Quasi-Optical Transmission Loss Extraction for High-Frequency Stackup Design

Foil

At millimetre-wave frequencies above 50 GHz, electromagnetic propagation collapses into a shallow outer skin layer of the conductor. The skin depth of copper contracts to 0.29 µm at 50 GHz, 0.19 µm at 110 GHz, and 0.13 µm at 220 GHz. At these physical dimensions, the micro-profile roughness of the foil interface dominates total channel loss.

Skin depth shrinks. Standard circuit extraction techniques, including multiline Thru-Reflect-Line test patterns and resonant ring structures, extract total insertion loss across the combined conductor and dielectric structure. When numerical de-embedding models attempt to isolate dielectric dissipation factor from these composite measurements, copper surface micro-roughness artifacts distort the result.

Conductor loss gets misattributed to dielectric loss, creating artificial inflation in the extracted loss tangent of the substrate material.

Copper foil manufacturing employs varied treatment profiles to anchor metal to resin matrices. Electrodeposited foil features treatment ridges up to 3.5 µm high, whereas hyper-very-low-profile electrodeposited foils and rolled-annealed foils hold profile roughness figures below 0.5 µm. High-frequency signal energy propagating along a rough interface travels a longer effective path length along the local topography while generating localized eddy currents in the microscopic copper peaks.

Thinner foil lowers loss. Standard analytical surface models, such as single-variable Hammerstad or hemispherical Huray models, underpredict attenuation above 60 GHz because spatial frequency distributions of rough copper deviate from ideal spherical geometries. Quasi-optical extraction bypasses this physical interface entirely by measuring unclad laminate core slabs before foil bonding occurs.

Roughness degrades phase accuracy. Extracting intrinsic material loss tangents from clad laminates introduces systematic error when numerical models fit measured S-parameters using uncalibrated roughness variables. If a mathematical solver assumes a smooth copper surface during parameter extraction, the model adjusts the dielectric loss tangent upward to match observed total insertion loss.

The true substrate loss tangent remains hidden beneath surface profile effects. Isolating bulk material attenuation from interface topography requires non-contact, unclad dielectric characterization methodologies.

At 140 GHz, standard electrodeposited copper foil with an Rz roughness of 3.2 µm doubles conductor attenuation compared to smooth rolled foil.

Designing mmWave stackups using total loss figures extracted from rough copper coupon tests forces buyers to over-specify low-loss laminate laminates, inflating panel costs without recovering channel margin.

Resonator

Quasi-optical extraction isolates intrinsic substrate properties by propagating focused Gaussian electromagnetic beams through unclad material samples in free space. The test apparatus uses horn antennas coupled with PTFE lenses or off-axis parabolic mirrors to shape vector network analyzer signals into tightly bounded beams. Eliminating direct contact, coax-to-microstrip transitions, and metallic ground planes removes conductor attenuation and launch parasitic inductances from the measurement path.

Focused beams probe raw resin. Open-cavity Fabry-Pérot resonators elevate extraction precision by placing dielectric slabs between highly reflective spherical mirrors, creating sharp standing-wave resonances that yield elevated quality factors exceeding five thousand.

Substrate Alignment and Beam Waist Calibration

Measurement accuracy depends on positioning the dielectric slab precisely at the beam waist where wavefront curvature flattens to a plane wave. Diffraction occurs at sample edges when panel size falls below three times the spot size diameter, leaking electromagnetic energy into surrounding space. Substrate alignment determines phase accuracy.

Reflection limits dynamic range. Calibrating the quasi-optical bench involves free-space Thru-Reflect-Line protocols paired with time-domain gating to truncate internal multipath reflections within the fixture frame.

1. Mount the raw substrate slab at the beam waist between focal horn antennas to minimize diffraction edge scattering.
2. Run vector network analyzer calibration using free-space reflection-thru-line sequences to establish phase reference planes.
3. Measure complex transmission coefficient amplitude and phase delay across the target frequency sweep.
4. Execute numerical inversion algorithms to compute complex permittivity and loss tangent independent of metallic boundary conditions.

De-Embedding Free-Space Reflection Coefficients

Extracting relative permittivity and loss tangent from quasi-optical scattering parameters involves solving Fabry-Pérot cavity transfer equations. Transmitted amplitude peaks correspond to half-wavelength constructive interference inside the dielectric slab thickness. Numerical root-finding routines calculate complex dielectric constant by matching phase shift and attenuation peaks across wide band sweeps.

Substrate thickness uniformity across the illuminated area governs phase extraction accuracy, where a thickness deviation of 5 µm causes measurable frequency shifts in dielectric resonance peaks above 100 GHz.

Quasi-optical extraction accuracy improves when substrate thickness equals an integer multiple of half the dielectric wavelength.

Laminate vendors frequently present quasi-optical Df data collected from bare resin matrices to claim ultralow dissipation factors, omitting the loss penalties added when foil treatment bonding layers are pressed into the substrate.

Anisotropy

Woven glass reinforced laminates exhibit dielectric anisotropy due to mechanical structure differences between continuous glass fibers and surrounding resin matrices. E-glass features a relative permittivity near 6.6, while hydrocarbon or fluoropolymer resin matrices exhibit permittivity values between 2.1 and 3.0. In free-space quasi-optical configurations, the transverse electromagnetic wave travels perpendicular or oblique to the laminate surface, aligning the electric field vector along the in-plane x-y axis of the panel.

Microstrip and stripline traces establish primary electric field lines perpendicular to the panel plane along the z-axis. Dielectric anisotropy alters velocity. Glass orientation alters capacitance.

Is Free Space Dielectric Loss Directly Applicable to Stripline?

Translating free-space in-plane dielectric properties directly into stripline modeling tools introduces impedance calculation errors. The in-plane permittivity of a 1078 glass-reinforced core measured via quasi-optical bench sits higher than the z-axis permittivity experienced by a stripline trace. Woven glass bundle density along the x-axis and y-axis creates orthogonal anisotropy ratios ranging from 1.05 to 1.18 depending on resin content percentages.

High resin fraction cores exhibit reduced directional variance because isotropic resin dominates the overall volumetric field distribution.

Converting quasi-optical measurement data to controlled impedance stackups demands an explicit tensor translation. Electromagnetic field solvers convert planar transverse properties into z-axis equivalent permittivity by applying Maxwell-Garnett effective medium approximations. Ignoring spatial anisotropy causes target 50-ohm stripline geometries to build between 44 ohms and 47 ohms on finished panels, altering reflection losses and phase timing in phased-array radar feeds.

IPC-4101 slash sheet declarations report single-axis permittivity figures, leaving spatial anisotropy variations unquantified for multi-layer stackup calculations.

Engineers continue to debate whether localized weave field distortions in sub-THz channels can be modeled through averaged anisotropy tensors or if full three-dimensional mesh models are required for every glass bundle intersection.

Stackup

Integrating quasi-optically derived dielectric constants and measured copper profiles into manufacturing stackup documentation fixes impedance profiles, conductor losses, and fabrication yield limits. The draughtsman converts extracted material tensors into explicit trace-and-space dimensions while adjusting target core thicknesses for resin displacement during lamination cycles. Phase velocity alters timing.

Etch factor alters impedance.

Worked Construction of a 77 GHz Radar Core Stackup

Consider a 77 GHz Automotive Radar PCB specification requiring a 50-ohm microstrip antenna feed line. Theoretical calculations using uncorrected datasheet dielectric figures (nominal relative permittivity of 3.00) specify a trace width of 185 µm on a 100 µm unclad core. Applying quasi-optical extraction and anisotropy tensor corrections reveals the actual z-axis relative permittivity sits at 2.84 due to high resin volume fraction, while copper surface etching creates a trapezoidal sidewall angle of 70 degrees with a base reduction allowance of 12 µm.

Recalculating with anisotropic parameters yields an optimized trace width requirement of 172 µm to maintain 50-ohm characteristic impedance within a +/- 2-ohm tolerance band.

Fabrication Note Engineering Requirements

Fabrication drawings control process tolerances by specifying raw material slash sheets, copper foil profile categories, and post-lamination core heights on the stackup drawing sheet.

• Permittivity directionality specification ~ Fabrication notes state whether impedance targets draw from z-axis or in-plane dielectric constant extractions.
• Etch factor compensation factors ~ Phototool compensation allowance matches target conductor sidewall slope profiles for fine line copper traces.
• Pressed prepreg thickness window ~ Resin flow parameters determine target dielectric height after multilayer pressing cycles.
• Foil roughness profile limits ~ Fabrication drawings constrain maximum allowed foil surface micro-roughness on high-speed inner layers.

Fine-line mmWave traces experience significant impedance shifts when chemical etching alters conductor sidewalls from ideal vertical geometries into trapezoidal cross-sections. Top-to-bottom trace width differentials on 1/2 oz copper layers frequently reach 15 µm in production shops. Incorporating sidewall slope allowances directly into phototool compensation steps maintains target impedance without triggering manufacturing drawing queries.

Etch factor variations alter transmission line characteristic impedance more severely on thin core mmWave layers than on standard digital stackups.

Incorporating IPC-6012 Class 3 tight-tolerance impedance notes directly onto the master assembly drawing shifts manufacturing risk to the fabricator, triggering mandatory TDR coupon testing on every production panel.

Yield

Selecting exotic laminates with ultra-low dissipation factors impacts panel production mechanics, lamination cycle parameters, and landed PCB unit cost. Hydrocarbon and fluoropolymer cores require high lamination press temperatures up to 220 °C, increasing thermal stress and core movement across large production panels. Yield losses drive total panel cost.

Tight tolerances raise cost. Array layout fixes yield.

Panel utilization governs landed unit cost. Standard production panels measuring 18 x 24 inches lose outer margin areas due to tooling pin placement, registration targets, and plating thief rails. Highly specialized mmWave substrates operating at 140 GHz demand thin cores down to 50 µm thickness, which buckle under standard mechanical handling equipment unless panel array layouts incorporate rigid perimeter support frame borders.

Scoring ultra-thin high-frequency core materials creates micro-cracks along dielectric boundaries; tab-routing with clean laser trimming prevents mechanical edge delamination.

Surface finish selection modifies effective surface conductivity at sub-THz frequencies. Electroless Nickel Immersion Gold (ENIG) deposits a nickel barrier layer beneath gold plating. Nickel possesses poor magnetic permeability and high electrical resistivity, increasing microstrip loss by up to 1.2 dB/cm at 77 GHz compared to bare copper.

Specifying Immersion Silver or Organic Solderability Preservatives (OSP) eliminates the resistive nickel layer, preserving signal integrity across high-frequency transmission channels.

Panel utilization determines bare board unit cost more directly than raw laminate material pricing.