Test and Measurement Equipment Instrument Selection Informational

What is the phase noise requirement for a signal generator used to test a 5G NR device?

Q: Can I test 5G NR with a budget signal generator?

For QPSK and 16QAM (most IoT and data channels): yes. The EVM requirement is relaxed (12.5-17.5%), and budget generators ($5,000-20,000) provide sufficient EVM. For 64QAM (most practical 5G deployments): mid-range generators ($20,000-60,000) are adequate. For 256QAM (high-throughput FR1/FR2): only high-end generators ($80,000+) provide the required EVM and phase noise. Match the generator to the modulation order you need to test.

Q: How do I compensate for high generator phase noise?

Option 1: average multiple measurements (the phase noise contribution averages out over time, improving EVM measurement accuracy). Option 2: use a reference loop (lock the DUT and generator to the same reference oscillator to cancel common-mode phase noise). Option 3: post-processing correction (measure the generator EVM separately and subtract it from the total measured EVM). Option 4: reduce the test bandwidth (narrower BW integrates less phase noise, improving EVM). These are workarounds; the preferred approach is to use a generator with sufficiently low phase noise.

Q: What about residual EVM from IQ impairments?

In addition to phase noise, the generator EVM includes: IQ gain imbalance: < 0.1 dB (high-end), 0.2-0.5 dB (mid-range). IQ phase imbalance: < 0.3° (high-end), 0.5-2° (mid-range). DC offset: < -50 dBc (high-end), -30 to -40 dBc (mid-range). DAC quantization noise: < -55 dB (14-bit DAC), < -48 dB (12-bit). For 256QAM at 5G: all of these must be tightly controlled. The generator residual EVM (all impairments combined) is typically: high-end: 0.3-0.5% (-46 to -50 dB). Mid-range: 0.8-2.0% (-34 to -42 dB). Entry-level: 2-5% (-26 to -34 dB).

What is the phase noise requirement for a signal generator used to test a 5G NR device? The signal generator phase noise directly contributes to the measured EVM of the device under test, so the generator must have significantly lower phase noise than the DUT specification allows: (1) 5G NR EVM requirements by modulation: QPSK: 17.5% EVM (-15.1 dB). 16QAM: 12.5% EVM (-18.1 dB). 64QAM: 8% EVM (-21.9 dB). 256QAM: 3.5% EVM (-29.1 dB). The generator EVM (including phase noise) must be at least 10 dB better than the DUT requirement to avoid corrupting the measurement. For 256QAM testing: generator EVM must be < 1.1% (-39 dB). (2) Phase noise contribution to EVM: the integrated phase noise (IPN) over the signal bandwidth contributes directly to the EVM: EVM_phase_noise (%) ≈ IPN_rms_degrees × π / 180. For a 256QAM test: maximum allowable EVM from phase noise ≈ 0.5% → IPN < 0.3° RMS. The integration range: from 100 Hz (or 1 kHz) to the signal bandwidth (e.g., 100 MHz for 5G NR FR1). (3) Phase noise specification at key offsets: at 10 kHz offset from carrier: requirement depends on carrier frequency. At 3.5 GHz (FR1 n78): generator should have < -110 dBc/Hz at 10 kHz. At 28 GHz (FR2 n257): generator should have < -95 dBc/Hz at 10 kHz. At 100 kHz offset: FR1: < -120 dBc/Hz. FR2: < -105 dBc/Hz. At 1 MHz offset: FR1: < -130 dBc/Hz. FR2: < -115 dBc/Hz. These values ensure the integrated phase noise contribution is < 0.5% EVM. (4) Phase noise scaling with frequency: phase noise degrades by approximately 20 log(f2/f1) when the carrier frequency increases (for generators using frequency multiplication). Example: a generator with -120 dBc/Hz at 10 kHz offset at 1 GHz will have approximately -120 + 20 log(28) = -120 + 29 = -91 dBc/Hz at 28 GHz. This may be insufficient for 256QAM testing at 28 GHz. High-end generators use low-division-ratio PLL architectures and DDS to minimize this degradation. (5) Signal generator phase noise specs: Keysight M9384B VXG: -121 dBc/Hz at 10 kHz offset at 10 GHz (excellent). R&S SMW200A: -137 dBc/Hz at 10 kHz offset at 1 GHz, scales to about -108 at 28 GHz (excellent). Keysight E8267D PSG: -110 dBc/Hz at 10 kHz offset at 10 GHz. Lesser generators: -95 to -105 dBc/Hz at 10 kHz offset at 10 GHz (may be insufficient for 256QAM at FR2).

Category: Test and Measurement Equipment

Updated: April 2026

Product Tie-In: VNAs, Spectrum Analyzers, Signal Generators

5G Signal Generator Phase Noise

Phase noise is often the single most important specification when selecting a signal generator for 5G NR testing, particularly at FR2 frequencies where the phase noise naturally degrades.

EVM Budget Analysis

(1) The total EVM measured at the DUT output is a combination (RSS) of: DUT EVM (the actual device performance). Generator EVM (including phase noise, IQ errors, DAC quantization). Measured EVM = sqrt(DUT_EVM² + Generator_EVM²). For the generator to not significantly affect the measurement: Generator_EVM < DUT_EVM / 3 (contributes < 5% to the total measured EVM). For 256QAM (DUT EVM 3.5%): Generator EVM < 1.2%. This is achievable with high-end instruments but challenging with mid-range equipment. (2) Phase noise measurement: verify the generator phase noise using a spectrum analyzer with phase noise measurement capability. Measure at the actual test frequency (not a lower frequency with scaling). Compare the measured integrated phase noise to the EVM budget requirement. If the generator phase noise is too high: reduce the carrier frequency if possible (test at IF with a downconverter for DUT), or use a higher-quality generator.

Phase Noise Requirements

256QAM: 3.5% EVM; gen must be < 1.1% (-39 dB)
IPN < 0.3° RMS for 256QAM margin
Phase noise at 3.5 GHz: < -110 dBc/Hz @ 10 kHz
Phase noise at 28 GHz: < -95 dBc/Hz @ 10 kHz
Scaling: +20 log(f₂/f₁) degradation

Common Questions

Frequently Asked Questions

Can I test 5G NR with a budget signal generator?

For QPSK and 16QAM (most IoT and data channels): yes. The EVM requirement is relaxed (12.5-17.5%), and budget generators ($5,000-20,000) provide sufficient EVM. For 64QAM (most practical 5G deployments): mid-range generators ($20,000-60,000) are adequate. For 256QAM (high-throughput FR1/FR2): only high-end generators ($80,000+) provide the required EVM and phase noise. Match the generator to the modulation order you need to test.

How do I compensate for high generator phase noise?

Option 1: average multiple measurements (the phase noise contribution averages out over time, improving EVM measurement accuracy). Option 2: use a reference loop (lock the DUT and generator to the same reference oscillator to cancel common-mode phase noise). Option 3: post-processing correction (measure the generator EVM separately and subtract it from the total measured EVM). Option 4: reduce the test bandwidth (narrower BW integrates less phase noise, improving EVM). These are workarounds; the preferred approach is to use a generator with sufficiently low phase noise.

What about residual EVM from IQ impairments?

In addition to phase noise, the generator EVM includes: IQ gain imbalance: < 0.1 dB (high-end), 0.2-0.5 dB (mid-range). IQ phase imbalance: < 0.3° (high-end), 0.5-2° (mid-range). DC offset: < -50 dBc (high-end), -30 to -40 dBc (mid-range). DAC quantization noise: < -55 dB (14-bit DAC), < -48 dB (12-bit). For 256QAM at 5G: all of these must be tightly controlled. The generator residual EVM (all impairments combined) is typically: high-end: 0.3-0.5% (-46 to -50 dB). Mid-range: 0.8-2.0% (-34 to -42 dB). Entry-level: 2-5% (-26 to -34 dB).

Keep Reading