The RT/duroid 6002 laminate data sheet is short: one dielectric constant, one dissipation factor, three thermal expansion coefficients. What it does not say is that each of those numbers is a property of a construction — PTFE resin, ceramic filler and woven glass in a fixed ratio — measured under a stated condition on a stated coupon, and reproducible only if the thickness, the copper and the bonding film are what the designer assumed. This article takes the 6002 parameter set apart in that order, because the design-side consequences are already covered in our parameter-by-parameter cost analysis, and the unanswered questions are usually about provenance and procurement rather than physics.
A Parameter Is a Construction, Not a Number
RT/duroid 6002 is a ceramic-filled PTFE composite reinforced with woven glass. The three ingredients set three different families of properties, and almost every argument about this laminate is really an argument about the ratio between them:
- The PTFE resin sets the loss floor and the chemical inertness. It also sets the two problems every fabricator knows: a thermal expansion in Z that is large in absolute terms, and a surface that no conventional wet chemistry wants to wet.
- The ceramic filler raises Dk above neat PTFE (2.1) to 2.94, and does it without adding the moisture sensitivity that a polar resin would bring — water absorption is 0.02 %.
- The woven glass holds dimensional stability and controls in-plane expansion, which is why X and Y come out at 16 ppm/°C, just under copper’s 17.
That is the reason the data sheet can be short. It is not a list of independent knobs; it is one recipe measured in nine ways. Change the thickness and you have changed the recipe’s geometry, not its chemistry — which is exactly why the electrical numbers stay put while the impedance does not.
The 6002 Parameter Set, With Conditions Attached
Below is the published set. The third column matters as much as the second: a number without its test method and its frequency is a marketing claim, not a specification. Note in particular that process Dk is quoted at 10 GHz by clamped stripline, and that design Dk is a separate measurement over 8–40 GHz by the differential phase length method.
| Property | Published value | Condition / method | What it governs in a design |
|---|---|---|---|
| Dk (process) | 2.94 ±0.04 | 10 GHz, 23 °C, IPC-TM-650 2.5.5.5 clamped stripline | Line width for a target impedance, and electrical length |
| Dk (design) | 2.94 | Differential phase length method over 8–40 GHz | What a broadband model should use |
| Df | 0.0012 | 10 GHz, 23 °C, clamped stripline | Dielectric share of insertion loss |
| TCDk | +12 ppm/°C | −50 to +150 °C | Phase drift over temperature |
| CTE, X / Y | 16 / 16 ppm/°C | IPC-TM-650 2.4.41, −55 to +288 °C | Layer-to-layer registration, warp |
| CTE, Z | 24 ppm/°C | Same method, through thickness | Plated barrel strain per thermal cycle |
| Thermal conductivity | 0.60 W/m·K | 50 °C, ASTM D5470 | Heat spreading through the dielectric |
| Water absorption | 0.02 % | D48/50 %, IPC-TM-650 2.6.2.1 | Dk stability in humid service, bake cycles |
| Peel strength | 8.9 lb/in (1.6 N/mm) | 1 oz ED foil, after solder float, IPC-TM-650 2.4.8 | Survives rework and connector soldering |
| Density | 2.1 g/cm³ | ASTM D792 | Weight budgets in airborne arrays |
| Flammability | UL 94 V-0 | UL file listing | Enclosure and rail compliance |
| Lead-free compatible | Yes | Supplier declaration | Reflow at 245 °C without delamination |
Two entries in that table are routinely misquoted in circulation. The first is Dk: a value of 3.0 or 2.94 ±0.05 shows up in supplier pages, both of which describe a different tolerance than the laminate’s. The second is CTE: X and Y are 16, not 13 or 17, and Z is 24 — the low Z figure is the actual selling point of this grade and it is the one most often replaced with the 46 or 237 of a neighbouring material.
Design Dk Equals Process Dk, and That Is Almost Unique
In most high-frequency laminates the two Dk numbers differ, and the difference is a trap: the process value is what a coupon measured at one frequency, while the design value is what a broadband model should use over the band the circuit actually works in. The gap runs about 5 % across the Rogers RF families. 6002 is the exception — both numbers are 2.94.
| Laminate | Process Dk (10 GHz) | Design Dk | Gap | Width error if design uses the wrong one |
|---|---|---|---|---|
| RT/duroid 6002 | 2.94 ±0.04 | 2.94 | 0 % | None |
| RO4003C | 3.38 ±0.05 | 3.55 | +5.0 % | ~2.5 % narrower line at 50 Ω |
| RO4350B | 3.48 ±0.05 | 3.66 | +5.2 % | ~2.5 % narrower line at 50 Ω |
| RT/duroid 6006 | 6.15 ±0.15 | 6.45 | +4.9 % | ~2.4 % narrower line |
| RT/duroid 6010.2LM | 10.20 ±0.25 | 10.7 | +4.9 % | ~2.4 % narrower line |
The practical consequence is small but it is free: on 6002 there is no argument about which Dk a model should carry, so the first-pass width is the pass-2 width. On a 3.48/3.66 material, a team that designs from the process value and then tunes on a test coupon is correcting a 2.5 % error it introduced itself — and, worse, is doing it in a state that looks like normal iteration rather than like a modelling mistake.
Thickness Is the Parameter That Sets Everything Else
Once Dk is fixed, the dielectric thickness h is the only free geometric variable, and it moves three things at once: the 50 Ω width, the conductor loss, and the sensitivity of impedance to everything else. Electrical length is not one of them. For a fixed width-to-height ratio, the effective permittivity is scale-invariant, so λg at 10 GHz stays at 19.4 mm whether the core is 0.127 mm or 1.27 mm thick.
| Core thickness | 50 Ω microstrip width | Dk / εeff | Electrical length at 10 GHz | Dielectric loss / 100 mm | Copper loss / 100 mm (1 oz) | Total / 100 mm |
|---|---|---|---|---|---|---|
| 0.127 mm | 0.33 mm | 2.94 / 2.38 | 19.4 mm (λ/2 at 7.7 GHz) | 0.21 dB | 1.31 dB | 1.52 dB |
| 0.254 mm | 0.65 mm | 2.94 / 2.38 | 19.4 mm | 0.21 dB | 0.67 dB | 0.88 dB |
| 0.508 mm | 1.30 mm | 2.94 / 2.38 | 19.4 mm | 0.21 dB | 0.33 dB | 0.54 dB |
| 0.762 mm | 1.95 mm | 2.94 / 2.38 | 19.4 mm | 0.21 dB | 0.22 dB | 0.43 dB |
| 1.270 mm | 3.25 mm | 2.94 / 2.38 | 19.4 mm | 0.21 dB | 0.13 dB | 0.34 dB |
The dielectric column is derived from αd ≈ 2.6·fGHz·√Dk·Df in dB/in, which at 10 GHz and Df 0.0012 gives 0.054 dB/in — and no choice of thickness changes it, because total dielectric loss depends on Df and electrical length only. The copper column is where thickness earns its keep: conductor attenuation falls roughly as 1/W, so doubling h halves it. On a 0.127 mm core, copper is six times the dielectric loss; on a 1.27 mm core the two are nearly equal.
Thickness is also the dominant impedance tolerance. A ±10 % window on h moves a 50 Ω line to 46.8 or 53.6 Ω — about ±6 % — which alone consumes most of a ±10 % impedance specification. That is why the thickness tolerance on the order matters more than the Dk tolerance: Dk ±0.04 is ±1.4 % on Dk, or about ±0.7 % on impedance.
Bonding Film Changes h, and h Is What You Designed
In a single-layer board the dielectric under the line is the core. In a bonded assembly it is the core plus the bonding film, and the bonding film is typically 0.038 to 0.10 mm of a PTFE composite close to the laminate in Dk. Designers who take h from the core thickness and then bond a cover layer or a second sub-assembly routinely find the line 10 to 23 % high — and the direction of the error is always upward, because adding dielectric height raises impedance.
| Dielectric under the line | h | W/h for the same 0.65 mm line | Resulting impedance | Error |
|---|---|---|---|---|
| 0.254 mm core only | 0.254 mm | 2.56 | 50.0 Ω | design intent |
| + 0.038 mm film | 0.292 mm | 2.23 | 54.8 Ω | +9.6 % |
| + 0.050 mm film | 0.304 mm | 2.14 | 56.2 Ω | +12 % |
| + 0.10 mm film | 0.354 mm | 1.84 | 61.5 Ω | +23 % |
The fix is not to tune the line; it is to put the whole stack in the model. Rogers 3001 bonding film is the matching material for this family (Dk 2.94, Df 0.0021), and its thickness belongs in the impedance calculation exactly like the core’s. Where the coupling medium differs in Dk — a lower-Dk film, or an FR-4 prepreg in a hybrid — the effective Dk of the gap has to be weighted rather than copied. Our stack-up design walkthrough covers the weighting method, and the lamination article covers what happens to that gap during the press.
Copper: Peel Strength Is Quoted, Roughness Is Not
The data sheet quotes peel strength because it is a pass/fail property that a fabricator can verify: 8.9 lb/in (1.6 N/mm) on 1 oz electrodeposited foil after solder float. It does not quote foil roughness, and roughness is what decides conductor loss above 5 GHz. At 10 GHz the skin depth in copper is 0.65 µm; a standard ED foil tooth profile is two to three times that, so the current follows a longer path than the geometry suggests.
Three consequences follow for a 6002 build:
- Standard ED foil is the default and the cheapest. It is right for anything below roughly 10 GHz where the dielectric still dominates the loss budget, as the 0.254 mm row of the thickness table shows (0.21 dB dielectric against 0.67 dB copper — here copper already dominates, so foil choice matters even at 10 GHz).
- Reverse-treated and low-profile foils trade adhesion for loss. They reduce the effective path length, at the cost of the mechanical key that peel strength measures. On a PTFE surface that is already hard to bond, this is a real trade rather than a free upgrade.
- Copper weight is a loss decision, not only a current decision. 1 oz is standard at 6002’s usual thicknesses; heavier copper widens the minimum feature and complicates fine-line etching, while 0.5 oz helps fine lines at the cost of current capacity and of handling damage before lamination.
What a Coupon Proves, and What It Does Not
Acceptance testing on a 6002 lot normally means a coupon built on the same panel, measured by clamped stripline at 10 GHz, plus a micrometer check on five points of the panel. That combination proves three things: the Dk of that lot sits inside 2.94 ±0.04, the thickness sits inside its tolerance band, and the process did not change between the qualification build and this one. It does not prove the impedance of a specific trace.
The reason is that coupon Dk and trace impedance are separated by every geometric variable in the stack: finished line width after etch, the actual dielectric height after the press, the copper thickness after plating, and the presence of solder mask over the line. Each of those is a manufacturing tolerance, and they add in root-sum-square. A coupon that reads 49.5 Ω is evidence that the material and the process are in family; it is not evidence that the 50 Ω line on the working panel is at 50.
Two practical rules follow. First, request the coupon from the production panel, not from a separate qualification panel — otherwise the thickness and the press are unverified. Second, state the Dk window and the measurement method on the purchase order, because “Dk 2.94” without a method is compatible with several readings. The testing sequence article sets out where coupon, TDR and VNA each fit.
TCDk 12 and Z-CTE 24 in Working Numbers
These are the two parameters that justify 6002 over a cheaper hydrocarbon laminate in a phased array or a satellite feed, and both can be turned into numbers a system engineer can use.
Phase drift. TCDk of +12 ppm/°C over a 125 K excursion (25 to 150 °C) changes Dk by 0.15 %, which changes electrical length by about 0.075 %. On a 100 mm line at 10 GHz that is 5.15 wavelengths, or 1854° of electrical length — so the drift is 1.4°. Against that, a laminate with TCDk of −125 ppm/°C drifts 14.5° and one at −425 ppm/°C drifts 49° over the same span.
| Laminate | TCDk | Dk change over 125 K | Phase drift, 100 mm at 10 GHz | Z-CTE | Barrel-to-copper differential, 1.6 mm / 215 K |
|---|---|---|---|---|---|
| RT/duroid 6002 | +12 ppm/°C | 0.15 % | 1.4° | 24 ppm/°C | 2.4 µm |
| RO4003C | +40 ppm/°C | 0.50 % | 4.6° | 46 ppm/°C | 10 µm |
| RT/duroid 5880 | -125 ppm/°C | 1.56 % | 14.5° | 237 ppm/°C | 76 µm |
| RT/duroid 6010.2LM | -425 ppm/°C | 5.31 % | 49° | 47 ppm/°C | 10 µm |
Barrel strain. Z-CTE of 24 ppm/°C is only 1.4 times copper’s 17. Over a 1.6 mm board and a 215 K excursion to reflow, the laminate grows 8.3 µm while the plated barrel grows 5.8 µm: a 2.4 µm differential that ductile copper absorbs without argument. Move to a pure-PTFE laminate with Z-CTE of 237 and the same arithmetic gives 76 µm of differential, which is why those materials need high-extension plating, aspect ratios at or below 6:1 and a hard look at every reflow. On 6002 the Z-axis is simply not the reliability problem it is on 5880.
Moisture, Outgassing and Chemistry Left in the Hole
Water absorption of 0.02 % is close to the floor for any laminate, and it is one of the reasons 6002 appears in space-qualified bills of material: low outgassing follows from low absorbed volatiles. But the moisture number describes the laminate as delivered, not the laminate after processing. A ceramic-filled PTFE is porous on the scale of the filler interface, and plating chemistry, etchant residue and resist stripper can be absorbed into that interface during fabrication.
Rogers documents the problem directly in its fabrication guidance for ceramic-filled PTFE composites, and the control is unglamorous: adequate rinse, an oven bake before the steps that trap chemistry, and a dry-storage regime between drilling and plating. For a space or high-reliability build, the bake schedule and the storage humidity belong on the process traveller, not in a general note. The drilling article and the lamination article carry the related window tables.
Four Process Windows That Follow From the Data Sheet
The parameter set does not just describe the material; it dictates four process windows. Each of the numbers below is a consequence of a number in the table above, which is why a fabricator who only knows “it is PTFE” will still get the windows wrong.
| Step | Window | Why this laminate needs it |
|---|---|---|
| Drilling | Sharp carbide, high spindle speed, low infeed, aspect ratio ≤6:1, entry and back-up material mandatory | Soft matrix with abrasive ceramic and glass; smear and hole-wall roughness are the failure modes, and Z-CTE 24 means the barrel is not the weak link |
| Hole-wall preparation | Sodium-naphthalene or plasma activation before metallization | PTFE does not wet conventional activation chemistry; without treatment plating passes inspection and fails thermal cycling |
| Bonding | 3001 film or FEP film at 290–315 °C, or low-flow prepreg at 180–190 °C; never in the same press cycle as FR-4 | Two resin systems with incompatible cure or melt windows; one cycle cannot serve both |
| After etch | Stress-relief bake per supplier guidance; no mechanical shock before bake | Etching releases the built-in orientation of the PTFE; unrelieved panels move during subsequent lamination |
For the metallisation side, the plating article gives the barrel and surface coverage numbers; for the material-family context, the RO3000 overview shows what changes when the ceramic loading, rather than the glass, carries the Dk.
What to Put on the Purchase Order
Most of the disputes about a 6002 build are settled or created before the panel is cut, in the purchase order. A laminate called out as “RO6002, 20 mil, 1 oz” leaves the Dk window, the copper type, the panel size, the bonding film and the acceptance method unstated.
| PO field | Example clause | What it prevents |
|---|---|---|
| Laminate | RT/duroid 6002, ceramic-filled PTFE with woven glass | Substitution by a lower-cost ceramic-filled PTFE with no glass |
| Thickness | 0.254 mm ±10 %, measured by micrometer at five points per panel | Impedance drift of ±6 % absorbed as “normal” build variation |
| Dk verification | 2.94 ±0.04 at 10 GHz, IPC-TM-650 2.5.5.5, coupon from the production panel | A Dk value quoted without a method |
| Copper | 1 oz ED, type stated (standard or reverse-treated) | Foil substitution changing conductor loss at 10 GHz and above |
| Bonding film | Rogers 3001, thickness stated; included in the impedance stack model | the +10 to +23 % impedance error of an unmodelled gap |
| Panel size | Stated panel and array size, with grain direction recorded | Registration and warp surprises from using the wrong panel format |
| Documentation | Certificate of conformance per lot, thickness and Dk data attached | No traceability when a lot fails in test |
Where 6002 Sits in Its Own Family
The RT/duroid 6000 family is short, and two of its most-quoted members do not exist. There is no 6202, no 6302 and no 6500; those designations circulate in copied articles and, when they appear, the parameters attached to them are usually borrowed from another family. There is also a persistent confusion between 6002 and 3001: 3001 is a bonding film, not a laminate, and quoting it as a board is a reliable sign that the source has been copied rather than read.
| Material | Dk | Df | CTE X/Y/Z | Role |
|---|---|---|---|---|
| RT/duroid 6002 | 2.94 ±0.04 | 0.0012 | 16 / 16 / 24 | Low Dk, copper-matched expansion, space-grade outgassing |
| RT/duroid 6002HT | 2.94 ±0.04 | 0.0012 | as 6002 | Same electricals, high-temperature variant |
| RT/duroid 6006 | 6.15 ±0.15 | 0.0027 | 47 / 34 / 117 | High Dk for size reduction below X-band |
| RT/duroid 6010.2LM | 10.20 ±0.25 | 0.0023 | 24 / 24 / 47 | Highest Dk in the family; filters and compact antennas |
| RT/duroid 6035HTC | 3.50 ±0.05 | 0.0013 | 19 / 19 / 39 | Thermal variant: k 1.44 W/m·K for high-power dissipation |
| Rogers 3001 | 2.94 | 0.0021 | n/a | Bonding film used with 6002 — not a laminate |
Read down the Dk column and the family logic is visible: one PTFE platform, four ceramic loadings, one bondply. Moving from 6002 to 6006 cuts linear dimensions by about 45 % for the same electrical length (√6.15/√2.94 = 1.45) and raises dielectric loss by 2.25 times, because at the same electrical length only Df decides. That trade, and the more aggressive one at Dk 10.2, are covered in the 6010.2LM analysis and the 6035HTC analysis.
When 6002 Is the Wrong Choice
Three honest exclusions:
- When the lowest possible loss is the requirement. Df 0.0012 is good, not best: 5880 and TLY-5 sit at 0.0009, which is 25 % lower, and at 10 GHz that is 0.054 against 0.035 dB/in of dielectric loss. For a long mmWave feed run the difference is real.
- When board area is the constraint. Dk 2.94 forces wider lines than 2.20 for the same impedance — 0.65 mm against 0.79 mm at 0.254 mm thickness, which is 20 % more width per line and a correspondingly larger combiner or corporate feed. Above roughly 30 GHz the phase stability of 6002 may still win, but the area has to be paid.
- When the budget is the constraint and the frequency is below 6 GHz. A hydrocarbon laminate at a third of the price will meet a ±10 % impedance spec at 3 GHz, because copper loss dominates the budget there. 6002 earns its price through TCDk 12, CTE 16/16 and outgassing — properties that matter in arrays, in space and across temperature, and do not show up in a room-temperature loss measurement. The RO4000 platform analysis sets out the other side of that trade.
Conclusion
The RT/duroid 6002 parameter set is short because the material is one recipe: PTFE for loss, ceramic for Dk, glass for stability. What the data sheet cannot convey is that every number has a condition and every condition has a geometric dependency. Process Dk and design Dk coincide at 2.94, which removes the most common source of first-pass width error in this material class; thickness then decides width, copper loss and impedance sensitivity; and the bonding film, if left out of the model, adds 10 to 23 % to every line it covers.
For a procurement or qualification engineer, the useful translation is that a 6002 order is a set of windows rather than a part number: thickness with a tolerance, Dk with a method, copper with a foil type, film with a thickness, and a coupon from the production panel. Get those five onto the order and the laminate behaves as specified. Leave them off and the material is still 6002 — only the board is no longer the one that was designed.
Frequently Asked Questions
What is the dielectric constant of RT/duroid 6002?
Why does adding bonding film change my 50 ohm impedance?
Is RT/duroid 6002 a PTFE material?
Does 6002 have low Z-axis CTE?
What thickness tolerance should I specify on the order?
Do RT/duroid 6202, 6302 or 6500 exist?
Specifying RT/duroid 6002 With a Dk Window and a Coupon?
Send us your Gerber files and requirements for a manufacturability review and a tailored quotation.
sales1@xcepcb.com
+86 180 3341 4657