Are there any “gold parameter” optimization tips for high-capacity LTE/NR events to improve AFR, throughput, and NSA/SA performance without using the typical coverage-reduction methods (such as changing QrxlevMin, CRS gain, or antenna tilts)? I’m looking for parameter optimizations that increase capacity while minimizing coverage degradation.
1 Like
Most “event optimization” advice is just coverage shrinking in disguise — and shrinking coverage at a stadium usually moves the problem to the next cell instead of solving it. The real gold at high-capacity events is almost never spectral efficiency. It’s the control plane. When 30,000 people hit a few sectors, you run out of RACH, PDCCH, PUCCH and RRC capacity long before you run out of PRBs. So here’s where I’d actually look, Ericsson-flavored:
1. Connection lifecycle — the cheapest capacity you’ll ever find. Shorten the UE inactivity timer (tInactivityTimer) for the event. Default values keep users sitting in connected mode holding PUCCH/SR resources doing nothing. Cutting it frees RRC and PUCCH capacity instantly. Trade-off: more RRC setups and paging, so watch your RACH load — these two levers fight each other, tune them together.
2. PDCCH — the invisible throughput killer. At events, cells often show mediocre throughput with plenty of free PRBs, and everyone blames radio. Check CCE utilization first. If PDCCH is congested, the scheduler physically can’t grant what the air interface could carry. Make sure CFI can reach 3 symbols (pdcchCfiMode set to auto/max) rather than pinned low for capacity you’re not using.
3. PUCCH/SR/CQI dimensioning. Default SR and CQI user dimensioning is built for normal cells, not events. Undersized PUCCH shows up as access failures and scheduling latency that look like RF problems. Increase the SR/CQI user allocations on the event sectors.
4. RACH. More PRACH opportunities (prachConfigIndex) for the density, and re-verify root sequence planning — high UE density plus tight reuse gives you ghost preambles that inflate your AFR with failures that never came from a real user.
5. UL power control instead of tilts. A dense crowd raises the uplink noise floor for everyone. Retuning p0/alpha (slightly lower pZeroNominalPusch, higher alpha) reduces IoT and often lifts aggregate UL throughput — that’s interference management without touching a single coverage parameter.
6. Spread before you optimize. IFLB thresholds (lbThreshold/lbCeiling) tightened for the event, plus idle-mode distribution at release so UEs land balanced across layers before the rush. A perfectly tuned congested cell still loses to a mediocre balanced one.
7. NSA specifics: get UEs onto the NR leg faster (slightly more generous B1) — every GB on NR is load off your LTE anchor. And keep the anchor layer clean: anchoring on your most congested LTE carrier is self-sabotage; if you have the layers, dedicate the cleanest one as anchor.
8. Break-glass only: access class barring and admission control staged and ready, not active. That’s damage control, not optimization.
One honest caveat: exact attribute names, ranges and defaults shift between L-releases, so verify each against ALEX/CPI for your release before the change window — never tune from a forum post, mine included. If you want, send your release, layer setup and the event sector list to fkamanzi@gmail.com and I’ll help you turn this into a CPI-verified parameter set with a rollback plan — events are unforgiving, and you want the rollback written before the gates open.
2 Likes
Thank you!