Coded Backscattering Communication with LTE Pilots as Ambient Signal

The Ambient Internet of Things (AIoT)  [1] presents a demand for communication solutions tailored to Zero Energy Devices (ZEDs). AIoT relies on harvesting environmental energy to meet its power requirements. The 6G flagship project Hexa-X II  [2], in continuity of Hexa-X-I  [3], is committed to designing such zero-energy communication solutions in general, and to developing and trialing on the field one particularly promising technique called Ambient Backscatter Communication (AmBC) [4]. In the AmBC approach, backscatter devices (BDs) do not generate their own signals; instead, they rely on ambient radio frequency (RF) signals such as television broadcast and cellular signals for illumination. Information in AmBC is conveyed through modulation of the propagation channel, as BDs either reflect or absorb ambient RF signals.

The primary advantage of utilizing backscatter communications in AIoT resides in its exceptional energy efficiency. When compared to active transmitters, backscatter devices demonstrate a remarkable reduction in energy consumption, approximately 1000 times less, as discussed in [5]. This substantial reduction in energy consumption lays the foundation for the deployment of ZED, which can operate without batteries. Such devices have valuable practical applications, including asset tracking [6] and proximity-based localization [7].

A significant challenge in AmBC is dealing with direct path interference (DPI) caused by the higher power of the direct path signal compared to the scattered signal. Another challenge, especially in the case of ambient cellular signals, is the intermittent and bursty nature of the signals.

In long term evolutio(LTE) downlink, the receiver at the user equipment (UE) first synchronizes with the Base Station (BS, a.k.a eNB) using primary and secondary synchronization signals (PSS and SSS) and then employs the downlink Cell-Specific Reference Signals (CRS) to estimate the channel [8]. In [9, 10], an in-band ambient Frequency Shift Keying (FSK) backscatter communication system was proposed for LTE downlink, leveraging the UE receiver for demodulating the backscatter-generated FSK modulated messages. There reason for FSK modulation scheme is that FSK is a better modulation scheme than BPSK in AmBC [11].

The LTE channel estimator is engineered to withstand significant Doppler shifts, typically encountered in high-mobility scenarios. However, it’s essential to emphasize that in the majority of indoor and pedestrian deployments, the Doppler shift remains considerably below the maximum threshold. The proposed in-band LTE backscatter system effectively harnesses this excess bandwidth to facilitate backscatter communication. The BD employs FSK to introduce an artificial Doppler shift, which is larger than the natural Doppler within the channel yet smaller than the maximum Doppler tolerated by the channel estimator. This strategic choice enables the receiver to discern BD-modulated messages from the natural multi-path components in the Doppler frequency domain, thereby resolving the DPI issue. Additionally, CRS symbols are transmitted with fixed and often boosted power, even during cell idle periods, enabling the receiver to independently demodulate BD messages without dependence on LTE downlink traffic.

In prior studies concerning in-band LTE backscatter systems, the primary emphasis has been on uncoded modulation as the predominant operational mode for the BD. Empirical measurements [12] have unveiled that the system’s bit-error rate performance is poor in many practical scenarios. In response to this challenge, the integration of forward error correction (FEC) coding within the BD is proposed in this paper. To optimize the utilization of existing receiver components within the UE, the adoption of the same convolutional code for BD messages as is conventionally employed for downlink control messages is recommended. We validate the proposed system with over the air measurements by using ambient signal from commercial BS and processing the BD signal as it would be if an LTE UE would be the BD signal reader.

The rest of paper is organized as follows: Sec. II discusses AmBC model and how BD signal is visible on UE side. Then, Sec. III outlines the AmBC receiver signal processing flows. Next, Sec. IV shows the experiment measurements under commercial LTE BS. Finally, Sec. V draws conclusions of this study.

The top-level of in-band LTE backscattering based AmBC system consists three parts, BS, UE and BD. BS continuously boardcasts LTE downlink pilots. Those ubiquitous continuous pilots enlighten BD. UE always estimates the channel base on pilot CRS signals. BD modulates the channel blindly, without know any ambient information. Once UE is close to BD, it can recognize the channel changes. In this way, information is sent from BD to UE.

As Fig. 4 illustrates. BS sends CRS to UE. BD reflects or absorbs the incident RF ambient signal, toggles states with different frequency and generates artificial Doppler frequencies. Those channel changes are picked up by UE, i.e. UE as AmBC receiver demodulates BD information based on the channel estimates.

The AmBC receiver efficiently repurposes the LTE Rx channel estimation, convolutional coding algorithm, and cyclic redundancy check (CRC). These components are intentionally designed with identical parameters to minimize receiver cost allowing the existing software components in the modem to be reused. The block diagram of the proposed backscatter receiver is illustrated in Fig. 1. The frame packet structure is organized as Fig. 2. In this section, we explain the FSK demodulator and convolutional decoder in more details.

To evaluate the feasibility of the proposed system, an experiment was conducted using real commercial LTE downlink signals within the indoor environment of the second floor of the Aalto University TUAS building in Espoo, Finland. Fig. 4 illustrates the wireless LTE downlink measurement setup. Tab. I presents some essential parameters related to the BS, BD, and UE. The experiment uses a commercial BS, providing a reference signal received power (RSRP) of approximately -70 dBm at the measurement location. The distance between the BD and UE is half a meter.

The BD operates create its packets without any knowledge of the LTE downlink signals. The BD was a Rasperry Pi connected to a diode based backscatter module. The generated frame was 13.08 s. Each frame was followed by half a second sleep after which BD restarts transmission. This half-second silence serves the purpose of separating two consecutive frames, allowing the receiver to synchronize with the next frame using its synchronization header.

On the reader end, the UE equipment is emulated as a software-defined radio (SDR) with computer and B210 USRP frontend from National Instrument. This emulated UE synchronizes with the commercial LTE downlink, performs channel estimation, and subsequently transmits the channel estimates to a laptop via the User Datagram Protocol (UDP).

To counteract clock drifting in B210 the frame length is deliberately kept short, and synchronization occurs frequently. This approach helps prevent the accumulation of clock errors, ensuring that symbols can be demodulated accurately within their correct symbol duration. Although there are many good channel coding for long packet, but that clock drifting prevents to design long packet.

The received channel estimates are post processed in MATLAB program, which serves as the AmBC receiver module depicted in Fig. 1.

It’s worth noting that, in comparison to mass-produced UE transceiver RF front-ends, the USRP is equipped with a 12-bit analog-to-digital converter (ADC) and is not using automatic gain control (AGC). To achieve an optimal balance between the dynamic range, quantization error, and signal clipping, the gain settings are selected by hand.

The measuements we done by keeping measurement setup in Fig. 4 in location. The measurement campaign was carried out during two days. During that time the channel varies due to users movement in surrounding. During the campaign we collected 6050 packets. For each packet we recorded the signal-to-noise ratio (SNR) and amount of errors. In BER computation we split the SNR into 0.25 dB intervals. The final bit error rate (BER) is computed over all the packets that are in the same interval Fig. 5.

In Fig. 5 we give also uncoded and convolutional coded FSK theoretical performance curves. The theoretical noncoherent FSK BER $P_{b}$ is [21, Eq. 4.5-46]

where $\mathcal{E}_{b}$ energy per bit and $N_{0}$ is noise power density. The convolutional coding reference curve is evaluated by Monte Carlo simulation. For a convolutional code we can estimate a coding gain as

For the used convolutional code and BPSK modulation the coding gain upper bound is 6.99 dB [21, Tab. 8.6-1]. In Fig. 5 the coding gain at BER $10^{-3}$ is about 6.28 dB.

At high SNR the measured results align well with the theoretical curve. At low SNR the measured performance is better than theory predicts. This discrepancy is explained by synchronization failure at low SNR. In our measurements we ignored packets that we could not synchronize to. That biased the measurements. Conversely, at high end of SNR the outliers are due to the lack of measured packet samples in this area.

The measurement results demonstrate that AmBC is viable even under a 5 dB SNR, achieving a BER of less than $10^{-3}$. This represents a gain of at least 6 dB when compared to uncoded alone. Furthermore, the implemented AmBC receiver takes advantage of LTE UE receiver. It reuses the UE existing modules such as channel estimation, convolutional channel decoder, and CRC.

In this paper, we have integrated convolutional channel coding and CRC into an AmBC system that relies on LTE downlink pilots. Our experiments conducted in a real commercial LTE environment validate the feasibility of AmBC with the proposed receiver. The introduction of convolutional channel coding has led to a noteworthy 6 dB improvement in BER performance, particularly at a challenging SNR of 5 dB. The results consistently show that this AmBC system can achieve a BER of at least $10^{-3}$ under the specified 5 dB SNR conditions, which are relevant for a BD and UE separation distance of half a meter. In summary, our research suggests the potential applicability of our AmBC solution in the context of 6G ZED communication, representing a notable advancement in the field.

This work is in part supported by the European Project Hexa-X II under (grant 101095759) and Business Finland Project eMTC (Dnro 8028/31/2022).