Solder Mask on High-Frequency PCB: A Dielectric Layer That Happens to Be Green

Solder mask is specified as a protective coating, and it is bought as a coating: colour, gloss, thickness, cure. Electrically it is something else. Twenty-five micrometres of a material with a dielectric constant near 3.8 sits directly on top of every trace, and how much that matters depends entirely on what is underneath it — on 0.127 mm PTFE it can move a 50 Ω line by six per cent, on 0.508 mm laminate barely two. Getting this right is a two-line decision in the stack-up, or a whole lot of re-tuning after the first impedance coupon.

Solder Mask - high-frequency PCB manufactured by XCE PCB

Two Specifications Wearing One Name

Every solder mask is described by the same two lines in a purchase order: a colour and a standard. That covers the protective function — solder containment, insulation, marking, and protection from handling and contamination — and says nothing about the electrical one.

Electrically, the mask is a thin dielectric covering the whole board. Over a microstrip it changes the effective dielectric constant of the field above the trace; between traces it adds a leakage path with its own loss tangent; over a launch or a filter it detunes what was simulated in air. All three effects scale with thickness over dielectric height, which is why the same mask process is harmless on one build and significant on another.

This is the step after plating in the twelve-step sequence, and the impedance budget it consumes is part of the same calculation set out in the design step guide.

The Numbers, Since None Are Usually Quoted

PropertyTypical value for liquid photoimageable maskWhy it matters
Thickness over a trace15–25 µmSets how much of the field above the trace is in mask
Dielectric constant3.5–4.2 at 1 MHz1.7× a PTFE substrate; 0.9× an FR-4 one
Loss tangent0.02–0.03Twenty to thirty times the Df of 5880
ColourAnyPigment changes nothing electrically at RF
Final cure≈150 °C, 60 minBelow the PTFE service range; safe on all families

The dielectric constant row is where high-frequency work differs from the general case. On FR-4 at Dk 4.3 a 3.8 mask is slightly lower than the substrate, so covering a trace barely changes the field. On 5880 at Dk 2.20 the same mask is 1.7 times the substrate, so the field above the trace is suddenly in a much denser dielectric.

What 25 µm of Mask Does to Impedance

The effect is a first-order calculation, and the useful part is the trend rather than the third digit. Treating the mask as an additional capacitance over the trace, the shift in effective Dk is roughly proportional to (Dkmask − 1) × t / h. Thickness over height, not thickness alone.

Build (25 µm mask, Dk 3.8)t / h50 Ω line beforeAfter maskShift
5880, 0.127 mm0.2050.0 Ω46.7 Ω−3.3 Ω (−6.6%)
5880, 0.254 mm0.1050.0 Ω48.3 Ω−1.7 Ω (−3.5%)
5880, 0.508 mm0.0550.0 Ω49.1 Ω−0.9 Ω (−1.8%)
4350B, 0.254 mm0.1050.0 Ω48.8 Ω−1.2 Ω (−2.3%)
4350B, 0.508 mm0.0550.0 Ω49.4 Ω−0.6 Ω (−1.2%)

Three practical readings. A thin PTFE build with mask over the RF lines can spend a third of a ±10% budget on the mask alone. A modest 0.508 mm build spends almost nothing. And the direction is always the same — the mask lowers impedance — so a coupon measured before mask reads high, which is exactly the discrepancy that shows up between a fabricator’s in-process coupon and the customer’s final measurement.

The design answers are all cheap: widen the trace by the calculated amount, open the mask over the RF line (mask-defined keep-out, common on filter and antenna structures), or specify the mask thickness in the stack-up so the coupon is measured on the finished article. The expensive answer is discovering it after the first lot.

Step 1 — Preparation, and the PTFE Adhesion Problem

Cleaning is routine: remove oxide, contamination and resist residue, then dry. The step that is not routine is adhesion promotion, and on PTFE it is the whole battle.

Surface energy is the number behind it. FR-4 sits around 40 mN/m, which most liquid masks wet readily. PTFE is one of the lowest-energy surfaces in engineering, in the high teens to about 20 mN/m, so a mask that cures perfectly well on FR-4 will form a film on PTFE that looks correct, passes a visual inspection, and lifts at the edge of a large copper pour after the first reflow or the first mechanical shock.

Three approaches are used, in increasing order of reliability. A plasma treatment before coating, which raises surface energy and adds polar groups. An adhesion promoter designed for fluoropolymers, applied as a thin primer. Or keeping the mask off the fluoropolymer altogether where the geometry allows — which is often the honest answer on an RF board, where large areas of 5880 are better left bare with a defined finish than covered with a film that may peel.

Step 2 — Coating: Screen, Spray or Curtain

Liquid photoimageable mask is applied by screen printing, spray coating or curtain coating. The choice is made on geometry and volume, and it determines the thickness distribution that the impedance calculation above assumed.

MethodThickness controlBest forWeakness
Screen printingGood, defined by mesh and emulsionHigh volume, flat boardsThin over tall features; mesh marks
Spray coatingModerate, program dependentMixed heights, prototypesOverspray, edge build-up
Curtain coatingVery uniform on flat panelsFine-line, high layer countNeeds a flat panel; waste at start-up

What matters for RF is thickness over the trace, not average thickness on the panel. Mask thins over the top of a trace and pools in the gaps beside it, so a process that measures 20 µm on a coupon may be 12 µm on a 35 µm-high trace and 30 µm in the trench next to it. That is one more reason the impedance effect is measured, not assumed.

Step 3 — Tack Dry, Exposure and the Resolution Limit

Tack dry drives the solvent out so the film can be handled and exposed without sticking to the artwork. Typical conditions are 70–80 °C for a period set by the coating weight, and the control point is the residual solvent: under-dried film sticks and marks; over-dried film does not develop cleanly and leaves scum in the openings.

Exposure cross-links the film through a phototool or by direct imaging. The resolution limit is what sets the design rules: a solder dam between two pads needs a finite width to survive development and to stay put through reflow, and 0.10–0.15 mm is the practical range for production. Below that the dam is defined but fragile, and it fails as a bridged solder joint rather than as a mask defect.

Registration to the copper is the second-half of the tolerance. The mask opening is normally 0.05–0.1 mm larger than the pad per side, which protects the pad from mask encroachment at the cost of a smaller dam. On a fine-pitch device the two requirements — opening big enough, dam wide enough — can be mutually exclusive, and that is a design decision, not a process one.

Step 4 — Development and Cure

Development removes the unexposed film in a dilute alkaline solution, typically around 1% sodium carbonate, at a controlled temperature and conveyor speed. Under-development leaves a scum in the openings that blocks wetting and causes poor solder joints; over-development attacks the dam edges and lifts resist at the interface with copper.

Final cure is a bake, usually around 150 °C for 30–60 minutes, which completes the cross-linking and drives adhesion. This is the step where an under-cured mask reveals itself later: it passes a tape test when new and fails after a reflow cycle, because the cure that was missing is the one that survives thermal excursion.

On high-frequency laminate, cure temperature is not the constraint — 150 °C is inside the service range of every family discussed here, including PTFE. What is a constraint is the number of thermal excursions: each reflow adds a cycle to a mask on a low-surface-energy substrate, and a mask that is marginal on adhesion degrades with each one.

Step 5 — Loss, Not Just Impedance

A mask with a loss tangent of 0.02–0.03 is not a neutral covering at millimetre wave. Only the fringing field above the trace passes through it, so the contribution is a fraction of the total loss — but on a 5880 line whose dielectric loss is already at the floor, that fraction can dominate the difference between two otherwise identical boards.

The practical consequence is that the loss figure quoted for a laminate is measured on bare material, and the board that ships has mask over it. If a link budget was built on the datasheet number, the mask is one of the items that was not in it — along with copper roughness and finish, as set out in the plating step.

Where Solder Mask Goes Wrong

SymptomRoot cause in this stepFix
Mask lifts at the edge of a copper pour on PTFESurface energy never raised; no plasma or primerPlasma or fluoropolymer promoter; or leave the area bare
Impedance 2–3 Ω low on a thin PTFE buildMask thickness not in the model (t/h high)Widen the trace, open the mask, or model it
Coupon reads 50 Ω, customer measures 48Coupon measured before maskingMeasure after cure, on the finished article
Scum in mask openings, poor solder wettingUnder-exposure or over-dried filmExposure energy step test; control tack dry by weight
Mask bridges between fine-pitch padsDam narrower than the process can hold0.10–0.15 mm minimum dam, or mask-defined pads
Loss higher than the laminate datasheet predictsMask Df 20–30× the substrate over the RF lineOpen the mask over RF; re-run the link budget

Solder Mask Checklist

#CheckPass condition
1Mask Dk and thickness recorded3.5–4.2 and 15–25 µm, used in the model
2t/h checked for thin builds≤0.10, or the trace widened to compensate
3RF lines mask-open where loss mattersDefined in the data package, not decided in the shop
4Surface energy raised on PTFEPlasma or promoter before coating
5Coating method matched to topographyCurtain or spray for mixed heights
6Tack dry controlled by residual solventNo stick, no scum
7Dam width within process capability0.10–0.15 mm minimum
8Opening 0.05–0.1 mm per side over padNo encroachment, no exposed trace
9Final cure verified≈150 °C, 30–60 min, tape test after reflow
10Impedance coupon measured after maskFinished-article value inside ±10%

Conclusion

Solder mask is bought as a coating and behaves as a dielectric. On a high-frequency board the two facts that decide whether it is harmless are the ratio of mask Dk to substrate Dk and the ratio of mask thickness to dielectric height — which is why the same process is a non-event on 0.508 mm 4350B and a 6% impedance shift on 0.127 mm PTFE.

The claims worth discarding are that mask is electrically neutral at RF (it is not, in either Dk or Df), that a mask that adheres to FR-4 will adhere to PTFE (surface energy differs by a factor of two), and that impedance is settled at the coupon stage (a coupon measured before cure is measured before the mask is in the stack).

XCE PCB applies liquid photoimageable mask with plasma preparation on fluoropolymer, mask-open RF lines where the loss budget needs them, and impedance coupons measured after final cure: 2–72 layers, impedance ±10%, panels to 600 × 1200 mm, 5,000 m² of production space and 80+ engineers, with samples in as fast as 24 hours.

Frequently Asked Questions

Does solder mask affect impedance?
Yes. The mask sits over every trace with a dielectric constant of 3.5–4.2, so it raises the effective Dk of the field above the conductor and lowers impedance. The shift scales with mask thickness divided by dielectric height: 25 µm of mask moves a 50 Ω line to 46.7 Ω on 0.127 mm PTFE but only to 49.4 Ω on 0.508 mm 4350B.
How thick is solder mask?
15–25 µm over a trace for liquid photoimageable mask, and the value over the trace matters rather than the average on the panel: mask thins over copper features and pools between them, so a coupon that reads 20 µm may be 12 µm on a 35 µm-high trace.
Why does solder mask peel off PTFE?
Surface energy. FR-4 is around 40 mN/m and wets readily; PTFE is in the high teens to about 20 mN/m, so the mask forms a film that looks correct and lifts after reflow. Plasma treatment or a fluoropolymer adhesion promoter is needed, and on large RF areas leaving the laminate bare is often the better answer.
Should RF traces be covered with mask?
Only if the link budget includes it. Mask has a loss tangent of 0.02–0.03, twenty to thirty times that of 5880, so over an RF line it adds loss that the laminate datasheet does not describe. Many RF designs open the mask over the line and specify the finish instead.
Why does my coupon read 50 Ω but the finished board 48 Ω?
Because most in-process coupons are measured before mask. The mask lowers impedance, so the finished article reads low against a bare coupon. The fix is to measure after final cure, or to widen the trace by the calculated mask effect.
What is the minimum solder dam width?
About 0.10–0.15 mm in production. Below that the dam is defined but fragile, and it fails as a solder bridge between pads rather than as a visible mask defect — which makes it a design rule rather than a process tolerance.
RF & Microwave PCB Fabrication

Need Mask That Does Not Move Impedance?

Send us your Gerber files and requirements for a manufacturability review and a tailored quotation.

2–72 layers ±10% impedance tolerance Manufacturing since 2009