TERAhertz links using SOlid state devices and photoNICs
Final Report Abstract
In the early phase of this project it was shown in the frame of a joint measurement campaign in Lille and with transceiver modules from a previous project that high data rate communication with up to 100 Gbit/s is possible with an optical transmitter and an electrical receiver in the WR3.4 band (220-330 GHz). Due to cuts in the funding for the originally planned waveguide packaging of the electric receiver, much unscheduled time and effort was necessary to establish a new PCB-based workflow at ILH to assemble the 300 GHz receiver MMICs. Two receiver versions were designed, each with breakouts to measure the functional components. The first receiver version is an improved version of a previous designed resistive mixer and performs as simulated. It was published with on-wafer measurements. The integration of an antenna-on-chip (AoC) and a low-noise amplifier wouldn’t have been possible because the chip area and aspect ratio is no longer tolerable. The second generation of the receiver uses an active mixer topology that needs less local oscillator signal power and does no longer need a buffer amplifier, which allows the monolithic integration of the remaining components including the AoC. By post-production removal of the AoC successful operation could be verified in on-wafer measurements also for this receiver variant. However, this second version of the 300 GHz receiver had severe yield and manufacturing problems in the semiconductor foundry. This led to a fallback solution with the first receiver, which was not intended to be packaged, since it is missing an antenna and amplifier stage, but was now integrated in the form of a multichip module along with a stand-alone 300 GHz amplifier and AoC chip. Both receiver versions contain custom designed DC supply networks and on-board wideband impedance matching networks, which were designed by comprehensive electro-magnetic field simulations. Two PCBs of both versions were assembled. In the experimental verification the conversion gain of both receiver modules was unfortunately found to be too low. They showed an offset to the simulation of -50 dB for the version with AoC and -28 dB for the multi-chip-module. The main cause for the conversion gain offset of version with AoC could be traced to chip fabrication issues and missing metal bridges in the tandem-x couplers. This leads to nonsufficient local oscillator signal power at the mixer and attenuation of the RF signal. The reason for the conversion gain offset for the multi-chip-module is most likely a combination of the influence of the chip-to-chip bond-wire transitions and higher than expected antenna losses. Besides the low conversion gain of the receiver modules, they performed well and a reliable Chip-on-PCB workflow was established. The DC networks function as expected and the stability of the chips is no issue except for the amplifier, for which instability is likely to be caused by the poor impedance matching at the inand output. The connections to the coaxial cables or the IF and local oscillator connections are matched well. The stepped impedance filters to compensate the bond-wire at the IF and LO ports worked as well. The low conversion gain of the modules ( -48 dB and -42 dB), however, made further measurements in combination with the photonic receiver obsolete, since superior measurements results have been collected with the previous modules, in the initial stage of the project. The project partner in Lille will conduct antenna measurements of the AoC, which however have not yet taken place at the time of writing this report.
