BGA RF
Understanding BGA RF
BGA packages dominate RF IC design above 2–3 GHz because their short, wide ball interconnects provide 5–20× lower parasitic inductance than leaded packages. At 5 GHz, reducing package inductance from 5 nH (QFP) to 0.5 nH (BGA) cuts the reactive impedance perturbation from j157 Ω to j15.7 Ω, directly improving return loss. The distributed ground ball array beneath the die provides a low-impedance reference plane, and the exposed die-attach pad offers thermal resistance 4–10× better than peripheral-lead alternatives.
Impedance control through the BGA transition requires coordinated design across die-to-substrate (wire bond or flip chip), substrate routing (stripline/microstrip on BT or LTCC), and ball-to-PCB transition (ground ring geometry, antipad tuning, via-in-pad). At mmWave frequencies (24–77 GHz), LTCC substrates with tanδ < 0.003 replace lossy BT resin, and flip-chip bumps (30–80 pH) replace wire bonds (0.7–1.2 nH/mm).
Parasitic Inductance Budget
Lball ≈ 0.3–1.0 nH (h = 0.2–0.5 mm)
XL = 2πfL = j31.4 Ω at 5 GHz per nH
Wire Bond vs. Flip Chip:
Wire bond: 0.7–1.2 nH/mm (25 μm Au wire)
Flip chip: 30–80 pH (50–100 μm bump)
Coaxial Ball Impedance:
Z ≈ 377/(2π√εr) × ln(Douter/Dball)
1.0 mm pitch, 0.5 mm ball: Z ≈ 50–70 Ω
BGA Substrate Materials for RF
| Material | εr | tanδ (10 GHz) | Max Freq | Application |
|---|---|---|---|---|
| BT resin | 3.3–3.8 | 0.008–0.015 | ~10 GHz | Wi-Fi, BLE, GPS, cellular |
| LTCC | 5–9 | 0.001–0.003 | 77 GHz+ | Automotive radar, 5G mmW |
| HTCC (alumina) | 9–10 | 0.0001–0.001 | 100 GHz+ | Mil/space |
| Megtron 6/7 | 3.3 | 0.002 | 40 GHz+ | 5G AiP |
BGA vs. QFP for RF
| Parameter | BGA | QFP |
|---|---|---|
| Lead inductance | 0.3–1.0 nH | 2–8 nH |
| Ground plane | Distributed (under die) | Peripheral only |
| θJB | 2–10°C/W | 20–50°C/W |
| Inspection | X-ray required | Visual/AOI |
| Rework | Complex (station + nozzle) | Simple (hot air/iron) |
| Max frequency | 77 GHz+ (LTCC) | ~3–5 GHz |
Frequently Asked Questions
Why BGA for RF ICs?
5–20× lower inductance than QFP (0.3–1.0 nH vs. 2–8 nH). At 5 GHz, 1 nH = j31.4 Ω. Distributed ground ball plane, not peripheral-only. Exposed pad θJB 2–10°C/W (vs. QFP 20–50°C/W). Dominant above 2–3 GHz. Trade-off: X-ray inspection required, complex rework.
Impedance control through BGA?
Three-level design: die-to-substrate (flip chip 30–80 pH above 10 GHz, wire bond 0.7–1.2 nH/mm below), substrate routing (50 Ω stripline on BT/LTCC), ball-to-PCB (ground ring coaxial geometry, antipad tuning). Via-in-pad eliminates stub resonance at fres = c/(4Lstub√εr).
Substrate materials?
BT resin: εr 3.5, tanδ 0.01, up to ~10 GHz (Wi-Fi/BLE/GPS). LTCC: εr 5–9, tanδ 0.002, through 77 GHz+ (radar, 5G mmW), embeds passives. HTCC: lowest loss (tanδ 0.0001), highest cost, mil/space. Megtron 6/7: organic alternative for 5G AiP at 28–39 GHz.