Study and performance characterization of two key RF hardware subsystems: microwave divide-by-two frequency prescalers and low noise amplifiers

Abstract

This thesis elaborates on the theory and art of the design of two key RF radio hardware subsystems: analog Frequency Dividers and Low Noise Amplifiers (LNAs). Specifically, the design and analysis of two Injection Locked Frequency Dividers (ILFDs), one Regenerative Frequency Divider (RFD), and two different LNAs are documented. In addition to deriving equations for various performance metrics and topology-specific optimization criterion, measurement data and software simulations are presented to quantify several parameters of interest. Also, a study of the design of LNAs is discussed, based on three “regimes:” impedance matching, transconductance-boosting, and active noise cancelling (ANC). For the ILFDs, a study of injection-locked synchronization and phase noise reduction is offered, based on previous works. As the need for low power, high frequency radio devices continues to be driven by the mobile phone industry, Frequency Dividers that are used as prescalars in phase locked loop frequency synthesizers (PLLs) must too become capable of operation at higher frequencies while consuming little power. Not only should they be low power devices, but a wide “Locking Range” (LR) is also desired. The LR is the bandwidth of signals that a Frequency Divider is capable of dividing. As such, this thesis documents the design and analysis of two ILFDs: a Tail-ILFD and a Quench-ILFD. Both of these ILFDs are implemented on the same oscillator circuit, which consumes 2.28 mW, nominally. Measurements of the Tail and Quench-ILFDs’ LRs are plotted, including one representing the Quench-ILFD operating at “very low” power. Also, an RFD is detailed in this thesis, which consumes 410 μW. This thesis documents Locking Ranges for the Tail and Quench-ILFDs of 12% and 3.7% of 6.4 GHz respectively, during nominal operation. In “very low” power mode, the Quench-ILFD has a LR of 4.8% while consuming 219.6 μW of power. For the RFD, simulations report a LR of 16.7% while consuming 410 μW. Recently in 2011, a wideband LNA topology by Nozahi et al., which employs Partial Noise Cancelling (PNC) of the thermal noise generated by active devices, was presented and claimed to achieve a minimum and maximum NF of 1.4 dB and 1.7 dB (from 100 MHz to 2.3 GHz), while consuming 18 mW from a 1.8 V supply. This thesis details the theory, design, and simulation results of a narrowband version of this PNC LNA. In order to compare the largesignal performance of this narrowband LNA to that of a well-known implementation, an LNA employing inductive source-degeneration (referred to as a “S-L LNA”) is designed and analyzed through simulation. The PNC LNA operates at a frequency of 2.3 GHz while the S-L LNA operates at 2.8 GHz. Simulations report a NF of 1.76 dB for the PNC LNA and 2.3 dB for the SL LNA, at their respective operating frequencies. Both LNAs consume roughly 15 mW of quiescent power from a 1.8 V supply. Lastly, a case for the suspected design and layout faults, which caused fabricated versions of the RFD and two LNAs documented in this thesis to fail, is presented. First, measurements of the two LNAs are shown, which display the input impedance of the S-L LNA and the s₂₁ responses for both. Then, general layout concerns are addressed, followed by topology-specific circuit design flaws.