The standard FR4 vs low loss FR4 decision is not a choice between glass-epoxy and an exotic RF laminate. Both families run on a normal multilayer FR4 line. What changes is the resin system, glass style, and how stable Dk and Df stay as frequency climbs. Commodity FR4 is the better manufacturing choice when traces are short, edges are modest, impedance is ±10%, and unit cost plus stock availability matter more than a few tenths of a decibel. Low-loss FR4 is the better choice when insertion loss, lot-to-lot Dk, and a tighter impedance coupon start to eat first-pass yield on multi-Gbps or multi-GHz channels.
From a fabrication standpoint, "low loss FR4 vs FR4" is an enhanced-FR4 conversation: mid-loss grades such as FR408HR / 370HR-class systems, and low-loss FR4-processable systems such as I-Speed, I-Tera, Megtron 4/6-class constructions that still drill, oxide, and press like FR4. It is not automatically Rogers, PTFE, or a ceramic-filled hydrocarbon. We normally recommend staying on standard FR4 until channel length and Nyquist frequency make the Df gap visible on the loss budget. Specifying "low-loss FR4" with no named slash sheet is how CAM substitutes a mid-loss sheet you did not intend — or a Megtron construction you did not budget.
What Separates Commodity FR4 from Low-Loss FR4 on the Floor
| Factor | Standard FR4 | Low-loss / mid-loss FR4 |
|---|---|---|
| Typical Dk @ 1-10 GHz | ~4.2-4.8; drops as frequency rises | ~3.3-4.0 on many grades; flatter vs frequency |
| Typical Df @ 1-10 GHz | ~0.015-0.025 | Mid-loss ~0.008-0.012; low-loss ~0.003-0.007 |
| Dk lot-to-lot / resin content | Moves with glass style and resin %; coupon scatter is common | Tighter published windows; still must lock glass and resin % |
| Practical frequency / rate window | DC-~1-2 GHz; short multi-Gbps links if length is small | Multi-GHz digital and longer SerDes; still not mmWave PTFE |
| Impedance control in production | ±10% is the default yield window | ±8% or ±5% is realistic if the stack is named |
| Process on a standard FR4 line | Default drill, desmear, oxide, press | Same line; drill smear and press recipe need a qualified grade |
| Laminate cost vs standard FR4 | 1x baseline; cores and prepreg in every crib | ~1.5-3x mid-loss; ~3-6x true low-loss FR4-processable |
| Lead time | Stock almost everywhere | Common mid-loss is stocked in Asia; odd cores and Megtron-class wait |
| Typical applications | Consumer, industrial control, short MCU / USB 2.0 class links | USB3 / PCIe / SATA / HDMI length, Ethernet PHY, modest RF digital |
Datasheet Dk/Df numbers are not interchangeable. Test frequency, resin content, and copper foil change the printed value. During CAM review we lock a named core and prepreg, not the phrase "low-loss FR4."
Which Grade Wins When the Priority Changes
| If your priority is... | Better choice | Why factories lean that way |
|---|---|---|
| Lowest board cost and fastest proto | Standard FR4 | Stock cores, default recipes, no allocation wait |
| Lowest insertion loss on a long pair | Low-loss FR4 | Df is the dielectric term you cannot layout away once length is fixed |
| ±5% impedance yield | Low-loss FR4 | Flatter Dk vs frequency and tighter published windows |
| Short USB / MCU traces on a 4-layer book | Standard FR4 | Loss budget is dominated by connectors and length, not Df |
| Mass production of a cost-driven industrial board | Standard FR4 | Paying 2-4x laminate on every layer does not change the product |
| Hybrid stack to save money | Low-loss outers, standard inners | Put the expensive resin where the fast traces live; keep planes on commodity FR4 |
| mmWave / very low Df RF | Neither — leave the FR4 family | Low-loss FR4 is not a substitute for hydrocarbon-ceramic or PTFE |
Dk and Df Are the First Numbers That Actually Diverge
Standard FR4 Dk is a moving target. At 1 MHz it often sits near 4.5-4.8. By a few GHz it has already dropped, and it keeps moving with resin content and glass style (106, 1080, 2116, 7628 do not share one Dk). Df on commodity FR4 is typically 0.015-0.025 in the GHz range. That is acceptable on a 50 mm clock. It is expensive on a 200 mm 10 Gbps pair.
Low-loss FR4 grades are formulated so Dk is lower and flatter, and Df is cut by roughly 2-5x depending on whether you bought mid-loss or true low-loss. The electrical point of the upgrade is not "a nicer datasheet." It is less dielectric loss per inch and less coupon surprise when the same stack is built from a different resin lot. We still model from the vendor's frequency-dependent table for that exact construction — not from a generic "FR4 Dk = 4.2" note on the drawing.
Glass weave and foil matter on both materials. A low-loss resin on heavy 7628 with standard foil will not beat a well-chosen standard FR4 stack that uses spread glass and RTF/VLP foil on the critical layers. The low loss FR4 vs FR4 comparison is resin plus glass plus copper, not resin alone.
Where Signal Loss Shows Up in Production, Not in a Simulator Screenshot
Insertion loss on a finished board is dielectric loss plus copper loss plus geometry. Standard FR4 loses more in the dielectric term as soon as the pair is long and the harmonic content sits in the GHz. Low-loss FR4 cuts that term. It does not remove skin-effect loss, rough copper, via stubs, or a connector. If the channel fails because of a stub or a cheap foil, changing only the resin is a wasted adder.
In production we see the difference first on long SerDes, backplane-style spans, and boards that must hit an eye mask with little EQ left. Short MCU-to-PHY runs on a 100 mm industrial board rarely justify the premium. The trade-off here is simple: buy lower Df when length x frequency makes dielectric loss a first-order term; keep standard FR4 when loss is second-order to routing and connectors.
Hybrid books are the factory compromise. Fast pairs on low-loss outer cores or prepreg, power and low-speed inners on standard or high-Tg FR4. Press compatibility has to be checked — not every mid-loss prepreg likes every commodity core — but this is how most shops keep material cost from following the highest-Df-critical net across every ply.
Frequency and Data Rate Are Length Problems, Not Marketing Bands
A 5 Gbps link that is 30 mm long can live on standard FR4. The same rate at 250 mm often cannot. We normally look at rise time, Nyquist, and routed length before we agree that "this is a high-speed board, so it must be low-loss FR4." USB 2.0, CAN, and most MCU GPIO do not need the upgrade. USB 3.x, PCIe Gen3+, HDMI at length, 10G Ethernet, and similar channels are where the low-loss conversation starts — still inside the FR4 process family if the loss budget closes.
Once you are into true RF front-end work, tight phase matching at many GHz, or mmWave, low-loss FR4 is the wrong comparison. That is the FR4 versus Rogers / PTFE discussion. Mixing those two RFQs is how a customer pays Megtron money and still does not get a 77 GHz stack.
Where Material Cost Starts to Diverge — and How Shops Contain It
Commodity FR4 cores and prepreg are the cheapest multilayer dielectric you can buy that still yields. Mid-loss FR4 typically adds about 50-200% on the laminate line depending on brand and thickness. True low-loss FR4-processable grades can run three to six times the commodity sheet. Finished-board price does not scale 1:1 with that, because drill, plate, finish, and test stay similar — but on a 10- or 12-layer book the resin adder is no longer a rounding error.
Lead time follows allocation, not just price. Standard FR4 is in the rack. Odd low-loss cores (thin, heavy copper, uncommon glass) wait. If the prototype needs to ship in five days, we will tell you what is on the shelf. A drawing that only says "low-loss FR4, Df < 0.008" without a construction often becomes a mid-loss substitute or a week of material lead time.
Cost also shows up in yield. Tighter impedance on standard FR4 means more coupon failures and more scrap. Sometimes the mid-loss sheet is cheaper than fighting ±5% on commodity Dk. That is a production argument, not a marketing one.
What High-Speed Design Actually Changes in CAM and on the Press
A high-speed design on standard FR4 still needs reference planes, stub control, and a named impedance stack. Switching the resin does not replace those. What CAM changes for low-loss FR4 is the construction library: exact core/prepreg part numbers, glass style, copper foil type, and a press recipe the shop has already run. Drill smear behavior and desmear can differ from commodity FR4; a shop that has never qualified that grade will not treat it as a drop-in.
Registration, oxide, and sequential lamination risk are closer to high-Tg FR4 than to Rogers. That is the manufacturing advantage of staying inside low-loss FR4 instead of jumping to PTFE. Yield on a qualified mid-loss grade is usually high. Yield drops when the customer mixes unmatched prepregs, asks for ±5% on an unnamed stack, or puts low-loss resin on every ply of a board that only has two fast pairs.
During DFM we ask for: target impedance and tolerance, which nets are loss-critical, length of those nets, foil type if already chosen, and whether a hybrid stack is allowed. "Make it low-loss FR4" without that list is not a stack-up.
How Fab Houses Decide Which FR4 Family to Release
Most manufacturers default to standard or high-Tg FR4 unless the drawing names a grade or a Df/Dk window that commodity material cannot hit. We do not upgrade a board to low-loss FR4 because the schematic contains a USB3 connector. We upgrade when the loss budget or the impedance yield says the commodity sheet will not close.
Panel utilization is similar. Process risk is similar if the grade is qualified. The factory preference is still standard FR4 for anything that does not need the electrical window, because scrap on an expensive sheet costs more and because material substitution is harder when the AVL is a single branded low-loss system.
We normally recommend mid-loss FR4 as the first step off commodity FR4, not the lowest-Df sheet in the catalog. Jumping straight to Megtron-class resin "to be safe" is the same class of overspend as specifying Rogers on a 200 mm digital pair that a mid-loss FR4 stack would have passed.
Which FR4 Grade to Put on the Drawing
Choose standard FR4 if you:
- Have short digital links, modest edges, and no tight loss budget
- Can live with ±10% impedance on a shop-standard stack
- Need the lowest laminate cost and the shortest material lead time
- Are building industrial control, consumer, or MCU boards where connectors dominate loss
- Do not want to qualify a new resin system on an existing product
Choose low-loss (or mid-loss) FR4 if you:
- Have long multi-Gbps pairs or a channel that already failed the eye on commodity FR4
- Need ±5-8% impedance with less Dk wander versus frequency
- Can name the laminate / prepreg (or accept a hybrid with low-loss only on the fast layers)
- Want to stay on an FR4 process line instead of moving to hydrocarbon-ceramic or PTFE
- Have checked that drill, press, and foil for that grade are already qualified at the shop
There is no single winner in the standard FR4 vs low loss FR4 comparison. Commodity FR4 is the manufacturing default. Low-loss FR4 is the next resin step when dielectric loss and Dk stability become first-order. It is not a substitute for layout, foil, via geometry — and it is not a substitute for Rogers when you have left the FR4 frequency window.
Questions That Show Up on Standard FR4 vs Low-Loss FR4 Quotes
Q1: Is low-loss FR4 the same as Rogers or Megtron 6?
A1: No. Low-loss FR4 here means an enhanced glass-resin system that still processes like FR4. Rogers hydrocarbon-ceramic and PTFE are a different family. Megtron 6-class material is a low-loss digital laminate some shops still run on an FR4-like line, but it must be named — it is not "standard low-loss FR4."
Q2: How much more does low-loss FR4 cost than standard FR4?
A2: Mid-loss sheets often run about 1.5-3x commodity FR4 laminate. True low-loss FR4-processable grades are commonly 3-6x. Finished-board price rises less than that because process steps stay similar, unless every ply is the expensive resin.
Q3: Can I keep standard FR4 and still run PCIe or USB 3?
A3: Yes, if the routed length is short and the stack has a real reference plane. Once the pair gets long or the generation climbs, dielectric loss on commodity Df starts to close the eye. Length and rate decide, not the interface name alone.
Q4: Does low-loss FR4 automatically give tighter impedance?
A4: It helps because Dk is usually flatter and better documented. You still need a locked construction, controlled etch, and the right foil. ±5% on an unnamed "low-loss FR4" stack will not yield.
Q5: Should the whole stack be low-loss FR4?
A5: Usually no. Put the low-loss resin on the layers that carry the fast pairs. Keep power and low-speed inners on standard or high-Tg FR4 if the press recipe allows it. That is how factories contain the cost comparison.
Q6: What must be on the drawing so we do not substitute the wrong sheet?
A6: A named laminate and prepreg (or an allowed equivalent list), Dk/Df at a stated frequency, impedance and tolerance, foil type on the fast layers, and whether a hybrid stack is acceptable. "Low-loss FR4" alone is not a spec.
