Why 5G Optimization Differs from LTE

5G NR optimization inherits many LTE principles but introduces new dimensions. Beamforming with Massive MIMO means that traditional sector-level tilt and power adjustments interact with beam codebook design. PCI planning must account for SSB beam indexing. RACH configuration spans more formats and is tied to beam correspondence. And the wider bandwidths (up to 100 MHz in FR1) make interference management more complex.

The optimization engineer's toolkit now includes: mechanical/electrical tilt, SSB beam configuration, PCI and PSS/SSS planning, PRACH format and root sequence selection, CIO and A3/A5 handover parameters, and carrier aggregation and BWP optimization.

Optimization Parameter Reference

PCI Planning: Mod-3 and Mod-30 Rules

Why PCI Planning Matters

In 5G NR, the Physical Cell Identity (PCI) ranges from 0 to 1007 and is derived from:

• PCI = 3 * N_ID_1 + N_ID_2
• Where N_ID_1 ranges from 0--335 and N_ID_2 ranges from 0--2

N_ID_2 determines the PSS (Primary Synchronization Signal) sequence -- there are only 3 PSS sequences. If two neighboring cells share the same N_ID_2 (i.e., PCI mod 3 is the same), their PSS sequences are identical, causing PSS collision. This degrades cell search performance and increases cell detection time.

Mod-3 Rule

Ensure that no two first-tier neighbors (cells whose coverage overlaps) share the same PCI mod 3 value. This guarantees distinct PSS sequences.

Example of violation: Site A (PCI 102), Site B (PCI 105), Site C (PCI 108)

• 102 mod 3 = 0, 105 mod 3 = 0, 108 mod 3 = 0 -- all three have the same PSS. This is a collision.
• Fix: Change Site B to PCI 106 (mod 3 = 1) and Site C to PCI 107 (mod 3 = 2).

Mod-30 Rule

The SSB DMRS sequence is derived from N_ID_cell (which equals PCI in most configurations). Cells with the same PCI mod 30 produce correlated DMRS sequences, which can degrade SSB DMRS detection. This is particularly impactful in dense deployments where multiple SSB beams from different cells overlap.

Rule: No first-tier or second-tier neighbors should share PCI mod 30.

PCI Planning Algorithm

1. Build a neighbor adjacency graph from drive test or MDT data.
2. Apply graph coloring with 3 colors for mod-3 constraint (first-tier).
3. Extend to 30 colors for mod-30 constraint (first + second tier).
4. Verify no co-PCI or mod-3/mod-30 conflicts using automated tools (e.g., Ericsson Network Manager, Nokia NetAct).
5. Additionally check that intra-site sectors (typically 3 sectors per site) use consecutive PCIs with different mod 3 values: e.g., PCI 300 (mod 3 = 0), 301 (mod 3 = 1), 302 (mod 3 = 2).

Worked Example: Tilt Optimization

Scenario: Site Alpha, sector 1, serves a suburban area with 500 m inter-site distance. Drive test shows overshooting -- the cell dominates beyond the intended 250 m radius, causing interference at neighboring Site Beta. Before optimization: Calculation -- Required tilt increase:

For a macro cell at height h = 30 m serving target distance d = 250 m:

• Current total tilt: 4 degrees
• Geometric tilt to 250 m: arctan(30/250) = 6.8 degrees
• Rule of thumb: Set total tilt = geometric tilt + 1--2 degrees for antenna pattern rolloff
• Target total tilt: 8 degrees
• Action: Increase electrical tilt from 2 to 6 degrees (total becomes 8 degrees)

After optimization:

The 4-degree electrical tilt increase reduced overshooting beyond 250 m by 14 dB, dramatically improving SINR at both the serving cell edge and the neighbor cell.

RACH Dimensioning

PRACH Format Selection

5G NR defines short and long PRACH preamble formats in TS 38.211 Table 6.3.3.1-1/2. Key considerations:

• Long preamble (Format 0-3): Supports cell radius up to 100 km (Format 3). Used for large cells.
• Short preamble (Format A1-C2): Lower overhead, supports up to 15--30 km radius. Preferred for urban deployments.
• Format B4: Commonly used for FR2 with 120 kHz SCS.

RACH Capacity Calculation

Worked example for an urban macro cell:

Given:

• PRACH configuration index: 27 (1 RACH occasion per frame, i.e., every 10 ms)
• Preambles per occasion: 64
• Contention-based preambles: 54 (10 reserved for contention-free HO)
• Target preamble collision probability: < 1%

Preamble collision probability for M simultaneous UEs attempting on N = 54 preambles:

P_collision = 1 - (1 - 1/N)^(M-1)

For P_collision < 0.01: M < 1 + ln(0.01) / ln(1 - 1/54) = 1 + (-4.605)/(-0.01887) = 1 + 244 = not realistic per occasion

The real constraint is RACH load per second. With 1 occasion per 10 ms = 100 occasions/second:

• RACH capacity = 100 * 54 = 5,400 preamble transmissions per second
• At 1% collision target with Poisson arrivals: practical limit is approximately 0.54 * 100 = 54 successful accesses per second per cell

For a cell with 500 connected users generating 0.05 RACH attempts/second/user = 25 RACH/sec -- well within capacity. But during mass events (stadium, concert), 5,000 UEs with 0.1 RACH/sec = 500 RACH/sec -- requires increasing PRACH occasions.

Solution: Change PRACH configuration index to support 10 RACH occasions per frame (10 * 54 = 540 preambles per 10 ms = practical capacity of 540 accesses per second).

Capacity vs Coverage Trade-Off

Real Deployment Data

Jio (India) optimized their n78 (3.5 GHz) macro network across 200,000+ sites using automated tilt optimization. Starting from planning defaults, the automated RET adjustment:

• Improved 95th percentile SS-SINR from 4.2 dB to 8.7 dB (+4.5 dB)
• Reduced inter-site interference by 6 dB on average
• Increased median cell throughput from 350 Mbps to 520 Mbps
• Required 3 optimization cycles over 8 weeks

SK Telecom applied ML-based PCI and PRACH replanning across their Seoul metropolitan n78 deployment (12,000 cells). The ML engine identified 847 mod-3 violations and 2,100 mod-30 violations that manual planning had missed. After correction:

• RACH success rate improved from 97.2% to 99.4%
• Cell detection time (from UE logs) decreased from 180 ms to 95 ms
• Handover failure rate dropped from 0.8% to 0.3%

Key Takeaway: 5G cell optimization is a multi-dimensional problem where tilt, power, PCI, and RACH parameters interact with Massive MIMO beam management. Start with PCI mod-3/mod-30 hygiene and geometric tilt calculations, then iterate using drive test data. Automated RET optimization consistently delivers 3--6 dB SINR improvement, and ML-based PCI replanning catches conflicts that manual methods miss. Always validate changes with post-optimization drive tests.