Inband Full-Duplex Transceiver Architecture

Propagation Domain

Analog Domain

Digital Domain

OR

SI

DAC PA

ADC LNA

Coding, Modulating Decoding, Demodulating Digital Interference

Cancellation SI

Receive bits

Transmit bits a)

b)

Figure 3.1. The architecture of the common inband full-duplex terminal utilizing either separate antennas or a shared transmit/receive antenna.

As Fig. 3.1 illustrates, the IBFD transmitter [4,12,43,65] attempts to transmit bits which are coded and modulated through the digital domain. Signal is then converted to analog with digital-to-analog converter (DAC), upconverted to high carrier frequency, amplified using power amplifier (PA), and radiated using the transmit antenna. The antenna can be separated shown in Fig. 3.1 case (a) or a shared, shown in Fig. 3.1 case (b). This transmission process introduces several non-idealities to the signal. Such examples of the non-idealities could be amplifier distortion, phase noise, and DAC quantization noise. Because of the STAR capability, the terminal functions as a receiver over the same frequency band. The STAR capability means that the signal gathered with the separate or the shared-antenna is carried over low-noise amplifier (LNA), downconverter, and analog-to-digital converter (ADC).

Hereafter, the signal is processed in the digital domain which includes demodulation, bit decoding, and interference cancellation to receive desirable bitstream. Next, we discuss the propagation domain, the analog domain, and the digital domain separately as a part of the STAR process.

Propagation Domain

In the propagation domain, the transmitter and the receiver block are separated or isolated using electromagnetic properties [48, 77]. Here, the goal is to suppress the SI before it reaches receiver’s hardware [43,65]. The problem of the SI suppression in the propagation domain is that the desired signal is also suppressed often in the process [77]. The suppression

usually happens because the SI near the receiver is commonly very powerful and therefore the capability of the receiver’s hardware can easily be exceeded [48]. The primary advantage of performing the SI suppression in the propagation domain stage is that later, the receiver’s hardware does not need to process the received signal accurately with enormous dynamic range [65].

In the separate-antenna systems shown in Fig. 3.1 case (a), the propagation domain isola- tion is done by using combination of path-loss, cross-polarization, antenna directionality, multiantenna-based system, and antenna placement while in the shared antenna systems shown in Fig. 3.1 case (b) it is accomplished using a circulator [36,43,48,65,77]. A first technique, the combination of the path loss [65], means that the path loss between the receive and the transmit antennas can be increased by placing absorptive shielding between these two antennas and increasing spacing between them. Of course, the problem here is that with small devices, there is less room to implement such technique. The second technique, the cross-polarization [43, 48, 65], offers additional mechanism to isolate the transmit and the receive antennas. An example of such technique used could be that the IBFD terminal’s transmit antenna transmits only vertically polarized signals and the re- ceive antenna receives only horizontally polarized signals which minimize the interference between them.

The additional techniques in separate-antenna systems, the use of directional antennas, the utilization of multi-antenna systems, and the different antenna placements offer ways of reducing interference between receive and transmit antennas [48,65,77]. For example, the polarization decoupling techniques in the multiantenna-based system enable both of the antennas operate with the aid of orthogonal vertical and horizontal polarizations aiming their coupling reduction [65]. Also, it has been shown experimentally that up to 72 dB can be suppressed with the directional antennas [36]. The shared-antenna systems utilize a circulator [36, 43, 65] which is a ferrite devise that routes the transmit signal from the transmitter to the antenna. Also, it routes signal received on the antenna to the receiver while isolating the transmitter and receiver blocks from each other. According to [39], the SI can be suppressed roughly 20 dB by using the circulator in 2.4 GHz industrial, scientific, and medical (ISM) band.

Analog Domain

Besides that there are many ways of implementing SI cancellation in analog domain, the basic principle of analog cancellation [36, 43, 48, 65, 77] here is to first to match the reference/SI-inverse signal corresponding to the transmitted signal by properly delaying and attenuating it as Fig. 3.2 block (a) and 3.2 block (b) show. It has been observed that the perfect signal inversion can be obtained at the central frequency. Then, the signal is subtracted from receive feed as Fig. 3.2 block (c) illustrates. Because the transmitted signal experiences attenuation, gain, delays, and phase change in all practical scenarios, the identification of such phenomena have to be applied to the inverted SI. Then, SI-

inverse signal is combined with the signal at the receive antenna. Figure 3.2 illustrates the fundamental principle of SI cancellation in the analog domain.

Analog Domain

Digi tal Doma in

Prop ag ation Domai n

a) Creation of

SI-inverse signal

c) Combining

SI and its inverse b)

Delay and attenuation adjustment

Figure 3.2. Self-interference cancellation in the analog domain.

To work efficiently, sufficient amount of SI suppression must be done in the propagation- and the analog domain because ADC’s dynamic range limits the amount of the possible digital SI reduction. Thus, the primary objective is to prevent saturation of the ADC so that the digital cancellation can handle the remainder. The SI suppression must be done before the ADC, and it may occur either before or after the LNA [77]. However, other options [43,65,77] exist, for example, at first tapping the transmit signal in digital domain.

Then, applying necessary delay/gain/phase adjustments digitally and converting it into the analog domain to use it in the SI cancellation. It has been shown [21], that with the 20 cm antenna separation and analog cancellation, the SI can be suppressed up to 72 dB.

Digital Domain

The fundamental objective of the digital-domain cancellation [36, 43, 48, 65, 77] is that it aims to cancel SI after ADC by applying digital signal processing techniques to the received signal. In practice, the SI channel is at first estimated in the digital domain. The SI channel contains both leakage over through the analog domain, which is also known as the residual SI, and delayed versions of the SI signal from environment. The residual SI after analog cancellation can be subdivided into linear and nonlinear components. The linear component constitutes majority of SI power and, and hence it can be estimated by using existent algorithms, such as minimum mean square error-based and least-square techniques. The linear component alone is not sufficient therefore requiring models where also the non-linearities are taken into account. The PA mostly causes nonlinearities in the transceiver, however, an I/Q imbalance and low-cost radio transceivers can also be signif- icant issues [56]. The nonlinearity of the channel must be accurately estimated because of

high cancellation requirements in the digital domain. Finally, by using linear or nonlinear estimation on the known transmit signal, the SI is subtracted from the received signal.

According to [39], by utilizing the nonlinear signal models in the digital cancellation, 25 dB of SI suppression can be obtained.