Noise properties of cryogenic microwave amplifiers and relevance to oscillator frequency stabilization

Abstract

We studied noise properties of a new generation of energy-efficient microwave low-noise high-electron-mobilitytransistor amplifiers. The noise measurements were conducted both at room and cryogenic temperature with the amplifiers strongly decoupled from the environment to reduce the influence of ambient temperature fluctuations on their phase noise spectra. Our results show the importance of keeping the amplifiers operating in the small-signal regime if they are to be used for applications such as precision electromagnetic measurements and oscillator frequency stabilization.

Technical noise in K‐band low‐noise cryogenic amplifier

“…At 1 Hz offset the phase noise decreased to −100 dBc/Hz (∼9 dB improvement) and at 1 kHz offset it was reduced to −132 dBc/Hz (∼15 dBc/Hz) for the lower input powers. These observations are consistent with measurements from a lower frequency cryogenic LNA, which also saw a decrease in phase noise intensity at room temperature when the drain voltage was increased [6]. The slope of the phase noise has increased, following −14 dB per decade for lower offsets (<10 Hz) before transitioning to −9 dB per decade (10 Hz-1 kHz).…”

“…For all measurements the amplifier was housed in the vacuum chamber of the cryostat, which is necessary for cooling to cryogenic temperatures but also suppresses the sensitivity of the amplifier to ambient temperature fluctuations, making it possible to confidently resolve the very close‐to‐carrier phase noise of the amplifier [6]. For example, at 100 mHz offset the closest any measurement came to the system noise floor was 10 dB.…”

“…For example, at 100 mHz offset the closest any measurement came to the system noise floor was 10 dB. As outlined elsewhere [6], data for frequency offsets 1 Hz and above are recorded on a fast Fourier transform (FFT) vector signal analyser, while a digital voltmeter (DVM) is used to sample the time‐dependent voltage output at a rate of 8 Hz. The voltages recorded by the DVM can then be converted into a PSD by taking a subset of 2 n points, where T obs is the observation time and N pts is the number of data points (2 n ).…”

“…The availability of ‘full spectrum’ noise information becomes increasingly elusive as one moves to higher and higher carrier frequencies. Yet, amplifier 1/ F noise ( α 1 ) is of paramount importance in oscillators [5, 6] and related technologies such as frequency synthesis [7, 8]. The physical causes of these fluctuations are technical in origin; it is generally accepted to be due to the DC power biasing of the amplifier [4], hence it is often referred to as technical noise.…”