“Rogers boards are hard to process” is a sentence that hides three different problems. A sheet of RO4350B runs down the same line as FR-4 at the same press temperature; a sheet of RO3003 has no glass, no melt point below 327 °C and no adhesion unless the surface is chemically altered; a sheet of RT/duroid 5880 moves 82 µm in the Z direction across one reflow while the copper around it moves 6 µm. Same brand name, three different factories inside one factory. This article takes the difficulty apart family by family, states the physical cause of each one, puts a number on it, and says who owns it — because about half of the “difficulty” is designed in before the first panel is cut.
One Brand, Three Process Families
The phrase “Rogers high-frequency board” covers laminates that share a brand and a purpose but not a process. They differ in resin chemistry, in glass content, in melt behaviour and therefore in what the factory has to do to them. Any list of difficulties that does not start here ends up describing the hardest family and applying it to all of them, which is why quotations for the same outline can differ by a factor of three.
| Family | Chemistry | Press window | Hole wall | Where it breaks |
|---|---|---|---|---|
| RO4000 series (4003C, 4350B, 4835) | Hydrocarbon / ceramic thermoset, glass reinforced | 190–200 °C with 4450B bondply | Standard desmear; conventional electroless copper | Dk tolerance, etch tolerance, black pad on ENIG |
| RO3000 series (3003, 3006, 3010) | Ceramic-filled PTFE, no glass | 260–300 °C fusion, or 180–190 °C low-flow prepreg | Sodium naphthalene or plasma; FEP bonding film | Adhesion, cold flow, dimensional movement |
| RT/duroid 5880 and 5880LZ | PTFE with micro-glass, 2.20 Dk | 260–300 °C fusion; no reflow-compatible bondply in the same cycle | Sodium naphthalene mandatory; plasma alone insufficient | Z-CTE 237, drilling smear, layer-to-layer registration |
| 6010.2LM / 6035HTC | Ceramic-filled PTFE, high Dk or high k | PTFE route, but filled heavily — abrasive | As PTFE, plus filler pull-out | Tool wear, Dk 10.2 tolerance ±0.25 |
The practical consequence: a factory that is good at 4350B is not automatically good at 5880, and a factory that is good at 5880 will still lose money on 4350B if it runs it through the PTFE route. When a supplier says a stack is difficult, the first useful question is which family it belongs to — the rest of the answer follows from that.
The Difficulty Ledger: Ten Problems, Their Numbers, Their Owner
Below is the whole list in one place. The “owner” column is the part that matters commercially: a difficulty owned by the design is free to remove before tooling and expensive to remove afterwards, while a difficulty owned by the process is a price, not a defect.
| Difficulty | Physical cause | Number that defines it | Owner |
|---|---|---|---|
| Press incompatibility | Thermoset cure 190 °C vs PTFE fusion 260–300 °C | Cannot be co-cured; two cycles minimum | Design (stack) |
| Cold flow | PTFE deforms plastically under press load | 4 % thinning turns 50.0 into 48.8 Ω, irreversibly | Process (pressure) + design (thickness) |
| Z-axis expansion | PTFE Z-CTE 217–237 ppm/°C vs copper 17 | 82 µm over 215 K on 1.6 mm, vs 6 µm for copper | Process (plating ductility) |
| Hole wall adhesion | PTFE is non-polar and non-porous | Peel 5.5 N/mm only after sodium naphthalene or plasma | Process |
| Drill smear | Low modulus, low melting point, high filler abrasion | 80–120 krpm, burr ≤ 25 µm | Process |
| Copper roughness vs loss | Adhesion wants roughness; skin depth is 0.65 µm at 10 GHz | HVLP foil adds ~0.1 dB/in at 10 GHz | Design (foil choice) |
| Dk tolerance | Process value vs design value differ | 4003C 3.38 process / 3.55 design; ±0.05 Dk = ±0.74 % Z | Material |
| Layer registration | Thermal expansion + prepreg shrink | 600 mm panel moves 0.12–0.31 mm over 20 K | Process + design (panel utilisation) |
| Moisture | Not the PTFE — the bondply, the FR-4, the FEP film | PTFE absorbs 0.02 %; FR-4 absorbs 10× more | Storage |
| Late detection | Electrical failure only appears at final test | Scrap at final test costs 10–20× scrap at inner layer | Process (test plan) |
Difficulty One: The Press Window, and Why Hybrid Stacks Cost Twice
Lamination is where the material family first asserts itself. FR-4 cures between 170 and 190 °C. RO4000 with 4450B bondply cures at 190–200 °C, close enough that a plant can run both on one press with a different recipe. PTFE does not cure at all — it fuses. Fusion happens at 260–300 °C with FEP film, or at 180–190 °C with a low-flow prepreg if the stack allows it.
Those two windows cannot be opened at the same time. A stack that contains both an FR-4 core and a PTFE core is therefore built in two cycles: the PTFE sub-assembly first at high temperature, then the FR-4 layers bonded to it at low temperature. That is a real cost, and it shows up in three places — press time, registration error, and yield. Each additional cycle adds roughly ±25 µm of positional error, which on a 100 µm microvia landing on a 250 µm pad leaves only 75 µm of annular ring before any other error is counted.
The second half of the press difficulty is that PTFE does not resist load. Under the 150–250 psi that a normal cycle applies, a PTFE core thins by about 4 %, and the thinning is plastic: it does not come back when the load is removed. Four per cent of a 0.254 mm dielectric is 10 µm, and on a 0.56 mm microstrip that is enough to move 50.0 Ω to 48.8 Ω — outside a ±2 % window, permanently. The countermeasures are a lower pressure, a thicker starting laminate and a stack-up that does not rely on the pressed thickness being nominal. The mechanics of that are set out in lamination in high-frequency PCB manufacturing.
Difficulty Two: The Hole Wall Is Where PTFE Punishes You
A plated through hole is a copper cylinder bonded to a material that expands fourteen times faster than copper. That single sentence contains most of the reliability problem in PTFE multilayer work.
| Material | Z-CTE (ppm/°C) | Expansion over 215 K, 1.6 mm | Relative to copper (17) |
|---|---|---|---|
| RT/duroid 5880 | 237 | 82 µm | 14× |
| Taconic TLY-5 | 217 | 75 µm | 12.8× |
| RO4003C | 46 | 16 µm | 2.7× |
| RO4350B | 32 | 11 µm | 1.9× |
| RO3010 | 16 | 6 µm | below copper |
Three rules follow, and they are not optional on 5880. First, plating ductility: the copper in the barrel has to stretch 82 µm without cracking, which means an elongation of at least 20 % and no brittle chemistry in the bath. Second, aspect ratio: keep the finished hole depth-to-diameter at 6:1 or better, because the deeper the barrel the worse the plating uniformity and the higher the strain at the top and bottom corners. Third, adhesion: PTFE is non-polar and non-porous, so a permanganate desmear that works on epoxy does nothing to it. Sodium naphthalene is the standard route, plasma is the alternative, and the two are not interchangeable in throughput or in the uniformity of the attack. The effect of getting that wrong is a hole that passes backlight and pulls out at thermal cycling, which is exactly the failure documented in multilayer Rogers 5880 processing.
One more number belongs here because it is misquoted constantly: PTFE has no glass transition temperature in any useful sense. Its Dk is stable from −100 °C to +150 °C not because it is below Tg but because the dipole mechanism does not change until the material melts at about 327 °C. Tg, as a specification, is an epoxy concept and says nothing about a PTFE board — a point worth knowing before a customer rejects a certificate for the wrong reason.
Difficulty Three: Cutting and Drilling a Soft Abrasive
Machining is the second place where the families separate, and the reason is a pair of properties that rarely appear together: PTFE is soft and it is extremely abrasive when filled with ceramic. A cutter that survives one does not necessarily survive the other.
| Operation | Window | Failure if missed |
|---|---|---|
| Drilling PTFE | 80–120 krpm, point angle 130–140°, helix 35–45°, stack ≤ 3 panels | Smear, burr, nail-head, barrel crack |
| Router / cutter | Diamond-coated for ceramic-filled grades; sharp single-flute for pure PTFE | Frayed edge, delamination, filler pull-out |
| Laser cutting | Not permitted on PTFE | Above ~400 °C PTFE decomposes and releases HF and carbonyl fluoride |
| Water jet | Discouraged | Water and garnet driven into a soft, open edge |
| Panel tolerance | 600 mm panel, 20 K swing: 0.31 mm (TLY-5), 0.37 mm (5880) | Cut dimensions taken at the wrong temperature |
The laser prohibition is worth stating plainly because it appears in sales material as a capability. Laser drilling works on epoxy because the resin absorbs the beam and vaporises; on PTFE the decomposition products are hydrofluoric acid and carbonyl fluoride, the edge carbonises, and the carbonised layer does not bond. A shop that offers “laser drilling on Rogers” is offering it on the FR-4 part of a hybrid, or it is about to learn this. The drilling parameter set for 5880 is set out in 5880 drilling parameter optimisation.
Difficulty Four: Adhesion and Loss Want Opposite Surfaces
The fourth difficulty is a genuine conflict rather than a skill gap. Copper sticks to a rough surface; high-frequency current wants a smooth one. At 10 GHz the skin depth in copper is 0.65 µm, so a foil with 2 µm of profile forces the current to follow a path measurably longer than the trace — about +0.1 dB/in on a 0.56 mm line, which is a third of the dielectric loss on a 5880 board.
On FR-4 the conflict is settled in favour of adhesion without much argument, because the dielectric loss is so large that copper roughness is a rounding error. On a 2.20 Dk, 0.0009 Df laminate the order reverses: conductor loss becomes the dominant term, and the factory has to achieve 5.5 N/mm of peel on a smooth foil. The route is chemical, not mechanical: sodium naphthalene or plasma to activate the PTFE surface, plus a FEP bonding film or a low-flow prepreg that melts into the micro-texture it creates. Mechanical polishing before solder mask is not available either — it destroys the activation and the ink will not hold. The five routes and their trade-offs are compared in surface treatment for copper-to-PTFE adhesion.
Difficulty Five: The Impedance Promise, and Why ±2 % Is Not a Tolerance
Impedance control is the difficulty most often quoted and least often quantified. A tolerance is a statement about a distribution of measured values; on a high-frequency board four independent terms feed that distribution, and only one of them is under the operator’s control.
| Term | Variation | Effect on 50 Ω | Who controls it |
|---|---|---|---|
| Dielectric thickness | ±10 % (standard laminate) | ±6 % — the dominant term | Material + press |
| Etch (line width) | ±25 µm | ±2.8 % on 0.254 mm, ±5.5 % on 0.127 mm | Process |
| Dk | ±0.05 | ±0.6–0.7 % | Material batch |
| Solder mask | −1 to −3 Ω over bare copper | −2 to −6 % | Design (mask opening) |
| Copper thickness | ±10 % | < 1 % on microstrip | Process |
Combining the first two randomly gives a root-sum-square of about ±6.6 %, and that is before any of the systematic errors — a stack that forgets the bondply, for instance, where 0.10 mm of extra dielectric with Dk 3.5 moves a 0.56 mm line from 50.0 to 61.1 Ω, a 22 % error that no amount of process control can fix. So a ±2 % claim is not a manufacturing tolerance; it is a statement that the supplier will select material, measure every layer with a coupon, and scrap what falls outside. That can be bought, at a price, on a specific stack. A blanket ±2 % on a catalogue page is a marketing number.
The realistic figure for a standard high-frequency build is ±10 % against a coupon measured per layer, with ±5 % available on selected laminate and a slower press. The tolerance budget is worked through in impedance control in high-frequency PCB manufacturing.
Difficulty Six: Moisture — and Why the Blame Is Usually Misplaced
Almost every list of high-frequency difficulties includes “PTFE absorbs moisture.” It does not: PTFE takes up about 0.02 % by weight, the lowest of any laminate in common use, which is one of the reasons it is specified for humid environments. The moisture problem is real, but it lives in the other materials in the stack.
| Material in the stack | Water absorption | What it does on the press |
|---|---|---|
| PTFE core (5880, TLY-5) | 0.02 % | Nothing measurable |
| RO4000 core (4003C, 4350B) | 0.05–0.06 % | Small Dk drift, blister risk if stored badly |
| Bondply / prepreg | Typically higher than the core | Steam expands at 200 °C — voids and delamination |
| FR-4 hybrid layers | 0.1–0.2 % | The usual source of blistering in a hybrid |
The control that follows is storage, not material selection: 23±2 °C and 40±5 % RH, first-in-first-out, vacuum sealing after opening with the date marked, and a bake before lamination whose profile depends on the family — not one bake for everything. Baking a PTFE core to dry it accomplishes nothing; baking the FR-4 and the prepreg in the same stack accomplishes a great deal.
Difficulty Seven: Two Families on One Board
Hybrid construction concentrates every difficulty above into a single part number. The electrical reason for doing it is sound — put the low-loss material only on the layers that carry the field, and use FR-4 for the rest. The process reason is that the two halves want different press cycles, different hole-wall chemistry and, in the worst case, different drilling parameters.
Three rules keep a hybrid buildable. Keep the RF material on the outside where the field is, so that the FR-4 sits symmetrically in the middle and the board does not bow. Use one bonding system throughout; a stack that needs 4450B for one interface and FEP for another is a stack that needs three press cycles. And define the drill file by span rather than by layer number, because the span is determined by the press order, not by the drawing’s layer count. The data package that carries these decisions is described in mixed-pressure PCB design files.
Difficulty Eight: Finding Out Late
The last difficulty is not physical, it is economic, and on high-frequency work it is the largest. A defect that is caught at inner-layer AOI costs one core. The same defect caught at final electrical test costs every operation performed in between, on a panel whose material may be worth twenty times an FR-4 panel.
| Detection point | Value already spent | What it can still tell you |
|---|---|---|
| Incoming material | 1× (material only) | Dk, Df, thickness, copper peel — on a coupon, not on your board |
| Inner layer AOI | 2–3× | Line width, spacing, opens — not impedance |
| After lamination / X-ray | 5–8× | Registration and voids — not dielectric thickness |
| Coupon TDR / VNA | 10× | Impedance and loss on a coupon, at a defined frequency |
| Final electrical + microsection | 15–20× | Everything, too late to do anything but scrap |
This is why the test plan is part of the difficulty, not an afterthought. A coupon designed with the stack, measured per layer at a stated frequency, moves the discovery of an impedance error from the end of the line to after lamination — a difference of an order of magnitude in the cost of the same mistake. What each test stage can and cannot prove is set out in high-frequency PCB testing.
What Each Difficulty Adds to a Quote
Buyers usually see difficulties as a single uplift. They are not: each one has a different cost mechanism, and separating them is what makes a quote comparable.
| Difficulty | Cost mechanism | Typical effect | Removable by |
|---|---|---|---|
| Second press cycle | Press time + registration risk | 15–30 % of unit cost | One bonding system, RF only where the field is |
| PTFE hole wall | Sodium naphthalene line, longer cycle, higher scrap | 10–20 % | Nothing — intrinsic to the material |
| Tight impedance window | Selected laminate, slower press, more coupon points | 5–15 % | Accept ±10 % where the system allows |
| Small batch | Fixed engineering over few boards | Dominant below ~20 boards | Consolidate part numbers |
| Exotic thickness | Non-stock laminate, minimum quantity | Material premium + lead time | Adjust W/h instead of h |
The ranking is consistently the same: the stack is more expensive than the process, and the process is more expensive than the material. A buyer who attacks the laminate price on a two-cycle PTFE stack is negotiating over the smallest term in the equation.
Six Decisions That Remove Difficulty Before Tooling
Half of the list above can be designed out. All six of these are free at schematic review and expensive after the first panel.
1. Confine PTFE to the layers that carry the field. A four-layer hybrid with 4350B outer layers and an FR-4 core avoids the second press cycle and cuts material cost by roughly two thirds for the same outline. 2. Choose the thickness from the stocked set. Adjust W/h to reach 50 Ω rather than ordering a non-stock dielectric. 3. State the impedance window with the coupon geometry and the test frequency. A number without a measurement method is not a specification. 4. Keep the aspect ratio at or below 6:1 on PTFE. Above that, plating uniformity, not skill, becomes the limit. 5. Specify the foil by profile, not by weight. 1 oz HVLP behaves like 1 oz standard electrically and very differently at 10 GHz. 6. Pick the finish before the assembly is fixed. ENIG puts 3–6 µm of nickel over the copper, and at 28 GHz that nickel is where the current flows; a nickel-free finish is a design decision, not an upgrade.
None of these reduces performance. Each of them removes a term from the difficulty ledger, and the ones that remain — the press window, the hole wall, the Z-CTE — are the ones a buyer should expect to pay for, because they are properties of the material rather than of the drawing.
Conclusion
The difficulties in manufacturing Rogers high-frequency boards are specific, and they are specific to a family. RO4000 runs on a thermoset line and fails on tolerance, not on chemistry. RO3000 and 5880 run on a fusion line and fail on adhesion, cold flow and Z expansion. A hybrid fails on the seam between the two. Once the family is identified, each difficulty has a number attached: 4 % cold flow for 1.2 Ω, 82 µm of Z movement against 6 µm for copper, ±10 % thickness for ±6 % impedance, 25 µm of etch for 2.8 % on a 0.254 mm line.
The useful conclusion for a buyer is that difficulty and price are not the same axis. Part of the difficulty is intrinsic and should be paid for without argument; part of it is a consequence of the stack, the tolerance and the batch size, and that part is negotiable before tooling and not afterwards. The useful conclusion for a designer is that the cheapest high-frequency board is the one that never asks the factory to do two incompatible things in one press cycle.
XCE PCB builds across all three families — hydrocarbon and ceramic thermosets, ceramic-filled PTFE and micro-glass PTFE — on 2 to 72 layers, panels to 600×1200 mm, impedance ±10 % with coupon verification, samples in as fast as 24 hours, inspected to IPC-A-600G and IPC-6015. Send the stack that you were told was difficult and we will tell you which of these ten items applies to it. The process map for the whole route is in high-frequency PCB manufacturing process.
Frequently Asked Questions
Which Rogers materials are actually difficult to manufacture?
Why is ±2 % impedance control not realistic as a standard tolerance?
Does PTFE absorb moisture?
Why can a PTFE multilayer not be laminated in one press cycle?
Can PTFE be laser drilled or laser cut?
How much Z-axis expansion does a 1.6 mm 5880 board really see?
Send Us the Stack You Were Told Was Difficult
Send us your Gerber files and requirements for a manufacturability review and a tailored quotation.
sales1@xcepcb.com
+86 180 3341 4657