Delta-L testing is an insertion-loss measurement method proposed by Intel for evaluating signal attenuation in high-speed PCB transmission lines. It is widely used in the validation of PCBs and materials for AI servers, data centers, high-speed communications, and high-frequency applications.
The method is designed to determine the loss contributed by a defined length of transmission line while minimizing the influence of probes, connectors, vias, and test fixtures. This makes Delta-L particularly useful when PCB materials and manufacturing processes must be evaluated for high-speed design requirements.
What Is the Purpose of Delta-L Testing?
The primary purpose of Delta-L testing is to measure the insertion loss (IL) of a transmission line and express the attenuation per unit length, such as dB/in or dB/cm. High-speed interfaces, including PCIe, Ethernet, USB, and DDR, are sensitive to transmission loss. As frequency increases, the effects of conductor loss, dielectric loss, copper surface roughness, and manufacturing variation become increasingly significant.
Delta-L testing provides a practical way to evaluate whether a PCB transmission structure meets the loss requirements of a particular high-speed design. It can be used during material evaluation, process qualification, PCB fabrication verification, and comparison of different stackups or manufacturing conditions.
How Does the Delta-L Method Work?
Delta-L uses a differential comparison principle. Two test structures are designed with nearly identical electrical and physical configurations. The primary difference between them is the length of the signal path. One structure is the longer line and the other is the shorter line.
A vector network analyzer (VNA) measures the S-parameters of both structures. The loss associated with the additional length is obtained by subtracting the loss of the shorter line from the loss of the longer line:
Loss of the longer line - loss of the shorter line = loss caused by the additional line length
The resulting loss is then divided by the difference in line length to obtain the loss per unit length. In simplified form:
Loss per unit length = (ILlong - ILshort) / ΔL
Because the two structures share most of the same launch, probe, via, pad, connector, and fixture characteristics, their common errors are largely canceled during the comparison. The remaining result more closely represents the loss of the PCB transmission line itself.
This comparison approach is important because a direct measurement does not contain only the loss of the PCB trace. It also includes loss and discontinuities introduced by the probe interface, test pads, vias, connectors, and fixture. If these contributions are not separated from the transmission-line loss, the measured result may be higher than the actual loss of the line under evaluation.
Why Not Measure the Transmission Line Directly?
A direct insertion-loss measurement includes the combined effects of the probe, test fixture, vias, pads, and PCB transmission line. These elements may introduce their own insertion loss and impedance discontinuities. At high frequencies, even a relatively small discontinuity can create a measurable reflection and influence the S-parameter result.
Delta-L uses two structures with different lengths but otherwise matching configurations. The common measurement contributions are therefore reduced through subtraction. This does not eliminate every possible source of error, but it provides a more accurate way to estimate the loss per unit length of the PCB transmission structure.
The accuracy of this method depends on how closely the long and short coupons match. If their stackups, pad structures, via configurations, reference planes, trace dimensions, or manufacturing conditions differ, the subtraction may remove or retain the wrong effects. In that case, the calculated loss can include errors unrelated to the additional line length.
Differences Between Delta-L 3.0 and Delta-L 4.0
Two commonly encountered versions are Delta-L 3.0 and Delta-L 4.0. Their main differences are related to the supported frequency range and the design of the measurement coupon.
- Delta-L 3.0: The test frequency commonly extends to approximately 20 GHz.
- Delta-L 4.0: The test frequency can extend to approximately 40 GHz and is intended for higher-speed applications, such as PCIe Gen5, PCIe Gen6, and 800G networking. Its coupon and launch structures are also more demanding.
Delta-L 4.0 places greater emphasis on reducing discontinuities in the measurement launch. The pad dimensions are reduced to limit their effect on impedance and reduce return loss. The probe structure is redesigned with a smaller pitch, a shorter and smoother contact region, and a more continuous impedance transition between the probe and the transmission line.
The via structure is also optimized. Via stubs receive greater attention, and the coupon may use back drilling or VIPPO (via-in-pad plated over) structures to reduce high-frequency reflections. The long-line and short-line configurations still follow the basic Delta-L comparison principle, but the requirements for length selection and impedance consistency are more stringent so that regression errors are reduced.
Key Considerations for Delta-L Coupon Design
A Delta-L coupon is not simply a long PCB trace. Its design must account for signal integrity, design for manufacturability (DFM), and measurement requirements. The objective is to ensure that the measured result reflects the transmission line rather than the pad, via, probe, or layout structure.
1. Control the Differential Impedance
The entire differential pair should maintain the target impedance, such as 85 Ω or 100 Ω, depending on the structure being evaluated. Trace width, trace spacing, dielectric thickness, copper thickness, and etching uniformity all affect impedance. Any local impedance change can generate return loss and influence the Delta-L calculation.
2. Make the Long and Short Coupons Equivalent
The basic assumption is that the only intentional difference is the additional line length, ΔL. The stackup, via count, pad dimensions, reference planes, copper surface condition, trace width, trace spacing, probe configuration, and launch geometry should otherwise be matched.
The differential pair should also maintain the same length, width, spacing, via count, and bend characteristics. One line should not contain a different serpentine pattern, copper proximity condition, or bend geometry from the other. Such differences can introduce delay variation, impedance imbalance, and additional loss that will distort the comparison.
3. Optimize the Probe Launch
The test pad is one of the locations most likely to affect the measurement. A pad that is too large or that widens abruptly can create an impedance discontinuity. The trace should enter the pad smoothly, and unnecessary stubs should be avoided.
The launch region is especially important at higher frequencies. A poorly controlled launch may produce return loss that is incorrectly interpreted as transmission-line loss. This is why Delta-L 4.0 gives greater attention to pad size, probe pitch, and the transition from the probe to the coupon trace.
4. Keep Via Stubs as Short as Possible
If the coupon must change layers, back drilling is preferred where applicable. At high frequencies, a via stub can behave as a resonant structure and produce additional reflection. The resulting discontinuity can affect both the measured S-parameters and the calculated loss per unit length.
5. Maintain a Continuous Ground Plane
The reference plane should remain continuous and should not be cut or interrupted near the differential pair. Isolation pads should not be unnecessarily large. Otherwise, the signal return path may be altered, changing the local impedance and introducing measurement error.
6. Configure the Coupon Length Carefully
The long and short coupons must provide a sufficient length difference for the additional transmission-line loss to be distinguished from the common measurement uncertainty. A length difference greater than 10 cm is used in the described coupon configuration to reduce the relative influence of multiple reflections.
This value should be treated as a configuration requirement for the applicable test design rather than as a universal rule for every PCB structure. The selected lengths must remain compatible with the frequency range, coupon geometry, fixture arrangement, and intended regression analysis.
7. Control Manufacturing Variation
Stable line-width and line-spacing tolerances, consistent board thickness, stable dielectric constant (Dk) and dissipation factor (Df), and controlled copper surface roughness are all important. Delta-L is often used to evaluate PCB materials, but excessive process variation makes it difficult to determine whether the measured loss is caused by the material or by fabrication variation.
The reliability of the result therefore depends on both measurement technique and coupon manufacturing consistency. A well-designed coupon with poorly controlled fabrication may still produce an ambiguous result, while a stable process allows the measured insertion loss to be related more confidently to the transmission-line structure and material system.
Summary
Delta-L testing determines PCB transmission-line loss by comparing two otherwise equivalent structures with different line lengths. Subtracting the loss of the shorter line from that of the longer line reduces common errors from probes, pads, vias, and fixtures, allowing the loss per unit length to be estimated more accurately.
As operating frequencies increase, coupon design becomes as important as the measurement instrument itself. Impedance control, matched long and short structures, optimized probe launches, short via stubs, continuous reference planes, consistent differential pairs, suitable coupon lengths, and stable manufacturing conditions are all essential to obtaining meaningful Delta-L results.
