How do I design a millimeter wave front end module for a 5G user equipment device?
5G UE mmWave FEM Design
The mmWave FEM is the most critical module in a 5G UE device. Its performance directly determines the device's data rate, range, and battery life.
Module Architecture
(1) Single-chip RFIC + separate FEM: the transceiver RFIC (e.g., Qualcomm QTM series) integrates the LO, mixer, phase shifters, and baseband interface. The FEM (a separate die or module) contains the PA, LNA, and switch. The two are connected by transmission lines on the AiP substrate. Advantage: flexibility (the FEM can be from a different vendor or technology than the RFIC). Disadvantage: additional package complexity and interconnect loss. (2) Fully integrated RFIC: the PA, LNA, switch, and transceiver are on a single SiGe or SOI CMOS die. Advantage: lowest cost, smallest size, and no inter-die transitions. Disadvantage: compromise in PA performance (CMOS/SiGe PAs have lower efficiency than GaAs or GaN at mmWave). Qualcomm QTM525 uses this approach (fully integrated in SiGe). (3) Multi-chip module (MCM): the RFIC and FEM are separate dies bonded into the same module (on the same substrate). The inter-die connections are made by short traces on the substrate (< 2 mm). This combines the advantage of a specialized FEM technology (GaAs PA for higher efficiency) with the integration of the transceiver RFIC.
Power and Thermal
(1) Power consumption: the FEM power budget per module: PA (TX mode): 100-400 mW per element at peak output. For 4 elements: 400-1600 mW. LNA (RX mode): 10-30 mW per element. For 4 elements: 40-120 mW. Switch + phase shifter: 5-20 mW per element. Total per module: TX mode: 500-1800 mW. RX mode: 60-180 mW. For a phone with 3 modules (only 1 active at a time): maximum power = 1.8 W (TX) or 0.2 W (RX). (2) Battery impact: at full TX power: the mmWave FEM draws 5× more power than a sub-6 GHz PA (which is why 5G mmWave drains the battery faster). The 5G NR specification manages this: discontinuous TX (the UE only transmits when it has data), TX power control (the gNB commands the UE to reduce power when close), and beam management (the UE activates only the minimum number of modules needed for the link). (3) Thermal: the mmWave modules are located at the phone edge (near the antenna). The heat must be conducted to the phone chassis through the PCB copper and a thermal pad. Without adequate thermal management: the PA performance degrades (gain decreases 0.02-0.04 dB/°C, PAE drops) and the phone surface becomes uncomfortably warm.
P_sens = -174 + 10log₁₀(BW) + NF dBm
4-element EIRP: +16 to +21 dBm
PA per element: P1dB ≈ +15 dBm
FEM power: TX 500-1800mW, RX 60-180mW
Frequently Asked Questions
SiGe or CMOS for the FEM?
SiGe BiCMOS: higher fT/fmax (> 300 GHz), better PA efficiency (+3-5% PAE advantage at 28 GHz), and a more mature mmWave design library. Most current 5G mmWave modules use SiGe (Qualcomm, Samsung, Murata). SOI CMOS (advanced RF SOI): lower cost at very high volumes. Improving rapidly. The PA efficiency gap with SiGe is narrowing with each technology node. Apple and some Chinese OEMs are evaluating CMOS-based mmWave FEMs for cost reduction. GaAs HBT: highest PA efficiency at 28 GHz (PAE ≈ 25-30% vs 15-20% for SiGe). But: requires a separate die (not integrated with the transceiver), adding module complexity and cost. Used in some high-performance gNB (base station) FEMs where efficiency is critical for reducing the heat load in large arrays.
How is the phase shifter implemented?
In the RFIC or FEM die: (1) Switched-line phase shifter: multiple transmission line segments of different lengths, selected by switches. Each bit selects between two path lengths (Δ length = Δ phase / (360° / wavelength)). For 5-bit at 28 GHz: the smallest bit (11.25°) requires ΔL ≈ 0.17 mm (very short, but repeatable with IC lithography). Advantage: wideband (the phase shift is approximately constant vs frequency, since it is based on physical length). (2) Loaded-line / reflective-type phase shifters: use varactors or switched capacitors to change the transmission line phase velocity. More compact than switched-line. (3) Vector modulator: decomposes the signal into I and Q components, applies variable gain to each, and recombines. Provides continuous phase control (infinite resolution) and simultaneous amplitude control. Used in advanced beamforming ICs (Analog Devices ADAR1000, Anokiwave AWMF series).
What 3GPP power class applies to 5G UE mmWave?
Power Class 1: EIRP_max = +43 dBm (for fixed wireless access CPE devices). Power Class 2: EIRP_max = +35 dBm (for vehicle-mounted UE). Power Class 3: EIRP_max = +23 dBm (for handheld UE, the standard smartphone power class). Power Class 4: EIRP_max = +10 dBm (for IoT/wearable devices). For a smartphone (PC3) at 28 GHz with 3 modules (one active): each module must provide up to +23 dBm EIRP. With 4 elements at +6 dBi: PA output per element = +23 - 6 = +17 dBm (before array gain). With 8 elements at +9 dBi: PA output per element = +23 - 9 = +14 dBm. More elements = lower PA power requirement per element = lower cost and power consumption.