2,000 5G Small Cell Boards Shipped with Zero RF Performance Failures
A London telecom startup needed high-frequency RF boards that held impedance in outdoor enclosures. After two assemblers failed on Rogers 4350B, they came to Queen EMS.
A Seed-Stage Telecom Startup, London, UK
This 9-person engineering team designs compact 5G small cell units for dense urban deployment — lamp posts, building facades, and transit hubs across the UK and EU. Their boards run at 24–29.5 GHz and require Rogers 4350B laminate to meet link budget at millimetre-wave frequencies.
Product Type
5G mmWave small cell unit for urban outdoor deployment. Single RF board on Rogers 4350B (Dk 3.66), 8-layer stackup with laser-drilled microvias and back-drilled through-vias to eliminate stub resonance at 28 GHz.
Technical Complexity
Controlled impedance on 50Ω, 75Ω, and differential 100Ω lines. Back-drilling to ±50 µm depth. Selective conformal coating masking antenna feed zones. Per-board VNA S-parameter test at 24–30 GHz required on every unit.
Production Volume
Scaling from 50 prototype boards to 200 units/month, totalling 2,000 units across 10 production batches — deployed across 47 urban small cell sites in the UK
What They Needed
A PCBA partner experienced with Rogers laminates, controlled-impedance fabrication, back-drilling for via stub elimination, and 100% per-board RF testing at millimetre-wave frequencies
What Went Wrong with Their Previous Suppliers
High-frequency RF assembly punishes every shortcut. The previous two assemblers made three mistakes that were invisible on standard AOI — but lethal to 5G link performance in the field.
Impedance Modelled at FR4 Dk — Boards Came Back 8Ω Out of Spec
Neither previous assembler had a controlled-impedance program for Rogers material. Trace widths were modelled using FR4's dielectric constant (Dk 4.3) instead of Rogers 4350B's actual Dk of 3.66 — a systematic error that shifted all 50Ω lines to 56–58Ω. The resulting ±16% impedance error caused standing waves on the antenna feed lines that degraded gain uniformity across every deployment site. Two full batches were shipped before the RF team identified the root cause.
Via Stubs Not Back-Drilled — Insertion Loss Spike at 26 GHz
Through-hole vias create resonant stubs at frequencies where the stub length equals a quarter wavelength. At 28 GHz, a 1.2mm stub resonates at approximately 25–26 GHz — right inside the n258 band. Both previous assemblers skipped back-drilling because their equipment lacked ±50 µm depth control. The stub resonance created a 3.5 dB insertion loss spike across a 400 MHz window, effectively blinding the small cell to a portion of its operating band.
Conformal Coat Applied Over Antenna Feed Points — RF Performance Degraded
Outdoor deployment requires conformal coating for moisture protection. The first assembler coated the entire board without masking — the dielectric constant of the coating shifted the antenna resonant frequency by 180 MHz. The second assembler masked inconsistently using manual tape, leaving coating voids over BGAs that caused delamination during humidity testing. The RF team's own measurements confirmed GPS acquisition time increased from 45 seconds to over 3 minutes on boards from both assemblers.
No RF Testing at Frequency — Defective Boards Shipped to Sites
Neither previous assembler had VNA capability above 6 GHz. Standard functional tests — continuity, power rails, basic digital comms — passed on every board. But the impedance errors and stub resonances that degraded 5G link performance were completely invisible without S-parameter measurement at 28 GHz. The RF team discovered the issues only after boards were already installed on lamp posts, requiring site engineers to climb and swap units — at a cost of £600–900 per swap.
"We'd been through two CMs who simply ran the files we sent. Neither mentioned Rogers material, neither suggested back-drilling, neither had a VNA that could test above 6 GHz. We were discovering assembly problems at 15 metres up a lamp post. That's when we knew we needed a fundamentally different kind of supplier."
Why They Chose Queen EMS
After two assembler failures and a costly field swap programme, the team needed a partner who understood RF from stackup to S-parameter — not just from Gerber to shipping box.
Rogers-Specific Impedance Control Program
Queen EMS calculates trace widths against the actual Rogers 4350B Dk for each laminate lot, not FR4 nominal values. Coupon TDR (Time Domain Reflectometry) measurements are run on three coupons per panel, with results logged and provided in shipment documentation. No board ships without a measured impedance report.
Back-Drilling to ±50 µm Depth Control
Queen EMS's fabrication partner runs dedicated CNC back-drilling with laser-verified depth control at ±50 µm — eliminating via stubs in the RF signal path. TDR measurements on drilled coupons confirm stub removal before boards are released to assembly. This step is standard on all mmWave Rogers jobs, not an option.
100% Per-Board VNA Test at 28 GHz
Every assembled board is tested with a VNA at 24–30 GHz before packing. S11 and S21 are measured at the antenna input port against agreed pass/fail criteria. Boards outside the ±3.5% impedance band or with insertion loss exceeding budget are quarantined and root-cause analysed. Test data sheets are included in every shipment.
"Queen EMS were the first people who actually read our Gerbers. Within 48 hours of receiving our files, their RF engineer had identified a ground pour coupling issue near the PA output that would have shifted our n258 gain by 1.8 dB. We hadn't caught it. Neither had our previous two CMs. That one email saved us at minimum one board spin — probably four months of delay."
How We Engineered the Build for 28 GHz Deployment
A four-step RF-specific process built from the ground up — not a generic FR4 workflow adapted for Rogers material at the last minute.
Controlled-impedance redesign against actual Rogers Dk
Before any fab release, Queen EMS's RF engineer audited the 8-layer stackup against the Rogers 4350B datasheet at the actual panel Dk (3.66 ±0.05). A revised trace width table was issued within 36 hours — correcting 50Ω, 75Ω, and differential 100Ω lines. The customer's EE signed off before fabrication began.
Back-drilling to ±50 µm with TDR coupon verification
All through-hole vias in the RF signal path were back-drilled to within 50 µm of target. TDR measurements on three coupons per panel confirmed stub removal. Coupon results are logged per panel and included in the shipment QC pack — not sampled, not estimated.
Low-residue flux + selective robotic conformal coating
A no-clean, low-residue flux profile was specified for Rogers 4350B compatibility. Post-reflow, conformal coating was applied by selective robotic dispensing — masking all antenna feed pads and SMA launch areas to within 0.3 mm. No manual tape; mask boundaries defined by robot program to remove human variation.
Per-board VNA S-parameter measurement at 24–30 GHz
Every board tested: S11 return loss and S21 insertion loss at antenna input. Pass criteria: impedance within ±3.5%, insertion loss ≤0.4 dB above nominal. Boards outside criteria quarantined for root-cause analysis before disposition. Test data sheet included with each shipped unit.
Ground pour coupling flagged before fabrication
During Layer 3 stackup review, Queen EMS identified a missing clearance window around a via cluster near the PA output stage. At 28 GHz, unintended coupling would have acted as a parasitic capacitor, rolling off n258 gain by 1.8 dB. A marked-up PDF was issued the same evening — the fix required a 20-minute layout change, not a board spin.
DDP London in 12 calendar days
Duty Delivered and cleared through UK customs — no VAT surprises, no broker delays. Queen EMS manages the full logistics chain from our factory to the customer's London address. 12 calendar days from Gerber approval to boards on the engineering team's desk.
From Gerber Upload to Boards in London
12-day turnkey delivery including Rogers stackup review, back-drilling, controlled-impedance verification, and 100% per-board VNA testing at 28 GHz.
RF DFM Review
Day 1–2
BOM Sourcing
Day 1–4
Rogers Fab + Back-Drill
Day 3–6
TDR Coupon Verify
Day 6–7
SMT + Selective Coat
Day 7–9
VNA Test @ 28 GHz
Day 9–11
Ship DDP London
Day 12
Measurable Impact After 9 Months in the Field
2,000 units deployed across 47 urban small cell sites in the UK. Every Queen EMS-assembled board is performing within RF spec — no site engineer climbs required.
| Metric | Before Queen EMS | After Queen EMS |
|---|---|---|
| 📐 50Ω Impedance Tolerance | ±8Ω (±16%) — FR4 Dk model error | ±1.75Ω (±3.5%) — Rogers Dk corrected |
| 🔩 Via Stub Back-Drilling | Not performed — 3.5 dB loss spike at 26 GHz | ±50 µm depth control, TDR verified per panel |
| 🛡️ Conformal Coating | Manual / inconsistent — antenna resonance shifted | Selective robotic, ±0.3 mm boundary accuracy |
| 📡 VNA Test Coverage | 0% — no RF test at frequency | 100% per board at 24–30 GHz |
| 📉 Insertion Loss (S21 @ 28 GHz) | Unmeasured — estimated 3–4 dB over budget | ≤0.4 dB above nominal — within link budget |
| 🏙️ Field RF Failures | 11% return rate (batch 1 + 2) | 0 returns in 9 months / 2,000 units |
| 🚀 Lead Time (Gerber → DDP London) | 19–24 days | 12 days — 37% faster |
"We'd been burned by two CMs who couldn't handle Rogers material. Queen EMS understood RF from day one — they caught a Gerber error before fabrication, held impedance to ±3.5% across 2,000 boards, and delivered DDP London in 12 days. For millimetre-wave work, there is no one else we would use."
Is This Approach Right for Your Project?
This engagement model works best for teams building 5G infrastructure, mmWave radar, satellite communications, or any RF product where controlled impedance and per-board frequency testing are non-negotiable.
✅ Good Fit If You…
- Build RF boards operating above 6 GHz on Rogers or PTFE laminates
- Need controlled impedance verified by TDR coupon — not estimated from nominal
- Require back-drilling for via stub elimination in mmWave bands
- Need 100% per-board VNA S-parameter testing at operating frequency
- Need selective conformal coating with antenna keep-out masking
- Require DDP delivery to the UK or EU with customs documentation
🔍 What You Should Ask Us
- How do you calculate trace widths for Rogers 4350B vs FR4?
- What back-drilling depth accuracy do you achieve, and how is it verified?
- Do you perform TDR coupon measurements per panel or per batch?
- What VNA frequency range can you test, and what pass/fail criteria do you use?
- How do you mask antenna keep-out zones during conformal coating?
- What DFM checks do you run specifically for RF and mmWave designs?
Have an RF Board That Needs Controlled Impedance?
Send us your Gerbers and stackup. Our RF process engineer will review impedance requirements, flag DFM concerns, and confirm back-drilling needs within 48 hours — at no cost and no obligation.