Regulación de la eNOS

This section will briefly look into five of the state of the art integrated radiators with on-chip antennas and without external modification.

In [33–36], the distributed active radiator (DAR) is presented. The DAR consists of a traveling wave oscillator, where the signal line is arranged in a figure-8 fashion that keeps differential pairs of transistors near each other (Figure 2.6). At the fundamental frequency of oscillation, this also means that differential signals are always traveling next to each other on the two signal lines, acting as a differential transmission line, and thus the fundamental frequency has minimal radiation. If these transistors are driven hard into their nonlinear regions, a second harmonic signal is also produced, which is in phase for each differential pair. This means that currents on the two signal lines will be going in the same direction at any instant and thus will radiate effectively. Coupling to substrate modes is also kept to a minimum by the placement of several of the DAR structures in an array in such a way that the substrate modes from various radiators cancel each

other out. Several versions of the DAR have been published, culminating in a 4x4 array at 280 GHz. This array achieves a +9.5 dBm effective isotropic radiated power (EIRP) with total radiated power of 190µW, and demonstrates beam steering of 80◦. The DAR is a self-oscillating radiating structure, which means that the reactive elements that determine the frequency of oscillation (the tank) are also the same elements used to radiate. This means that there is a danger of the DAR picking up signals that are near in frequency and injecting those signals back into the oscillator, which might pull, or in the worst case lock, the radiator to an undesired frequency. This concern can be alleviated by designing radiators that are driven unilaterally, where the reactive elements that determine the oscillation are separated from the radiating elements by unilateral amplification stages, ensuring that any external signals do not get coupled into the tank of the oscillators.

Figure 2.6: Conceptual depiction of a single distributed active radiator (DAR), and the block diagram of the 4x4 DAR phased array from [34].

In [37], a 210 transceiver is implemented using a 2x2 array of dipole antennas (Figure 2.7) in 32 nm CMOS SOI. The transmitter is made up of a 210 GHz VCO whose output can be modulated with on-off keying (OOK). The signal is amplified and split into 4 and is amplified again before being sent into the antennas. This work achieves a 5.13 dBm EIRP, with a saturated power amplifier (PA) power of 4.6 dBm per PA, or 10.6 dBm for the system. This leads to an antenna gain from the output of the PA to the far field electric field of -5.47 dBi, which means there is still significant room for improvement in the performance of the antennas.

In [8], an array of 4 leaky wave antennas are used both as receive as well as transmit antennas at 260 GHz (Figure 2.8). The leaky wave antennas are implemented as microstrip transmission lines with a bottom ground plane that is still above the silicon substrate, shielding the signal from the lossy substrate and preventing surface modes from being excited. However this type of arrangement means that mirror return currents will appear in the bottom ground plane and will cancel out a significant portion of the radiation. This means that very large currents are required to get a reasonable total radiated power, which often leads to lower antenna efficiency, as the metal losses at 260 GHz on integrated processes are often quite high. This can be seen in

Figure 2.7: Block diagram of the transmitter and receiver, and the die photo of the transmitter showing the radiator array from [37].

this work, as the PA provides +13 dBm of output power, and the reported EIRP is +5 dBm at 260 GHz, which means that the antenna gain from the output of the PA to the far field is -8 dBi. The total radiated power is not reported. This work also focuses on non-coherent (OOK) modulation to create a wireless fatal-link over a 40 mm range at 14 Gb/s. The leaky wave antennas are also 1250µm long, which at 260 GHz is more than one wavelength in air, which means that compared to their wavelength, these antennas require a large footprint, and thus become very costly.

Figure 2.8: 3-dimensional depiction of the four leaky wave transmission line antennas, their use for both transmitter and receiver, and a die photo showing the transceiver from [8].

In [38], tapered dipoles are used for both the transmit and receive antennas in a transceiver at 164 GHz (Figure 2.9). The radiation was picked up by a standard RF probe hovering over the top of the chip. Because the antenna properties of the probe were never measured, no standardized metrics of the radiator performance are presented. The polarization of the probe being used as an antenna is also not discussed, which means that

comparisons between different antennas measured in this setup may be due to the performance of the antenna itself, or possibly due to better polarization matching between some of the antennas and the probe. Another concern is that the probe is likely located in the near field of the on-chip dipole, which means that the near- field coupling between the probe and dipole may not only be interfering with the radiation of the dipole, but also may be changing its input impedance, which affects the output power of the driver PA. Still this setup is able to confirm that the chip is capable of radiating some power at 164 GHz.

Figure 2.9: Block diagram of the transceiver and die photo showing transmit and receive antennas from [38].

2.3

Performance Degradation due to Process and Environmental Vari-