4G LTE & 5G NR Cell Capacity & Subscriber Dimensioning Calculator

The Paradigm Shift: From Erlangs (Voice) to Megabits (Mobile Broadband)

In legacy second-generation (2G GSM) and third-generation (3G UMTS) cellular systems, radio dimensioning was strictly deterministic: one voice call occupied exactly one time slot or one orthogonal spreading code. Network engineers sized trunk groups and cellular sectors using the classic Erlang B or Erlang C queuing formulas formulated by Danish mathematician Agner Krarup Erlang in 1909. In that circuit-switched paradigm, blocking probability (“Grade of Service”) was the singular design metric: if all eight time slots on a GSM carrier were occupied, the ninth incoming call was rejected immediately.

Modern 4G LTE-Advanced (3GPP Rel 10–14) and 5G New Radio (3GPP Rel 15–18) operate on a fundamentally different physical foundation. Voice, video, augmented reality, and background telemetry are entirely packet-switched over Orthogonal Frequency Division Multiple Access (OFDMA). Radio resources are not locked to individual users for minutes; instead, the cellular scheduler inside the gNodeB baseband distributes physical resource blocks (PRBs) every Transmission Time Interval (TTI) of 1.0 ms (LTE) down to 0.5 ms or 0.25 ms (5G numerologies μ=1 and μ=2).

Deconstructing the Busy Hour Subscriber Traffic Model

Cellular networks are dimensioned for the peak continuous 60-minute window of user activity, known as the Busy Hour (BH). To translate a monthly billing plan quota (e.g., 30 Gigabytes per subscriber per month) into an active physical layer bit rate, engineers apply standard 3GPP traffic conversion models:

Where:

• D_month: Average monthly data consumption per active subscriber in Gigabytes (GB).
• 8 × 1000: Conversion factor from Gigabytes to Megabits (using decimal telecommunications notation: 1 GB = 8,000 Megabits).
• K_BH: Busy Hour concentration factor (typically 0.08 to 0.12, representing that 8% to 12% of a subscriber's daily data transfer occurs during the single peak busy hour).
• D_days: Effective working or billing days per calendar month (typically 25 to 30 days depending on whether weekend traffic drops or rises).
• 3600: The duration of the busy hour in seconds.

For a typical 30 GB/month subscriber with K_BH = 10% and D_days = 25 days:
R_{sub_raw} = (30 × 8,000 × 0.10) / (25 × 3,600) = 24,000 Mb / 90,000 s = 0.2667 Mbps = 266.7 kbps.

However, human users do not stream at a constant 266 kbps for 60 minutes straight. A user downloading a webpage or buffering a short video clip pulls data at 25 to 100 Mbps for 1.5 seconds, then reads or watches for 15 to 45 seconds while generating zero network traffic. To model this statistical burstiness, capacity planners multiply the raw busy hour rate by an Active Packet Concurrency Factor (typically 15% to 20%), yielding an effective continuous busy hour load of 40 to 50 kbps per active provisioned subscriber.

3GPP Physical Layer Overheads & TDD Duty Cycle Impact

Equipment vendors frequently promote theoretical peak cell throughputs calculated at maximum 256-QAM modulation, maximum MIMO rank (e.g., 8 spatial layers), and 0% control overhead (claiming “1.5 Gbps per 100 MHz sector”). In practical operational deployments, real-world cell capacity is substantially lower due to three fundamental physical constraints:

• Layer 1 / Layer 2 Control Plane Overheads: Critical synchronization signals (PSS, SSS), Primary Broadcast Channel (PBCH), Demodulation Reference Signals (DMRS), Channel State Information (CSI-RS), Physical Downlink Control Channel (PDCCH), and PUCCH scheduling grants consume between 18% and 24% of all available Resource Elements (REs) in the time-frequency grid.
• TDD Frame Structure Slot Allocation: In 5G Time Division Duplexing (TDD) bands such as C-Band (3.5 GHz n78) and 2.5 GHz n41, the base station must periodically halt downlink transmission to permit mobile terminal uplink transmissions. Under the industry standard 3GPP Frame Structure (2.5 ms periodicity, 4:1 Downlink-to-Uplink slot ratio), the downlink transmitter is active for only 74.3% of total frame time. Under a 3:1 ratio, downlink duty cycle drops to 65.2%.
• Cell Utilization Headroom (The Bufferbloat Boundary): Operating a cell at 95% to 100% continuous resource block utilization creates catastrophic queue buildup in the gNodeB RLC/MAC buffers. Round-trip ping latency skyrockets from 12 ms to over 400 ms (bufferbloat), TCP window growth collapses due to packet drops, and user throughput degrades exponentially. Telecom operators enforce an engineering ceiling of 70% to 80% maximum utilization to ensure pristine Quality of Experience (QoE).

User Plane vs. Control Plane Bottlenecks (RRC Connected Limits)

A common mistake in RAN dimensioning is assuming that radio frequency (RF) bandwidth is the only limit on cell capacity. In high-density environments (such as stadiums, transportation hubs, or shopping districts), the baseband processing hardware frequently runs out of Radio Resource Control (RRC) Connection capacity long before the air interface spectrum is saturated.

Every mobile phone attached to a cell exists in one of three 3GPP RRC states:

1. RRC_IDLE: The device is registered in the core network (AMF/MME) and listens for paging messages, but maintains no dedicated signaling or data bearers with the gNodeB. Idle users consume zero gNodeB baseband scheduling resources.
2. RRC_INACTIVE: Introduced in 5G NR, this intermediate state allows the UE to suspend user plane bearers while retaining gNodeB context, enabling sub-10 ms resumption without full core signaling.
3. RRC_CONNECTED: The device has established Signaling Radio Bearers (SRB1/SRB2) and Data Radio Bearers (DRB). The gNodeB must allocate dedicated memory, track channel state feedback (CQI, PMI, RI), and schedule periodic SRS sounding.

Commercial macro baseband units (BBUs) feature hardware and software license limits typically capped at 1,000 to 2,000 concurrent RRC_CONNECTED sessions per sector. If 20,000 users in a stadium generate background push notifications that trigger RRC transitions simultaneously, the base station will reject connection attempts even if the 100 MHz RF channel has 80% unused PRBs.