arXiv:2610.23810
Experimental Demonstration of a Balanced Travelling Wave Parametric Amplifier
Nikita Klimovich, Peter K. Day, Henry G. Leduc, Boon-Kok Tan
Abstract
Superconducting travelling-wave parametric amplifiers (TWPAs) provide broadband gain, low added noise, and high dynamic range, making them attractive for applications requiring wide instantaneous bandwidth. This is crucial for amplifying multiplexed narrow band signals spread over a wide frequency range, such as qubit resonator arrays, or amplifying inherently wide band signals, such as in spectrometers or radiometers. In many cases, however, the useful bandwidth of a TWPA is limited by the production of the idler tone at an image frequency. For three-wave-mixing (3WM) TWPAs, for example, an input signal at $f_s$ is indistinguishable from one at $f_p - f_s$, which can produce spectral ambiguity when the input spectrum is broadband or not known in advance. We experimentally demonstrate a balanced architecture that doubles the instantaneous image-rejected bandwidth available to a 3WM TWPA by spatially separating the amplified signal from the output pump and generated idler, which resolves the sideband degeneracy. The architecture consists of two amplifiers connected between a pair of 180$^{\circ}$ hybrid couplers, with the signal and pump injected and output through the sum and difference ports, respectively. Using two kinetic-inductance TWPAs, we demonstrate approximately 20 dB of signal gain across nearly an octave of bandwidth, together with pump rejection below --20 dB and an average image rejection of approximately --15 dB over the operating band. These results establish balanced TWPAs as a practical approach to broadband parametric amplification, enabling improved contamination free bandwidth and eliminating the need for complex filtering.
AI Summary
Conventional three-wave-mixing TWPAs emit the amplified signal, pump, and idler through the same port, creating image ambiguity for broadband or unknown input spectra. This paper experimentally implements a previously proposed balanced architecture that routes the signal and generated idler to different physical outputs, aiming to double the image-free usable bandwidth. Two nominally identical NbTiN kinetic-inductance TWPAs are connected between two $180^\circ$ hybrids. Injecting the signal through the sum port and the pump through the difference port produces in-phase signals but opposite-phase pumps and idlers in the two arms. The output hybrid then separates these components by interference. Frequency-resolved measurements at both outputs, with a $4.75\,\mathrm{GHz}$ pump, show approximately $20\,\mathrm{dB}$ signal gain over the defined $1.8$–$3.2\,\mathrm{GHz}$ operating band, average image leakage near $-15\,\mathrm{dB}$ relative to the desired output, and pump leakage of $-21.37\,\mathrm{dB}$. Individual-device and back-to-back-hybrid measurements characterize gain and imbalance. A multiple-reflection model also explains reduced gain ripple through differential arm phases. Performance is limited by hybrid bandwidth and amplitude/phase imbalance, together with unequal TWPA electrical lengths. The inferred kinetic-inductance mismatch is approximately $0.05\%$; closer wafer placement is proposed to improve matching. At $f_p/2=2.375\,\mathrm{GHz}$, signal and idler coincide spectrally and comparable powers appear at both outputs, giving approximately $0\,\mathrm{dB}$ rejection under the measurement definition. The demonstration does not measure added noise or dynamic range; its experimental performance metrics are gain and pump/image rejection.
Error found: The stated 1.8–3.2 GHz operating band includes signals above 3.05 GHz whose idlers fall below the hybrids’ reported 1.7 GHz leakage threshold, qualifying the claimed image-rejected bandwidth. The abstract’s claim that the architecture eliminates complex filtering is broader than the demonstrated finite suppression and the argument that residual pump power is below typical HEMT saturation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.