Demonstration of a 1820 channel multiplexer for transition-edge sensor bolometers

Abstract

The scalability of most transition-edge sensor arrays is limited by the multiplexing technology which combines their signals over a reduced number of wires and amplifiers. In this Letter, we present and demonstrate a multiplexer design optimized for transition-edge sensor bolometers with 1820 sensors per readout unit, a factor of two more than the previous state-of-the-art. The design is optimized for cosmic microwave background imaging applications, and it builds on previous microwave superconducting quantum interference device multiplexers by doubling the available readout bandwidth to the full 4-8 GHz octave. Evaluating the key performance metrics of yield, sensitivity, and crosstalk through laboratory testing, we find an end-to-end operable detector yield of 78%, a typical nearest-neighbor crosstalk amplitude of ~0.4%, and a median white noise level of 83 pA/rtHz due to the multiplexer, corresponding to an estimated contribution of 4% to the total system noise for a ground-based cosmic microwave background telescope. Additionally, we identify a possible path toward reducing resonator loss for future designs with reduced noise. We expect these developments to alleviate the system complexity, cryogenic requirements, and cost of future large arrays of low temperature detectors.

Figures (5)

• Figure 1: Schematic representation of the microwave SQUID multiplexer which also depicts the general method for lithographically adjusting the resonator frequency and its couplings to the SQUID and feedline. The strength of the capacitive coupling $C_c$ between the resonator and the feedline is adjusted via the finger length in an interdigital capacitor, and the strength of the inductive coupling $M_c$ between the resonator termination and the SQUID is tuned by adjusting the overlap area of their respective inductive loops. The resonator spacings are tuned through the number of "wiggles" $n_w$, the number of "sliders" $n_s$, and the indent of the sliders $\delta_s$. In the example resonator shown, $n_w = 11$ and $n_s = 2$. The ranges these circuit parameters take are given in Table \ref{['tab:parameters']}.
• Figure 2: Physical implementation of the multiplexer. a) Micrograph of a few microwave SQUID channels. b) Photograph of an individual chip. c) Photograph of the 28-chip multiplexer assembly, integrated vertically with the TES wafer via wire bonds along its perimeter. d) Exploded CAD model of the full detector and readout module, showing (i) an optical low-pass filter MetalMeshFilters, (ii) a feedhorn arraySiliconPlateletFeedhornArray, (iii) the stack of optical coupling wafers and the TES wafer, (iv) the multiplexer assembly, and (v) the spring-loaded copper lid which encloses the resonators in an electrically small volume.
• Figure 3: Low-level characterization of the fully assembled multiplexer. Top: Measured readout line transmission for the fully integrated readout and detector module, normalized at 4 GHz and measured with a fixed sweep power that corresponds to $\approx$-70 dBm on the mux chip feedline at 6 GHz. Center left: Target and measured pairwise channel spacings, extracted from the above transmission measurement. Bottom left: Target and measured channel bandwidths, measured using the SMuRF readout system under nominal operating conditions. Bottom right: Measured saturation powers of 1366 out of a maximum possible 1748 detectors, extracted from measurements of the TES I-V relationship. In all panels, dark and light shaded bands contain the central 50% and 90% of the measured channels in each frequency bin, respectively.
• Figure 4: Key performance metrics. Left: Measured nearest-frequency-neighbor crosstalk amplitudes for 3 channel pairs from single-chip tests, along with the modeled distribution across all channels in the multiplexer. The direct measurements are consistent with expectations given the particular spacing of each pair. We model that 82% of neighboring channel pairs will experience crosstalk <$10^{-2}$. Center: Measured white noise level of the readout, referenced to an equivalent current fluctuation through the TES. Right: Measured distributions of the readout and detector thermal fluctuation white noise contributions to the total noise, referenced to an equivalent power fluctuation measured by the TES bolometer. For further comparison, projected photon noise levels for an example CMB telescope salatino2020design are also shown. In all panels, the boxes and whiskers contain the central 50% and 90% of channels.
• Figure 5: Left: Measured $Q_i$ of the resonators under nominal power before and after wire bonding the inputs of the multiplexer chips to the detectors and intermediate routing traces. All other wire bonds (flux ramp, feedline, and perimeter ground bonds) were present for both measurements. Right: Measured readout noise from a set of 100 channels before wire bonding the inputs of the multiplexer chips to the detectors and intermediate routing traces, along with another set of 100 channels measured afterwards. The shaded regions are histograms of per-channel the readout white noise level, and dots indicate the median white noise level for each configuration. After connecting the multiplexer chip inputs, the noise level is elevated due to the reduced $Q_i$ and becomes slightly dependent on the number of channels simultaneously operating due to the impact of 3$^{\mathrm{rd}}$-order intermodulation distortion products.