The project has made advances in designing RF backscatter tag to tag link where both backscatter modulator and backscatter receiver are passive devices. An external RF signal from an exciter provides both the power as well as the carrier for backscatter. This is a significant departure from conventional RFID where one end of the link uses a powerful active device (reader). The external RF signal could be an ambient RF signal if it can provide high enough power to the tags. Otherwise, intentionally deployed exciters need to be used (more common case). The project has so far developed multiple generations of tag prototypes, evaluated them experimentally and developed several modeling and analysis methods. The project also developed multiple innovative ideas related to tag-to-tag backscattering.

Tag Prototyping

Prototype tag designs have been completed and its performance has been evaluated. The design follows a modular approach. The excitation signal is derived form an external continuous wave (CW) signal generator operating at 915MHz. We also have tested the tags with SDR generated TV signal in this band. Much of the basic design and related analysis are covered in publications [1-5]. One specific challenge we face in this design is to be able to demodulate the backscatter signals with very low modulation index using entirely passive techniques.

The major modules implemented on PCB using discrete components are below.

1. Receive Section:

This module consists of a matching circuit designed to maximize power transfer, a passive envelope detector, an RC filter to cut off unwanted frequencies, leaving only baseband signal, and a comparator to digitize the analog signal. The detector is built using a multi-stage diode detector circuit. The goal is to optimize the ability of the detector to reliably detect backscatter signals from other tags in the network in the presence of external excitation signal. Various analysis and advanced ideas are discussed in publications [1,3] below.

2. Backscatter Modulator Section:

We constructed a backscatter modulator module using an RF SPDT switch and associated circuitry so that the switch can change the antenna load impedances thus varying its reflection coefficient. The choice of right load impedance levels is critical for avoiding the phase cancellation problem [4] that arises in tag-to-tag backscatter networks. Phase cancellation happens when the excitation signal and backscattered signal combine in opposite phase to produce signal nulls at the receiving tags making it very hard for a passive detection circuit to detect the backscatter. This is addressed using a technique called multiphase backscatter where the tags have the ability to change the phase of the backscattered signal by changing the antenna load impedance [2,4]. Availability of such phase diversity can help in other applications too, a topic of our current work.

3. Digital Section:

We implemented the digital section module using a low-power microcontroller. The microcontroller reads and decodes the digitized signal from the Receive Section and drives the RF switch for the Modulator Section. It also performs the necessary link and network layer functions. The microcontroller is generally in a low-power idle mode while tags are in listen mode. It wakes up only when a pilot tone is received that is inserted as a preamble to each frame.

4. Prototype Implementation:

The prototypes are made of 2-layer FR4 PCB in thickness of 31 mils with discrete off-the-shelf components on both sides. The PCB was designed using Altium designer software. The antenna is implemented directly on a separate piece of PCB that is attached to the main PCB using SMA connector. This modular design helps future experiments with different types of antennas. The design is available by contacting the project PIs.

5. Power Consumption:

The power consumption depends on three components on the tag - comparator in the receive section, switch in the modulator section, and the microcontroller. Discounting the microcontroller (as is the practice in related literature), the power consumption is roughly 140 microwatts when using discrete components, primarily due to the comparator. While lower power comparators are possible with limited design changes, this will also limit the design in terms of data rate and range. However, in an IC implementation, the power consumption can reduce to below1 microwatt.

Communication Performance

1. We are able to demonstrate tag-to-tag links up to about 3 m at 5Kbps when the available excitation power is about -20 dBm. We get a significantly better range (up to about 10 m though with some gaps) at the same rate when the power increases to about -15dBm.

2. We are able to demonstrate multihop operation for up to 4 hops in a static routing set up, thus enabling tag-to-tag communication over about 12 m at 5 Kbps with only -20 dBm available excitation power. See [2].

3. Our work in [1] has demonstrated that simulated tags implemented using 45nm CMOS technology are able to demodulate ASK backscatters with very low modulation index (0.6%) consuming only 1.2 microwatts at a data rate of 10Kbps. Our later work shows potential of bringing down the power consumtion to the order of 100nW (e.g., [8]).

Addressing Phase Cancellation

The phase cancellation problem occurs at the receiving tag and results from the superposition of the backscatter signal from the transmit tag and the external excitation signal used for backscattering. These signals could combine in opposite phases depending on the nature of the multipath channel and tag/exciter locations leading to signal cancellation and zero or very poor modulation index [4,5]. While there are various avenues to address this, we have explored use of phase diversity at the backscatter modulator as a practical mechanism.

This is done by using a multi-port backscatter modulator that provides the tag the ability to backscatter in one of several phase channels. Each channel is characterized by a different reflection coefficient used for backscatter, and adjacent channels are separated by a fixed and deterministic phase differences. The basic idea is to have the ability to transmit a message over multiple phase channels available on the tag. Then, at the receiving tag, even if the message is canceled or severely attenuated in one channel, it will still be detected in another channel. Choosing the appropriate values for the reflection coefficient and hence the load impedances for the different phase channels is an important step in the implementation of the proposed scheme. We developed a systematic methodology for doing this and developed prototype tags capable of backscattering in up to 2 phase channels (a larger tag with more channel options is in the works). Experimental results and demonstration of how this diversity actually improves data communication performance is available in [2].

Characterizing Backscatter Channel using Multiphase Backscatter

The amplitudes of the received signal at different phase channels could be used to characterize the wireless backscatter channel between the Tx and Rx tags. Changes in the channel due to change the surroundings or movement of tags also change this amplitude. Now, if a sizable number of phase channels are used (say, 4 to 8) the vector of these amplitudes could be considered as a fingerprint of the channel. This fingerprint (called Backscatter Channel State Information or BCSI) changes in a predictable fashion due to specific changes in the surroundings. This property makes BCSI an effective tool to recognize human activities around the tags. This observation is very powerful. In essence, BCSI provides a similar power to measure wireless channels and use it for activity recognition as regular CSI-based activity recognition that has been popular in recent literature. But the latter must use active, high power radios. Using BCSI we can do the same using passive RF tags. The overall vision is explained in our work in [6]. In the early experimental results reported in [6] with 9 participants and 8 activities of daily living, the average error for activity recognition was only 6%.

Passive Wireless Channel Estimation

We develop passive RF tags with the ability to estimate the RF parameters of the wireless channel between pairs of communicating tags. This ability is the key to detect interactions and events. With such capability, the network of RF tags can provide a real-time, precise, fine-grained RF fingerprint of the environment. As the tags continuously sense the parameters of the RF channel (amplitude and phase) between neighboring tags, they can detect and classify activities in their environment. We have previously reported this capability in our work in ACM Mobisys 2018 [6] and IEEE ICASSP 2019 [7]. More refined techniques for isolating the tag-to-tag channel has been recently developed in our work in ACM DFHS Workshop 2019 (with ACM Sensys/Buildsys) [16].

Sampling the received baseband signal at different reflecting phases at the backscattering (Tx) tag enables estimation of amplitude and phase of the tag-to-tag channel. The low-power implementation of the channel estimator, after envelope detection, integrates amplification and filtering of the baseband signal that is followed by analog-to-digital conversion.

The backscattering Tx tag switches between a set of terminating impedances that represent different reflecting phases in order to compute amplitude and phase of the tag-to-tag channel. The channel estimator of Rx tag measures the amplitude of the baseband signal at each reflecting phase to determine the amplitude and phase of the tag-to-tag channel. The technical details of this available in our ACM Mobisys 2018 [6] and IEEE ICASSP 2019 [7] papers. The quantification of the baseband signal is challenging due to the low modulation index of the received RF signal and low sensitivity of envelope detector, as well as wide range of the possible input RF powers and modulation indexes.

We developed the architecture and circuit implementation of the channel estimator such that it is able to operate in a very low power budget. The expected power budget of our tags is in the order of 100s of nW. This enables the RF tag operate with the targeted input RF power of -25 dBm. The detailed description of the design is available in our paper in IEEE ISCAS 2019 [8].

We have designed and implemented the channel estimator in 65 nm CMOS technology, has sensitivity of -45 dBm at 2.5% modulation index and consumes 122 nW. The circuit simulation results are available in the IEEE ISCAS 2019 paper [8].

Doppler Shift Measurements

We used the channel characterization approach above for Doppler shift measurements to be done entirely on the passive tags. While such measurements have been shown to be possible on self-powered RF tags, previous work used regular RFID tags and such measurements have been done on active readers. We are now able to do the same without the involvement of any active device. The method exploits the concept of multiphase probing and, from it, inferring information about changing distances between two communicating tags. We show that we can measure Doppler shifts on passive tags almost as accurately as with costly conventional RFID readers. With our experiments, we demonstrate that with two tags in an office environment the median tracking error is about 2.5cm. This is comparable to previous work that requires the use of RFID readers. The modeling approach and experimental results are reported in IEEE ICASSP 2020 [13].

Application of Passive Channel Estimation to Ranging and Localization

We further refined the tag-to-tag backscatter channel estimation technique and developed a methodology to estimate the range between two communicating tags. The range estimation method is then used for tag localization.

All known backscatter tag localization techniques rely on active receivers (RFID readers or similar devices) for measuring and characterizing the received signal. As a result, they cannot be directly applied to passive tag-to-tag networks as in PNET. Our work effectively overcomes this challenge by developing a localization technique for such passive networks by using phase-based ranging in passive receivers. This method allows pairs of passive tags to collaboratively determine the inter-tag channel phase, while effectively minimizing the effects of multipath and noise in the surrounding environment. The determination of the channel phase is based on our previously developed multiphase probing technique. We have shown that the phase measurements are accurate enough to be able to provide a range estimate between two tags modulo the wavelength. The `modulo wavelength' factor is due to the phase wrap-around every 2 pi radians. The accuracy of the range estimate here (median error < 1cm) with our tag hardware is comparable to active radio-based techniques (see attached figure). We validated that the nature of the error is very similar to measurements with Gen2 RFID tags and commercial RFID readers in a similar environment (of course, our scheme avoids the need for bulky and expensive RFID readers).

Building on this accomplishment, we have developed a localization technique that benefits from the large link diversity, which is uniquely available in a passive tag-to-tag network. Traditional tag localization methods are limited only to measurements on tag-to-active- receiver links. Since the number of active receivers is expected to be far smaller than the number of tags (active devices are larger, more expensive and power-hungry), tag-to-tag measurements provide a clear advantage of simply having measurements on many more links. Our technique does not require active receivers. When ranging estimates are formed on all possible tag-to-tag links, the localization technique benefits tremendously from the richness and diversity of many links. The localization technique is likelihood-based - it finds the most likely locations of all tags in the network that are consistent with the available range measurements.

Our measurement and analysis have shown that this approach is indeed able to provide much better localization accuracy when compared to an equivalent active receiver-based technique with no tag-to-tag measurements available. Overall, our median localization error is about 2 cm. A paper describing the entire technique is under preparation.

Receiver Circuit Implementation

The resolution of the channel estimation is limited by the resolution of the envelope detector based receiver. The main challenge in the design of the receiver is resolving the weak received reflected signal that is superimposed on the continuous wave signal directly received from the excitation source under stringent power limit. We have explored an architecture of the receiver that includes an active amplification stage in the envelope detector. We have demonstrated that a novel self- biased common-source based envelope detector provides sufficient conversion gain and at the same time operates with a low power consumption. We have also studied the feasibility of application of the tag-to-tag link for the communication between free floating mm-sized brain implantable devices, as well as the use of the proposed demodulator in this application case. The feasibility study is reported in the IEEE UEMCON 2019 paper [14]. The circuit details and simulation results of the proposed modulator are reported in the IEEE ISCAS 2020 paper [15].

Energy Harvesting and Management for RF Tags

The RF tags work in an extremely power-constrained regime using the energy harvested from the CW signal emitted by the exciter for their operation. As the RF tags operate on the power budgets as low as 10s of nWs, optimization of the energy harvesting and management poses some unique challenges compared to the implementations used in the conventional sensory nodes. These implementations use DC-DC converters that hinder extremely low-power operation and suffer from bulky off-chip passive components. We have devised an architecture of energy harvester circuit, along with the management strategy, that is tailored for the passive near-zero power devices like RF tags. The energy harvester efficiently operates over a wide range of incident RF power leveraging a novel adaptive charging circuit. The adaptive charging circuit optimizes the output impedance of the rectifier that converts the incident AC voltage to DC voltage and controls the delivered energy to a low-capacity storage element. This preserves a high efficiency of the energy conversion and enables operation of the RF tag at low incident RF power level. The energy harvester implementation in 65 nm CMOS process demonstrates the power efficiency of the energy harvester higher than 50% over the input power range from -25 dBm up to -5 dBm. The energy harvester implementation and energy management are further described in a conference paper published in ISCAS 2021 [19].

Routing, MAC, and Scheduling in RIBBN (at UT-Dallas)

Specific Objectives:

In the first year of our work on this project, we have concentrated on following objectives of RIBBN: (1) design of scheduling algorithms RIBBN with different hardware capabilities of the tags; (2) design of a distributed MAC protocol for RIBBN, and (3) design of an algorithm for energy transfer protocol for RIBBN.

Significant Results:

To unleash the potential of RIBBN communication, we proposed a novel backscattering operation, which allows implementation of a multihop RIBBN, improving the network coverage and the distances of the communicating tags by orders of magnitude. Due to the asymmetric communication links and interference among tags’ transmissions in a RIBBN, the routing protocol design has become one of the main technical challenges. In this work, we designed different RIBBN routing schemes for three distinct types of tags with different hardware capabilities. We evaluated the performances of the proposed protocols, and we study the impacts of several network parameters on the network capacity. We also compare the proposed routing protocols, as to investigate the performance gains due to the tags’ different hardware capabilities.

Other achievements:

Due to the asymmetry of communication links in RIBBN, existing routing protocols cannot be used. To address this shortcoming, we proposed a basic protocol to identify tags and their multiple-hop uplink routing paths. Then we considered three types of passive RFID tags with different capabilities (i.e., received power measurements and transmission power attenuation), we proposed three corresponding routing protocols to schedule transmissions, as to maximize the network throughput. Computer simulations verified the effectiveness of the proposed protocols in improving the concurrency of the transmissions (i.e., the throughput) due to the added tag capabilities. As one example of the results of our study, we concluded that power measurement capability combined with transmission attenuation of the tags can significantly improve the overall network throughput especially for adaptable network topologies.