Millimeter Wave Specific Challenges 5G and mmWave Communications Informational

What are the linearity requirements for a 5G NR millimeter wave power amplifier?

5G NR mmWave power amplifiers must meet stringent linearity requirements because of the wide channel bandwidths and high-order modulation used. Key specifications: (1) ACLR (Adjacent Channel Leakage Ratio): the ratio of power in the desired channel to the power leaked into the adjacent channel. 3GPP requirement: ACLR > 28 dB for the first adjacent channel (for FR2 BS). ACLR > 26 dB for UE. In practice: operators require ACLR > 33-40 dB (the 3GPP minimum is insufficient for clean network operation). (2) EVM (Error Vector Magnitude): the distortion of the modulated signal constellation. 3GPP EVM limits: QPSK: < 17.5%. 16-QAM: < 12.5%. 64-QAM: < 8%. 256-QAM: < 3.5%. The PA must maintain these EVM levels at the rated output power. For 256-QAM (the highest modulation in 5G NR): the PA must be extremely linear (3.5% EVM corresponds to approximately P1dB_out - 6 to -8 dB backoff). (3) PAPR (Peak-to-Average Power Ratio): 5G NR uses CP-OFDM (same as LTE) with high PAPR. For a 400 MHz channel with 264 subcarriers: PAPR = 8-12 dB (depending on the waveform, number of carriers, and CCDF probability). The PA must operate at an average power 8-12 dB below its saturated output to avoid clipping the signal peaks. This means: for a required average output power of +30 dBm: the PA must have P_sat > +38 to +42 dBm (the PA is significantly oversized compared to the average power). (4) In-band emissions: the spectral flatness of the transmitted signal within the channel bandwidth. 3GPP specifies < ±5 dB ripple across the channel (EVM-based floor). The PA AM-AM (gain compression) and AM-PM (phase shift with power) distortion degrade the in-band flatness, especially at the channel edges.

Category: Millimeter Wave Specific Challenges

Updated: April 2026

Product Tie-In: 5G Components, Phased Arrays, Front End Modules

mmWave PA Linearity

Linearity is the dominant design challenge for mmWave PAs in 5G NR systems, directly trading off against efficiency and output power.

Technical Considerations

(1) Class A PA: most linear (IP3 > P1dB + 10 dB). Maximum efficiency: 50% (theoretical) at P_sat, but at the 8-10 dB backoff needed for 5G NR: efficiency drops to 5-10%. At 28 GHz with GaAs pHEMT: PAE = 5-8% at the backed-off average power. Unsuitable for most applications due to excessive power consumption and heat. (2) Class AB PA: moderate linearity (IP3 > P1dB + 8 dB). Maximum efficiency: 35-45% at P_sat. At backoff: PAE = 8-15%. Most common for mmWave 5G PAs. GaAs: PAE = 10-20% at backed-off average power. SiGe: PAE = 5-12%. CMOS: PAE = 3-8%. (3) Doherty PA: a load modulation technique that improves efficiency at backoff. Two PAs (main + auxiliary): the main PA operates at full power, and the auxiliary PA activates at peaks to extend the dynamic range. At 6 dB backoff: a Doherty PA achieves 25-35% PAE (vs 10-15% for Class AB). At mmWave: Doherty designs are challenging due to the tight tolerances on the quarter-wave impedance inverter (lambda/4 at 28 GHz = 2.7 mm). GaAs and GaN Doherty PAs at 28 GHz have been demonstrated with 25-30% PAE at 8 dB backoff. (4) Outphasing (Chireix) PA: two constant-envelope PAs with output signal combining. Theoretically maintains high efficiency across all power levels. Very complex at mmWave (requires precise amplitude and phase control at 28 GHz). Research-stage for mmWave 5G.

Performance Analysis

DPD is the standard technique for linearizing sub-6 GHz PAs, but its application at mmWave faces unique challenges: (1) Bandwidth: DPD requires the observation bandwidth (feedback path) to be 3-5× the signal bandwidth. For a 400 MHz 5G NR signal: DPD observation BW = 1.2-2.0 GHz. The feedback ADC must sample at > 2.4-4.0 Gsps with 10+ ENOB. This is extremely demanding (ADCs with > 3 Gsps and > 10 ENOB are expensive and power-hungry). (2) Processing: the DPD algorithm (memory polynomial or Volterra series) must process 3-5× the signal bandwidth in real time. For 2 GHz observation BW: approximately 100 GMAC/s computational load. Requires dedicated DPD hardware (ASIC or FPGA). (3) Observation path: the mmWave signal must be downconverted and digitized for DPD processing. The observation receiver adds cost, power, and complexity to the PA module. At sub-6 GHz: DPD is standard (all base station PAs use it). At mmWave: DPD is used in base stations but not in UE (the UE PA power is too low to justify the DPD overhead). UE mmWave PAs operate with sufficient backoff to meet linearity without DPD.

Performance verification: confirm specifications against the application requirements before finalizing the design
Environmental factors: temperature range, humidity, and vibration affect long-term reliability and parameter drift
Cost vs. performance: evaluate whether the application demands premium components or standard commercial grades

Interface compatibility: verify impedance, connector type, and mechanical form factor match the system architecture
Margin allocation: include sufficient design margin to account for manufacturing tolerances and aging effects

Design Guidelines

(1) GaAs pHEMT: the mature technology for 28/39 GHz PAs. P_sat: 20-28 dBm (100-600 mW per die). PAE at P_sat: 25-40%. f_T: 80-150 GHz. Used in: UE front-end modules (Qualcomm QTM, Murata). (2) GaN HEMT: higher power than GaAs. P_sat: 30-40 dBm (1-10 W per die) at 28 GHz. PAE at P_sat: 30-45%. Rugged (can withstand VSWR mismatch). Used in: base station PAs (Wolfspeed, MACOM, Qorvo). GaN is replacing GaAs for BS mmWave PA due to higher power and efficiency. (3) SiGe BiCMOS: integrated with transceiver on the same die (or same process). P_sat: 15-20 dBm (30-100 mW). PAE: 15-25% at P_sat. Lower power than GaAs/GaN but more integrated. Used in: integrated transceiver SoCs with on-chip PA. (4) CMOS (14-28 nm): lowest power but most integrated. P_sat: 10-15 dBm (10-30 mW per PA cell). PAE: 10-20% at P_sat. Requires power combining of many PA cells to achieve useful output power. Used in: radar SoCs, research-stage 5G SoCs.

Common Questions

Frequently Asked Questions

Why is DPD harder at mmWave than at sub-6 GHz?

Three reasons: (1) Wider bandwidth: at sub-6 GHz: 5G NR channels are 20-100 MHz. DPD observation BW: 60-500 MHz (feasible with standard ADCs). At mmWave: channels are 100-400 MHz. DPD observation BW: 300 MHz - 2 GHz (requires very high-speed, high-resolution ADCs that are expensive and power-hungry). (2) More elements: a mmWave BS has 64-256 PA elements. Each may need its own DPD (group DPD can reduce this, but calibration is still needed per element). At sub-6 GHz: typically 4-32 PA elements. (3) Antenna-PA integration: at mmWave, the PA is often integrated with the antenna in the same package (AiP). Adding a coupler for DPD feedback is physically challenging at the small wavelengths. Alternative: beam-space DPD (observe the combined far-field output with a single receiver, apply DPD in the digital beamforming domain). This is an active research area.

What efficiency can I expect from a 5G mmWave PA?

At the backed-off average power (accounting for PAPR and linearity requirements): UE PA (GaAs, +15 to +20 dBm average): PAE = 10-18% (no DPD). BS PA (GaN, +25 to +30 dBm average per element): PAE = 15-25% (with DPD and Doherty). These are significantly lower than sub-6 GHz PAs (which achieve 30-50% PAE with DPD at similar backoff) because: (1) The active device (transistor) has lower gain at mmWave, requiring more power from earlier stages. (2) The matching network losses are higher (0.5-1.5 dB at 28 GHz vs 0.1-0.3 dB at 2 GHz). (3) The power combining losses for multi-cell PAs are higher at mmWave.

Can I use envelope tracking at mmWave?

Envelope tracking (ET) modulates the PA supply voltage to track the signal envelope, improving efficiency at backoff. At sub-6 GHz: ET is standard in handset PAs (Qualcomm QET, Qorvo ET products), providing 5-10% PAE improvement. At mmWave: ET is much more challenging because: (1) The signal bandwidth is much wider (100-400 MHz vs 20-100 MHz). The ET modulator must track the envelope at 2-3× the signal bandwidth (200-1200 MHz tracking bandwidth). This requires an extremely fast, high-efficiency voltage modulator. (2) The PA supply current is lower (the PA power is distributed across many elements). The ET modulator efficiency advantage decreases for lower currents. (3) The interconnect inductance between the ET modulator and the PA (even 0.5 nH) creates supply ripple at the modulator switching frequency. Status: ET at mmWave is in research/early development. Not yet commercially deployed. Doherty provides a simpler path to improved backoff efficiency at mmWave.

Keep Reading