SWIPT Receiver Architectures

• SWIPT receiver structures are designs that split or integrate RF signals to both decode information and harvest energy, addressing the rate–energy trade-off.
• Architectures like power splitting, time switching, integrated, and diplexer-based systems differ in hardware complexity, nonlinearity exploitation, and performance under various SNR regimes.
• Advanced designs including PPM-integrated receivers, dynamic antenna allocation, and superimposed chirp methods demonstrate significant gains in energy harvesting and reliable data decoding for IoT applications.

Simultaneous Wireless Information and Power Transfer (SWIPT) receiver structures are a foundational aspect of wireless systems supporting energy-constrained or batteryless devices. SWIPT receivers are engineered to exploit incident RF signals to both extract information and harvest usable DC power. Their architectures are tightly coupled with the physical-layer modulation, RF circuit models, and waveform design to manage the fundamental rate–energy trade-off in the presence of rectifier nonlinearity, hardware constraints, and diverse deployment scenarios.

1. Canonical SWIPT Receiver Architectures

The primary receiver structures in SWIPT can be classified as:

• Separated Architecture (Power Splitting, PS): The incident RF signal is first divided via a power splitter, directing proportions $\rho$ and $1-\rho$ to the information decoder (ID) and the energy harvester (EH), respectively. The ID path includes conventional RF front-ends (mixer, LNA), while the EH path drives a rectifier and DC management circuitry. Channel state information and estimation, as well as dynamic $\rho$ adaptation, play critical roles in maximizing ergodic capacity under energy constraints [14.05.4623, (Jameel et al., 2018)].
• Integrated Architecture (IntRx): The incoming RF is rectified directly via a Schottky diode. The output DC waveform is simultaneously used for EH storage and information recovery, usually by an ADC and DSP operating on the rectified envelope. There are no RF mixers or local oscillators, minimizing receiver energy cost (Kim et al., 2021, Jameel et al., 2018).
• Time-Switching (TS): The receiver alternates between ID and EH modes on a per-frame or per-symbol basis, using an RF switch. For each frame interval, a fraction $\tau$ is dedicated to EH and $1-\tau$ to ID. TS is simple to implement but introduces idle intervals for both functionalities (Janatian et al., 2016, Kim et al., 2019).
• Diplexer-Based Integrated Receiver (DIR): The rectified output current is split by an RF diplexer into DC (EH) and high-frequency (ID) paths. The band-pass branch enables information decoding from mixing products or modulated subcarriers without explicit power splitting, improving the energy–rate feasibility region (Roy et al., 2023, Qin et al., 2016).
• Antenna Partitioning/Antenna-Switching: In multi-antenna systems, each receive antenna can be dynamically allocated to either EH or ID at each symbol and subcarrier, yielding a combinatorial trade-off between MIMO capacity and harvesting efficiency. Dynamic allocation is solved via greedy algorithms exploiting the submodularity of mutual information under matroid constraints (Vaze et al., 2014, Jalali, 2020, Mukherjee et al., 2023).

2. Physical-Layer Signal and Rectifier Models

The efficacy of receiver architectures depends critically on the interaction between signal design and rectifier nonlinearity:

• Nonlinear Diode Rectification: The RF rectifier is accurately modeled via a Taylor-series expansion (typically up to the fourth-order) of the diode IV relationship. The delivered DC power is thus

$$P_{\rm DC} = k_2\,R_{\rm ant}\,{\mathbb E}[y_{\rm RF}^2] + k_4\,R_{\rm ant}^2\,{\mathbb E}[y_{\rm RF}^4],$$

with $k_2$ and $k_4$ set by the diode physics and load (Kim et al., 2021, Mukherjee et al., 2023, Roy et al., 2023, Clerckx, 2016).

• Waveform-Dependent Harvesting: High peak-to-average power ratio (PAPR) signals (e.g., multisine, PPM, chirp superposition) are substantially more effective for energy harvesting due to the $P^2$ and high-order dependence in the rectifier output, compared to CW or QAM (Kim et al., 2021, Roy et al., 2023, Clerckx, 2016).
• Multisine and Superimposed Chirp: Multisine and superimposed chirp designs are optimal for leveraging diode nonlinearity. Superposition of high-PAPR energy waveforms with data-carrying signals enables efficient trade-off navigation (Clerckx, 2016, Roy et al., 2023).

3. Advanced / Modular Receiver Designs

Recent research has introduced structurally novel and highly tailored receiver structures, including:

• Pulse-Position-Modulation Integrated Receiver (PPM-IntRx): All RF is first rectified to DC; high-PAPR pulse-position modulated signals inject strong envelope fluctuations. The ID chain samples the DC output and demodulates by locating the pulse position, achieving both high EH yield and simple baseband ID, with no mixers or LOs. Demonstrated harvested DC power exceeds continuous-wave benchmarks by 100–250%, and BER $\approx 10^{-4}$ is achieved at moderate SNRs for typical IoT rates (Kim et al., 2021).
• Differential Chaos Shift Keying (DCSK) Receivers with Dynamic Antenna Allocation: Each receive antenna is dynamically toggled between ID and EH branches per frame, with closed-form expressions for BER and harvested energy. The SR–z_DC “trade-off region” is parametrically scanned by waveform parameters and antenna assignment (Mukherjee et al., 2023).
• Superimposed Chirp with DIR: Chirp waveform superposition over multiple subbands (with subband selection) is paired to a DIR. The diplexer allows full RF energy to reach the rectifier, while the band-pass path enables demodulation—without a power splitter. Integration enables simultaneous PAPR gain (for EH) and frequency diversity (for ID), yielding a strictly larger rate–energy region compared to power-splitting and multisine-based PSR (Roy et al., 2023).
• AC-Computing-Enabled Architectures: Post-splitting, a fraction of harvested energy is routed directly—unrectified—to on-chip AC logic. This bypasses AC→DC conversion loss and supports a lower operating threshold for the computing block, significantly enlarging the achievable rate–energy region (Tran et al., 2019).

4. Resource Allocation, Optimization, and Signal Adaptation

Receiver structure optimizations are typically enforced jointly with transmit-side waveform, power, beamformer, and scheduling design:

• Pareto Region Tracing: Multi-objective and robust optimization is widely adopted—weighted Chebycheff or utility-scalarized SDPs yield trade-off frontiers, with optimal schedules for splitting or TS parameters per user (Janatian et al., 2016, Mohjazi et al., 2017, Vaze et al., 2014).
• GNN-Based and Learning-Aided SWIPT Receivers: End-to-end graph neural network models can learn resource allocation policies (splitting ratios, beamformers) for both PS and TS receivers under arbitrary topology and QoS constraints. Single-type outputs (beamformers only) with analytic mappings for splitting facilitate efficient and scalable learning (Han et al., 6 Feb 2025).
• Dynamic Markovian and HARQ Receivers: In energy-constrained HARQ systems, the optimal SWIPT receiver policy is a per-slot binary action (harvest or accumulate ID) and can be solved via absorbing Markov chain analysis. For i.i.d. channels, the simple “harvest-first-store-later” family is optimal; in time-correlated channels, the policy is a 2D table lookup (Abad et al., 2017).

5. Comparative Analysis of Receiver Structures

Architecture	Simultaneity	Hardware Complexity	Nonlinearity Exploited	EH–ID Control	Achievable Region
Power Splitting (PS)	Yes	Splitter + 2 chains	Partial	Variable $\rho$	Good at high SNR, flexible
Time Switching (TS)	No	RF switch + 1 chain	None during ID	Variable $\tau$	Good at low SNR, but incurs idle
Integrated (IntRx)	Yes*	1 rectifier, no LOs	Strong (full DC input)	Buffer/timing	High EH, typically lower rate
Diplexer-Based (DIR)	Yes	1 rectifier + diplex	High (no power loss)	Frequency-split	Strictly larger than PS/TS region
Antenna Partitioning	Yes	M-antenna switch	Configurable	Greedy/table-based	Maximal in MIMO settings
PPM-IntRx	Yes	1 diode/no mixers	Max (high-PAPR PPM)	PPM symbol tuning	Joint optimization, IoT-centric
DCSK multi-antenna	Yes	N x {switch+EH/ID}	Explicit, chaos-tuned	Per-antenna switching	SR–z_DC region parametrically
AC Computing	Yes	PS + AC/DC split	Bypasses DC loss	$(\rho, \phi)$ optimization	Region enlarged vs. DC-only

*IntRx enables simultaneous EH/ID in terms of circuit flow, but the ID chain recovers information from the rectified DC envelope, potentially limiting data rate compared to separated PS/TS.

6. Experimental and Implementation Insights

• Prototyped IntRx-PPM achieves 100–250% harvested DC gain over CW, BER $\leq 10^{-4}$ at moderate SNR, with zero LO/mixer power, making this architecture highly attractive for low-power IoT (Kim et al., 2021).
• In practical PS vs. TS receivers, PS achieves a larger E–T (energy–throughput) trade-off region if the information decoder operates effectively at low powers; TS is preferable at low SNR or when switching simplifies the RF chain (Kim et al., 2019, Mohjazi et al., 2017).
• Diplexer-based receivers deliver a fixed (0.5,0.5) split of input power to EH and ID, but when deployed with waveform engineering (e.g., chirp superposition), the region can surpass any fixed PS (Roy et al., 2023, Qin et al., 2016).
• In multi-antenna settings, optimized partitioning or antenna-switching delivers MIMO diversity gains, submodular maximization, and linear complexity resource allocation at scale (Vaze et al., 2014, Jalali, 2020).

7. Practical Design Guidelines and Future Directions

• The optimal receiver architecture must be chosen according to target SNR regime, device power budget, modulation, and system scale.
• For ultra-low-power IoT, integrated PPM receivers with high-PAPR modulation and no RF frequency conversion yield maximal EH with modest rates (Kim et al., 2021).
• For flexible and throughput-demanding scenarios, power splitting or diplexer-based receivers combined with superposition-modulated waveforms deliver superior trade-offs.
• Antenna/antenna-mode partitioning is preferred in multi-antenna, spatially flexible deployments with the required switching infrastructure (Vaze et al., 2014, Jalali, 2020, Mukherjee et al., 2023).
• Learning-aided (e.g., GNN) frameworks support real-time, scalable resource allocation for both PS/TS modes and can leverage transfer learning between modes (Han et al., 6 Feb 2025).
• Practical constraints such as rectifier sensitivity, circuit losses, CSI uncertainty, and the nonlinearity of realistic diode models must be explicitly incorporated.
• Future work includes full integration of waveform, mode-switching, and beamforming optimization—especially for wideband, multi-user, or frequency-selective channels—and the deployment of on-chip ultra-low-power architectures for embedded sensor networks (Roy et al., 2023, Kim et al., 2021).

---

Principal References:

• "Wireless Information and Power Transfer for IoT: Pulse Position Modulation, Integrated Receiver, and Experimental Validation" (Kim et al., 2021)
• "SWIPT using Hybrid ARQ over Time Varying Channels" (Abad et al., 2017)
• "Chaotic Waveform-based Signal Design for Noncoherent SWIPT Receivers" (Mukherjee et al., 2023)
• "Superimposed Chirp Waveforms for SWIPT with Diplexer-based Integrated Receivers" (Roy et al., 2023)
• "Experimental Analysis of Harvested Energy and Throughput Trade-off in a Realistic SWIPT System" (Kim et al., 2019)
• "Receiver Antenna Partitioning for Simultaneous Wireless Information and Power Transfer" (Vaze et al., 2014)