The short answer: if your RF traces carry signals above 5 GHz, JLCPCB is not a viable production path. The controlled variables that determine RF performance — dielectric constant stability, copper surface roughness, via stub resonance, and etch factor consistency — are simply not managed in a budget fabrication flow.
Quick Decision: JLCPCB or Custom Manufacturer for RF PCBs?
| Frequency Band | JLCPCB Viability | Custom Manufacturer Advantage |
|---|---|---|
| Below 2.4 GHz | Adequate for prototyping | Overkill unless production |
| 2.4-5 GHz | Marginal | Recommended for production |
| 5-10 GHz | High failure risk | Essential |
| 10-28 GHz (5G FR2) | Not viable | Required |
| 28-77 GHz (mmWave) | Impossible | Specialized process mandatory |
The Physics of Why Budget PCB Services Fail at RF
The fundamental challenge with RF PCB fabrication above 5 GHz is that every fabrication variable that a budget service leaves uncontrolled becomes a first-order contributor to signal degradation. At 1 GHz, a dielectric constant variation of +/-0.2 shifts your 50-ohm microstrip by about 2 ohms — annoying but manageable. At 28 GHz, that same Dk variation creates impedance discontinuities that generate standing waves, increase insertion loss by 1-2 dB per inch, and destroy your noise figure budget.
In our facility, we track incoming laminate Dk at 10 GHz using a split-post dielectric resonator. We have measured lot-to-lot variation in standard FR-4 of 0.25 in Dk — which at 28 GHz corresponds to approximately 8% impedance shift on a microstrip line. Rogers 4350B, by contrast, arrives with certified Dk values at multiple frequencies, and our incoming QC confirms the laminate meets its +/-0.04 specification before any panel enters production.
The second critical factor is copper surface roughness. Standard FR-4 from budget fabricators uses electrodeposited copper with an Rz roughness of 5-8 um on the treated side. At 5 GHz, skin depth in copper drops to approximately 0.9 um — meaning the current flows almost entirely within the surface roughness peaks and valleys. The additional conductor loss from rough copper adds 0.3-0.5 dB/inch at 10 GHz compared to HVLP copper that RF-focused manufacturers specify.
Via Transitions: The Silent RF Performance Killer
Above 5 GHz, the via stub — the unused portion of a through-hole via extending beyond the signal layer — becomes a resonant antenna. A 1.0mm stub resonates at approximately 37 GHz, with its first quarter-wave null creating a notch in the passband. Even at 10 GHz, this stub contributes 1-2 dB of insertion loss.
The solution is backdrilling — mechanically removing the via stub after plating. JLCPCB does not offer backdrilling on standard orders. Our process engineers control backdrill depth to +/-3 mil using laser depth measurement on each panel. For 77 GHz automotive radar boards, we routinely backdrill to within 4 mil of the target signal layer. The difference between 4-mil and 12-mil residual stub at 77 GHz is approximately 4 dB of insertion loss improvement per via transition.
Material System Control: What JLCPCB Cannot Guarantee
When you order a "Rogers 4350B" board from JLCPCB, you receive a board fabricated on Rogers 4350B laminate. What you do not receive is any guarantee about which lot that material came from, what the actual measured Dk was, or whether it was stored within humidity specifications before lamination.
A custom RF PCB manufacturer manages the complete material chain. We maintain Rogers material inventory in humidity-controlled storage (below 50% RH, 23+/-2C) and track each panel back to the specific laminate lot. When an engineer specifies Dk=3.48 at 10 GHz, we verify that the material measures within +/-0.04 of that value.
The hybrid stackup is where budget services create the most unpredictable results. The prepreg between Rogers and FR-4 layers must be characterized for its contribution to the overall effective dielectric constant. In our process, we run test vehicles for each new hybrid stackup combination and document the effective Dk at 1, 5, 10, and 20 GHz.
The True Cost of RF Board Failure
Engineers choosing JLCPCB for RF prototypes often frame the decision as "$50 vs $300 per board." That arithmetic ignores the cost of failure — which for RF boards above 5 GHz is typically "the board sort-of works, inconsistently, and you spend three weeks debugging phantom performance variations."
We have onboarded over 120 RF customers who came to us after exactly this scenario. The typical story: engineer designs a 5.8 GHz front-end, orders from JLCPCB, gets boards with -15 dB return loss on some channels and -8 dB on others. Third order from us with material control produces boards that consistently hit -20 dB return loss across all channels.
For production volume, a 1000-unit run at $45/board from a custom manufacturer vs $15/board from budget service is a $30,000 premium. One field failure requiring recall and re-qualification costs $150,000-500,000.
When JLCPCB Actually Makes Sense for RF
Not every RF project needs a custom manufacturer:
- Concept validation below 3 GHz — LoRa, BLE, sub-GHz IoT
- Antenna prototype iteration — relative improvement matters more than absolute performance
- Test fixtures — boards that interface to RF components but don't carry primary RF signals
The decision framework: if the RF performance of the PCB directly determines whether your product passes system-level specs, use a custom manufacturer.
RF Verification: What Your Test Report Should Include
A custom RF PCB manufacturer provides:
- TDR impedance data on every production panel (not just edge coupons)
- Insertion loss measurement at operating frequency on dedicated test structures
- Cross-section analysis documenting actual trace geometry and dielectric height
- Backdrill depth verification through X-ray or destructive cross-section
None of this data exists in a JLCPCB order.
Reviewed by AtlasPCB Engineering Team — 15+ years in advanced PCB fabrication for RF, HDI, and rigid-flex applications.
Further Reading:
Top comments (0)