Tg 150 degrees Celsius is enough for most four to eight layer boards. Go to Tg 170 when the finished board is thicker than 2.0 mm, carries more than ten layers, or takes a press-fit connector. Tg 130 is for thin, simple boards that see two reflows and no rework. That is our rule of thumb, and it is about thickness and layer count rather than temperature, for a reason that is below.
Tg is not a service temperature. It is the point where the resin stops behaving like glass and starts behaving like rubber, and the property that changes at that point is expansion in the z direction. Below Tg a typical FR-4 core expands about 16 ppm per degree in z. Above Tg the same core expands about 55 ppm per degree, more than three times faster, while the copper in the barrel is still at 17 ppm. The barrel is bonded to a resin that has just started moving three times faster than the copper it is plated onto.
What Tg actually controls
Run the arithmetic on a board that goes through a lead-free profile with a 245 degrees Celsius peak, starting from 25 degrees.
A 1.6 mm board expands about 12.8 micrometres through the profile on Tg 130 laminate, 11.6 on Tg 150 and 10.3 on Tg 170. A 3.2 mm board expands 25.6, 23.1 and 20.6 micrometres on the same three grades.
The gap between Tg 130 and Tg 170 on the 3.2 mm board is 5.0 micrometres, and our published barrel copper is at least 20 micrometres. The grade therefore moves the strain on the plating by a quarter of the wall thickness in a single pass. Multiply by the number of excursions: a double-sided assembly with one rework is four passes above 217 degrees.
That is the whole case for a higher Tg on a thick board. Not that Tg 130 laminate cannot survive 245 degrees, since thousands of boards do exactly that. The expansion budget left for the barrel is smaller, and a barrel that plates thin halfway down a 20:1 hole has nothing left to give. The same arithmetic explains why thickness matters more than the air temperature in the enclosure.
The number that matters more than Tg
Two laminates both labelled Tg 170 can differ by a factor of five in how long they survive at 288 degrees. The tests that separate them are T260 and T288, the minutes to delamination at 260 and 288 degrees, and Td, the temperature at which the laminate loses 5 percent of its mass. Ask for all three on the grade you are quoted, and ask for z-axis expansion below and above Tg, because that expansion is what the barrel actually feels.
For operating temperature, Tg is the wrong number as well. The rating that matters is the relative thermal index on the UL card for that laminate, or the maximum operating temperature the end product is qualified to, and a Tg 150 board can carry a 105 degree rating for years. Confusing the two produces both directions of error: a Tg 170 stack specified for a 60 degree enclosure that never needed it, and a Tg 130 stack specified for a 125 degree environment that will soften the resin over time.
Every excursion above 217 degrees spends part of the same budget, and a board that is assembled, reworked twice and reflowed again for a second side sees more passes than the drawing assumes. The count is in how many reflow cycles a board can take, and it belongs in the laminate decision.
Three ways the callout goes wrong
Tg read as a maximum temperature. A drawing that says Tg 170 minimum is a material callout, not a thermal rating. If the enclosure runs hot, the drawing needs the operating temperature and the relative thermal index; Tg follows from layer count and thickness.
A high Tg core bonded with low Tg prepreg. The stack expands as one body, so the weakest layer sets the z expansion and the interface is where delamination begins. A Tg 170 core pressed with Tg 140 prepreg behaves in a thermal cycle like a Tg 140 stack with an expensive core inside it. Order the whole stack from one laminate family, and where two have to be mixed for electrical reasons, ask the supplier for the expansion of the mixed stack.
Ignoring what the higher grade costs in registration. A Tg 170 laminate presses at a higher temperature and flows less, so layer-to-layer movement is harder to hold on a 20 layer stack. Our published registration is plus or minus 0.025 to 0.06 mm depending on layer count and stack, and the top of that band is where thick high-Tg work lives. If inner layers carry 0.10 mm line and space, the laminate grade and the registration budget have to be decided together, not one after the other.
A fourth item is worth a line. Higher Tg helps against conductive anodic filament, so a Tg 170 stack with tight spacing and no CAF data is not safer than a Tg 150 stack that was tested. The geometry side is in what causes conductive anodic filament.
What SUPERB PCB holds on laminate
Published: A-grade FR-4 from Shengyi, Jiantao and Jin'an Guoji with Tg 130, 150 and 170 degrees Celsius; CTI at least 175 V; dielectric breakdown at least 4,000 V; UL-94 V0; standard thickness 0.6 to 3.2 mm with 0.1 to 12 mm on the extended line; thickness tolerance plus or minus 0.10 mm below 1.0 mm and plus or minus 10 percent above; up to 68 layers on the high-layer line; layer-to-layer registration plus or minus 0.025 to 0.06 mm; warpage 0.75 percent or less for assembly and 0.5 percent or less on HDI and rigid-flex; monthly capacity 2.05 million square feet. The figures sit on the FR-4 process parameters page and the capability page.
The honest boundary: we quote a laminate grade, and we do not publish a T260 or T288 number of our own for each grade, because those come from the laminate supplier and move with the batch. Ask for them with the stack and we return the supplier datasheet for the exact grade quoted. Where a customer needs Tg 170 and 0.075 mm line width at the same time, the answer is usually a different stack, not a different laminate.
SUPERB Tech builds the multilayer side of this on the multilayer PCB line, where the press cycle count and the registration budget are fixed before the first panel. For power boards, where copper weight drives the stack, the choice is in copper weight for power stages.
What to send with the stackup
Layer count, finished thickness, copper weight per layer, whether the board carries a press-fit connector or a heavy through-hole part, the number of reflow passes including rework, the operating temperature the end product has to hold and its relative thermal index, whether CAF or biased humidity testing is in the qualification plan, and the surface finish. Seven lines, and the grade becomes a conclusion instead of a preference.