Network Device PIM Cancellation Using Segmented Nonlinear Model

The patent segments the PIM cancellation process into two distinct phases: an offline training phase where polynomial coefficients are determined using pilot signals, and an online cancellation phase where the pre-determined coefficients are applied to cancel PIM in actual data signals. This segmentation allows complex polynomial modeling to be performed only when necessary, reducing real-time computational complexity while maintaining cancellation accuracy.

The patent performs preliminary action by determining the polynomial coefficients during a training phase before actual data transmission. The coefficients are calculated in advance using known pilot signals and stored for reuse. During data transmission, only simple polynomial evaluation and subtraction are needed, avoiding the need to recompute complex polynomial models for each data symbol, thus significantly reducing processing power requirements.

The patent applies partial action by using a truncated polynomial model that includes only the necessary terms up to a certain order (e.g., third-order terms) rather than modeling all possible polynomial interactions. The patent demonstrates that including terms up to third order provides sufficient accuracy for PIM cancellation in practical scenarios, avoiding the computational burden of higher-order terms while maintaining adequate cancellation performance.

The patent changes the parameter of polynomial order dynamically by adapting the model complexity based on signal conditions. The system can adjust the maximum polynomial order used in cancellation based on the observed PIM signal characteristics and channel conditions, using lower-order models when sufficient and higher-order models only when necessary, thus optimizing the balance between accuracy and computational efficiency.

The patent implements periodic action by updating the polynomial coefficients at specific intervals using periodic pilot signals embedded in the transmission stream. Rather than continuously tracking and updating coefficients for every data symbol, the system periodically re-estimates coefficients using known pilot sequences, which provides channel adaptation while significantly reducing the computational frequency and energy consumption compared to continuous updates.

The patent uses feedback mechanisms where the receiver monitors the residual PIM signals after cancellation and feeds this information back to adjust the polynomial coefficients. This feedback-driven adaptation allows the system to maintain accuracy by updating coefficients only when necessary based on actual cancellation performance, rather than using fixed update schedules or continuous adaptation, thus reducing unnecessary processing power consumption.