Joint Beam Steering and Direct Antenna Multimodulation Using Time-Modulated Arrays

42702 IEEE INTERNET OF THINGS JOURNAL, VOL. 12, NO. 20, 15 OCTOBER 2025 Joint Beam Steering and Direct Antenna Multimodulation Using Time-Modulated Arrays Roberto Maneiro-Catoira ,Member, IEEE, Grzegorz Bogdan , Member, IEEE, Konrad Godziszewski ,Member, IEEE, Julio C. Brégains ,Senior Member, IEEE, José A. García-Naya ,Senior Member, IEEE, Yevhen Yashchyshyn ,Senior Member, IEEE, and Luis Castedo ,Senior Member, IEEE Abstract—In this article, we introduce a novel architectureindependent technique for time-modulated arrays (TMAs) that enables simultaneous direct-antenna multimodulation (DAMM) and beam steering. We validated the approach across a remarkably broad variety of modulation constellations, confirming its ability to implement arbitrary direct-antenna IQ modulation. This flexibility makes the technique highly suitable for Internet of Things (IoT) applications spanning scenarios that demand either robust communication or high-spectral efficiency, all on a single energy-efficient, low-cost hardware platform. The proposed method operates within the available TMA bandwidth and achieves precise beam steering without relying on computationally intensive optimization algorithms. Unlike previous studies that evaluate TMAs in isolation, we design and characterize a fully integrated transmitter that combines DAMM and beam steering functionalities. The resulting TMA-based DAMM transmitter employs a reduced hardware architecture, delivering both cost and energy savings ideally matched to resource-constrained, low-latency IoT deployments. Index Terms—Antenna arrays, beam steering, modulation. I. INTRODUCTION TRADITIONAL wireless transmitters use quadrature modulators, which involves mixing the analog baseband signal components, in-phase (I) and quadrature-phase (Q), with two 90◦out-of-phase versions of the same carrier wave [1,Ch.5]. If in-phase and quadrature-phase (IQ) modulation and beam steering (BS) are to be performed, conventional transmitters offer two options, shown in Fig. 1(a) and (b), where the antenna (or antenna array) is fed by a Received 15 April 2025; revised 18 July 2025; accepted 26 July 2025. Date of publication 31 July 2025; date of current version 9 October 2025. This work was supported in part by the Xunta de Galicia under Grant ED431C 2024/18; in part by MCIN/AEI/10.13039/501100011033 under Grant PID2022137099NB-C42 (MADDIE); in part by MCIN/AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR under Project TED2021130240B-I00 (IVRY); and in part by the Open Access Charge: Universidade da Coruña/CISUG. (Corresponding author: José A. García-Naya.) Roberto Maneiro-Catoira, Julio C. Brégains, José A. García-Naya, and Luis Castedo are with the CITIC Research Center and the Department of Computer Engineering, Universidade da Coruña (University of A Coruña), 15071 A Coruña, Spain (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). Grzegorz Bogdan, Konrad Godziszewski, and Yevhen Yashchyshyn are with the Institute of Radioelectronics and Multimedia Technology, Warsaw University of Technology, 00-665 Warsaw, Poland (e-mail: grzegorz. [email protected]; [email protected]; yevhen.yashchyshyn@ pw.edu.pl). Digital Object Identifier 10.1109/JIOT.2025.3594557 modulated and up-converted radio frequency (RF) signal [2]. In conventional transmitters with BS capability, as shown in Fig. 1(a) and (b), the digital (binary) information must be first converted into its analog (baseband) representation, requiring two analog-to-digital converters (ADCs). The two analog components of the baseband signal, I and Q, are then fed into RF mixers, where they modulate the carrier wave and its orthogonal counterpart [2].Fig.1(a) illustrates a directconversion transmitter with a phased array. This approach benefits from requiring only one RF chain, but it comes at the cost of reduced flexibility in BS control. On the opposite side, Fig. 1(b) shows the digital beamforming solution, which provides large flexibility in BS control, but at the cost of high complexity and power consumption because it requires one RF chain per antenna element. A direct antenna modulation (DAM) method offers a significant simplification of the transmitter architecture by employing a fully digital modulation approach. Unlike conventional systems, the binary stream is not converted into its analog baseband representation. Instead, it is delivered directly to the antenna controller, as illustrated in Fig. 1(c), which shows a direct-conversion transmitter with a phased array utilizing variable-gain amplifiers (VGAs) and variable phase shifters (VPSs). This approach results in a simpler and more energy-efficient system architecture compared to conventional transmitters with IQ modulators, as it eliminates the need for frequency mixers and digital-to-analog converters (DACs). However, the use of VGAs and VPSs introduces other issues, such as phase quantization errors and additional insertion losses. Furthermore, these devices are both powerhungry and expensive. Notice that all these drawbacks are critical for battery-powered wireless sensors and Internet of Things (IoT) devices. To meet the requirements of the IoT domain, the aforementioned disadvantages must be overcome. Theoretical concepts of removing or replacing VGAs with RF switches were presented in [4] and [5], respectively. In this work, we go even further by proposing a DAMM transmitter with BS that is also VPS free. This results in significantly improved power efficiency and hardware complexity reduction. Our design, illustrated in Fig. 1(d), is based on a periodically switched beamforming network (BFN) controlled by periodic sequences –in other words, a time-modulated array (TMA) layout– which can be used to perform DAMM-BS, alleviating insertion c 2025 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ MANEIRO-CATOIRA et al.: JOINT BEAM STEERING AND DIRECT ANTENNA MULTIMODULATION 42703 Fig. 1. (a) Conventional “non-DAM” scheme with analog beamformer [3,Ch.8]; (b) conventional “non-DAM” scheme with digital beamformer [3,Ch.8]; (c) conventional “DAM” scheme with both VGAs and VPSs [2]; and (d) proposed architecture of our DAMM-TMA based on switches governed by periodic signals to jointly perform DAMM and beam steering. losses and quantization errors while minimizing the hardware complexity. A. State-of-the-Art DAM is reported quite prolifically in the literature from an experimental point of view. An illustrative example is a prototype which performs on–off keying (OOK) modulation realized by a microstrip patch with diodes placed between the antenna and the ground plane [6],[7] or by an on-off switch located in a feeder of a wire monopole antenna [8]. Utilization of DAM for such a simple modulation scheme brings a significant advantage because it allows for transmitting signals over a range of frequencies well beyond the antenna’s resonant bandwidth [9],[10]. Accordingly, [11] shows a method of generating a binary frequency-shift keying (BFSK) signal exceeding the loop antenna’s impedance bandwidth by using a complementary pair of high-electron mobility transistors (HEMTs). In [12], quadrature phase-shift keying (QPSK) and binary phase-shift keying (BPSK) modulated signals were generated in a monopole antenna driven by a simple circuit consisting of a switch, a transformer and a series inductance. In [13], a two-element antenna array with switchable feeding lines for BPSK DAM was presented, although without any signal analysis at the communication level. Direct generation of frequency modulated (FM) signals through varactor diode tuning was reported in [14]. In [6],[7],[8],[9],[10],[11],[12],[13], and [14],DAM takes place within the antenna array or its feeding structure. However, according to [15],[16],[17],[18],[19], and [20], DAM can take place also outside the antenna, for instance in its proximity. For example, in [15] a 60-GHz signal radiated by a dipole antenna is modulated by reconfigurable reflectors located under a hemispherical silicon lens. Another approach involves using frequency-selective surfaces (FSS) in the antenna proximity, which can control properties of radiated electromagnetic waves to generate high-order phaseshift keying (PSK) modulated signals [16],[17].In[18],an amplitude-shift keying (ASK) modulated signal is generated by a two-state gain control of a microstrip antenna with a reconfigurable metasurface layer placed at a distance of 70mm (0.58 λ). DAM can be realized even in the far-field region by means of reconfigurable intelligent surfaces (RISs) [19],[20], [21],[22],[23]. DAM and BS are distinct and well-established technologies which are not usually combined in a single antenna system. 42704 IEEE INTERNET OF THINGS JOURNAL, VOL. 12, NO. 20, 15 OCTOBER 2025 TABLE I COMPARISON TABLE OF DAM-TMA TRANSMITTERS WITH BS However, the recent literature reveals two joint DAM and BS examples. The first one is an array of DAM transmitters with VPSs that adjust the phase of each signal as in the phased array [15]. The second is based on the subset modulation of the phased array with RF switches and VPSs controlled at the symbol rate [5]. Nevertheless, there are still notable challenges and limitations regarding flexibility, insertion losses, complexity, and cost of both approaches due to VPSs used as a control element. For instance, a b-bit VPS can only offer 360◦/2bdiscrete phases, necessitating 2bsingle-pole dual-throw (SPDT) switches [31],[32] in its circuit topology, which increases power losses and system complexity. The need for simpler and more energy efficient circuits motivated the development of DAM-BS transmitters without VPSs, as the ASK transmitter with four switchable beams presented in [33]. Recent studies demonstrated the feasibility of DAM based on TMAs. Table Igathers the main parameters of direct antenna modulation with time-modulated array (DAM-TMA) transmitters with BS published over the last five years. In [24], a multicarrier DAMM approach with TMAs is explored, in which two harmonics are utilized for direct BPSK modulation. This DAMM method with TMAs relies on harnessing multiple harmonics, necessitating appropriate windowing of these harmonics. Consequently, the DAMM method becomes closely associated with and dependent on time-modulation efficiency. Furthermore, constructing signal constellations cannot be accomplished through closed formulas, requiring a time-consuming optimization method that conditions the low latency. Unfortunately, this approach only achieves efficiencies1below 40 % and significantly penalizes the TMA directivity, making it a less desirable option. Outstanding results were obtained with a time-modulated leaky-wave antenna (TM-LWA) [30],[35] fabricated in the substrate-integrated waveguide (SIW) technology with 164 p-i-n diodes embedded inside 82 slots (2 diodes per slot). TM-LWA operates at the center frequency of 23.5 GHz, which 1Notice that in Table I, only one reference (other than this work) presents a fabricated prototype with conceptual tests, but without rigorous efficiency measurements. Hence, the only fair way to compare efficiency is by means of theoretical calculations. Moreover, according to [34], theoretical and measured TMA efficiencies show a strong correlation. is designated for 5G communications. However, because of very low time-modulation frequency (f0=1.2MHz)only narrowband signals can be transmitted without overlapping with spectral replicas [36].TheDAMMin[35] and [30] is based on space translation introduced to the spatial amplitude envelope. Experiments were conducted for QPSK, 8-PSK, 16-APSK, and 16-QAM. However, the results presented in [37] are without BS. Moreover, p-i-n diodes consume substantial amount of energy and significantly limit the timemodulation frequency. Therefore, in this article the idea of using time-modulation for DAM is further developed with goals of improving the total efficiency and extending the signal bandwidth by proposing a design based on a simple and accessible fabrication technology. While the DAM transmitter with BS is applicable to various wireless communication systems, it is particularly well-suited for IoT devices operating in congested frequency bands. The key advantages of BS in IoT communication include enhanced antenna gain, reduced interference, improved communication reliability, and strengthened physical layer security [38],[39],[40]. Previous research has demonstrated BS antenna arrays without VPSs for IoT applications [41],[42]. However, these designs feature large dimensions and high energy consumption from p-i-n diodes, making them suitable only for relays rather than end devices. Conversely, alternative designs presented in [43] and [44] lack essential feeding networks, which are critical components for practical implementation. B. Contributions of This Work The TMA method proposed in this article offers three advantages not provided by existing approaches. First, it is truly suitable for low-latency IoT applications. The reasons are the following. 1) The prototype provides a significant higher bandwidth (50MHz) compared to existing alternatives in the related TMA literature; see Table Ifor a summary of competing approaches. Indeed, our proposal is the only one valid for low-latency IoT. Bandwidth is critical in IoT applications where delays can lead to security risks or process failures, such as advanced e-health, smart video surveillance, or industrial automation. MANEIRO-CATOIRA et al.: JOINT BEAM STEERING AND DIRECT ANTENNA MULTIMODULATION 42705 2) It does not require real-time optimization algorithms to perform the combined DAMM beam steering technique, unlike, for example, [24],[25],[28],[29]. 3) It includes a comprehensive experimental verification of the prototype following the measurement guidelines suggested in [45], unlike works, such as [5],[24],[25], [26], and [46]. Note that it is a low complexity, switchonly-based prototype that eliminates the need for RF mixers, DACs, VPSs, and VGAs, while demonstrating beam steering with relatively high angular resolution (1.44◦). Second, it enables arbitrary direct antenna IQ modulation, including high order quadrature amplitude modulation (QAM), amplitude and phase-shift keying (APSK), or golden angle modulation (GAM), thereby supporting a wide variety of IoT applications using the same costand energy-efficient hardware. More specifically 1) The DAMM-TMA technique allows for straightforward generation of spectral efficient modulation schemes, such as GAM or the hexagonal quadrature amplitude modulation (HQAM). Note that, at present, the use of modulation schemes other than frequency-shift keying (FSK) or PSK in IoT with conventional technology is not common, as it requires expensive, complex, and energyinefficient hardware. 2) It also allows the implementation of high order phase shift keying modulations on the same hardware, enabling high bandwidth applications in noisier channels, such as cameras, smart buildings, e-health, and more. Third, it is architecture-independent, allowing the realization of DAMM and beam steering with any existing TMA hardware without compromising the time modulation efficiency of the original architecture, since only the switching delays are adjusted at the symbol rate. II. DAM-BS WITH TMAS A. Efficiency and Radiation Pattern of TMAs TMAs are antenna arrays equipped with switched timemodulated feeding networks (TM-FNs), offering an attractive alternative to phased arrays that rely on VPSs [47].TMAs excel in beam steering capabilities by simply adjusting the onoff instants of the switches in their feeding network [34],[48], [49],[50],[51],[52],[53],[54],[55],[56],[57]. This inherent adaptability allows TMAs to significantly enhance signal-tonoise ratio (SNR) and error vector magnitude (EVM) not only in line-of-sight communication scenarios [58], but also in multiple-input–multiple-output (MIMO) setups [59],[60]. Let us consider a linear array comprising Nisotropic elements with unitary static excitations In=1, where n∈Γ= {1,...,N}. Each element’s excitation is modulated by a periodic pulsed signal hn(t)with a fundamental period T0, realized through a switched TM-FN, as illustrated in Fig. 1(d). In the design of TMAs, new considerations must be taken into account that do not arise in conventional phased arrays due to the presence of radiation harmonics. Consequently, a vital performance metric for TMAs is the time-modulation efficiency, defined as the ratio of the useful power radiated by TABLE II CHARACTERISTICS OF HIGH-EFFICIENCY TMASINTHELITERATURE IN TERMS OF SBRPEAK [SEE (5)]AND TIME-MODULATION EFFICIENCY ηTM [SEE (1)] the TMA (PTM U)to the total average power radiated (PTM R) and is given by [56] ηTM =PTM U PTM R .(1) To enhance clarity and without loss of generality2,we assume that TM-FNs can efficiently exploit only one of the first-order harmonics (q=−1orq=+1), while suppressing or strongly attenuating all other harmonics. In simpler terms, the technique presented could be applied to any of the highefficiency TMA architectures mentioned in the literature and shown in Table II. In this work, the same basic periodic modulating pulse h(t)will be utilized for all antenna elements. However, the excitation of the nth antenna element will be time-modulated using a time-shifted version of h(t), i.e., hn(t)=h(t−Dn)(2) where Dnis a controllable time-delay. Let Hq,q∈Z, represent the exponential Fourier series coefficients of the periodic signal h(t)with period T0. The exponential Fourier series coefficients of hn(t)are then given by Hqe−j2πf0Dn(3) where f0=1/T0and they have the same modulus as Hq.Ifwe focus on utilizing the harmonic q=+1, the time-modulation efficiency can be expressed as [66, Eq. (14)] ηTM =|H1|2 ∞ q=−∞|Hq|2(4) where |H1|is the magnitude of the Fourier coefficient of h(t)at the harmonic q=+1, and the sum is taken over all harmonics. Considering that Fq(θ) is the spatial array factor at frequency fc+qf0, with fcbeing the carrier frequency and θ∈(−90◦,90◦)is the angle measured with respect to the x(broadside) axis and on the x−zplane, then SBRPEAK measures the relative level between two patterns: 1) the maximum of the most significant unwanted harmonic (with index u), |Fu(θmax)|; and 2) the maximum of the useful harmonic, |F1(θmax)|, which we use as our reference pattern because it is stronger than the fundamental harmonic F0(θmax)[66].We can express this mathematically as SBRPEAK =20log    Fu(θmax) F1(θmax)    (5) 2It is important to note that this approach is applicable to any type of TMA architecture, regardless of its time-modulation efficiency ηTM. 42706 IEEE INTERNET OF THINGS JOURNAL, VOL. 12, NO. 20, 15 OCTOBER 2025 where θmax is the direction in which the array factor amplitude reaches its maximum value. For our analysis, we make two important assumptions about TMAs: 1) They operate with high time-modulation efficiency (ηTM); and 2) they effectively suppress unwanted harmonics, keeping SBRPEAK below certain limits. Table II shows several examples of TMA designs that meet these requirements: time-modulation efficiency (ηTM)greater than 91% and harmonic suppression (SBRPEAK)better than −13.96 dB. Notice that the modulating pulses h(t)used in these TMAs typically come in two forms: 1) quantized versions of sinusoidal waveforms and 2) pulsed signals with constant amplitude and quantized linear phase. For each TMA design, Table II also specifies the type and number of switches needed for their implementation. Since the majority of the total energy is transmitted over the first positive harmonic for any of the TMAs specified in Table II, the Fourier series expansion of hn(t)can be approximated as follows: hn(t)= ∞  q=−∞ Hqe−jq2πf0Dnejq2πf0t ≈H1ej2πf0(t−Dn).(6) Therefore, in the case of the TMA-DAM architecture proposed in Fig. 1(d), when transmitting the single-frequency signal s(t)=ej2πfct, where fcis the carrier frequency, the radiated signal can be expressed as follows: x(θ,t)= N  n=1 hn(t)s(t)= N  n=1 hn(t)ejβcznsinθej2πfct ≈ N  n=1 H1ej2πzn λcsinθ−f0Dnej2π(fc+f0)t(7) where znrepresents the position of the nth array element along the zaxis, and βc=2π/λcdenotes the wavenumber for a carrier wavelength λc=c/fc. After normalizing (7) with respect to H1, introducing the notation F0=fc+f0, we obtain x(θ,t)≈ N  n=1 ej2πF0t+zn λcsinθ−f0Dn.(8) The TMA technique allows us to achieve two important objectives simultaneously: 1) modulate a carrier signal ej2πF0t; and 2) point the maximum radiation pattern toward any desired angle θ0. This dual functionality is possible by adjusting the time delays Dnin the periodic modulating waveforms hn(t). These adjustments control the phase of the signals transmitted by each TMA element, which in turn determines the direction of maximum radiation power. B. Direct Generation of Phase-Shift Keying Modulations M-ary phase-shift keying (MPSK) modulation entails phaseshifting the carrier signal based on the symbol to be transmitted. If there are Mdifferent symbols si, where i∈ ={1,...,M}, a phase shift of π(2i−1)/Mis applied to the carrier signal during the symbol interval Ts, corresponding to the specific symbol si=ej([π(2i−1)]/M)being transmitted. To steer the TMA useful harmonic beampattern toward a specific direction θ0while transmitting a symbol sibelonging to an MPSK modulation alphabet (where i∈), we can adjust the time delays Dnfor each antenna element n∈Γover a symbol time Ts. Our method aims to achieve a twofold objective: 1) to direct the radiation beam toward a specific direction; and 2) synthesize and radiate a particular MPSK data symbol. By analyzing (7), if we adjust Dnaccording to the following equation: ej2πf0Dn =ej2πzn λcsinθ0−π(2i−1) M(9) and then substitute (9) into (8), taking into account (8),we can derive the transmitted signal over a symbol time (noting that Dndoes not change for this interval) x(θ,t)≈ej2πF0t+π(2i−1) MN  n=1 ej2πzn λc(sinθ−sin θ0) ≈siej2πF0tF(θ)(10) where 0≤t<Ts,TsT0.ThetermF(θ) represents the TMA‘s spatial array factor, given by F(θ)= N  n=1 ej2πzn λc(sinθ−sin θ0).(11) If the TMA exploits the harmonic q=−1, (10) remains valid with some minor changes. In this case, F0is given by F0= fc−f0, and (7) becomes x(θ,t)≈ N  n=1 H(−1)ej2πzn λcsinθ+f0Dnej2π(fc−f0)t.(12) As a result, the condition in (9) changes to ej2πf0Dn =ej−2πzn λcsinθ0+π(2i−1) M.(13) In each symbol interval, the useful harmonic beampattern at frequency F0can be steered toward a specific direction θ0, resulting in a maximum power radiated beampattern when transmitting an M-PSK symbol si. This steering is achieved by imposing the carrier signal ej2πF0tto have a phase shift π(2i−1)/M. The phase adjustment is determined by the normalized time values Dn=Dn/T0∈[0,1),asshownin(9). The number of possible discrete values of Dn, achieved by the TM-FN proposed in Fig. 1(d), is dependent on the clock speed of the control unit [54]. Specifically, if the clock period is Tck, then the number of possible phase steps will be T0/Tck, resulting in 360◦/(T0/Tck)different quantized phases. For example, setting T0=210Tck ≈1000Tck yields a resolution equivalent to that of a 10-bit VPS. C. Direct Generation of Arbitrary Amplitude and Phase-Shift Keying Modulation IoT standards, for example IEEE 802.15 (Bluetooth) and IEEE 802.15.4 (ZigBee), use low-order PSK modulations, MANEIRO-CATOIRA et al.: JOINT BEAM STEERING AND DIRECT ANTENNA MULTIMODULATION 42707 such as offset quadrature phase-shift keying (OQPSK) or π/4-DQPSK. However, if a higher spectral efficiency is required, then the corresponding improved performance is obtained with amplitude and phase-shift keying modulation schemes, such as APSK, QAM, and GAM [67]. APSK signals can be generated in DAM-TMA in a similar way as MPSK, i.e., by setting appropriate values of delays Dn. Finding these values from (7) is analytically intractable and requires nonlinear programming or global optimization algorithms. Alternatively, we used the full-search algorithm to calculate all possible configurations and select only the results which correspond to the required constellation diagram. Numerical results for 16-APSK, as a representative example of the amplitude and phase-shift keying modulation, are given in Section III-B. III. NUMERICAL EXAMPLES A. Quadrature Phase-Shift Keying Modulation With Beam Steering To bridge the gap between theoretical analysis and experimental measurements, we conducted numerical simulations to compare the theoretical modeling with the actual behavior of different TMA configurations that jointly perform QPSK DAM and beam steering. We considered a TMA with N= 4 isotropic elements spaced a half-wavelength (λc/2)apart, which are time-modulated using the high-efficiency TM-FN schemes listed in Table II. The carrier frequency is fc= 5.6GHz, and the fundamental frequency of the periodic pulsed signal is f0=50 MHz (hence, T0=20 ns). The TMA operates as a QPSK modulator (M=4)with a symbol period Ts= 100 μs. We consider a TMA that exploits the harmonic q=−1 and examine three different scenarios where the array transmits the symbols of the QPSK constellation in the maximum radiation pattern direction for three different angles: θ0=0◦,θ0= −30◦, and θ0=30◦. Considering (9), or the version given in (13) for q=−1, we calculate the corresponding time delays Dnapplied to the pulses at each antenna element to directly modulate each QPSK symbol in the TMA for each scenario. For example, if θ0=0◦, we obtain D1=2.5 ns, and hence, D1=D1/T0= 1/8. If θ0=−30◦, we obtain D1=7.5 ns, and hence, D1= D1/T0=3/8. Figs. 2–4plot the Dnvalues that enable QPSK DAM while performing beam steering toward the considered directions. B. Amplitude and Phase-Shift Keying Modulation With Beam Steering For this discussion, let us consider a DAM-TMA with n=4 switches and delays ranging from 0 ns to 20 ns in steps of 80 ps, which gives 2504possible configurations in total. The amplitude and phase of a signal generated with each configuration toward any angle θ0is calculated with (7). Results form a domain of all feasible discrete IQ symbols. As can be noticed from Fig. 5(a), the generation of virtually all symbols is possible, which is due to a small step size of the programmable delay lines (PDLs) (only 80 ps with respect (a) (b) Fig. 2. (a) Adjusted values of Dnto directly modulate each QPSK symbol siwhile simultaneously directing the maximum of its power radiated pattern toward the broadside direction θ0=0◦in a 4-element TMA; (b) relative power radiated pattern (with respect to the maximum of the useful harmonic, |F1(θmax)|)for all symbols siwhen Dnare set as in (a). to the modulation period T0=20 ns). PDLs are key devices in our DAM-TMA approach which are integrated inside the control unit of Fig. 1(d) and explained in detail in the next section, conveniently supported by Fig. 8. Notice that in our case PDLs is 3D3438Z-80 [68]. From the domain of all feasible IQ symbols we form a codebook by selecting only the points which correspond to the desired modulation scheme, such as QAM, APSK, and GAM. The codebook defines values of switching delays which cause: 1) generation of given modulation scheme toward θ0, 2) beam-steering toward θ0. Let us assume that θ0=0◦and the required modulation is 16-APSK as illustrated in Fig. 5(b). Fig. 6shows sixteen DAM-TMA radiation patterns, one per symbol. One may notice that the main beam is always pointing at θ=0◦, although its amplitude can be suppressed to facilitate the two-state amplitude modulation of 16-APSK. Corresponding normalized delays are illustrated in Fig. 7. IV. EXPERIMENTAL ASSESSMENT A. Measurement Setup The DAMM method presented in this article can be applied to any TMA architecture, as it relies solely on controlling time delays in the antenna’s periodic modulating pulses. To experimentally validate this approach, we implemented a TMA module incorporating patch antennas and SPDT switches, as illustrated in Fig. 8. Note that the antenna symbol in Fig. 8represents an arbitrary antenna or subarray. In fact, the prototype presented in Fig. 9(b) uses two series-fed patch antennas as a single element. Therefore, even though the 42708 IEEE INTERNET OF THINGS JOURNAL, VOL. 12, NO. 20, 15 OCTOBER 2025 (a) (b) Fig. 3. (a) Adjusted values of Dnto directly modulate each QPSK symbol siwhile simultaneously directing the maximum of its power radiated pattern toward the direction θ0=−30◦in a 4-element TMA; (b) relative power radiated pattern for all symbols siwhen Dnare set as in (a). prototype has 16 patches, it actually consists of eight elements. We selected this TMA design because it achieves the highest combination of signal bandwidth and realized gain among all experimental TMAs reported in the literature to date. For a detailed analysis of its performance, see [34]. Two types of signals are processed within the module: RF signals (in the gigahertz range) and digital control signals (in the megahertz range). Fig. 8illustrates these signal paths, where the RF signal paths are marked with thick lines, and digital control signal paths are represented by thin lines. Additionally, double lines indicate the serial interface between the microcontroller unit and the PDLs. 1) RF Signals: A5.6-GHz sinusoidal carrier wave from an external RF generator is delivered to a uniform fourbranch microwave power divider. Each output of the power divider is connected to an SPDT RF switch. The module utilizes four ADRF5020 ultrafast switches [69], selected for their very short rise/fall time of 2ns. These switches feature alternatively controlled outputs, meaning that only one output is active at a time, while the other remains terminated to a matched load. The switches are used to alternate the RF signal between pairs of oppositely polarized patch antennas. As a result, each antenna element is excited with a pulsed RF signal, where the start and end moments of the RF pulses serve as degrees of freedom, enabling flexible control over the radiation pattern. 2) Digital Signals: The RF switches are controlled by periodic rectangular waveforms (modulating signals) with 50 % duty cycles. According to the TMA theory, (a) (b) Fig. 4. (a) Adjusted values of Dnto directly modulate each QPSK symbol siwhile simultaneously directing the maximum of its power radiated pattern toward the direction θ0=30◦in a 4-element TMA; (b) relative power radiated pattern for all symbols siwhen Dnare set as in (a). (a) (b) Fig. 5. Complex diagram showing: (a) all feasible symbols (b) selected symbols forming 16-APSK modulation scheme. Calculated for θ=0. the pulse repetition frequency should be higher than twice the transmitted signal bandwidth [70]. Therefore, for most IoT applications, even the frequency of 1MHz would be sufficient. However, in our module an excessive modulation frequency of 50MHz is chosen to demonstrate the full potential of the ultrafast AD5020 RF switch. In consequence, the distance between sidebands illustrated in Fig. 10 is broadened, allowing for potential transmission of wideband signals without any aliasing of the spectral replicas. To enable independent command of the switches, the control signal is split into four paths using a fan-out clock buffer. Each output of the buffer gives a copy of the 50 MHz clock signal, which is delayed in PDL by a value ranging from 0 ns to 20.4 ns, with a step of 80 ps. The PDLs are programmed by a microcontroller unit, with their settings updated MANEIRO-CATOIRA et al.: JOINT BEAM STEERING AND DIRECT ANTENNA MULTIMODULATION 42709 (a) (b) (c) (d) (e) (f) (g) (h) Fig. 6. Complex radiation patterns of DAMM-TMA (amplitude and phase plotted with different lines) used for the generation of 16-APSK symbols toward θ0=0◦. (a)–(h) Correspond to symbol labels in Fig. 5(b). (a) (b) (c) (d) (e) (f) (g) (h) Fig. 7. Values of Dnresulting in radiation pattern presented in Fig. 6. at the symbol rate. The independently delayed clock signals control the operation of the switches: a low-state enables the first RF output and a high-state enables the second RF output. Hence, the antennas are excited with pulsed RF signals maintaining the abovementioned 50 % duty cycle, while the pulse positions in the time domain are delayed by Dn. The DAMM-TMA module operates at the center frequency fc=5.6GHz with the time-modulation period of T0= 20ns. PDLs embedded inside the DAM-TMA module allow a minimal delay step of Tck =80 ps, resulting in a minimal feasible phase-change of 360◦/(T0/Tck)=1.44◦, which yields a resolution equivalent to that of a 16-bit phaseshifter. Considering that the maximal quantization error of the selected PDLs is ±0.72◦[68], the DAM-TMA can be used to generate BPSK, QPSK, 8-PSK, 16-PSK, 32-PSK, 64-PSK, and 128-PSK signals, assuming a noise-free channel and a perfect carrier synchronization between the transmitter and the receiver. The experimental setup is shown in Fig. 9(a). The transmitter includes a signal generator (Rohde & Schwarz SMBV100A) and the DAM-TMA. The receiver utilizes a Fig. 8. Diagram of DAMM-TMA transmitter under test. signal analyzer (Agilent E4440A) and a vector signal analyzer software running on a personal computer. During the experiments, the signal generator served as the source of a 5.6-GHz single-tone signal. The spectrum analyzer was used to down convert the received the signal and to demodulate it to IQ baseband samples. Fig. 10 displays the measured power spectrum of a signal generated by the DAMM-TMA under test. Due to the modulation period of 20 ns, harmonics appeared at multiples of 50MHz around the carrier frequency of 5.6GHz. In order to demodulate a single harmonic, the spectrum analyzer was configured with a frequency span narrower than the time modulation frequency and its center frequency was 42710 IEEE INTERNET OF THINGS JOURNAL, VOL. 12, NO. 20, 15 OCTOBER 2025 Fig. 9. Experimental setup. (a) Overview of the equipment. (b) Anechoic chamber. Fig. 10. Signal spectrum measured at the output of DAMM-TMA excited by 5.6GHz carrier. Harmonics are not symmetric in respect to the center frequency when the BS is enabled. A higher level of q=−1 with respect to q=+1 indicates that the angular direction of the main beam associated to q=−1 is closer to the direction of the signal source. set to 5.55 GHz or 5.65GHz to demodulate the first negative (q=−1)or positive (q=+1)harmonic, respectively. The vector signal analyzer software was used for digital carrier synchronization, symbol synchronization, and symbolsto-bits demapping, allowing for accurate evaluation of the experimental results. The DAMM-TMA transmitter was thoroughly tested to evaluate its capability to generate various amplitude and phaseshift keying modulation schemes. Fig. 11 shows a few selected results obtained for modulation schemes which are typically used in IoT, i.e., QPSK, 8-PSK, 16-PSK, and 16-APSK. The observed results positively verify the theoretical concepts drawn in Sections II-B and II-C. A detailed description of the measurement methodology is given in subsequent Sections IV-B and IV-C for QPSK with beam steering disabled and enabled, respectively. Fig. 11. IQ samples generated by DAMM-TMA transmitter after demodulation: (a) QPSK, (b) 8-PSK, (c) 16-PSK, and (d) 16-APSK. TABLE III CONFIGURATION OF PDLSFORQPSK AND THE BROADSIDE BEAM B. Broadside Beam and Direct Generation of QPSK To generate QPSK symbols and fix the antenna beam toward the broadside direction, the delays were configured according to Fig. 2(a). The actual values applied in the experimental trials are given in Table III and match Fig. 2 with two differences: 1) all normalized delays were shifted by −1/8; and 2) due to the hardware related constraints, delays D1and D3were set to (5.04 ns instead of 5 ns, and 15.04 ns instead of 15 ns), respectively. The first difference simplifies the hardware configuration and has virtually no impact on the transmission, as the demodulator performs automatic phase-shift compensation and detects symbols based on the relative phase change. The second difference introduces an inaccuracy of 40ps, which is less than 1 % of T0=20 ns and can be considered as negligible. During the experimental assessment the binary sequence 00011110 was repeatedly transmitted by changing the delays according to Table III. PDLs were updated every 102µs, resulting in a symbol rate of 9.8ksymbol/s. The baseband symbols received at frequencies corresponding to q=−1 and q=+1(5.55 GHz and 5.65 GHz, respectively) are presented on complex planes in Fig. 12. Although they seem similar, the order of the symbols is different. At 5.55 GHz (q=−1), the phases of the received symbols were changing consecutively with a π/2 step, yielding the symbol sequence (s1,s2,s3,s4)which corresponds to 00011110, as expected. However, according to Fig. 13, which shows 16 symbols received during 1.6msat5.65 GHz, the sequence at the harmonic q=+1was(s2,s1,s4,s3). This observation reveals an interesting relationship between the symbols generated over the harmonics q=−1 and q= +1, which are complex conjugates. Hence, only the q=−1 harmonic is correctly modulated and can be faithfully received.