PaperlayerView exact revision on arXiv

Source-linked AI summary

Towards 1 Gbps/UE in Cellular Systems: Understanding Ultra-Dense Small Cell Deployments

David Lopez-Perez, Ming Ding, Holger Claussen, Amir H. Jafari

arXiv:1503.03912v1cs.NI

TL;DR

Projected traffic demands require substantially greater network capacity, motivating analysis of densification, higher-frequency bands, and multi-antenna techniques. The paper evaluates their gains and limitations alongside idle mode, UE density, scheduling, and energy considerations. Simulations find the largest cell-edge throughput gains from densification, while practical cost-effectiveness and energy efficiency remain unresolved.

  • Problem

    Future services may require up to 1 Gbps per UE, while networks are targeted to achieve 100× or more capacity growth.

  • Method

    The paper analyses network densification, higher frequency bands, and multi-antenna techniques together with idle mode, UE density, scheduling, and energy-efficiency effects.

  • Results

    Network densification provides cell-edge UE throughput gains up to 48x, higher frequency bands up to 5x network-capacity gains, and beamforming around 1.49x cell-edge gains.

  • Takeaways & Limitations

    An average 1 Gbps per UE is possible at roughly 35 m ISD, 250 MHz bandwidth, and four antennas per small-cell BS, but deployment cost-effectiveness and energy efficiency remain challenges.

Abstract

from arXiv · show

Todays heterogeneous networks comprised of mostly macrocells and indoor small cells will not be able to meet the upcoming traffic demands. Indeed, it is forecasted that at least a 100x network capacity increase will be required to meet the traffic demands in 2020. As a result, vendors and operators are now looking at using every tool at hand to improve network capacity. In this epic campaign, three paradigms are noteworthy, i.e., network densification, the use of higher frequency bands and spectral efficiency enhancement techniques. This paper aims at bringing further common understanding and analysing the potential gains and limitations of these three paradigms, together with the impact of idle mode capabilities at the small cells as well as the user equipment density and distribution in outdoor scenarios. Special attention is paid to network densification and its implications when transitioning to ultra-dense small cell deployments. Simulation results show that network densification with an average inter site distance of 35 m can increase the cell- edge UE throughput by up to 48x, while the use of the 10GHz band with a 500MHz bandwidth can increase the network capacity up to 5x. The use of beamforming with up to 4 antennas per small cell base station lacks behind with cell-edge throughput gains of up to 1.49x. Our study also shows how network densifications reduces multi-user diversity, and thus proportional fair alike schedulers start losing their advantages with respect to round robin ones. The energy efficiency of these ultra-dense small cell deployments is also analysed, indicating the need for energy harvesting approaches to make these deployments energy- efficient. Finally, the top ten challenges to be addressed to bring ultra-dense small cell deployments to reality are also discussed.

I. INTRODUCTION

Future traffic demands motivate a 100×-or-more capacity increase, pursued through densification, higher frequencies, and spectral-efficiency techniques. The paper examines their gains and limitations, including small-cell tiers and deployment trade-offs.

  • Future services such as augmented reality, 3D visualisation, and online gaming may require up to 1 Gbps per UE.
  • 100× or more network-capacity growth is targeted over the next 20 years, motivating the use of multiple capacity-enhancement tools.
  • The three capacity paradigms are spatial reuse through network densification, larger bandwidths through higher carrier frequencies, and improved spectral efficiency.
  • These approaches have distinct limitations: densification complicates deployment and backhaul, higher frequencies incur greater path loss, and spectral-efficiency methods require synchronization and complex processing.
  • The paper analyses the gains and limitations of these paradigms, idle-mode capabilities, and UE density and distribution to identify configurations targeting over 1 Gbps average throughput per UE.
  • HetNets use small cells for local spectrum reuse and capacity while macrocells provide blanket coverage, but co-channel operation introduces inter-tier interference and coverage issues.
  • Dense orthogonal small-cell deployments are envisioned for extensive spatial reuse and primarily target static UEs, while dual connectivity can split traffic across macro and small-cell tiers.

III. WHY ARE TODAY’S SMALL CELLS NOT PRACTICAL TO MEET FUTURE CAPACITY DEMANDS?

Co-channel small cells face interference, coverage, handover, and coordination problems that complicate practical deployment. CRE and eICIC mitigate some issues, but their joint optimisation becomes increasingly complex as cell density grows.

  • Pilot-RSS association can favor macrocells, limiting small-cell offloading and causing severe uplink interference to nearby small cells.The large transmission-power difference makes small-cell coverage appear smaller near macrocells.
  • CRE expands small-cell downlink coverage and facilitates offloading, but larger range-expansion bias increases co-channel downlink interference.
  • ABSs mitigate interference by suppressing most macrocell control and data transmissions while retaining reference signals.Macrocell ABSs protect range-expanded small-cell UEs; small-cell ABSs can protect mobile macro UEs.
  • Effective interference mitigation requires dynamic adaptation of range-expansion bias and ABS patterns to cell density and traffic conditions.Larger range expansion requires a larger macrocell ABS duty cycle for expanded-region UEs.
  • Joint optimisation of range expansion, ABSs, power reduction, scheduling, and inter-BS coordination becomes more complex with cell count, limiting suitability for ultra-dense deployments.

IV. SYSTEM MODEL

The system model evaluates outdoor ultra-dense small-cell deployments across cell spacing, UE density and distribution, carrier frequency, antenna count, and idle-mode operation. It uses a 500-by-500 m scenario with hexagonally deployed small cells and specified propagation and scheduling assumptions.

  • The model places outdoor small-cell BSs on a uniform hexagonal grid in a 500-by-500 m scenario with ISDs from 200 m to 5 m.These spacings correspond to 29 to 46189 small-cell BSs per square kilometre.
  • Three active-UE densities—600, 300, and 100 per square kilometre—are evaluated under uniform and non-uniform distributions.The non-uniform case combines uniform UEs with 40 m-radius hotspots containing 20 UEs each.
  • Carrier frequencies are 2.0, 3.5, 5.0, and 10 GHz, with bandwidths equal to 5% of carrier frequency.The resulting bandwidths are 100, 175, 250, and 500 MHz.
  • Each small-cell BS uses 1, 2, or 4 horizontal-array antennas with quantised MRT beamforming, while each UE uses one antenna.The beamforming targets maximum received signal strength without inter-BS coordination.
  • The analysis uses strongest-pilot association subject to a pilot-SINR threshold and assumes round-robin scheduling with multipath fading omitted from the main analysis.
  • A BS with no associated UE is switched off through the adopted idle-mode capability.

V. NETWORK DENSIFICATION

Network densification raises capacity through spatial reuse, greater per-UE bandwidth, and improved signal quality. However, once deployments become denser than the active-UE population, bandwidth gains saturate and further improvements become logarithmic.

  • Network densification increases capacity through simultaneous bandwidth reuse by more geographically separated BSs.
  • Smaller cells reduce the number of connected UEs per BS, increasing the bandwidth available to each UE.This produces a capacity increase proportional to the number of offloaded UEs.
  • Shorter serving distances improve UE signal quality, yielding a logarithmic capacity increase with SINR.
  • When active UEs per cell reach Um ≤1, cell splitting cannot further increase per-UE bandwidth, leaving signal-quality improvement as the slower remaining gain.The paper identifies one UE per cell as an operational densification sweet spot because further investment can produce diminishing returns.

A. Idle mode capability and the 1 UE per cell concept

Idle mode lets ultra-dense deployments adapt active BSs to UE demand, reducing interference and transmit power. Around an ISD of 35 m, deployments approach one active UE per cell, while denser operation faces diminishing returns and propagation-dependent SINR effects.

  • A. Idle mode capability and the 1 UE per cell concept: Surplus BSs without active UEs can be switched off, reducing both interference and energy consumption.
  • A. Idle mode capability and the 1 UE per cell concept: Lower UE density activates fewer BSs, while uniform UE distributions require more active BSs than non-uniform distributions.
  • A. Idle mode capability and the 1 UE per cell concept: An ISD of 35 m achieves an average of 1.1 active UEs per active BS or fewer, approaching the spatial-reuse limit.Further densification requires exponentially greater investment for diminishing logarithmic gains.
  • 1) Transmit Power:: Ultra-dense idle-mode operation significantly reduces network transmit power because per-cell power savings outweigh the increased number of active cells.The studied uniform and non-uniform distributions show reductions of up to 43 dB in network transmit power.
  • 2) UE SINR Distribution:: With idle mode disabled, denser deployments lower UE SINR because interfering signals increasingly transition from NLOS to LOS.Interference power therefore increases faster than signal power, producing slower capacity growth and potentially diminishing returns.
  • 2) UE SINR Distribution:: With idle mode enabled, denser deployments improve UE SINR by turning off more interfering cells.Median SINR improvement reaches around 8.76 dB at 35 m ISD and 20.62 dB at 10 m ISD.

C. Transition from Interference to Noise Limited Scenarios

Ultra-dense deployments generally remain interference limited rather than transitioning to noise-limited operation, while densification and wider bandwidth provide substantial throughput gains with diminishing returns and higher power costs.

  • Interference and noise limits: Transmit-power changes affect SINR only at extreme densities and low UE densities; realistic deployments do not become noise limited.The observed decoupling occurs only in the high-SINR regime of non-cluster UEs.
  • Network densification: Densification produces rapidly increasing throughput up to 35 m ISD, followed by diminishing gains as spatial reuse approaches one UE per cell.Proximity to serving BSs and idle-mode interference mitigation continue improving SINR after the transition, but more slowly.
  • Network densification: 5 m ISD yields average and cell-edge throughput gains of 17.56× and 48.00× over the 200 m ISD, 100 MHz baseline.The corresponding 35 m ISD gains are 7.56× and 5.80×.
  • Higher frequency bands: 500 MHz bandwidth yields average and cell-edge gains of 5.31× and 5.17× over the 35 m ISD, 100 MHz baseline.A 250 MHz bandwidth provides gains of 2.59× and 2.58×.
  • Higher frequency bands: Higher-frequency capacity gains increase transmit power by up to 24.34 dB, though ultra-dense small-cell BSs can remain below 20 dBm.Higher bands also incur larger path losses and more expensive equipment.
  • Combined configurations: The 1 Gbps-per-UE target is reachable with 50 m ISD and 500 MHz bandwidth, or 20 m ISD and 250 MHz bandwidth.These configurations achieve averages of 1.27 Gbps and 1.01 Gbps per UE, respectively.

VII. MULTI-ANTENNA TECHNIQUES AND BEAMFORMING

The paper evaluates LTE-codebook MRT beamforming with one, two, or four antennas per small-cell BS, finding useful but diminishing gains that remain smaller than densification and bandwidth gains.

  • Beamforming approach: The analysis focuses on quantised MRT beamforming using the standardised LTE codebook because ultra-dense cells may produce highly correlated channels.Spatial multiplexing is left for future research because correlation may limit available degrees of freedom.
  • Beamforming gains: Beamforming gains increase diminishingly with antenna count, consistent with logarithmic antenna-gain scaling.The study assumes horizontal arrays with 1, 2, or 4 elements and one antenna per UE.
  • Beamforming gains: At 35 m ISD and 500 MHz, increasing from one to four antennas raises average UE throughput by 18.92%.The incremental gains from one to two and two to four antennas are 13.77% and 13.05%.
  • Beamforming gains: At 35 m ISD and 500 MHz, increasing from one to four antennas raises cell-edge UE throughput by 48.96%.The gains from one to two and two to four antennas are 46.67% and 38.63%.
  • Target throughput: The 1 Gbps-per-UE target is reachable with 75 m ISD, 500 MHz, and four antennas, or 35 m ISD, 250 MHz, and four antennas.Overall beamforming gains are estimated at up to 1.49×, below densification and higher-frequency gains.

VIII. SCHEDULING

The scheduling analysis compares RR and PF under densification and finds that shrinking cells reduce channel variability, eroding PF’s multi-user-diversity advantage and making RR attractive at low ISDs.

  • PF scheduling: PF ranks UEs using instantaneous performance relative to average performance across time-domain and frequency-domain scheduling stages.The time-domain stage selects candidate UEs before frequency-domain RB allocation.
  • UE throughput: Reducing ISD from 150 m to 40 m and 20 m lowers PF 5%-tile UE throughput by approximately 40.8% and 36.7%.At 20 m ISD, PF’s 5%-tile gain over RR is approximately 9%.
  • Scheduler choice: At low ISDs, PF’s minor gains over RR suggest RR may be preferable because its implementation complexity is lower.PF evaluates UEs on each RB, with exhaustive search substantially increasing complexity.

IX. ENERGY-EFFICIENCY

The energy analysis shows that ultra-dense deployments require careful idle-mode and antenna-power design: lower idle consumption improves efficiency, while beamforming antennas can reduce it.

  • Energy scaling: Deploying 50 million femtocells consuming 12 W each would use 5.2 TWh/a, so this approach does not scale energetically.The paper therefore treats energy efficiency as necessary for sustainable ultra-dense networks.
  • Power model: The study estimates BS energy use with the GreenTouch power model across active, slow-idle, and shut-down states for one, two, and four antennas.The analysis uses a 2020 small-cell BS type and 20 MHz bandwidth.
  • Antenna impact: For any idle mode, adding antennas decreases energy efficiency because beamforming gains do not offset the added antenna-chain power.This conclusion may differ for spatial multiplexing when channel degrees of freedom permit stronger capacity scaling.
  • Idle modes: Lower idle-mode power produces higher network energy efficiency.With GreenTouch idle modes, efficiency decreases with densification; futuristic lower-power modes can reverse that trend over part of the density range.
  • Idle modes: Zero-consumption idle mode makes energy efficiency increase with densification, motivating advanced idle capabilities and energy harvesting.Energy harvesting is proposed to keep idle small-cell BSs operational without grid power.

X. WHAT IS DIFFERENT IN ULTRA-DENSE SMALL CELL DEPLOYMENTS

Ultra-dense small cell deployments differ from regular HetNets in BS-to-UE density, propagation conditions, and UE diversity, requiring revised models and network procedures. Their realization also faces challenges spanning backhaul, mobility, cost, planning, idle mode, scheduling, modulation, spatial multiplexing, dynamic TDD, and WiFi coexistence.

  • Network differences: Ultra-dense HetNets can have fewer UEs than BSs, making it necessary to power off BSs without active UEs to reduce interference and conserve power.
  • Propagation and diversity: LoS interferers and spatially correlated channels become more prevalent, so regular-HetNet assumptions and spatial multiplexing approaches cannot be transferred directly.The paper states that spatial correlation can limit available degrees of freedom and render spatial multiplexing less useful.
  • Radio resource management: Ultra-dense deployments reduce UE diversity, causing proportional-fair scheduling to lose advantages and motivating simpler scheduling procedures.
  • Deployment challenges: Backhaul is a major deployment constraint: 96 % of operators identify it as important for small cells, and ultra-dense deployments likely require wired connectivity in dense urban areas.
  • Deployment challenges: Other challenges include mobility bearer management, low-cost hardware, hotspot-based location planning, higher-order modulation, dynamic TDD, and coexistence with WiFi.The paper notes that LTE interference can cause WiFi nodes to remain in listening mode, while higher-order QAM raises CSI, EVM, and PAPR concerns.
  • Deployment challenges: Efficient idle mode is needed to mitigate inter-cell interference and save energy, potentially by switching off most BS modules and waking on detected uplink signalling.The described sniffing approach does not support selective wake-ups, where only the best cell in an idle cluster activates.

XII. CONCLUSION

The conclusion compares network densification, bandwidth expansion, and beamforming, while also identifying scheduling, energy, and deployment constraints. Simulations indicate that densification provides the largest throughput gains but that cost-effective and energy-efficient ultra-dense deployments remain unresolved.

  • Up to 48x cell-edge UE throughput comes from network densification, compared with up to 5x network capacity from higher-frequency bandwidth and around 1.49x cell-edge beamforming gains.
  • Efficient small-cell idle mode mitigates interference and saves energy, while one UE per cell is identified as the fundamental limit of spatial reuse.UE density and distribution therefore need to be understood before deployment.
  • Network densification reduces multiuser diversity, leaving proportional-fair scheduling only around 5 % ahead of round robin at an ISD of 20 m.The paper indicates that round robin may be more suitable because of its lower complexity.
  • For 300 active UE per square km, roughly 35 m ISD, 250 MHz bandwidth, and 4 antennas per BS can achieve an average 1 Gbps per UE in simulation.The paper states that these deployments are currently neither cost-effective nor energy efficient.

APPENDIX

The appendix defines the simulated gain, propagation, antenna, transmit-power, SINR, and throughput calculations used to evaluate small-cell deployments. It combines path, shadow-fading, antenna, and scheduling models across repeated simulation runs.

  • The simulated environment represents channel gains as two-dimensional matrices, with each BS-location overall gain formed by summing individual gains in decibels.
  • Overall gain combines antenna gain, path gain, and shadow-fading gain for each BS and UE location.
  • The antenna model uses four vertical half-wave dipoles spaced by 0.6 λc, while horizontal arrays use LTE-codebook beamforming weights.
  • The path-gain model combines line-of-sight and non-line-of-sight gains weighted by distance-dependent LoS probability, using 3GPP urban-micro models.Shadow fading is spatially correlated, has 6 dB standard deviation, and has inter-BS correlation 0.5.
  • Transmit power is selected to achieve a cell-edge target SNR of γedge[dB] = 9, 12 or 15 dB using noise power and cell-edge path gain.
  • UE throughput is derived from SINR using a round-robin allocation, an SINR-to-throughput mapping, and the number of UEs served by the BS.Statistics use 150 simulation runs with independent UE locations and shadow fading, plus extra external tiers to reduce border effects.
Loading 1503.03912v1…