Variable-temperature attenuator calibration method for on-wafer microwave noise characterization of low-noise amplifiers

TL;DR

The paper addresses the challenge of accurately measuring cryogenic noise temperatures of on-wafer low-noise microwave devices, where input losses and temperature gradients introduce large uncertainties. It introduces a variable-temperature attenuator calibration based on Y-factor measurements and S-parameters, deriving a mismatch-corrected Y-factor and demonstrating how attenuators at two physical temperatures can extract the input and backend noise terms $T_\text{in}^\text{c}$, $T_\text{in}^\text{h}$, and $T_{50}^\text{BE}$ for a two-port DUT. The method is validated with electromagnetic simulations and experimentally on InP HEMTs in the 4–8 GHz range, showing good agreement with prior measurements and highlighting practical considerations such as attenuator heating and impedance mismatches. The technique offers a pathway to more robust, impedance-aware noise characterization with potential extension to broader temperature ranges via on-chip, temperature-controlled attenuators, benefiting quantum computing and radio-astronomy instrumentation.

Abstract

Low-noise cryogenic microwave amplifiers are widely used in applications such as radio astronomy and quantum computing. On-wafer noise characterization of cryogenic low-noise transistors is desirable because it facilitates more rapid characterization of devices prior to packaging, but obtaining accurate noise measurements is difficult due to the uncertainty arising from the input loss and temperature gradients prior to the device-under-test (DUT). Here, we report a calibration method that enables the simultaneous determination of the backend noise temperature and effective-noise-ratio at the input plane of the DUT. The method is based on measuring the S-parameters and noise power of a series of attenuators at two or more distinct physical temperatures. We validate our method by measuring the noise temperature of InP HEMTs in 4-8 GHz. The calibration method can be generalized to measure the microwave noise temperature of any two-port device so long as a series of attenuators can be measured at two or more distinct physical temperatures.

Figures (7)

• Figure 1: Schematic of the CPS showing all lumped component parameters. The color gradient represents the temperature gradient from room temperature to cryogenic temperatures. The significance of each parameter is described in the text.
• Figure 2: (a) Y-factor vs attenuator loss data simulated assuming perfect impedance matching. Simulated data points at physical temperatures of 20 K (blue) and 80 K (red) are shown. Solid lines show the fit to \ref{['eq:YfacSimple']}(b) Experimentally measured Y-factor vs attenuator loss. Solid lines show the fit to \ref{['eq:YFactAtten']} with interpolation between points to account for discrete changes in the correction coefficients. The non-monotonic line-shape is due to changes in impedance mismatch between devices of different losses. Error bars are the size of the symbols.
• Figure 3: (a) Microscope image of the attenuator chip used in this study. Zoomed image shows two characteristic devices. (b) S-parameters of pictured devices. The measured loss differs from the design by a few dB, which is attributed to a thinner than intended NiCr layer.
• Figure 4: (a) Schematic depiction of the CPS, including relevant electronics. (b) Schematic of the room temperature backend receiver used for noise measurements. The numbered components are listed in \ref{['tab:RTbackend']}. (c) Picture of the CPS with relevant electronics and microwave components in focus. The backend receiver is thermally insulated within a box to minimize 1/f noise from ambient environment fluctuations.
• Figure 5: (a) Time series of the first 400 ms of a pulsed noise source, frequency swept Y-factor dataset for a calibration attenuator. Raw noise voltage data (black) shows noise source pulsing and trigger voltage (yellow) indicates a switch between swept parameter values, in this case frequency. Shown are the end of the calibration period for the first 0.12 ms, three frequency steps, and several hot/cold noise source pulses. Colored sections guide the eye between frequency steps. (b) Extracted Y-factor versus frequency. The dashed line is a guide for the eye.