Rogers 5000 Series Laminates: the Benchmark Measured, Band by Band

The Rogers 5000 series is the PTFE end of the catalogue: three grades — RT/duroid 5870, 5880 and 5880LZ — that share a dissipation factor between 0.0009 and 0.0019 and a dielectric constant between 1.96 and 2.33. That is a real benchmark, and it is worth stating in units a link budget can use rather than in adjectives. This article converts the benchmark into decibels per 100 mm and decibels per wavelength at 10, 24, 39, 77 and 110 GHz, then checks the claims that circulate about the family against the three data sheets — because six of the eight claims in the most-shared description are wrong, and four of them are wrong in the direction that costs money.

The 5000 series in the Rogers catalogue is not a single material. It is three PTFE composite laminates that bracket the lowest dielectric constants on the market: 5870 at 2.33, 5880 at 2.20 and 5880LZ at 1.96. All three are PTFE with micro-glass reinforcement — 5880LZ adds ceramic filler to bring its Z-axis expansion and density down — and all three are specified at 10 GHz with a dissipation factor between 0.0009 and 0.0019.

That is the substance of the word benchmark, and it should be expressed in a unit the design can absorb. A dissipation factor is not a specification an engineer can compare with a link budget; decibels per 100 mm and decibels per wavelength are. Everything below is that conversion, and the conversion is unforgiving: it shows where the material’s advantage is decisive, where it is marginal, and where the surrounding copper has already taken over.

One naming note before the numbers. The series is RT/duroid 5000, and it does not include every part number that starts with a 5. The 6000 series (6002, 6010.2LM, 6035HTC) is a different family with a different filler system. The verified model list, and the three part numbers that circulate but do not exist, are set out in the RT/duroid 5000 family verification; this article stays on performance.

These are the data-sheet values that decide whether the family suits a design. The columns that matter most are the last three: TCDk, Z-axis CTE and thermal conductivity, because those are where PTFE normally extracts its price.

Two rows are worth reading twice. The Z-axis CTE column runs 41.5 to 237 ppm/°C. Against copper at 17 that is 2.4× for 5880LZ and 14× for 5880, and it is the reason plated-hole rules on this family are stricter than on any FR-4 variant: aspect ratio at or below 6:1, plated copper of at least 30 µm, elongation of 20 % or better. The thermal conductivity column runs 0.20 to 0.33 W/m·K, which is a quarter of a hydrocarbon laminate and about a fifteenth of a thermally conductive one.

Turn Df into loss and the family separates from everything else in the catalogue. Using αd ≈ 2.6 · f_GHz · √Dk · Df dB/inch for the dielectric term, plus the conductor term for a 1 oz smooth-copper microstrip on a 0.254 mm core, the table gives the total at 10 GHz.

The gap that justifies the price is between 5880 and RO4350B: 0.71 against 1.49 dB per 100 mm, a factor of 2.1. Against FR-4 it is a factor of 7.2. And between the three PTFE grades the difference is small — 0.71, 0.78, 0.80 — which is the first useful conclusion: do not choose within the family on loss. Choose on Z-axis CTE (5880LZ at 41.5), on density (1.4 for weight-sensitive airborne hardware), or on the tolerance of the Dk itself.

The second conclusion is less comfortable. At 10 GHz the conductor term is already four times the dielectric term on 5880: 0.145 against 0.035 dB/inch. That ratio is why a 0.0009 Df does not produce a 0.7 dB/100 mm board in practice. Copper roughness, plating thickness and surface finish add more than the dielectric difference between grades, which is why the finish choice matters as much as the laminate on this family — the mechanism is set out in PCB surface treatment processes.

Dielectric loss rises linearly with frequency and conductor loss with the square root, so the balance between the two shifts as the band goes up. The table keeps the same 0.254 mm stack and gives both terms, the total, and the guided wavelength that governs how long a matching element has to be.

Read the last two columns and the practical difficulty of the upper bands becomes concrete. At 77 GHz a quarter-wave stub on this material is 0.71 mm long and the trace is 0.79 mm wide, so the element is shorter than it is wide: the layout is no longer a two-dimensional problem, and the discontinuity at the stub matters as much as the stub itself. At 110 GHz the quarter wave is half a millimetre, which is under two trace widths, and every via, every connector launch and every solder mask opening is electrically large.

The same table explains the honest limit. Between 10 and 77 GHz the dielectric term grows by a factor of 7.6 while the conductor term grows by 2.8, so by 77 GHz the two are nearly equal and the laminate choice buys at most a 20 % reduction in total loss. Beyond that point the return on a lower Df collapses: 0.0009 versus 0.0019 at 110 GHz is a 0.12 dB/100 mm difference, which a single 0.1 mm of via stub or one extra connector will erase.

A PTFE laminate is not an expensive FR-4; it is a different process route, and the route has four fixed costs that a quotation will not show separately. They are worth tabulating because they are the reason two shops can quote the same material 40 % apart.

There is a fifth cost that only shows up in the field: PTFE creeps. A clamped or bolted assembly compresses the dielectric, and 4 % of compression turns a 50 Ω line into 48.8 Ω and does not recover when the clamp is released. Connector launches, heat-sink mounting and board-to-board standoffs all have to be designed for a soft substrate. The mechanical side, including the Z-axis rules that follow from the 237 ppm figure, is covered in multilayer RT/duroid 5880 processing.

The description of this series most often reproduced online is written as prose and contains numbers that do not exist in any Rogers document. Each line below pairs the claim with the data sheet value and the decision it distorts. The material is not at fault in any of them; the copy is.

Two of these deserve a longer note. The thermal conductivity claim is the one that most often reaches a customer as a reason to skip thermal vias, and the arithmetic is simple: 0.254 mm of 5880 is 12.7 K·cm²/W, so a 10 W/cm² die without vias sees a 127 K rise through the board. The CTE claim is the one that survives longest because 31 ppm in X really is close to copper; the number that fails is Z, and Z is the axis a plated barrel lives in.

One claim does hold up and is worth keeping. Water absorption of 0.02 % is genuinely low, and it is why the family holds its Dk in humid conditions where a hydrocarbon laminate at 0.06 to 0.08 % drifts. The comparison against those hydrocarbon grades is tabulated in the RO4000 family overview.

There are only three choices and each has one dominant reason. 5880 is the reference grade: lowest Df, 0.02 % absorption, 31 ppm X-axis CTE, and the widest thickness range. 5870 trades 33 % more dielectric loss for a slightly higher Dk, which shortens a resonator by 3 % — a difference that rarely decides anything, so 5870 is usually selected for availability rather than for performance.

5880LZ is the interesting one. Its Z-axis CTE is 41.5 ppm/°C against 237, which is a 5.7× improvement and turns the plated-hole problem from the hardest part of the build into a routine one. Its density is 1.4 against 2.2, a 36 % weight saving on the substrate. The price is a Df of 0.0019 and water absorption of 0.22 %, so it is not the grade for a phase-critical array in a humid enclosure; it is the grade for large, light, hole-dense, thermally cycled structures.

Because the family’s grades are close in electrical performance and far apart in mechanical behaviour, most of the risk in procurement is ambiguity rather than quality. Seven items close it.

1. The exact grade and the Dk with its tolerance, written as 2.20 ±0.02 rather than “5880”. 2. The Df and the test method and frequency, since a supplier can quote 0.0009 at 10 GHz and deliver a value measured at 2 GHz. 3. Cladding: electrodeposited or rolled annealed, and the weight, because RA foil costs more and changes the conductor loss. 4. Thickness in millimetres with a tolerance, and whether the value is core thickness or finished dielectric. 5. Panel size, since the FEP bonding route constrains it. 6. PTH requirements: aspect ratio, plated thickness and elongation, and the activation method to be used. 7. For space or high-altitude use, the outgassing figures — TML 0.03 % and CVCM 0.00 are published for the PTFE grades and are not automatic for a substitute.

The reason to be this explicit is that “Rogers 5000” on a drawing can legitimately be filled with three different materials whose Z-axis CTE differs by a factor of nearly six. XCE PCB works from the grade and the numbers: 2 to 72 layers, impedance to ±10 % on coupons cut from the centre and edge of every panel, panels to 600×1200 mm, samples in as fast as 24 hours, inspected to IPC-A-600G and IPC-6015 — request a quote or WhatsApp +86 180 3818 8537.

The claim that this family represents “the pinnacle” has been stable for a decade because the physics has been stable: a PTFE composite with micro-glass can reach 0.0007 to 0.0009 Df and no lower without changing the resin chemistry. What is changing is everything around the laminate.

Three developments matter more than another decimal place in Df. First, copper foil: at 77 GHz the conductor term in the table above is 0.403 dB/inch against 0.267 for the dielectric, so a foil with a third of the roughness profile is worth more than a change of laminate grade. Second, Z-axis engineering: the 5880LZ approach — ceramic filler traded against Df and absorption — is the direction that makes large PTFE boards manufacturable, and it is where the next real improvement sits. Third, transition design: at 110 GHz the connector launch and the via stub dominate, and a board that is specified to 0.0009 Df but built with a 1.6 mm through via will not show it.

That ordering is the useful takeaway from a “benchmark” article. The laminate sets the floor; the copper, the plating, the stack-up and the transitions decide whether the board gets anywhere near it. The stack-up side of that argument is in high-frequency PCB stack-up design, and the measured side in ten critical parameters for high-frequency PCBs.

The Rogers 5000 series is three PTFE grades — 5870 at 2.33 Dk and 0.0012 Df, 5880 at 2.20 and 0.0009, 5880LZ at 1.96 and 0.0019 — and the benchmark they set is 0.71 to 0.80 dB per 100 mm at 10 GHz on a 0.254 mm core against 1.49 for RO4350B and 5.10 for FR-4. That is a factor of 2.1 and 7.2, and it is the entire commercial case.

The case weakens as the band rises, because conductor loss grows more slowly than dielectric loss but starts four times higher and never stops mattering: at 77 GHz the total is 2.64 dB per 100 mm and the dielectric share is only 40 %. It also has four fixed process costs — a 260–300 °C bonding route, sodium-naphthalene or plasma hole activation, a 6:1 aspect-ratio ceiling from 237 ppm Z expansion, and 0.20 W/m·K thermal conductivity that demands vias.

What the family does not do is match copper in three axes, conduct 0.8 W/m·K, or process like FR-4, and a specification built on those three statements will fail in exactly that order: first the thermal budget, then the plated holes, then the press schedule. Built on the real numbers, it is still the lowest-loss laminate in volume production, and the rest of the catalogue is measured the same way in ultra-low loss PCB materials and the RT/duroid 5880 parameter set.