Material Selection Series: From Isola and Rogers to EMC — A Field Map of 10 CCL Makers and 7 Laminate Families
As AI servers push signaling toward 224G PAM4 and automotive radar settles at 77 GHz, choosing a laminate is no longer just a Dk/Df lookup — you are simultaneously choosing a supplier, a lead time, and a yield-risk profile. Notably for engineers in supply chain, Taiwanese CCL makers are rapidly gaining ground in this materials race. This article condenses the major laminate OEMs and seven material families into one map, so you can get aligned quickly in your next stackup review.
Why choosing a material is really choosing a supplier
Two laminates can both claim 'low loss' yet differ widely in process window, lot‑to‑lot consistency, and fabricator stocking. A Df 0.002 material your board house doesn't stock can turn a 7‑day prototype into a 5‑week wait, while a Df 0.004 material they laminate every day gives predictable yield and delivery. The right sequence: define your electrical requirement band first, then ask which part numbers within that band your fabricator can actually source and process fluently.
The global CCL landscape: who is good at what
| OEM | Strength | Representative materials |
|---|---|---|
| Isola | High-speed digital, servers | FR408HR, Tachyon, Astra MT77 |
| Panasonic | Ultra-low-loss, AI/HPC | Megtron 6, Megtron 7 |
| Rogers | RF/microwave/mmWave | RO4350B, RO3003, RT/duroid 5880 |
| Taconic (AGC) | Cost-effective RF, PTFE | RF-35, TLY series |
| Nelco (AGC) | High-layer-count networking, PI | N4000-13, N7000-2HT |
| DuPont | Flex and rigid-flex | Pyralux AP/LF/FR |
| Ventec | Aerospace, automotive, thermal | VT-47, VT-901, VT-5A2 |
| Arlon (part of EMC group) | Military-grade, high-temp PI | 85N, 49N |
Elite Material Co. (EMC) and Nan Ya Plastics are already fixtures on global top‑10 lists. EMC's EM‑528 (Tg 250 °C, Df 0.0061) and EM‑890K (Dk 3.0, Df 0.00254 @ 10 GHz) are among the mainstream low‑loss choices for AI servers and HDI accelerator cards, with rising penetration across high‑speed computing supply chains. Nan Ya's NP‑155F and NP‑175 anchor the high‑volume consumer and industrial segments, winning on supply consistency and cost.
Beyond that list, two more Taiwanese names matter in practice: Taiwan Union Technology (TUC) , whose TU‑863+ and TU‑933 series are widely used in networking switches and server boards, and ITEQ , whose IT‑968‑class low‑loss materials are a common second source for high‑speed networking programs. The shared advantage of the Taiwanese four: broad stocking across fabricators, stable lot supply, and short prototype lead times — which, when a programme is on the clock, often matters more than the third decimal place of Df.
Seven material families at a glance
| Family | Df range (@10 GHz) | Typical applications | Representative materials |
|---|---|---|---|
| Standard/high-Tg FR4 | 0.012–0.025 | Consumer, industrial | NP-155F, VT-47 |
| Low-loss FR4/epoxy | 0.006–0.0095 | 10–56G networking | FR408HR, EM-528 |
| Hydrocarbon-PPE ultra-low-loss | 0.0017–0.004 | 112G/224G, AI/HPC | Megtron 7, Astra MT77, EM-890K |
| Polyimide (PI) | 0.009–0.021 | Aerospace, high-temp | VT-901, 85N |
| PTFE | 0.0006–0.0015 | >10 GHz RF | RT/duroid 5880, TLY-5 |
| Ceramic-filled PTFE | 0.001–0.0013 | 77 GHz radar, antennas | RO3003, RT/duroid 6002 |
| Flex materials | 0.003–0.02 | Wearables, medical, automotive | Pyralux AP, Felios |
The high‑speed digital battleground: where low‑loss FR4 ends and hydrocarbon‑PPE begins
For the 10–56G generation, most designs close with low‑loss FR4 (the FR408HR/EM‑528 tier). Past 112G, hydrocarbon‑PPE becomes essentially the only answer: Astra MT77 pushes Df down to 0.0017 and Megtron 7 to roughly 0.003, paired with HVLP low‑profile copper to keep long‑trace insertion loss inside budget.
Engineering reminder: this tier can cost 2–3× low‑loss FR4. Work backward from your loss budget to identify which layers truly need ultra‑low loss — a hybrid stackup usually beats swapping the entire board.
RF and mmWave: the PTFE trade‑offs
Material selection is the first fork in any RF design. Hydrocarbon‑ceramic materials like RO4350B (Df 0.0037) run on standard multilayer processes and dominate 10–30 GHz. At 77 GHz automotive radar and mmWave antennas, you enter RO3003 and RT/duroid 5880 territory — Df around 0.001, but PTFE drilling, desmear, and lamination all require specialised process parameters that not every fabricator has mastered. For RF programs, verify that your fabricator has real production history with the exact part number.
Three selection pitfalls R&D teams hit most often
Reading datasheets without checking stocking: a beautiful spec sheet is a lead‑time black hole if the material isn't on your fabricator's stocked list — ask for that list before committing.
No second source: Isola‑to‑EMC and Panasonic‑to‑TUC have rough grade‑level equivalents; write approved alternates into your spec at design time so a shortage doesn't stall the programme.
Simulating impedance on material A and mass‑producing on material B: even same‑grade materials can differ by 0.1–0.3 in Dk, so any substitution requires re‑running impedance simulation and adjusting the stackup — never carry numbers over directly.
Closing: turn the material map into negotiating leverage
The laminate world looks crowded, but three axes — OEM landscape, material family, and fabricator stocking — locate any material quickly. With Taiwanese CCL makers now at the core of the AI supply chain, knowing how to cross‑reference international and Taiwanese grades is direct leverage on both cost and lead time.
Sunkey regularly supports customers on laminate selection and hybrid stackup planning, with hands‑on process experience across Isola, Panasonic, Rogers, and the Taiwanese CCL makers' part numbers. Evaluating materials for your next high‑speed or RF project? Talk to our engineering team to get selection guidance and a DFM pre‑review at the layout stage — before lead‑time and yield risks surface at prototyping.
