Quantified effect of copper roughness Ra on 28GHz microstrip loss (measured data)

Engineering Summary

At 28 GHz, copper roughness Ra has a directly measurable impact on insertion loss. Tests on 50 Ω microstrip lines (Rogers RO4350B, 8‑mil core) show: ED copper (Ra≈2.0 µm): 0.85 dB/cm; Low‑profile ED (Ra≈1.2 µm): 0.72 dB/cm; RTF (Ra≈0.8 µm): 0.61 dB/cm; HVLP (Ra≈0.4 µm): 0.52 dB/cm. Switching from ED to HVLP reduces loss by ~39% at 28 GHz.

Loss Summary Table (28 GHz, 50 Ω microstrip, RO4350B)

Test conditions: 8‑mil RO4350B core, 1/2‑oz starting copper, microstrip line width 0.42 mm (50 Ω), solder mask removed, measured by VNA with SOLT calibration, de‑embedded to line ends.

1. Why Copper Roughness Matters at 28 GHz

At 28 GHz, skin depth in copper is ≈0.39 µm – only about one‑fifth the thickness of a human hair. Current flows almost exclusively in an ultra‑thin surface layer. When the copper surface is rough:

• The effective current path lengthens (follows valleys and peaks)
• Local current density increases near peaks
• Result: higher resistance and higher insertion loss

Engineering fact

A rough surface can increase conductor loss by 20–50% compared to an ideally smooth surface at mmWave frequencies.

2. Measurement Method (Reproducible)

We used a delta‑loss (Δ‑loss) coupon design per IPC‑TM‑650‑2.5.5.14:

• Two identical 50 Ω microstrip lines, length 50 mm
• Reference line: HVLP copper (Ra ≈0.4 µm)
• Test lines: other copper types on same panel
• VNA measurement from 10 MHz to 40 GHz, after SOLT calibration
• Fixture de‑embedding using TRL (thru‑reflect‑line) to remove connector effects

Uncertainty: ±0.03 dB/cm (95% confidence)

3. Detailed Measured Data

Key insight

Even HVLP copper (0.4 µm) still adds 0.08 dB/cm over an ideal smooth surface – this is the residual “real‑world” penalty.

4. Engineering Implications for 28GHz Designs

Rule of thumb: If your line length exceeds 20 mm at 28 GHz, HVLP copper is strongly recommended.

5. How to Specify Copper Roughness on Your Fabrication Drawing

Do not just write “low‑loss copper”. Be explicit:

UltroNiu standard: We provide copper roughness measurement reports on request, using a white‑light interferometer (ISO 25178).

6. Cost vs Performance Trade‑off

ROI perspective

The extra copper cost is typically <5% of total board cost, but can improve link margin by 1–2 dB – often the difference between pass and fail.

7. Practical Checklist for Your Next Design

• Specify copper type on each layer (do not leave default)
• Request profilometry data from your fabricator for the production lot
• Design Δ‑loss coupons that match your critical line length and layer
• Simulate with roughness model (e.g., Hammerstad or Groisse) – not ideal copper
• Remove solder mask from all mmWave transmission lines
• Validate first article with TDR and VNA before full production

8. Frequently Asked Questions

Related Engineering Resources

References: IPC‑TM‑650‑2.5.5.14 (Δ‑Loss), ISO 25178 (profilometry), Rogers RO4350B datasheet.