Digital frequency domain multiplexing readout: design and performance of the SPT-3G instrument and LiteBIRD

This thesis presents my work in areas critical to the development and success of the South Pole Telescope (SPT-G) and Polarbear-instruments, and to the design of the LiteBIRD telescope.These efforts cannot be contextualized without the work of others within these collaborations, nor do they exist in a vacuum independent of that labour.Where it may otherwise be ambiguous within the text I indicate work done solely or dominantly by myself using the first person; work done in collaboration with others in the third person; and work done primarily by others through citations or by directly referencing those people.This section provides a general guide to my specific contributions in each chapter, and highlights work from others that I have drawn from or built off of.Chapter provides the scientific and theoretical background and context, and largely invokes historical work cited in the text.Chapter describes the SPT-G and LiteBIRD instruments.Sections . .and . .are devoted specifically to detailing the scope and content of my contributions to both collaborations.Chapter describes well-established TES detector theory, although I introduce some new formalism to present results relevant to readout performance.Chapter presents the theoretical foundation for the present implementation of Digital Frequency Domain Multiplexing (DfMUX).Some of this theory is well-established, or has already been presented in papers I co-authored (specifically, [ ]).Other elements are presented here for the first time, such as the theoretical description of Digital Active Nulling in the presence of digital latency (Section .).This was a collaboration between myself and Graeme Smecher.I authored the resulting algorithm for operating the Digital Active Nulling feedback in this regime, which is used on SPT-G and Polarbear-, and is being developed for LiteBIRD.Chapter presents a detailed analytic description of crosstalk mechanisms in DfMUX systems.These derivations are my own, and improve upon the previous theoretical framework presented in Dobbs et al., [ ].This work distinguishes itself from Dobbs et al. in a number of ways that are now relevant in the higher multiplexing (and bandwidth) regime.Specifically, it addresses crosstalk cancellation mechanisms; consequences of non-ideal stray impedances; and incorporates projection effects from the complex demodulation.I also introduce a quantitative relationship between the crosstalk and scatter in the LC resonance fabrication, which will drive LC fabrication requirements for LiteBIRD.A key outcome of this study is that current efforts to reduce crosstalk by minimizing series stray inductance will spoil the cancellation between crosstalk mechanisms, and would lead to no-better or even worse crosstalk.Instead, I propose an alternative optimization strategy that can reduce crosstalk by a factor of three relative to present designs.Chapter presents an evaluation of SPT-G instrument parameters and performance relative to theoretical expectations.It also presents a theory to explain LC resonance scatter and parasitic series resistance by edge effects due to lithographic defects.The analysis presented, and formulation of the edge effect theory, are my own work; though, I relied on vi critical resources for understanding the cryogenic lithography geometry provided by Amy Lowitz and Gavin Noble.There are a few sets of measurements in this chapter that are not my own -including the characterization of the focal plane detector time constants (led by Zhaodi Pan); measurements of SPT-G crosstalk using optical sources (led by Jessica Avva and Amy Bender); and the initial theory for parasitic parallel capacitances in the LCs (by Daniel Dutcher).Any quantities such as those are credited to others directly in the text.Although I did not contributed directly to those analyses, the measurements were performed using tools I developed or co-developed as part of the pydfmux readout control software library.pydfmux is currently used to operate SQUIDs and tune detector arrays at dozens of laboratory test beds, as well as on the SPT-G and Polarbear-telescopes, and will be adapted for the LiteBIRD system.The pydfmux repository itself is approximately k lines of python and contains algorithms to perform the core functionality of detector and readout operation and provide a real-time interface to the electronics system and TES biases.It also tracks the book-keeping and performs the concurrency required to efficiently operate , detectors simultaneously.Within that repository are transfer functions calculated for the warm and cold electronics, and tools to track and evaluate instrument state and performance.Graeme Smecher developed the computing architecture, concurrency, and database models, while I wrote the algorithms that interface with cryogenic hardware, including those to operate and evaluate SQUIDs and TES detector arrays.In I hosted a summer-school style tutorial session over several weeks to introduce DfMUX users to the new ICE system and software, and traveled to support and teach users at national laboratories and universities throughout the next year.While I continue to be the primary author and manager of pydfmux, there have since been more than individual contributors, with several dedicated developers who continue to modernize and improve the repository.The requirements of modern systems such as SPT-G and Polarbear-mean that this algorithmic design is considerably more sophisticated than previous generations, however many of the algorithms I designed were informed by work done by my predecessors on the SPTpol, Polarbear, and EBEX instruments.These instruments used the previous generation of readout electronics and software, written by Tijmen de Haan, James Kennedy, and Kevin MacDermid, among others.In addition to pydfmux, I devised and wrote substantial portions of the on-board (compiled C) firmware responsible for lower-level interactions between the control software, cryogenic hardware, and data products.Chapter evaluates the SPT-G readout system noise performance, both in isolation and within the context of the SPT-G detector parameters.This work, and in particular the three main findings of this study, are the product of my own analyses.These are ( ), an additional source of readout noise due to a parasitic capacitive "current sharing" path (Section . .); ( ), the relationship between the 220 GHz detector performance and boosted responsivity due to parasitic series impedance (Section .), which resolves the question of why these detectors under-performed expectations; and ( ), the quantitative relationship between readout noise and SQUID dynamic impedance.The primary output of this chapter is a full electronic circuit and noise model, which captures all relevant warm and cryogenic dynamics for predicting instrument readout noise.I have indicated in the text where the above work draws upon specific earlier work of others.The initial discovery of (a separate form of) current sharing noise was a collaborative effort between a large segment of the SPT-G collaboration, and inspired my work to find different mechanisms [ ]. Daniel vii Dutcher provided invaluable early circuit models of capacitive parasitics in the LCs [ ].John Groh, while working on the Polarbear-instrument, first identified a relationship between SQUID output gain and SQUID dynamic impedance.In deriving the nuances in the amplifier noise contributions to the demodulation chain I benefited greatly from internal memos written by Amy Bender.Chapter applies the circuit and noise models validated on SPT-G in Chapter to forecast LiteBIRD readout noise performance and recommend specific detector and readout designs.This analysis is my own, but uses initial conditions provided by the LiteBIRD collaboration and calculated by others.In particular, the expected LiteBIRD photon noise, detector phonon noise, and radiative loading from the optical elements for each observing band are drawn from the LiteBIRD Sensitivity Calculation Version .

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.