🤖 AI Summary
Existing backscatter tags face two critical bottlenecks: large spatial overhead from single-sensor integration and synchronization challenges in multi-sensor multiplexing—requiring on-chip clocks or multiple modulation chains, which increase cost, size, and susceptibility to desynchronization. This work proposes a clockless, single-link multi-sensor backscatter tag architecture. It innovatively extracts a common reference clock from ambient RF signals to enable parallel, cross-sensor synchronization. A voltage-divider-based sensor multiplexing scheme and single-oscillator frequency-shift modulation eliminate the need for on-chip clocks and multiple modulation paths. Combined with RF-switch-based modulation, a frequency-tracking algorithm, and a finite-state machine, the design achieves efficient data multiplexing and demodulation. Implemented in ASIC, the tag consumes only 25.56 μW, supports synchronized acquisition from five 5-kHz or three 18-kHz sensors, and achieves an average reconstructed signal SNR exceeding 15 dB.
📝 Abstract
Backscatter tags provide a low-power solution for sensor applications, yet many real-world scenarios require multiple sensors-often of different types-for complex sensing tasks. However, existing designs support only a single sensor per tag, increasing spatial overhead. State-of-the-art approaches to multiplexing multiple sensor streams on a single tag rely on onboard clocks or multiple modulation chains, which add cost, enlarge form factor, and remain prone to timing drift-disrupting synchronization across sensors.
We present mmBack, a low-power, clock-free backscatter tag that enables synchronous multi-sensor data acquisition and multiplexing over a single modulation chain. mmBack synchronizes sensor inputs in parallel using a shared reference signal extracted from ambient RF excitation, eliminating the need for an onboard timing source. To efficiently multiplex sensor data, mmBack designs a voltage-division scheme to multiplex multiple sensor inputs as backscatter frequency shifts through a single oscillator and RF switch. At the receiver, mmBack develops a frequency tracking algorithm and a finite-state machine for accurate demultiplexing. mmBack's ASIC design consumes 25.56uW, while its prototype supports 5 concurrent sensor streams with bandwidths of up to 5kHz and 3 concurrent sensor streams with bandwidth of up to 18kHz. Evaluation shows that mmBack achieves an average SNR surpassing 15dB in signal reconstruction.