Successive Interference Cancellation (SIC)

SIC relies on a sequential decoding process where streams are decoded one by one, not simultaneously. The order is typically determined by signal-to-noise ratio (SNR) or channel gain.

• The strongest signal (highest SNR) is decoded first, as it is the most reliable to detect.
• This decoded stream is then re-encoded and reconstructed using the known channel state information (CSI).
• The reconstructed contribution is subtracted from the original received signal, effectively removing that layer of interference for the remaining, weaker streams.
• This process repeats until all spatial streams are decoded.

A critical vulnerability of SIC is error propagation. If the first stream is decoded incorrectly, the reconstructed signal will be flawed.

• Subtracting a flawed reconstruction from the received signal introduces residual interference and distorts the signal for subsequent stages.
• This error cascades, severely degrading the decoding performance of all later streams.
• To mitigate this, practical systems often combine SIC with channel coding and cyclic redundancy checks (CRC) to validate the decoded block before subtraction.

The sequence in which streams are decoded is the primary factor determining SIC performance. Optimal ordering maximizes the minimum post-processing SNR.

• Post-detection SNR: Streams are ordered based on the signal quality after linear nulling (e.g., Zero-Forcing or MMSE) is applied.
• Channel Matrix Norm: Ordering can be based on the column norms of the channel matrix, where a higher norm indicates a stronger spatial signature.
• Capacity-Driven Ordering: In some implementations, the order is chosen to maximize the sum-rate capacity of the system.

SIC introduces a processing latency penalty due to its sequential nature, making parallelization difficult.

• Each stage must wait for the previous stage to complete decoding, reconstruction, and subtraction.
• This creates a processing pipeline where latency scales linearly with the number of spatial streams.
• Hardware designs often use pipelined architectures to maintain throughput, where a new received vector enters the first stage as soon as it finishes processing the previous one.
• The computational complexity is higher than linear receivers (ZF/MMSE) but significantly lower than optimal Maximum Likelihood Detection (MLD).

SIC is typically combined with a linear front-end detector, most commonly MMSE-SIC or ZF-SIC.

• ZF-SIC: Uses a Zero-Forcing filter to null out undecoded streams. It completely eliminates interference but suffers from noise enhancement, especially in ill-conditioned channels.
• MMSE-SIC: Uses a Minimum Mean Square Error filter that balances interference suppression with noise amplification. It achieves a better error rate than ZF-SIC, particularly at low to moderate SNRs.
• MMSE-SIC is the practical standard for 5G and Wi-Fi advanced receivers because it approaches the capacity of the MIMO channel.

SIC is the spatial-domain equivalent of a Decision Feedback Equalizer (DFE) used in single-input single-output (SISO) channels.

• A DFE cancels inter-symbol interference (ISI) in the time domain by feeding back past decisions.
• SIC cancels inter-stream interference in the spatial domain by feeding back decisions from already-decoded antennas.
• The core principle is identical: use reliable past decisions to cancel known interference for future decisions, a concept known as successive decoding.