PTFE Laminate PCB Fabrication: Design Rules for RF Boards
Radio frequency boards are built on a different family of materials, and the fabrication that suits them is different too. A PTFE laminate has a dielectric constant close to that of air, an extremely low loss and a tendency to move under heat that ordinary FR4 does not have. Those properties are exactly what a microwave circuit needs and exactly what makes the board awkward to build, so the design and the process have to be planned together.
Why PTFE Is Used at Radio Frequency
Signal loss in a laminate comes from two places: the conductor and the dielectric. Above a few gigahertz the dielectric contribution dominates, and it scales with the loss tangent of the material and with frequency. A polytetrafluoroethylene based laminate has a loss tangent an order of magnitude lower than standard epoxy glass, which is the difference between a link that closes and one that does not.
The dielectric constant is also stable. A material whose dielectric constant varies with temperature and with frequency makes the impedance of a transmission line wander, and a filter designed on nominal values then behaves differently in the product than it did in simulation. PTFE laminates hold their dielectric constant over frequency and temperature far better than epoxy glass, and their moisture absorption is very low, so humidity does not shift the tuning.
What the Material Family Contains
Pure PTFE is soft and expands a great deal with temperature, so practical laminates reinforce it. The common constructions are PTFE with woven glass, PTFE with a microdisperse ceramic filler, and composites that combine both. The filler raises the dielectric constant, reduces expansion and improves dimensional stability, at some cost in loss. Choosing among them is a matter of matching the electrical requirement to the mechanical one.
The softness of the base material affects every process step. It is easy to mark, it deforms under pressure, and it expands more than the copper bonded to it, which sets up stress during lamination and during soldering. A fabricator experienced with these materials schedules the press cycle, the bake and the drilling to account for that behaviour; one without that experience will produce panels that warp or that fail on registration.
How Fabrication Differs from FR4
The differences start before drilling. PTFE laminates usually need a preparation step on the copper surface to give the resist and the plating something to hold on to, because the surface is chemically inert. Drilling requires sharp tooling, conservative parameters and careful entry and exit material, because the material tends to burr and to smear rather than cut cleanly.
Lamination is the step where the material’s expansion matters most. A low flow bonding material and a press cycle with slow heating and cooling keep the layers registered and prevent the voids and the resin starvation that appear when the stack is pressed too hard. The routing rules that apply to fast nets on such a board are described in high frequency trace and data bus routing.
Impedance Tolerance and Geometry
Impedance tolerance is specified much more tightly than on a digital board. A microwave design may call for plus or minus five percent or better, because the return loss of a matched network depends on how close the manufactured geometry is to the design. That places the burden on the dielectric thickness, the copper thickness, the trace width and the etching, all of which the fabricator controls.
The calculation must use the actual laminate the fabricator will press, with its measured dielectric constant rather than a nominal datasheet value, and the structure has to be defined for both microstrip and stripline because the same geometry produces different impedance in each. The principles behind those structures are explained in microstrip and stripline routing.
Vias, Back Drilling and Grounding
Vias behave differently on a thick RF stack because the stub left by a through via resonates. At ten gigahertz a stub of a few millimetres is enough to spoil the match, which is why back drilling is common: the unused portion of the barrel is drilled away after plating to leave a short, clean connection. Where the design allows, blind vias achieve the same result without the extra step.
Grounding also requires more structure than on a digital board. The ground return has to be continuous under the transmission line, the ground planes on opposite faces of the stack should be stitched with vias that are spaced closely enough to look like a continuous wall at the operating frequency, and connectors must be grounded with vias placed right at their pins. The options for connecting layers are set out in blind and buried via stack selection.
Copper, Surface Finish and Loss
The conductor contributes loss through resistance and through surface roughness. Smooth copper is preferable at high frequency, and the copper weight is chosen for the current and for the impedance rather than for habit, with one ounce outer layers being typical and heavier copper reserved for power sections. Very heavy copper changes the etching behaviour and the achievable geometry, so the requirement should be stated for each layer.
Surface finish matters as well, because a finish with high loss or with an uneven surface adds resistance to the outer conductor. Immersion gold and immersion silver are common choices for radio frequency work, while a hot air levelled tin is avoided because the finish thickness varies along the line. The reasoning behind the organic alternative is described in OSP as a solderability preservative.
Hybrid Stackups with FR4
PTFE is expensive, so most designs use it only where they need it. A hybrid stack places the radio frequency layers on PTFE and the digital and power layers on a conventional laminate, bonding the two together in one press. That keeps the cost and the board size manageable, and it introduces a new set of process considerations: the two materials expand differently, their flow characteristics differ, and the bond between them has to survive reflow and thermal cycling.
The hybrid board also needs a transition between the two material systems. Signals crossing from a microstrip on PTFE to a stripline in FR4 pass through a region where the dielectric constant changes, and the geometry has to be adjusted so that the impedance is continuous. Where a connector or a coax launch is involved, the transition deserves its own simulation rather than an estimate.
Testing and Verification
Verification of an RF board goes beyond continuity. Impedance is measured with time domain reflectometry on a coupon, and the finished circuit is measured with a vector network analyser to confirm return loss, insertion loss and the frequency response of any filter or matching network. These measurements are usually recorded per design rather than per board, with the coupon retained as the evidence.
Reliability testing confirms that the material survives the assembly process. Thermal shock and thermal cycling expose weak plated barrels and any separation at the bond between the laminate and the copper, both of which occur earlier on a soft material than on a rigid one. For a product in a harsh environment, the test programme should be defined before the stackup is fixed rather than after the first samples arrive.
FAQ
Can a PTFE board be assembled like an FR4 board? It is assembled with similar equipment, but the higher expansion and the softness of the material require a gentler reflow profile and careful support of the board through the process.
Is back drilling always necessary? For fast signals on a thick stack it is the usual way to remove the stub. Thin stacks or blind vias can avoid the need entirely.
Why is impedance tolerance so tight on RF boards? Because return loss depends on how well the manufactured line matches the design impedance, and at microwave frequencies a small error produces a large reflection.
