Digital frequency-division multiplexing
As described in the introduction to my more recent work on a different kind of multiplexer, the name of the game for extracting more physics from the cosmic microwave background is to pack more cryogenic detectors into the same observatory. Current telescopes are using ~10,000 sensors, and future projects have plans for another order of magnitude more. An important technology driver for this expansion is multiplexed readout, which reduces the number of wires that have to go into the cryostat holding all these sensors from tens of thousands to roughly one hundred. This reduction brings the system complexity to a manageable level, enables proper cooling of all the cryogenic sensors, and can even reduce the overall cost.
During my graduate work, I helped develop a new version of Digital Frequency-Division Multiplexing (often shortened to just "Dfmux") for use in the Simons Array and SPT-3G telescopes. This multiplexing scheme works in a manner similar to AM radio, in that indiviual detector signals amplitude modulate "carrier tones" which are at unique frequencies. Custom FPGA-based electronics generate these tones, a bank of superconducting LC bandpass filters ensure that only one tone is transmitted through each detector, a Series Superconducting Quantum Interference Device Array (Series SQUID Array or just SSA for short) amplifies the combined multiplexer output, and the same FPGA-based electronics demodulate to separate each detector's signal. Through this development work, we improved upon a previous version of the technology which could combine 16 channels on a single readout line by incorporating fully superconducting custom lithographed bandpass filters, moving to higher carrier tone frequencies of a few MHz, and implementing a new baseband feedback scheme to linearize the SSA. With these developments, we achieved a multiplexing factor of 68. These advances enabled the SPT-3G camera - with a detector count of 16,000 - and the Simons Array telescopes - with a total detector count of 21,000 - to proceed to construction.
Many of my contributions to the Dfmux project focused on testing, production, and deployment of cryogenic multiplexer circuitry for the Simons Array. I also performed detailed studies of the interactions of the complex SQUID and resonator components with their environment, including implications for the total system sensitivity. In the future, Dfmux readout will be incorporated in the LiteBIRD satellite.