Source-linked AI summary
NetScatter: Enabling Large-Scale Backscatter Networks
Mehrdad Hessar, Ali Najafi, Shyamnath Gollakota
TL;DR
Existing backscatter systems do not scale beyond small networks, motivating a protocol for hundreds of concurrent transmissions under weak-signal conditions. NetScatter uses distributed chirp spread spectrum coding with ON-OFF keying, plus mechanisms for near-far and synchronization challenges. In a 256-device deployment over 500 kHz, it reports 14–62x throughput gains and 15–67x latency reductions over prior approaches.
Problem
Prior backscatter systems were not designed to scale with device count, while the target network must reliably decode weak signals from hundreds to thousands of concurrent transmissions.
Method
NetScatter uses distributed chirp spread spectrum coding with ON-OFF keying, assigning cyclic shifts across devices and decoding concurrent transmissions with a single FFT while addressing near-far and synchronization issues.
Results
14–62x higher end-to-end link-layer data rates and 15–67x lower network latency than prior long-range backscatter systems are achieved in a 256-device deployment.
Takeaways & Limitations
The protocol demonstrates hundreds of concurrent backscatter transmissions over 500 kHz and supports a path to 1,000 concurrent devices using 2 MHz total bandwidth.
Takeaways & Limitations
Concurrent decoding requires all backscatter devices to be time synchronized.
Abstract
from arXiv · showhide
We present the first wireless protocol that scales to hundreds of concurrent transmissions from backscatter devices. Our key innovation is a distributed coding mechanism that works below the noise floor, operates on backscatter devices and can decode all the concurrent transmissions at the receiver using a single FFT operation. Our design addresses practical issues such as timing and frequency synchronization as well as the near-far problem. We deploy our design using a testbed of backscatter hardware and show that our protocol scales to concurrent transmissions from 256 devices using a bandwidth of only 500 kHz. Our results show throughput and latency improvements of 14--62x and 15--67x over existing approaches and 1--2 orders of magnitude higher transmission concurrency.
1 Introduction
NetScatter targets the scalability gap in low-power backscatter networks with distributed coding that supports hundreds of concurrent transmissions while addressing practical synchronization and near-far challenges.
- Prior long-range backscatter systems primarily operate at the link layer and were evaluated with only 1–2 devices, limiting network scalability.
- 256 devices concurrently transmit over 500 kHz, while the design scales to 1,000 devices using 2 MHz total bandwidth.
- The coding mechanism operates below the noise floor, runs on low-power backscatter devices, supports same-band concurrency, and decodes transmissions with one FFT.
- NetScatter introduces distributed chirp spread spectrum coding with ON-OFF keying, assigning each cyclic shift to a different concurrent device.Devices transmit a bit through the presence or absence of their assigned cyclic-shifted chirp.
- Near-far mitigation combines power-aware cyclic-shift allocation with independent SNR-based power calibration, supporting devices with up to 35 dB SNR difference.
- In a 256-device office deployment, NetScatter improves end-to-end link-layer data rates by 14–62x and reduces network latency by 15–67x over prior long-range systems.
2 CSS Primer & Existing Approaches
CSS encodes information through chirp shifts that appear as FFT peaks, but existing collision approaches impose bitrate, receiver-complexity, or concurrency limits.
- CSS Primer: CSS de-spreads upchirps with a downchirp, then uses an FFT peak’s bin index to identify the transmitted shift.
- CSS Primer: Cyclic time shifts preserve bandwidth while producing distinct FFT-bin peaks, allowing an N-point FFT to distinguish N shifts.
- CSS Primer: Increasing spreading factor improves sensitivity but decreases bitrate, while decreasing bandwidth also reduces bitrate.
- Existing Collision Approaches: Existing CSS backscatter systems do not support collision decoding, motivating collision-handling approaches developed for CSS radio systems.
- Existing Collision Approaches: Using different spreading factors requires multiple FFTs and offers too few high-rate, high-sensitivity configurations to support hundreds of devices on 500 kHz.
- Existing Collision Approaches: Choir exploits oscillator-induced fractional FFT shifts, but backscatter’s lower-frequency operation produces smaller frequency differences that limit this strategy.
- Existing Collision Approaches: NetScatter assigns distinct cyclic shifts to devices and satisfies differentiation, device association, and collision-avoidance constraints.
3.1 Distributed CSS Coding
NetScatter uses distributed CSS with ON-OFF keying to let many low-power backscatter devices transmit concurrently on the same band, while the receiver decodes their FFT-bin peaks efficiently.
- CSS modulation combined with ON-OFF keying assigns concurrent devices distinct cyclic shifts and represents bits through chirp presence or absence.
- 2^SF cyclic shifts can support 2^SF concurrent transmissions, with each device using one orthogonal FFT bin.
- A single despreading and FFT operation reveals peak presence or absence, so receiver complexity is nearly constant as device count grows.
- The aggregate throughput is BW, whereas LoRa’s corresponding throughput is BW/(2^SF·SF).
- Increasing SF raises concurrency and gain exponentially, but lowers each device’s bitrate, motivating bandwidth aggregation to preserve both.
3.2 Addressing Practical Issues
NetScatter addresses timing, frequency, and near-far impairments that can shift or mask FFT peaks, combining cyclic-shift spacing with adaptive power control.
- Timing Mismatch: Timing offsets shift demodulated peaks by ∆FFTbin = ∆tBW, so synchronization is required to prevent adjacent assignments from interfering.
- Timing Mismatch: Bandwidth-based cyclic-shift assignment leaves SKIP−1 empty adjacent bins, preventing variable hardware delays from causing neighboring-device interference.
- Timing Mismatch: Increasing bandwidth increases timing-induced FFT shifts and empty-bin requirements, trading total throughput against per-device bitrate and sensitivity.
- Frequency Mismatch: Frequency offsets cause FFT-bin shifts, but crystal tolerances translate to less than one bin under the paper’s bandwidth and spreading-factor settings.
- Near-Far Problem: Power-aware cyclic-shift allocation keeps BER unaffected for simulated power differences near 40 dB and tolerates up to 35 dB in practice.
- Near-Far Problem: Self-aware power adjustment uses the AP query’s signal strength and three device power levels to track changing SNR from mobility and fading.
- Near-Far Problem: NetScatter deliberately reduces high-SNR devices’ bitrate to enable more concurrent transmissions at a fixed bitrate.
3.3 NetScatter Protocol & Receiver Details
The protocol uses AP queries to synchronize devices, assign cyclic shifts, and support association alongside ongoing communication; shared preambles let the receiver detect and decode concurrent transmitters.
- Protocol Operation: The AP query synchronizes participating devices, conveys signal-strength-based cyclic-shift assignments, and supports device power fine-tuning.
- Packet Structure: Concurrent packets use six upchirps and two downchirps before payload and checksum, with shared preambles reducing per-device overhead.
- Receiver Operation: The receiver locates packet start from the upchirp–downchirp midpoint, then detects transmitters through repeated FFT peaks across preamble symbols.
- Receiver Operation: Average preamble peak power provides each device’s payload threshold: peaks above half that average decode as 1, otherwise 0.
- Association: Association reserves Nassoc cyclic shifts so incoming devices can request entry concurrently with existing communication rather than waiting for dedicated periods.
- Association: The AP assigns the incoming device’s cyclic shift and timing schedule after measuring its association signal, then piggybacks those assignments in a query.
- Query Message: The query includes a group ID for up to 256 concurrent devices, while association responses add an 8-bit network ID and 8-bit cyclic shift.
4 Evaluation
NetScatter’s evaluation demonstrates practical operation across hardware, synchronization, near-far, and network-deployment conditions. In a 256-device deployment, it achieves substantial throughput and latency gains over LoRa backscatter.
- Hardware frequency variations remain below 150 Hz, nearly 0.15 FFT bin at BW = 500 kHz and SF = 9.
- 256 devices achieve around 250 kbps aggregate throughput and around 1 kbps per-tag bitrate with SKIP = 2.
- 35 dB is the maximum practical power difference tolerated between two concurrent transmissions under power-aware cyclic-shift assignment.
- 6.8x and 26.2x are the PHY bit-rate increases over LoRa backscatter with and without rate adaptation, respectively, at 256 devices.
- 61.9x (50.9x) and 14.1x (11.6x) are NetScatter’s link-layer gains over LoRa backscatter without and with rate adaptation for config#1 (#2).
- 67.0x (55.1x) and 15.3x (12.6x) are NetScatter’s latency reductions over prior LoRa backscatter without and with rate adaptation for config#1 (#2).
5 Related Work
Related work includes Wi-Fi and LoRa-based backscatter systems, as well as RFID concurrent-decoding approaches. These systems differ from NetScatter in range, below-noise operation, or supported concurrency.
- Wi-Fi backscatter systems have receiver sensitivity of only -90 dBm, limiting range and cross-room operation unless the RF source is nearby.
- LoRa backscatter achieves long range by generating LoRa-compliant packets at the backscatter device, while pLoRa backscatters ambient LoRa signals.
- Prior long-range backscatter systems are evaluated in networks of only 1–2 devices.
- RFID concurrent-decoding systems exploit time-domain transitions or constellation changes but do not provide CSS-based long-range and below-noise operation.
6 Conclusion
NetScatter is a backscatter-network protocol designed to scale to hundreds of concurrent transmissions. It combines distributed chirp spread spectrum coding with practical solutions for interference and synchronization, demonstrated in a 256-device indoor deployment.
- NetScatter introduces distributed chirp spread spectrum coding using chirp spread spectrum modulation and ON-OFF keying.
- The system addresses the near-far problem and timing and frequency synchronization in an indoor deployment with 256 concurrent devices.