Spectral Characterization of a 90 GHz CLASS Pixel

TL;DR

This paper presents a spectral characterization of 90 GHz CLASS witness pixels, introducing a Fourier-transform spectroscopy approach and a spectral-ratio method to separate the on-chip bandpass from other microwave components. By testing multiple optical-filter configurations across three measurement runs, the authors demonstrate that the on-chip bandpass is relatively flat, with -3 dB edges near $80$ GHz and $108$ GHz, and they discuss possible mechanisms for a small low-edge shift. The methodology also enables isolation of differential transmission for quasi-optical filters and neutral-density filters, enhancing calibration and optimization for current and future CLASS detector arrays. The results provide a practical framework for component-level spectral diagnostics that improve spectral purity and inform instrument design and data analysis for CMB polarization studies.

Abstract

The Cosmology Large Angular Scale Surveyor (CLASS) is an experiment designed to measure the polarization of the cosmic microwave background on large angular scales to probe cosmic reionization and search for the inflationary $B$-mode signal. CLASS is a multi-frequency ensemble of telescopes with bands centered at 40, 90, 150, and 220 GHz. Each telescope has arrays of feedhorn-coupled transition edge sensor bolometers at the focal plane. The frequency response is primarily defined by the on-chip bandpass filter with additional contributions coming from the feedhorn, orthomode transducer, and 180-degree hybrid. In this study, we compare simulations and measurements of the frequency response of single pixel witness devices in the 90 GHz band with and without the bandpass filter. For the first time, we can separate the effects of the bandpass filter from the other microwave components using Fourier transform spectroscopy and design splits of the pixel. The results show that the -3 dB band edges are at 80 GHz and 108 GHz. The measurements demonstrate a robust method for characterizing the spectral response of individual components, which is crucial for optimizing the performance of future detector arrays.

Figures (9)

• Figure 1: Confocal microscope image of the pixel from a 3D optical profilometer with mm-wave components labeled. The mm-wave components for the orthogonal polarization can be seen on the upper left side of the image. Images of the TESs under higher magnification are inset in the corners. An image of the bandpass filter is shown on the bottom with the microstrip running horizontally and the five stubs extending above.
• Figure 2: Simulated transmission and reflection of the band pass filter design. The atmospheric transmission with 1 mm of precipitable water vapor is shown as the dotted line. The passband falls squarely within the atmospheric transmission window pardo2002atmospheric.
• Figure 3: Photograph of two 6$\times$6 witness pixels from the first of five fabricated arrays. The pixel on the left has on-chip bandpass and lowpass filters, while the pixel on the right has these features replaced by a run of microstrip transmission line.
• Figure 4: Spectra from both baseline and no on-chip filter pixels from run 1. The 10 and 11 quasi-optical low pass filters have distorted the spectrum near 100GHz. The ratio of each spectra yield the isolated on-chip filter response which has a much flatter passband in comparison. The ratio fluctuates outside the band edges due to low signal-to-noise ratio in the no on-chip filter pixel and is shaded gray in these regions.
• Figure 5: Frequency spectrum of the baseline pixel in both polarizations from measurement run 2, without the NDF, and from measurement run 3, with the NDF installed and spectral corrections applied. Overplotted is the atmospheric transmission with 1 of precipitable water vapor pardo2002atmospheric.