Frequency-dependent beam steering of LWAs

In general, LWAs employ a waveguide structure along which a guided electromagnetic wave propagates and controlled leakage is exploited to form radiation. For this reason, such antennas are also referred to as “traveling wave” antennas. For LWAs, the frequency dependence of the phase parameter is then typically used to realize the beam steering characteristics of the antenna as it can be seen from equation (1) for uniform LWAs [Reference Chen, Liu, Nakano, Qing and Zwick27].

(1)\begin{equation}{\theta _{\text{u}}} = \,\arcsin \left( {\frac{{\beta \left( \omega \right)}}{{{k_0}}}} \right)\end{equation}

Here ${\theta _{\text{u}}}$ is the steering angle of a uniform LWA, $\beta \left( \omega \right)$ the frequency-dependent phase parameter, and ${k_0}$ the vacuum wave number.

The introduction of periodically repeating unit-cells into the guiding structure, results in a leakage and thus a forced radiation of the wave at these interruptions by translating the non-radiating slow wave into a radiating fast wave. This also affects the phase parameter according to equation (2) as in principle an infinite number $n$ space harmonics are created. The antenna design is then optimized to ensure an unitary fast space harmonic of $n = - 1$. The incorporation of periodic structures offers a further advantage: in contrast to uniform LWAs, radiation is also possible at angles smaller than the broadside of the antenna. Thus, the steering angle ${\theta _{\text{p}}}$ that can be addressed in this way is significantly larger and can be approximated by equation (3) considering the desired first order Floquet wave ($n = - 1$) at the periodic disruptions with $p$ being the periodicity of the unit cell [Reference Jackson, Caloz and Itoh28].

(2)\begin{equation}{\beta _{\text{n}}}\, = {\beta _0} + \frac{{2\pi n}}{p}\end{equation}

(3)\begin{equation}{\theta _{\text{p}}} \approx \,\arcsin \left( {\frac{{{\beta _0}\left( \omega \right)}}{{{k_0}}} - \frac{{2\pi }}{{{k_0}\,p}}} \right)\end{equation}

Moreover, this distinctive frequency dependency of the radiation direction can be used to serve a communication scenario with several users, as each frequency can be radiated at any time – i.e., also simultaneously – and thus in different spatial directions. The detailed design of the utilized LWA is discussed in the following subsection.

THz LWA design

The antenna design utilized in this work is based on periodically altered microstrip line as it can be seen in Fig. 1 and is optimized for an operation within the WR3.4 band (230–330 GHz) using CST Studio Suite from Dassault Systems. Indium phosphide (InP) is used as a substrate and allows to enhance the steering range due to its high dielectric constant of ε _r = 12.4 as the guided wavelength is affected. The antenna is fabricated on a 50 µm thin InP substrate and features 20 unit-cells, one of which is shown in more detail in the inset highlighting its asymmetry. Here, by incorporating a matching stub the symmetry is eliminated and the typically occurring open stopband effect in periodic LWAs is reduced. As reported in paper [Reference Haddad, Biurrun-Quel, Lu, Kaya, Mohammad and Stöhr29], a significant increase in the radiated power is realized by introducing this asymmetry, resulting in a flat gain curve compared to a design without this asymmetry.

Figure 1. Schematic LWA design based on a periodically altered microstrip line. InP is used as a substrate to allow for enhanced beam steering angles due to its high dielectric constant. A zoom-in of a unit-cell is given on the right side. It highlights the asymmetry that is used to overcome the open-stopband effect.

The simulated steering behavior of the antenna for the complete WR3.4 band is depicted in Fig. 2. It shows a steering range of ∼90° in the frequency range of 230–330 GHz. Due to the fact that the antenna must be contacted with a ground-signal-ground (GSG) probe, not the entire steering range can be used in practice, as reflections occur due to the probe when radiating in the backfire direction. The angular coverage considered in this work is therefore limited to angles larger than the broadside at 273 GHz. A comparison of simulated and measured realized gain for the selected region of 280–320 GHz (∼5°–35°) is given in Fig. 2 on the right. To determine the radiation pattern a spherical mm-Wave/THz antenna measurement system was used. The measurement resolution of the system is 1° in both the elevation and azimuthal angle. In addition to a very good agreement between simulation and measurement, it also shows a flat course of the realized gain of ∼14 dBi. It is important to mention that the LWA not only enables the emission and control of a single beam, but that all frequencies and therefore all radiation angles can be addressed at the same time. This is particularly important for the point-to-multipoint communication link.

Figure 2. The simulated steering behavior of the LWA is given on the left for frequencies from 230 GHz to 330 GHz indicating a steering range of around 90°. A detailed view of the frequency and beam steering range used in the work (280–320 GHz) is shown in the graph on the right. The simulated and measured values show a good agreement and a constant peak realized gain, which is around 14 dBi for each frequency. The measurement was carried out with a spherical mm-Wave/THz antenna measurement system with 1° resolution.

It should be noted that only beam steering along one spatial angle (elevation) is discussed here. However, a concept for extending beam control with LWAs to the second dimension (azimuth) has already been shown in [Reference Haddad, Biurrun-Quel, Lu, Tebart, Sievert, Makhlouf, Grzeslo, Teniente, Del-Río and Stöhr21] by incorporating a linear phase distribution network at the input of a LWA array.

Photonic transmitter for THz beam steering

As can be seen in the schematic in Fig. 3a, the waveform radiated by the LWA in the range of 300 GHz is generated by optical heterodyne mixing in an uni-travelling carrier photodiode (UTC-PD, NTT J-Band photomixer). More precisely, the photodiode generates an output signal ${f_{{\text{THz}}}}$ which corresponds to the frequency difference of the two laser lines provided by tunable external cavity laser diodes (PurePhotonics PPCL200). Due to the enormous tunability of the lasers, it is possible to generate signals from low frequencies (MHz) up to the THz range. However, the limiting factor here is the frequency-converting component, i.e., the photodiode. In our setup, the optoelectronic conversion is optimized for the WR3.4 range, meaning the operational band of the LWA. In order to transmit a data signal in the wireless path, one of the two optical carriers is intensity-modulated with data using a Mach–Zehnder modulator resulting in two sidebands. The data sequence is in turn generated by a field programmable gate array (FPGA) in combination with a high-speed digital-to-analog converter (DAC) in the Keysight Universal Signal Processing Architecture (USPA). To optimize the performance and quality of the signal, it is first amplified using an erbium-doped fiber amplifier (EDFA, LiComm poA) and then bandpass filtered to minimize the out-of-band amplified spontaneous emission (ASE) noise of the EDFA. Both optical paths are combined together using a 3 dB coupler. Here, it is ensured that the power levels are adapted to each other for optimum modulation depth. In addition, to be able to generate a sufficiently high power with the photodiode, further amplification is then required and provided by a second EDFA (Thorlabs 100P) with again subsequent filtering to minimize the ASE noise. The feeding power is additionally increased to about 3 dBm using a THz medium power amplifier (Fraunhofer M155AMPH) and fed to the antenna by a GSG probe (FormFactor I325-T-GSG-100BT). Taking the GSG probe losses into account, the coupled power to the antenna is about −3 dBm.

Envelope receiver for THz signal reception

At the receiver, after wireless transmission, the THz signal is detected using a comparatively simple envelope detector architecture. This comprises a horn antenna with 25 dBi gain (VDI WR3.4 diagonal horn) for signal reception followed by a power amplifier (Fraunhofer M145ALNH) with an additional gain of 24 dB at the cost of a 10 dB noise figure. Down-conversion is then carried out by a zero-biased Schottky-barrier diode (VDI WR3.4ZBD-F40) which output is digitized via a digital storage oscilloscope (DSO) with 33 GHz analog bandwidth and 80 GSa/s sampling rate (Keysight DSA-X 93204A). To allow for angle-dependent measurements and thus verification of the beam steering properties, the THz receiver is mounted on a goniometer. The offline DSP used to evaluate the received signal is described in the following subsection.

Digital signal processing

Receiver DSP chain

For the reconstruction of the wirelessly transmitted signal, the DSP chain shown in Fig. 4 is used at the receiver side. In the first step, the received waveform is digitized using a DSO. The signal is resampled to two samples per symbol, as required by the subsequent timing recovery. Here, we apply a feed-forward timing recovery according to the algorithm proposed by Barton and Al-Jalili [Reference Barton and Al-Jalili30, Reference Matalla, Mahmud, Füllner, Freude and Randel31].

Figure 4. Digital signal processing chain to recover the transmitted waveform. After signal reception using a digital storage oscilloscope, resampling to two samples per symbol is carried out. The following feed-forward timing recovery and a subsequent blind Sato equalizer allow to reconstruct the transmit (Tx) symbols. Evaluation is then performed in terms of constellation signal to noise ratio and bit error ratio.

The next step in recovering the signal involves the application of a fully blind equalizer. This allows mitigation of channel impairments, e.g., transceiver bandwidth limitations and channel effects. The approach is based on a minimization of the cost function, which describes the distance between the expected and the actual symbol value, as proposed by Sato [Reference Sato32]. The necessary filter coefficients are then adapted blindly using the stochastic gradient descent method.

This brings advantages in terms of the transmitted net data rate compared to preamble-based equalizers, which determine the distortion of a channel by transmitting a known symbol sequence. In addition, down-sampling by a factor of two is performed after the Sato equalizer and, as a result, the transmitted symbols are recovered at the end of the DSP chain. An evaluation of the recovered signal is then carried out on the basis of constellation signal to noise ratio (CSNR) and bit error ratio (BER) to provide an indication of the transmission quality. The results are presented in the following subsections.

THz beam steering communications for single users

In the first step, to obtain information about the transmission quality in different spatial directions, the received signal was evaluated on the basis of the CSNR for two distances (20 cm and 50 cm) and symbol rates (15 Gbaud and 25 Gbaud) at different THz center frequencies. The results are illustrated in Fig. 5. A rather flat CSNR characteristic can be seen, particularly in the 290–310 GHz range (13°–28°), which confirms the expectations from the simulations with regard to the linear antenna gain over this frequency range.

Figure 5. CSNR as a function of the beam steering angle is depicted on the left. The drop at the edges of the investigated frequency band is explainable by the frequency roll-off of the THz components and misalignment of the lens. On the top right, the histogram at 13° steering angle (290 GHz) and a symbol rate of 15 Gbaud at 20 cm wireless distance is given. In contrast, the histogram down right corresponds to 25 Gbaud and 50 cm at the same steering angle.

However, when approaching the edges of the steering range considered here, a stronger drop in the CSNR can be observed, especially at higher distances and symbol rates. In our scenario, this can be explained partly by the frequency roll-off of the THz components (photodiode, Tx and Rx amplifier) and partly by a slight misalignment between Teflon lens and LWA, which leads to a broadening of the beam. In addition, the frequency-dependent beam steering behavior of the LWA must also be taken into account when considering larger distances, as this is becoming more significant. From the steering parameter (0.725°/GHz) and the 3 dB beamwidth (∼10°), it can be seen that the signal bandwidth that is detectable by a highly directive receiver must be reduced in this scenario.

The histogram after the linear equalizer and down-sampling for two exemplary measurements are given in the right part of Fig. 5. The upper histogram is measured at 15 Gbaud and 290 GHz, i.e., an angle of 13°, with a wireless distance of 20 cm. Here, the two signal levels are nicely separated leading to an error-free transmission. In contrast, a noticeable broadening of the symbol levels can be seen in the figure below. This corresponds to a measurement at the same angle but now with a 25 Gbaud symbol rate and a distance of 50 cm. In both cases, however, error-free transmission is still possible considering a hard-decision forward error correction (HD-FEC) limit of 3.8 × 10⁻³, as it is further indicated by Fig. 6 on the left. Here, we generally aim to analyze the BER behavior as a function of beam steering for different symbol rates at a fixed distance of 50 cm. As already expected from the consideration of the CSNR, higher BERs can be observed especially at the highest transmit frequency (320 GHz), so that the maximum achievable data rate here is 15 Gbit/s using PAM2 modulation. However, if the focus is set to the core frequency range of 280–310 GHz (>20° steering coverage), data rates of 25 Gbit/s are supported considering the soft-decision forward error correction (SD-FEC) limit of 2 × 10⁻². The best transmission characteristics are recorded with 30 Gbit/s (HD-FEC compliant) at 290 GHz.

Figure 6. On the left, the BER is given as a function of the beam steering angle and frequency at a wireless distance of 50 cm and symbol rates of 15–30 Gbaud. Transmitting PAM2 and taking the HD-FEC limit into account, up to 30 Gbit/s have been transmitted error-free at 290 GHz. On the right, the BER is shown as a function of wireless distance and fixed frequency of 280 GHz for different symbol rates. A maximum data rate of 35 Gbit/s was achieved at 30 cm with SD-FEC limit. At the highest distance examined (110 cm), the maximum data rate is 5 Gbit/s.

In addition to a demonstration of the beam steering behavior, measurements were carried out at a fixed angle of 6° (280 GHz) and variable distance in the range of 30–110 cm. The results are shown in Fig. 6 on the right. As can be seen from this, the maximum error-free transmitted data rate is 35 Gbit/s at a distance of 30 cm, assuming SD-FEC as the limit, which to the best of our knowledge is the highest reported value for THz beam steering antennas.

Simultaneous point-to-multipoint THz communications

Besides transmission to a single user, the multi-beam property of the LWA have been exploited for addressing two users simultaneously. The transmission experiment is also based on the communications architecture introduced in section “Wireless THz communications setup.” However, a slight change was made to the baseband signal generation. Instead of the previously used FPGA and DAC, an arbitrary waveform generator (AWG, Keysight M8194A) with a higher baseband bandwidth of 50 GHz was used here. In this way, it is possible to distribute individual user data over a wider spectrum already in the digital domain. With regard to the LWA, this offers the possibility of different data being transmitted in a wider angular coverage. Compared to the approach of using an additional laser that acts as a second local oscillator (LO)at a slightly offset wavelength and thus different steering angle, as shown in paper [Reference Lu, Haddad, Tebart, Steeg, Sievert, Lackmann, Rennings and Stöhr23], the system utilized here offers the possibility that individual data can be transmitted to different users simultaneously. Hereby, the angle at which the signal is radiated is set via the digital frequency up-conversion of the signal. This means that if the angle to the receiver is known, the frequency shift – and thus the radiation angle – can be adjusted in the digital domain so that the particular user is provided with individual data as long as they do not obscure each other. As there is no inherent tracking functionality in the presented system, the position of the receivers can be varied, but manual adjustment of the transmission signal is required.

A picture of the setup with two wireless receivers is depicted in Fig. 7 and Table 2 gives an overview on the main transmission properties. As can be seen, the receivers are located at a wireless distance of 50 cm and 70 cm, respectively and angles of 17° and 6°. The corresponding center frequencies are 300 GHz and 280 GHz. Once again, it should be noted that the 20 GHz offset between the two data signals was already generated in the transmission sequence in the digital domain.

Figure 7. The image shows the point-to-multipoint setup with two users that can simultaneously receive individual data at different locations.

Table 2. Summary of the main transmission properties for the point-to-multipoint experiment indicating the receiver locations and the maximum achievable HD-FEC compliant data rates

Overall, simultaneous data transmission of 5 Gbit/s for wireless receiver 1 (WR1) and 10 Gbit/s for wireless receiver 2 (WR2) was demonstrated while complying with the HD-FEC limit.

In the current system, both the transmitting and the receiving side influence the maximum number of supported users. From the transmitting side, the relatively high roll-off factor plays a particular role as neighboring receivers are affected by interference. On the other hand, the transmitting side is also influenced by the receiver architecture. By using an SBD receiver, it is necessary that the transmit waveform contains an additional carrier for down-conversion. This means that the maximum power level in each data signal is reduced. One way to improve this would be to implement a heterodyne receiver in which the required carrier for down-conversion is generated at the receiver. However, as shown in [Reference Dittmer, Füllner, Stöhr and Randel15], the use of an intradyne receiver is significantly better in terms system performance compared to a heterodyne receiver, which is why this approach will be pursued in upcoming systems. In [Reference Lu, Haddad, Tebart, Steeg, Sievert, Lackmann, Rennings and Stöhr23] it was shown with a similar LWA that a 1 GHz guard band is sufficient for the separation of two users. Under this assumption, the concept presented here can be used to generate seven simultaneous beams in the 280–320 GHz frequency range, each with a bandwidth of 5 GHz, which would correspond to a total throughput of 35 Gbit/s.