Purpose of 5G Drive Testing

Drive testing remains the primary method for validating real-world RF performance of 5G NR networks. While MDT (Minimization of Drive Tests) and crowdsourced data complement traditional approaches, controlled drive tests with calibrated equipment provide the ground truth that operators need for acceptance testing, interference hunting, and optimization validation.

5G NR introduces complexities that 4G drive testing did not face: beam-level measurements (SSB beams vs cell-level in LTE), FR2 measurements requiring specialized mmWave scanners, and EN-DC scenarios where LTE anchor and NR secondary cell must be measured simultaneously.

RF Parameter Reference

5G NR Drive Test KPI Thresholds

Note that SS-RSRP is measured per SSB beam index, not per cell as in LTE. The serving beam is the one with the highest SS-RSRP (SSB index = best beam). Beam-level reporting is defined in TS 38.331 Section 5.5.2, which specifies MeasObjectNR with nrofSS-BlocksToAverage.

Drive Test Tool Comparison

For FR2 (mmWave) measurements, both TEMS and Nemo require an external scanner such as the Rohde & Schwarz TSME6 or Keysight E7515B to capture beam-level SSB sweeps at 28 GHz or 39 GHz. The phone modem alone does not expose beam-level RF measurements at FR2 in most chipsets.

Drive Test Methodology

Pre-Test Checklist

1. Route planning: Define routes covering target cluster areas, sector boundaries, and known problem zones. Typical urban cluster size: 500 m x 500 m.
2. Equipment calibration: Verify GPS lock (HDOP < 2.0), phone firmware matches target network bands, scanner frequency list loaded.
3. Configuration: Set logging granularity to 100 ms or per-slot for detailed beam analysis. Enable SS-RSRP, SS-RSRQ, SS-SINR, CQI, BLER, throughput, and NAS/RRC event logging.
4. Reference conditions: Record time of day, weather (rain fade matters for FR2), and traffic conditions. Document antenna tilt/azimuth of target sites from network plan.
5. Idle + connected mode: Plan separate passes -- idle mode captures cell selection/reselection behavior; connected mode captures handover, throughput, and BLER.

During Test

• Maintain constant speed (30--50 km/h urban, 80--120 km/h highway) for consistent spatial sampling.
• Log GPS coordinates at every measurement sample -- aim for 1 sample per 3--5 m at urban speeds.
• Perform stationary measurements at sector boundaries and known weak spots (minimum 60 seconds per point).
• For throughput testing, use iPerf3 to a known server with TCP window size >= 4 MB and 8+ parallel streams to saturate the NR carrier.
• Record serving PCI, SSB index, and timing advance at each sample point.

Post-Test Processing

1. Export logs to CSV/KML format for GIS overlay.
2. Bin data into geographic grids (10 m x 10 m for urban, 50 m x 50 m for suburban).
3. Generate coverage plots: SS-RSRP, SS-SINR, throughput heatmaps.
4. Identify events: Handover failures (RLF with cause), RRC setup failures, abnormal RRC releases.
5. Cluster analysis: Group geographically adjacent poor-quality bins into optimization clusters.

Worked Example: Identifying a Coverage Hole

Scenario: Operator X deploys 5G NR at 3.5 GHz (n78) in a dense urban area. Drive test data shows a throughput drop to < 5 Mbps DL along a 200 m stretch of Main Street. Step 1 -- Extract RF data for the affected segment: `

Location: Main Street, bins 47--62

SS-RSRP range: -108 to -118 dBm (poor)

SS-SINR range: -3 to +2 dB (poor)

Serving PCI: 312 (SSB index 4)

Neighbor PCI: 318 (SSB index 2), SS-RSRP = -95 dBm

CQI: 3--5

BLER: 18--25%

` Step 2 -- Root cause analysis:

• SS-RSRP below -100 dBm indicates coverage hole -- signal strength is insufficient.
• SS-SINR of -3 dB with neighbor at -95 dBm and server at -112 dBm indicates the neighbor is 17 dB stronger -- this suggests the UE is camped on the wrong cell or handover is delayed.
• Check A3 event configuration: if A3 offset = 3 dB and TTT = 640 ms, the UE waits too long before handover. At 40 km/h, the UE travels 7.1 m during TTT, but the coverage hole is 200 m -- TTT is not the primary issue here.
• The real issue: buildings on Main Street create a shadow zone for PCI 312. The sector antenna (azimuth 120 degrees) faces away from this stretch.

Step 3 -- Recommended actions:

Cluster Analysis Methodology

Cluster analysis groups spatially correlated problem areas for prioritized optimization. The standard approach:

1. Define quality gates: For acceptance, typical thresholds are SS-RSRP > -100 dBm for 95% of samples and SS-SINR > 5 dB for 90% of samples.
2. Flag failing bins: Any 10 m x 10 m bin where the median measurement falls below the threshold is flagged.
3. Spatial clustering: Use DBSCAN (density-based spatial clustering) with epsilon = 50 m and min_samples = 5. This groups nearby failing bins into clusters.
4. Prioritize clusters: Rank by (a) number of affected bins, (b) traffic weight from subscriber density data, (c) severity (how far below threshold).
5. Root cause tagging: Assign each cluster a category: coverage hole, pilot pollution, overshooting, missing neighbor, handover failure, or interference.

Real-world example: Vodafone Germany applied this methodology across 15,000 drive test km in their n78 rollout. They identified 847 clusters, of which 62% were coverage holes resolved by tilt optimization, 21% were pilot pollution requiring power reduction, and 17% required new small cell sites. Post-optimization drive tests showed 95th percentile SS-RSRP improved from -104 dBm to -91 dBm.

T-Mobile US reported in their 2024 network update that systematic drive test analysis across 50 major cities on n41 (2.5 GHz) reduced handover failure rate from 1.8% to 0.4% and improved median DL throughput from 180 Mbps to 290 Mbps through iterative cluster-based optimization.

FR2 (mmWave) Drive Testing Specifics

FR2 drive testing requires additional considerations:

• Beam tracking latency: At 28 GHz with 120 kHz SCS, SSB sweeps occur every 20 ms. Scanner must capture all 64 SSB beams within the sweep period.
• Rain fade: At 28 GHz, rain attenuation is approximately 7 dB/km at 25 mm/h rain rate (ITU-R P.838). Test during dry conditions for baseline, then separately characterize rain impact.
• Body blockage: Human body attenuates 28 GHz by 20--35 dB. The scanner antenna must have clear line-of-sight to the gNB sector. Vehicle-mounted external antennas are mandatory.
• Beam index logging: FR2 measurements must log SSB beam index (0--63) and CSI-RS resource indicator to correlate beam quality with gNB beam codebook entries.

Common Mistakes to Avoid

• Logging at 1-second granularity -- NR beam sweeps happen every 20 ms; 1-second logging misses beam dynamics entirely.
• Testing only in connected mode -- Idle mode reselection problems cause call setup failures that connected-mode tests never reveal.
• Ignoring UL performance -- PUSCH power headroom, PRACH attempts, and UL SINR are critical for VoNR and low-latency services.
• Not correlating with MDT -- Drive test data should be cross-validated with UE-reported MDT data per TS 37.320 to verify spatial accuracy.

Key Takeaway: 5G drive testing demands beam-level measurement granularity, FR2-specific procedures, and systematic cluster analysis. Using calibrated tools at 100 ms logging intervals, binning data into geographic grids, and applying DBSCAN clustering transforms raw logs into actionable optimization tasks. Operators who combine drive test ground truth with MDT data consistently achieve the fastest network quality improvements.