PCB Laminate Dielectric Constant (Dk) and Loss Tangent (Df): A Practical Guide

Every PCB engineer has opened a laminate datasheet, looked at a table of Dk and Df values, and then had a quiet moment of uncertainty. The numbers vary by frequency, by test method, by resin content, and sometimes by which page of the datasheet you’re reading. Then you put them into your stack-up calculator and wonder whether the simulation will actually match hardware. It usually doesn’t match perfectly — and understanding why is the practical engineering lesson that most textbooks skip.

PCB laminate Dk Df explained properly means going beyond the definition and into how these numbers behave in a real PCB, why the same material reports different values from different manufacturers and labs, how Dk and Df interact with your signal budget at actual operating frequencies, and what the selection rules look like for the range of real-world applications you’re designing for. That’s what this guide covers.

What Dielectric Constant (Dk) Actually Means in a PCB Context

The dielectric constant, also called relative permittivity (εᵣ), is the ratio of how much electric energy a material can store compared to air or a vacuum. Air has a Dk of 1.0 by definition. FR-4 sits around 4.2–4.8. PTFE sits around 2.1. Rogers RO4003C is specified at 3.38 at 10 GHz. These numbers matter because Dk is the primary determinant of two critical design quantities: signal propagation velocity and characteristic impedance.

How Dk Controls Signal Propagation Velocity

The velocity at which an electromagnetic signal travels through a dielectric medium is inversely proportional to the square root of Dk. In free space, signals travel at the speed of light (approximately 300 mm/ns). In a PCB dielectric with Dk of 4.0, signals propagate at approximately 150 mm/ns — half the speed of light. The formula is:

v = c / √Dk

Where c is the speed of light and Dk is the relative permittivity of the medium. This has direct consequences for propagation delay, which scales as:

TD (ps/inch) ≈ 85 × √Dk

For a standard FR-4 with Dk of 4.2, propagation delay is approximately 174 ps/inch. For Rogers RO4003C with Dk of 3.38, it drops to approximately 156 ps/inch. That 18 ps/inch difference is noise-floor-level for low-speed designs, but becomes significant timing budget in a 1-meter differential pair on a high-speed backplane, where the accumulated delay difference reaches nearly 700 ps across the full trace length.

How Dk Controls Trace Impedance

For a microstrip or stripline trace, characteristic impedance depends on the geometry of the trace and the Dk of the surrounding dielectric. A lower Dk material produces a higher impedance for the same trace geometry, which means that to hit a 50Ω target on a low-Dk material you need a narrower trace than on a high-Dk material of the same dielectric thickness. This is not a minor effect — changing from FR-4 (Dk 4.2) to Rogers RO4350B (Dk 3.66) on the same stack-up geometry shifts your microstrip impedance by 5–8 Ω, which is easily enough to push a nominally 50Ω design outside the ±10% tolerance range.

The practical implication: any material substitution mid-design or mid-production that changes Dk requires recalculation of all controlled-impedance trace widths and re-verification of impedance test coupons on the first production panels.

Dk Is Not a Single Number — It Varies With Frequency, Resin Content, and Temperature

This is the point where most introductory explanations stop being useful. The Dk value printed on a material datasheet is the result of a specific test method at a specific frequency. For most PCB laminate materials, Dk decreases as frequency increases. This is a fundamental consequence of the polarization mechanisms in the polymer matrix: at low frequencies, multiple polarization mechanisms contribute to the dielectric response; at high frequencies, molecular dipoles cannot rotate fast enough to track the rapidly alternating electric field, so fewer mechanisms contribute and Dk falls.

Resin content also shifts Dk. The glass reinforcement in an FR-4 laminate has a different Dk than the epoxy resin — glass (E-glass) sits around Dk 6.3 and resin around Dk 3.5 depending on formulation. As the resin-to-glass ratio changes with different prepreg styles and target thicknesses, the net Dk shifts toward whichever component dominates. A resin-rich 106 prepreg will have a meaningfully different Dk than a glass-rich 7628 prepreg made from the same resin system. Isola publishes per-construction Dk/Df tables for their high-speed laminates precisely because a single value is inadequate for stack-up simulation.

Temperature adds further variation. As the board heats up through operation, the polymer matrix expands and changes its polarization response, shifting Dk by a measurable amount. For RF antenna designs or filters where resonant frequency is the critical parameter, the thermal coefficient of Dk (TCDk) is a specification that must be verified — Rogers explicitly calls this out as a key advantage of their ceramic-filled hydrocarbon materials over standard FR-4 in temperature-varying environments.

What Dissipation Factor (Df) / Loss Tangent (Tan δ) Actually Means

The dissipation factor (Df), also called loss tangent or tan δ, is the ratio of the energy lost (dissipated as heat) in the dielectric to the total energy stored in it per electromagnetic cycle. Mathematically it is the tangent of the loss angle — the phase angle between the real and imaginary parts of the complex permittivity. A higher Df means a greater fraction of signal power is converted to heat rather than arriving at the far-end receiver.

If Dk determines how fast your signal travels and where it needs to be to hit impedance targets, Df determines how much of your signal is left after it gets there.

Why Df Becomes the Dominant Concern at High Data Rates

At lower signal frequencies (below 1 GHz), conductor loss — the resistive loss in the copper traces due to skin effect — typically dominates total insertion loss. Dielectric loss is a secondary contributor. As frequency increases, dielectric attenuation grows proportionally with frequency while conductor attenuation grows with the square root of frequency. Somewhere between 1 and 5 GHz depending on trace geometry and material, the crossover happens and dielectric loss takes over as the primary loss mechanism.

By the time you’re at 28 Gbps SerDes (Nyquist at 14 GHz) or 56 Gbps PAM4 (Nyquist at 28 GHz), dielectric loss is the constraint that determines how long a trace can be before the far-end eye diagram collapses. At these frequencies, the difference between a Df of 0.020 (standard FR-4) and 0.002 (Panasonic Megtron 6) is not a minor performance delta — it’s the difference between a 10-inch channel that closes with margin and one that doesn’t close at all.

The approximate relationship for dielectric attenuation in a transmission line is:

α_d (dB/inch) ≈ 2.3 × Df × √Dk × f(GHz)

This shows dielectric loss increasing linearly with frequency, which is why datasheets that only publish Df at 1 GHz are nearly useless for high-speed simulation. Using a 1 GHz Df figure in a simulation for a 28 Gbps channel will give you an overly optimistic insertion loss prediction that will not match hardware.

Df Is Also Not a Single Number

Just as Dk varies with frequency, resin content, and temperature, Df is equally non-constant — and its variation with frequency is more pronounced. Df generally increases with frequency, and it increases faster in the resin-rich regions of a composite laminate than in the glass-reinforced regions. This is why single-point Df values are insufficient for wideband simulation, and why responsible laminate manufacturers like Rogers, Isola, and Panasonic publish multi-frequency Dk/Df tables rather than single values.

Humidity is another significant variable. FR-4 substrates can absorb up to 0.15% moisture by weight, and moisture absorption shifts both Dk and Df upward. For RF antenna designs where frequency accuracy matters, this moisture sensitivity is a real operational concern. Hydrophobic materials like PTFE maintain consistent Dk and Df under varying humidity, which is part of why they dominate precision RF applications despite their fabrication challenges.

The Measurement Problem: Why Datasheets Don’t Always Compare Directly

Here is the source of enormous confusion in material selection: two datasheets showing the same nominal Dk value for different materials do not necessarily represent equivalent materials for your application. The Dk value depends critically on the test method used to measure it.

The most common test method in the laminate industry is the IPC-TM-650 2.5.5.5 Clamped Stripline Resonator, typically run at X-band frequencies (8–12.5 GHz). The clamped stripline resonator evaluates the Z-axis (through-thickness) dielectric properties of the material. The material under test is not physically bonded in the test fixture — it’s clamped. This means small air gaps exist at the interfaces, which cause the measured Dk to be slightly lower than the actual in-laminate Dk where everything is bonded together. This artifact is well understood and documented — it’s why Dk numbers from datasheets tend to be slightly lower than what you’ll back-calculate from fabricated boards.

The Split-Post Dielectric Resonator (SPDR) method places the material sample between two shorted cylindrical cavity sections. The electric field in the SPDR is oriented in the X-Y plane (in-plane), not the Z-axis. Since glass-reinforced laminates are anisotropic — the glass fibers run in the X-Y plane, and the Dk in-plane differs from Dk through-thickness — the SPDR and clamped stripline methods will report different values for the same material. This is not measurement error; both values are physically correct for their respective electric field orientations. For a stripline routing application where signal propagates in the X-Y plane, the SPDR value is more representative of what your signal actually experiences.

The Full Sheet Resonator (FSR) method tests copper-clad laminate at lower frequencies (typically around 1 GHz) and has different sensitivity to resin content than the stripline method. For a datasheet that reports Dk = 3.48 at 10 GHz (clamped stripline) and Dk = 3.66 at 10 GHz (split-post resonator), both numbers are valid — they just measure different things. When comparing materials across datasheets, always check which test method was used. Without knowing the test method, Dk and Df numbers are not directly comparable.

Table 1: Dk/Df Test Method Comparison

Test Method	Standard	E-field Orientation	Typical Frequency	Copper Required?	Key Characteristic
Clamped Stripline Resonator	IPC-TM-650 2.5.5.5C	Z-axis (through-thickness)	8–12.5 GHz	No (etched off)	High-volume QC, slightly low Dk due to air gaps
Split-Post Dielectric Resonator (SPDR)	—	X-Y plane (in-plane)	1–15 GHz	No	Better for in-plane signal propagation; anisotropy visible
Full Sheet Resonator (FSR)	IPC-TM-650 2.5.5.6	Mixed	~1 GHz	Yes (copper-clad)	Good for copper-clad QC; lower frequency
Bereskin Stripline	IPC-TM-650 2.5.5.5.1	Z-axis	Multiple frequencies	No	Used by Isola exclusively; high accuracy
Traveling Wave / Ring Resonator	Fabricated test circuit	X-Y plane	Broadband	Yes (fabricated circuit)	Most representative of actual PCB behavior

The Fiber Weave Effect: The Dk Variation You Can’t See in a Datasheet

Any engineer who has debugged timing skew on differential pairs that appeared perfectly matched in layout should know about this phenomenon. A PCB laminate is not a uniform dielectric — it is a composite of woven E-glass cloth embedded in resin. E-glass bundles have Dk around 6.3; the resin between them has Dk around 3.5. Because the glass weave has a periodic structure with distinct glass-bundle regions and resin-pocket regions, a trace routed exactly over a glass bundle experiences a different effective Dk than a trace routed through a resin pocket. This difference can be 0.2–0.5 in Dk units depending on glass style.

For a differential pair where both lines need matched propagation delay, the fiber weave effect causes skew that no amount of length matching compensates for — because the length match was calculated against a uniform assumed Dk, and the actual Dk of each trace differs based on its spatial position relative to the glass weave pattern.

The solutions are material-level: spread weave glass (where the weave pattern is mechanically disrupted to spread glass bundles more uniformly), rotation of the glass angle relative to the board edge (typically 10° to 15°), or very-low-profile copper foil combined with materials where the resin-to-glass Dk contrast is minimized. High-speed laminate datasheets from Isola and Panasonic specify spread weave glass specifically to address this effect in differential pair routing.

Dk and Df by Material Class: The Full Performance Map

Table 2: PCB Laminate Dk/Df Material Class Overview

Material Class	Typical Dk @ 10 GHz	Typical Df @ 10 GHz	Usable Frequency Range	Typical Application
Standard FR-4	4.2–4.8	0.018–0.025	DC to 1–3 GHz	Consumer electronics, industrial, general purpose
High-Tg FR-4 (Isola 370HR, Ventec VT-47)	4.0–4.4	0.016–0.021	DC to 3–5 GHz	High-reliability multilayer, automotive body
High-Performance FR-4 (Isola FR408HR)	3.6–3.8	0.0085–0.012	DC to 8–10 GHz	Server backplanes, 5–10 Gbps interfaces
Low-Loss Modified Epoxy (Megtron 6, I-Speed)	3.3–3.7	0.002–0.007	DC to 20–30 GHz	10–56 Gbps high-speed digital
Ultra-Low Loss (Megtron 7/8, Tachyon 100G)	3.0–3.6	0.0012–0.002	DC to 40 GHz	56–112 Gbps PAM4, 800GbE
Hydrocarbon Ceramic (Rogers RO4350B)	3.48–3.66	0.0031–0.0037	DC to 30+ GHz	RF/microwave, 5G, automotive radar
PTFE (Rogers RT5880, Taconic TLX)	2.1–2.6	0.0005–0.002	DC to 60+ GHz	mmWave, aerospace, precision RF
High-Dk Ceramic (Rogers RO3010)	10.2	0.0022	DC to 30 GHz	Compact antenna, power dividers

Table 3: Key Laminate Dk/Df Values at Multiple Frequencies

Material	Dk @ 1 GHz	Dk @ 10 GHz	Df @ 1 GHz	Df @ 10 GHz	Notes
Isola 370HR	~4.04	~3.93	~0.021	~0.022	Standard high-Tg FR-4
Isola FR408HR	~3.72	~3.65	~0.009	~0.010	Enhanced FR-4, high-speed
Isola I-Speed	~3.62	~3.53	~0.007	~0.009	Mid-tier low-loss
Isola Tachyon 100G	~3.02	~3.02	~0.002	~0.003	Flat Dk, ultra-low loss
Panasonic Megtron 6	~3.34–3.70	~3.3–3.5	~0.002	~0.003	Leading high-speed epoxy
Panasonic Megtron 7	~3.6	~3.5	~0.0015	~0.002	Ultra-low loss
Rogers RO4350B	~3.66	~3.66	~0.003	~0.0037	Stable Dk, hydrocarbon ceramic
Rogers RT5880	~2.20	~2.20	~0.0005	~0.0009	PTFE, very stable Dk

Note: Values are representative of datasheet specifications. Always verify at your specific operating frequency using the supplier’s construction-level tables.

How Dk and Df Drive the Signal Budget in Real Designs

Impedance Tolerance and Dk Stability

Controlled impedance traces depend on Dk staying within tolerance through production. A Dk variation of ±0.2 across a laminate panel translates to approximately ±2–3% impedance variation on a 50Ω microstrip trace — acceptable for a ±10% tolerance spec, marginal for ±5%. Premium laminates specify Dk tolerance explicitly; Rogers RO4003C specifies ±0.05 at 10 GHz, while standard FR-4 may have no Dk tolerance specification at all, relying only on resin content tolerances from IPC-4101 slash sheet compliance.

Loss Budget and Df Selection

The signal budget for a high-speed channel is defined by the total allowable insertion loss (insertion loss budget) minus the contributions from via transitions, connectors, package, and receiver input return loss. What’s left is available for trace loss. The dielectric attenuation formula shows that Df × √Dk × frequency × trace length determines the dielectric contribution to trace loss. For a 40-inch differential pair at 14 GHz Nyquist (28 Gbps NRZ), the difference between using standard FR-4 (Df 0.020) and Megtron 6 (Df 0.002) is approximately 7 dB of additional insertion loss — an enormous difference that no equalization scheme can fully recover.

The rule of thumb that emerges from channel simulations across many real programs: for data rates below 5 Gbps, standard FR-4 variants with Df around 0.018–0.022 are acceptable for normal trace lengths. Above 10 Gbps, you should be looking at materials with Df ≤ 0.010. Above 25 Gbps, Df ≤ 0.004 is the practical minimum for any backplane-scale channel. Above 56 Gbps, you need Df ≤ 0.002.

The Copper Foil Roughness Interaction

Copper surface roughness is frequently overlooked as a Dk/Df modifier. The rough surface of copper foil at the copper-dielectric interface increases the effective path length of electromagnetic signals traveling along the trace, effectively increasing conductor loss. But it also changes the apparent Dk seen by the signal — rougher copper effectively increases the electric field confinement near the conductor surface, making the substrate appear slightly more lossy. This is why premium low-loss laminates are always paired with low-profile or hyper-very-low-profile (H-VLP) copper foil. Specifying Megtron 7 with standard electrodeposited copper partially defeats the purpose of the low-Df resin system.

Practical Material Selection: Matching Dk/Df to Your Application

Ventec PCB materials and other manufacturer options cover the full Dk/Df spectrum. The practical decision process:

Step 1 — Define your operating frequency or maximum data rate. Use the Nyquist frequency of your highest data rate channel as your frequency reference for Dk and Df lookup. For 25 Gbps NRZ, that’s 12.5 GHz. For 56 Gbps PAM4, it’s 28 GHz.

Step 2 — Estimate your worst-case trace length and budget insertion loss. Longer channels on more layers eat more of the loss budget. A 30-inch backplane trace needs a fundamentally different material than a 3-inch daughter card trace at the same data rate.

Step 3 — Work backward from your required insertion loss budget to maximum acceptable Df. Using the approximate dielectric attenuation formula or a full simulation, determine the maximum Df your channel can tolerate. Then select the most cost-effective material class that meets that Df at your operating frequency.

Step 4 — Verify Dk at your operating frequency for stack-up design. Use per-construction Dk tables from the manufacturer’s datasheet, not a single headline value. Confirm trace widths with your fabricator using the actual stack-up geometry.

Step 5 — Confirm copper foil grade is consistent with material tier. Low-loss laminates require low-profile copper to capture the full Df benefit. Confirm this specification with your fabricator during stack-up approval.

Table 4: Dk/Df Material Selection Cheatsheet by Application

Application	Max Data Rate	Max Useful Freq.	Required Max Df	Recommended Material Tier
Consumer electronics, general IoT	< 1 Gbps	< 500 MHz	< 0.025	Standard FR-4
Industrial controls, moderate digital	1–5 Gbps	1–2.5 GHz	< 0.020	High-Tg FR-4 (370HR, VT-47)
Server, networking	5–10 Gbps	2.5–5 GHz	< 0.012	High-performance FR-4 (FR408HR)
Switch fabric, 25G/40G Ethernet	10–25 Gbps	5–12.5 GHz	< 0.008	Low-loss modified epoxy (I-Speed, Megtron 6)
100G/400G datacenter	25–56 Gbps	12.5–28 GHz	< 0.003	Ultra-low loss (Megtron 7, I-Tera MT40)
800GbE, AI accelerator backplane	112 Gbps PAM4	up to 56 GHz	< 0.0015	Megtron 8, Tachyon 100G
5G base station, automotive radar	1–30 GHz RF	Up to 30 GHz	< 0.004	Hydrocarbon ceramic (Rogers RO4350B)
Satellite, mmWave, aerospace RF	Up to 60 GHz	Up to 60 GHz	< 0.001	PTFE (RT5880, RO3003)

Useful Resources: Datasheet Downloads and Reference Material

Every engineer working with Dk and Df values on real programs should have direct access to primary manufacturer data sources. Third-party aggregators are useful for initial screening but always fall behind the most current construction-level tables.

Resource	URL / Source
Isola Dk/Df Per-Construction Tables (I-Speed, FR408HR, Tachyon 100G)	isola-group.com — product pages → documentation
Isola “Making Sense of Laminate Dielectric Properties” (PDF)	isola-group.com/wp-content/uploads
Rogers Corporation Dk/Df Material Data & Application Notes	rogerscorp.com/resources
Rogers “Why Do Different Test Methods Yield Different Results” (PDF)	rogerscorp.com/articles
Rogers “Overview of Test Methods for Dk and Df” (PDF)	rogerscorp.com/articles
Panasonic MEGTRON Series Datasheets	industrial.panasonic.com/ww/products/pt/megtron
Ventec Group Product Datasheets (VT-47, VT-901, tec-thermal)	ventec-group.com/products
IPC-TM-650 Test Methods Manual	ipc.org/test-methods
Z-zero Blog — Dk/Df Characterization Methods	z-zero.com/blog-post/dk-and-df-characterization-methods
Altium Blog — Role of Dielectric Loss Tangent	resources.altium.com
PCBSync Dk Guide	pcbsync.com/pcb-dielectric-constant-guide
PCBSync Df Guide	pcbsync.com/pcb-dissipation-factor-df

5 FAQs About PCB Laminate Dk and Df

Q1. Why does my stack-up simulation predict a different impedance than what I measure on the fabricated board, even when I used the datasheet Dk value?

The most common reason is that the datasheet Dk was measured by the clamped stripline resonator method (IPC-TM-650 2.5.5.5), which uses unlaminated, unclamped samples with air gaps at the interfaces. Air gaps lower the measured Dk relative to the actual laminate Dk in a properly bonded fabricated board. The effective Dk in your fabricated board is typically 2–5% higher than the clamped stripline datasheet value. For precise stack-up simulation, use the traveling wave method or ring resonator data if published, or apply a correction factor based on your fabricator’s historical impedance coupon data. The fiber weave pattern and copper foil roughness also contribute to this discrepancy.

Q2. My laminate datasheet only shows Dk and Df at 1 GHz. I’m designing for a 10 Gbps interface. Is that data useful?

It’s a starting point but should not be used directly in simulation. At 1 GHz, the Dk will be higher and the Df will be lower than at 5 GHz (the Nyquist for 10 Gbps NRZ). Using 1 GHz values in a 10 Gbps channel simulation will predict optimistically low insertion loss and falsely close eyes. For any design above 2–3 Gbps, request the multi-frequency Dk/Df table from the laminate manufacturer, or use the per-construction tables published on their website. If the manufacturer only publishes a single-frequency Dk/Df and offers no multi-frequency data, treat the material as uncharacterized for high-speed use.

Q3. Is lower Dk always better for high-speed design?

Not always. Lower Dk means faster signal propagation, which helps with propagation delay budgets and can reduce board area for a given timing requirement. But Dk and Df are not independent — for some laminate systems, reducing Dk correlates with a small Df increase. The Df value is almost always the primary constraint for high-speed digital channels above 10 Gbps. There are also cases where higher Dk is specifically beneficial: power integrity applications benefit from higher Dk between power and ground planes (more plane capacitance = lower PDN impedance), and compact RF antenna designs use high-Dk materials to shrink physical dimensions. The right Dk depends on whether the application is limited by speed, loss, size, or power integrity.

Q4. What does “fiber weave effect” mean and when should I worry about it?

The fiber weave effect is a spatial Dk variation caused by the periodic structure of woven glass reinforcement in FR-4 and similar laminates. Where a trace passes over a glass bundle, it sees higher Dk (~6.3 for E-glass) than where it passes over the resin pocket between bundles (~3.5 for epoxy). For a differential pair, if one trace of the pair is predominantly over glass and the other over resin, the propagation velocities differ and you get intra-pair skew — despite perfect length matching. At data rates above 5 Gbps on differential pairs longer than a few inches, this is a measurable margin consumer. The solution is to specify spread-weave or non-woven glass construction, rotate the board at 10–15° relative to the glass weave, use materials specifically designed to minimize glass-resin Dk contrast, or route differential pairs at 45° to the weave axes.

Q5. Can I substitute one laminate for another with similar Dk/Df values mid-production without re-qualifying?

This requires caution. Dk and Df values at a single frequency from two different manufacturers are not sufficient evidence of equivalence because the frequency dependence, temperature dependence, resin content variation curves, and Z-axis CTE may all differ. A material substitution is a serious endeavor — a lack of understanding of the actual material properties can result in a costly and embarrassing mistake. At minimum, substituting a laminate mid-production requires confirming equivalent Dk at operating frequency, equivalent Df at operating frequency, equivalent CTE for via reliability, confirmation that the lamination press profile and desmear chemistry are compatible, and impedance coupon testing on the first panel run with the new material. For high-volume production or safety-critical applications, a formal re-qualification with updated stack-up models and impedance testing is warranted.

Conclusion: Using Dk and Df as Working Engineering Inputs, Not Brochure Numbers

PCB laminate Dk Df explained for working engineers comes down to this: both numbers vary with frequency, resin content, temperature, and humidity. The datasheet value at a single frequency from a single test method gives you a starting point for material screening, not a final answer for design simulation. Use per-construction, multi-frequency tables from the manufacturer’s technical documentation. Know which test method produced the values you’re using and understand the air-gap artifact in clamped stripline data. Account for the fiber weave effect when routing high-speed differential pairs on glass-reinforced laminates. And match your material tier to your actual operating frequency — over-specifying ultra-low-loss materials on designs that don’t need them adds cost without benefit, while under-specifying them on channels that do need them produces re-spins that cost far more than the laminate premium.

The discipline of material selection starts with understanding what Dk and Df are actually telling you — and what they’re not.