8.2 A 22.4-to-26.8GHz Dual-Path-Synchronized Quad-Core Oscillator Achieving −138dBc/Hz PN and 193.3dBc/Hz FoM at 10MHz Offset from 25.8GHz

The denser modulation scheme, employed by high-speed wireless communications, such as 5G, demands millimeter-wave (mm-wave) oscillators with low phase noise (PN). Generally, the PN of an LC oscillator can be reduced by downscaling its tank inductance. Yet, an over-shrunk single-turn spiral inductor suffers from
$Q$
degradation [1], impairing the figure-of-merit (FoM) and the minimum-achievable PN of the oscillator. Coupling multiple oscillators together is effective for PN reduction. Ideally, an N-core coupled oscillator can achieve 10log(N) lower PN when compared with a single-core oscillator. Nevertheless, the frequency mismatch between the oscillator cores tends to penalize the PN improvement and FoM of a coupled oscillator (Fig. 8.2.1 upper). Figure 8.2.1 (bottom) depicts two approaches to implement a quad-core oscillator by connecting four cores using: 1) spiral inductors (transformers) in a concentric way [2], [3], or 2) slab inductors (transformers) in a circular way [1], [4]. Since the outputs of either of the two cores in the first approach can be directly connected via a short metal trace with a small resistance, it benefits from a small PN penalty induced by the frequency mismatch. But the poor
$Q$
of the small spiral inductor limits its PN. On the contrary, the latter approach features a circular inductor (transformer) with an improved
$Q$
but suffers from a large PN penalty due to the frequency mismatch between two non-adjacent cores, i.e., cores #1 (#2) and #3(#4), since the non-adjacent cores are not directly connected and synchronized.