Substrate Materials — Choose on Data

FR-4 and the three Tg classes that matter

FR-4 is not one material — it is a family of woven-glass/epoxy laminates that all meet the UL 94 V-0 flammability rating. Within that family, the property that decides which one you get is the glass transition temperature (Tg): the point where the resin stops behaving like a rigid solid and starts to soften. Below Tg the board is dimensionally stable; above it, the Z-axis expansion rate multiplies and plated barrels get stressed.

• Standard Tg, 130–140 °C — fine for consumer products assembled with lead-free reflow once, at moderate board thickness, with low layer counts. The cheapest option, and still the right one for a lot of work.
• Mid-Tg, 150–160 °C — the sensible default for lead-free assembly. Leaves headroom between the peak reflow temperature (typically 235–245 °C) and the resin transition, which is what protects barrels during multiple reflow passes and rework.
• High-Tg, 170–180 °C — required when the stack is thick, the aspect ratio is high, or the product sees thermal cycling: automotive, industrial drives, power modules, backplanes. Also the right call when a board will be reworked more than once.

The other axis worth knowing is decomposition temperature (Td), where the resin chemically breaks down. Modern lead-free grades sit at 300 °C and above, which gives the margin that older laminates lacked. If a datasheet quotes Td below ~290 °C, treat the material as leaded-process only.

When to leave FR-4 behind

FR-4 was designed to be cheap, flame-retardant and mechanically decent — electrical high-frequency behaviour was never the goal. Two numbers decide where it stops working:

• Dissipation factor (Df, loss tangent) — how much signal turns into heat per wavelength. Standard FR-4 sits around 0.016–0.020; purpose-built RF laminates run 0.002–0.004. Loss climbs with frequency, so the penalty grows the faster your edges are.
• Dielectric constant stability (Dk) — FR-4's Dk wanders with resin content, glass style and frequency. A controlled-impedance design needs a Dk that the fabricator can hold across a panel, which is exactly what engineered laminates are sold on.

The practical thresholds we quote from: FR-4 is comfortable to a few GHz for typical edge rates; between that and ~10 Gb/s serial traffic, a mid-loss engineered laminate usually pays for itself in margin; above that — or for RF front ends, radar and anything length-matched at millimetre wavelengths — you want a true low-loss PTFE or ceramic-hydrocarbon system, often as a hybrid stack with FR-4 doing the non-critical layers to keep cost sane.

Hybrid stacks are routine here: an RF layer set on Rogers or Taconic bonded to FR-4 power/ground wings. The lamination cycle has to respect the CTE mismatch between the two resin systems, which is a process-window question — ask us before you freeze the stackup.

The comparison table

Typical ranges for the material classes we stock. Every design deserves the actual datasheet — these numbers are for narrowing the field, not for closing a stackup.

Dk and Df are typically specified at 1 MHz for FR-4 and at 1–10 GHz for RF laminates — never compare the two at face value without checking the test frequency.

Metal core and heavy copper: two different problems

Heat leaves a board two ways: through copper (traces, planes, vias) and through the dielectric into a metal plate. They solve different failures.

Metal-core PCB (MCPCB) puts an aluminium or copper plate behind a thin dielectric. The dielectric layer is the bottleneck — its conductivity (2–12 W/m·K typical, versus ~0.3 for FR-4) matters far more than the metal behind it. MCPCB wins for LED boards, on-board chargers and anything with a single dominant hot spot, usually as 1–4 layer single-sided or double-sided constructions.

Heavy copper attacks current, not conduction. At 3 oz and above, traces carry serious current and act as their own heat spreader; the process window (etch, plating, fill) is different from standard board, which is why it prices separately. We run finished copper from 1 oz to 13 oz, with outer base copper to 18 oz for bus-bar style work.

When both problems exist at once — high current and a hot spot — the usual answer is heavy copper on FR-4 with thermal vias under the source, rather than an MCPCB, because you keep the multilayer routing. Use the copper current calculator to size the traces before choosing the stack.

Polyimide and the bend-radius discipline

Flex failures are almost never material failures — they are mechanical design failures caught late. Polyimide itself handles continuous temperatures above 200 °C and bends indefinitely; the copper inside it does not, unless the bend geometry respects three rules:

• Bend radius — 10× total thickness for dynamic flex (folds in service), 5× for static flex (bends once at assembly). Herringbone or curved traces through the bend zone, never a straight trace across the outer radius.
• Copper as a neutral axis — keep copper centred in the stack for dynamic zones, single-layer where possible, and let coverlay — not solder mask — cover bend areas. Solder mask cracks; coverlay flexes.
• Adhesiveless base for thin, high-duty flex. The adhesive layer creeps and thickens the stack; adhesiveless PI buys reliability for two-sided and multilayer builds.

Rigid-flex combines both worlds and removes connectors — usually the reliability win that justifies its cost. Our rigid-flex window runs 2–20 layer rigid sections with controlled impedance carried across the transition. Full flex parameters are on the capabilities page.

What we stock and run routinely

Stocked grades quote faster because the laminate is on the shelf and the process window is proven:

If your drawing names a laminate not listed here, send the datasheet with the RFQ. We confirm availability and the process window before quoting — the honest answer sometimes is "that grade is a 3-week mill order", and you want to know that on day one.

Material questions we actually get