The Computational Overhead of Interference Cancellation

163 concepts145 papers1 proposal

To maximize the capacity of modern wireless networks, engineers have increasingly abandoned a long-standing rule of communications design: orthogonality. Traditional Orthogonal Multiple Access (OMA) systems ensure that users do not interfere with one another by assigning each device a unique slice of time, frequency, or code. While clean, this approach leaves significant spectral efficiency on the table.

Modern paradigms like Non-Orthogonal Multiple Access (NOMA) and Rate-Splitting Multiple Access (RSMA) intentionally overlap user signals within the same spectral band. By treating interference as something to be managed rather than avoided, these techniques dramatically increase network throughput. However, this spectral "free lunch" comes with a steep price tag: a massive shift in computational complexity from the transmitter to the receiver. To untangle overlapping signals, receivers must deploy advanced, iterative signal processing algorithms—most notably Successive Interference Cancellation (SIC)—that challenge the physical and economic limits of consumer hardware.

The Mechanics of the Receiver Bottleneck

To understand why non-orthogonal communication strains receiver hardware, we must look at how a signal is decoded. In a basic downlink NOMA scenario, a base station broadcasts a superimposed signal containing data for multiple users, differentiated only by power levels.

[Base Station] ---> Superimposed Signal (User 1 + User 2 + User 3) ---> [Receivers]
                                                                              |
   +--------------------------------------------------------------------------+
   |
   v
[User 3 Receiver (Weakest Channel)] ---> Decodes User 1 directly (Treats others as noise)
   |
   v
[User 1 Receiver (Strongest Channel)] 
   ├── Step 1: Decode User 3's signal
   ├── Step 2: Reconstruct and subtract User 3's signal from the total waveform (SIC)
   └── Step 3: Decode own signal from the clean residual waveform

In this architecture, a user with a strong channel condition cannot simply decode their own data. Instead, they must first decode the signals intended for users with weaker channel conditions, reconstruct those waveforms, and subtract them from the received signal. Only then can they decode their own message.

As Zhiguo Ding 0001 et al. demonstrated in their foundational analysis, On the Performance of Non-Orthogonal Multiple Access in 5G Systems with Randomly Deployed Users, this process requires significantly higher computational overhead and hardware sophistication than traditional orthogonal methods. The complexity scales poorly with the number of co-channel users. For a cluster of $N$ users, the strongest receiver must perform $N-1$ stages of decoding and subtraction.

This multi-stage processing introduces three distinct engineering hurdles:

  1. Processing Latency: Each decoding and subtraction step takes time. In a cross-layer study of NOMA systems, Yuanwei Liu et al. observed that this decoding latency can cause severe processing delays, directly degrading the user's Quality-of-Experience (QoE) in time-sensitive applications like video streaming.
  2. Error Propagation: If a receiver makes a single bit error while decoding a weaker user's signal, the reconstructed waveform will be flawed. Subtracting this flawed waveform introduces further distortion rather than removing it, a phenomenon noted by Wenchao Zhai et al. as a primary barrier to deploying rate-splitting strategies in real-time systems.
  3. Dynamic Sequence Calculation: In a mobile environment, channel conditions fluctuate constantly. Receivers cannot use a fixed decoding order; instead, they must compute the optimal decoding sequence in real time. As Guangtao Zheng et al. pointed out, calculating these sequences on the fly creates an acute processing burden in dense, dynamic multi-user environments.

The Asymmetry of Low-Power Devices

The computational burden of interference cancellation creates a severe deployment paradox for the Internet of Things (IoT) and massive Machine Type Communications (mMTC). Non-orthogonal access is theoretically ideal for IoT because it allows thousands of low-rate sensors to connect to a single base station simultaneously. However, the sensors themselves are precisely the devices least capable of handling the resulting processing overhead.

In low-power networks, hardware is constrained by strict battery budgets and low-cost processor architectures. Research by Rundong Jia et al. emphasizes that the mathematical operations required for multi-stage SIC frequently exceed the hardware capabilities of simple IoT sensors. If a device must drain its battery to execute complex matrix operations and signal subtraction, the spectral efficiency gains of the network are offset by shortened device lifespans.

This problem is not confined to the downlink. In uplink scenarios, where multiple devices transmit to a central base station simultaneously, the base station bears the brunt of the processing load. When scaling to massive numbers of devices, Garima Chopra showed that executing SIC at an ecosystem scale introduces massive computational bottlenecks and latency spikes at the access point, complicating real-time network orchestration.

A Ubiquitous Challenge Across Waveforms

While NOMA and RSMA are the most prominent examples, the tension between spectral efficiency and receiver-side complexity is a universal theme across next-generation physical layer designs.

  • Advanced Multi-Carrier Modulation: Schemes like Generalized Frequency-Division Multiplexing (GFDM) reduce out-of-band emissions to allow tighter subcarrier packing. However, Seungyul Han et al. showed that achieving these clean spectral boundaries requires non-linear interference cancellation receivers whose computational costs quickly exceed the power budgets of low-cost hardware. Similarly, in the chirp domain, Guangyao Liu et al. noted that index modulation detection mechanisms introduce a high level of computational intensity compared to standard orthogonal multicarrier receivers.
  • Interference Alignment and Neutralization: In multi-user interference channels, techniques like retrospective interference alignment can maximize the secure degrees of freedom. Yet, as Mohamed Seif et al. argued, the multi-phase transmissions and complex cancellation logic required make these schemes highly impractical for low-latency or hardware-constrained environments. Joan S. Pujol Roig et al. arrived at a similar conclusion regarding cached interference networks, noting that the underlying mathematical optimizations often exceed mobile hardware capabilities.
  • Cooperative Relaying: In cooperative networks where nodes act as relays for one another, non-orthogonal scheduling forces destination nodes to untangle overlapping relay signals. Information-theoretic work by Kambiz Azarian et al. proved that while non-orthogonal schedules optimize the diversity-multiplexing tradeoff, they demand sophisticated signal processing at the destination that can be prohibitive for real-time hardware execution.

Architectural Strategies for Mitigation

To bridge the gap between theoretical spectral capacity and practical hardware execution, researchers have developed several architectural workarounds that minimize or eliminate the need for intensive receiver-side processing.

Exploiting Alternative Degrees of Freedom

One of the most effective ways to reduce interference cancellation overhead is to separate users in domains other than pure signal power. For example, Caijun Zhong et al. proposed marrying Spatial Modulation (SM) with NOMA. By mapping different users' data directly to specific antenna indices and spatial paths, their framework eliminates intra-cluster interference entirely, bypassing the need for complex SIC at the end-user device.

Similarly, Arthur S. de Sena et al. introduced dual-polarized massive MIMO networks as a tool to combat imperfect SIC. By leveraging the polarization domain (separating signals into vertical and horizontal waves), they provided an extra degree of freedom that insulates low-power devices from the residual interference caused by hardware-level decoding errors.

Pre-Calculation and Algorithmic Approximations

Instead of forcing mobile devices to calculate optimal decoding paths in real time, some architectures shift the computational burden offline or to the cloud. The "Constrained Partial Group Decoding" framework designed by Guangtao Zheng et al. pre-calculates optimal decoding sequences into a compressed, static "decoding map." Receivers simply look up their pre-computed path based on current channel states, avoiding intensive real-time optimization.

Hardware-Friendly Equalization

In multicarrier systems where subcarriers overlap, researchers focus on restructuring the waveform itself to simplify the receiver's math. Andrea M Tonello et al. developed Cyclic Block Filtered Multitone (CB-FMT) modulation, which replaces complex linear convolutions with circular ones. This structural shift allows receivers to use simple, block-based frequency-domain equalization, capturing the spectral benefits of non-orthogonal subcarrier packing without the heavy computational footprint of traditional time-domain filtering.

Ultimately, the transition to non-orthogonal communication represents a fundamental trade-off in information theory and hardware design. While overlapping signals unlock unprecedented spectral efficiency, the true capacity of next-generation networks will not be determined solely by Shannon's formula, but by our ability to implement low-complexity, energy-efficient interference cancellation algorithms in silicon.

Key Foundational Work

Reduced complexity joint iterative equalization and multiuser detection in dispersive DS-CDMA channels (2005) — contributed 2 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 138 papers
  2. Version 1 · 2026-10-04 · 146 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_interference_cancellation_overhead_1c7fead4,
  title = {interference cancellation overhead},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/interference-cancellation-overhead/rdr_1c7fead4a6434be9af804df1793470ae},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 2, frozen text):

@misc{lacuna_interference_cancellation_overhead_1c7fead4_v2,
  title = {interference cancellation overhead},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/interference-cancellation-overhead/rdr_1c7fead4a6434be9af804df1793470ae/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake