In textbook wireless communications, Non-Orthogonal Multiple Access (NOMA) looks like a free lunch. By overlapping multiple users' signals in the same frequency and time slot and separating them purely by power levels, NOMA promises to squeeze vastly more capacity out of limited spectrum.
The magic engine behind this is Successive Interference Cancellation (SIC). The receiver decodes the strongest signal first, treats the weaker signals as background noise, and then subtracts this decoded component from the total received waveform. With the dominant interference cleanly peeled away, the receiver can then decode the weaker signal with a pristine signal-to-noise ratio.
In the real world, however, this subtraction is never perfect. Hardware impairments, channel estimation errors, and phase noise leave behind a "ghost" of the original signal. This phenomenon—imperfect successive interference cancellation (ipSIC)—fundamentally changes the behavior of multi-user wireless networks, introducing hard performance ceilings that cannot be overcome by simply turning up the transmit power.
Why Power Scaling Fails: The Intuition of the Error Floor
To understand why imperfect cancellation is so destructive, consider what happens when we increase the transmission power of a NOMA system.
In a conventional wireless link, increasing the signal-to-noise ratio (SNR) steadily drives down the probability of a transmission failure (the outage probability). If you want a more reliable connection, you simply pump more power into the antenna.
But in a system plagued by ipSIC, the remaining noise isn't just static thermal background noise; it is residual interference that scales proportionally with the power of the transmitted signals. If you double the power of the signals, you also double the power of the uncancelled residual interference.
Mathematically, the Signal-to-Interference-plus-Noise Ratio (SINR) for a weaker user looks something like this:
$$\text{SINR} = \frac{P_{\text{weak}}}{N_0 + \beta P_{\text{strong}}}$$
where $N_0$ is the background noise and $\beta$ represents the fraction of the strong user's signal left behind due to imperfect cancellation. As the system power scales up to infinity, the background noise $N_0$ becomes negligible ($N_0 \to 0$), reducing the equation to:
$$\text{SINR} \approx \frac{P_{\text{weak}}}{\beta P_{\text{strong}}}$$
Because both the numerator and denominator scale linearly with total power, the SINR hits a hard ceiling. No matter how much power you inject, the signal quality never improves past this ratio.
As Xinwei Yue et al. demonstrated in their analysis of a unified NOMA framework, this creates an irreducible error floor. Beyond a certain SNR threshold, the outage probability flattens completely, effectively dropping the system's diversity order to zero. In another work focusing on two-way relaying, Xinwei Yue confirmed that this residual noise floor can cripple maximum throughput, making idealized assumptions of perfect signal subtraction highly unrealistic for dense network deployments.
The Domino Effect: Error Propagation
The vulnerability of SIC is deeply tied to its sequential nature. Because signals are decoded and subtracted one after another, an error at any stage cascades catastrophically through every subsequent layer.
If the receiver incorrectly decodes the first user's message, it will subtract the wrong waveform. Instead of reducing interference for the second user, it actually adds more noise to the mix. As Amanat Kafizov et al. pointed out in the context of Visible Light Communication (VLC) systems, ignoring this risk of error propagation leads to overly optimistic models that gloss over the fragility of multi-user decoding chains.
This sequential failure mode is a universal bottleneck in communication theory. For instance, Wenbin Hu 0001 et al. observed a highly analogous behavior in sequential decoding for polar codes, where an incorrect decision on an early bit cascades through the remaining sequence, causing total message failure. In NOMA, this means that the reliability of the entire network is tightly bound to the accuracy of the very first decoding step. As highlighted by Yuanwei Liu et al. in The Road to Next-Generation Multiple Access, this extreme sensitivity to transceiver impairments creates a critical stability bottleneck that demands robust, error-resilient detection algorithms.
System-Wide Consequences
When residual interference is introduced, the theoretical advantages of non-orthogonal multiplexing begin to erode across various network architectures:
- Network Capacity Overestimation: In large-scale heterogeneous networks, models assuming perfect cancellation vastly overestimate throughput. Yuanwei Liu et al. showed that real-world hardware constraints and residual noise significantly degrade signal quality, altering the optimal power-sharing coefficients required to keep users covered (Non-orthogonal Multiple Access in Large-Scale Heterogeneous Networks).
- Severe Security Degradation: Physical Layer Security (PLS) relies on maintaining a signal quality advantage over an eavesdropper. Minh-Sang Van Nguyen et al. proved that residual interference from ipSIC causes a sharp degradation in secrecy performance (Exploiting Secrecy Performance of Uplink NOMA in Cellular Networks). Because the residual noise scales with power, legitimate receivers cannot out-power an eavesdropper, creating a persistent secrecy outage floor.
- Environmental Vulnerabilities: In satellite-to-ground networks, the combination of severe channel shadowing and ipSIC creates an unforgiving performance ceiling. Xinwei Yue et al. demonstrated that under realistic conditions, these combined factors force an early flattening of the outage probability curve, limiting NOMA's efficacy in satellite deployments (Outage Behaviors of NOMA-based Satellite Network over Shadowed-Rician Fading Channels).
Designing for Imperfection
To make non-orthogonal communication viable, modern research focuses on explicitly modeling hardware imperfections and designing mitigation strategies.
One approach is hardware-aware resource allocation. Instead of treating the receiver as an ideal mathematical operator, optimization frameworks can treat the receiver's sensitivity as a hard constraint. Abdulkadir Celik et al. proposed a framework that incorporates a minimum required power differential between signals and models the residual noise directly into the cluster formation process (Distributed Cluster Formation and Power-Bandwidth Allocation for Imperfect NOMA in DL-HetNets). By ensuring that signals are only multiplexed when the receiver has enough "power gap" to reliably separate them, NOMA can still maintain an edge over traditional orthogonal access.
Another frontier involves using external hardware, such as Reconfigurable Intelligent Surfaces (RIS), to physically alter the wireless environment. By dynamically adjusting the phase shifts of signals reflecting off a surface, researchers can artificially widen the channel quality gap between users. Xinwei Yue et al. demonstrated that integrating RIS into NOMA networks helps counteract the degradation caused by ipSIC, providing a crucial buffer that lowers the outage probability floor and restores system reliability (Performance Analysis of Intelligent Reflecting Surface Assisted NOMA Networks).
Ultimately, understanding why successive interference cancellation fails in practice has shifted the research direction away from idealized capacity proofs and toward robust, impairment-aware network design.