limited applicability in large backplane architectures.

HDI Microvias

Advantages: Minimal discontinuity. Limitations: Cost constraints, reliability considerations in some applications.

Backdrilling

Advantages: Significant stub reduction, compatible with conventional through-hole structures, lower cost than full HDI solutions.

For many 112G networking and server designs, backdrilling remains the most practical method of controlling via-induced reflections.

How Much Stub Is Acceptable?

A common misconception is that every stub must be completely removed.

In reality, the engineering objective is to reduce the stub below the point where it materially affects channel performance.

Acceptable residual stub length depends on operating frequency, channel budget, stack-up design, and compliance requirements.

The correct answer is determined through simulation and measurement—not assumptions.

This is why many high-speed programs define maximum residual stub requirements within fabrication drawings.

Validation: How to Prove Backdrilling Is Necessary

Backdrilling decisions should be based on evidence.

Several validation methods are commonly used.

Time Domain Reflectometry (TDR) – measures impedance discontinuities, reflection magnitude, and stub influence.

Vector Network Analysis (VNA) – measures return loss, insertion loss, and frequency-domain channel performance.

Eye Diagram Analysis – evaluates actual receiver margin and directly demonstrates the impact of stub-induced distortion.

Channel Operating Margin (COM) – used extensively for 56G PAM4, 112G PAM4, and 224G PAM4, providing quantitative evidence of whether backdrilling improves compliance margin.

Correlation Testing – simulation alone is insufficient. Engineering confidence comes from correlating simulation, fabrication, measurement, and compliance results.

Manufacturing Considerations

Backdrilling effectiveness depends on fabrication execution.

Critical parameters include residual stub targets, drill registration accuracy, layer alignment, hole position tolerance, and finished hole dimensions.

A poorly controlled backdrill process can leave excessive residual stubs and reduce the expected benefit.

For this reason, backdrill requirements should be clearly documented during the design-to-manufacturing handoff process.

UltroNiu’s special PCB and Rogers PCB services include backdrill as a standard capability.

Engineering Conclusions

112G PAM4 systems operate with significantly reduced signal integrity margin compared with previous generations.

Via stubs create resonant structures that generate reflections, degrade return loss, reduce eye openings, and consume valuable channel margin.

As channel speeds increase, these effects become increasingly difficult to compensate through equalization alone.

Backdrilling removes the unused portion of the via barrel and directly addresses one of the most common high-speed PCB failure mechanisms.

For many 112G channels, the question is not whether backdrilling improves performance. The question is whether sufficient compliance margin remains without it.

The answer should always be supported by TDR measurements, VNA characterization, eye diagram analysis, COM evaluation, and hardware validation.

Because a 112G design rarely fails from a single catastrophic issue. Most failures occur when multiple small impairments consume the available margin—and via stub resonance is one of the most common and preventable among them.

Related Engineering Questions

• When does backdrilling become mandatory in a high-speed PCB?
• How much residual stub is acceptable for 112G PAM4?
• Backdrilling vs blind vias: which provides better channel performance?
• Can low-loss materials replace the need for backdrilling?
• How is via stub resonance measured and validated?
• What is the impact of backdrilling on COM margin?

For deeper insights, explore our 112G loss budget allocation and capability matrix articles.