Mod-Cam: CCAT/FYST Cryogenic Receiver

• Mod-Cam is a single-module cryogenic receiver for CCAT/FYST, hosting a 280 GHz broadband camera with over 10,000 kinetic inductance detectors.
• Its design features innovative off-axis dilution refrigeration, multiple cryogenic stages, and flexible stripline interconnects for effective thermal and electrical management.
• Mod-Cam’s successful integration of cryogenic performance and RFSoC-based readout demonstrates a scalable path toward Prime-Cam’s future deployment of ~100,000 KIDs.

Mod-Cam is the CCAT/FYST program’s single-module cryogenic receiver used as the telescope’s first-light and commissioning instrument and, simultaneously, as an in-lab systems testbed for CCAT’s first deployable instrument module. In its current configuration, it hosts a 280 GHz broadband camera with more than 10,000 kinetic inductance detectors (KIDs) across three detector arrays, 18 RF chains, and six 46 cm low-thermal-conductivity flexible stripline assemblies between 4 K and 300 K. Within the CCAT instrumentation roadmap, Mod-Cam is the pathfinder for Prime-Cam, which is intended to scale the same module and readout concepts to roughly $10^5$ KIDs across seven instrument modules (Keller et al., 29 Sep 2025).

1. Programmatic role within CCAT/FYST

Mod-Cam is embedded in the instrumentation plan of the Fred Young Submillimeter Telescope (FYST), the 6 m crossed-Dragone telescope being built for Cerro Chajnantor in Chile at 5600 m elevation. In that plan, Prime-Cam is the full first-generation science receiver, designed to host up to seven independently developed instrument modules, while Mod-Cam accepts one Prime-Cam-compatible module at a time and is used both for early on-sky deployment and for laboratory qualification of module hardware (Vavagiakis et al., 2022).

The programmatic role of Mod-Cam has been stable across successive design and status papers, while the project maturity has changed. An early design/status paper presented Mod-Cam as the first-light cryostat for a 280 GHz MKID array, with deployment targeted for late 2022 (Duell et al., 2020). A later design paper described Mod-Cam deployment on FYST with a 280 GHz instrument module as planned for early science observations in 2024, and formalized its longer-term role as the module testbed for Prime-Cam (Vavagiakis et al., 2022). The 2025 readout-overview paper then described Mod-Cam as nearing readiness for deployment, with integrated in-lab validation already underway (Keller et al., 29 Sep 2025).

That pathfinder function is central rather than incidental. Mod-Cam is used to validate cryogenic interfaces, cold and warm readout, module packaging, flexible stripline harnessing, and full-chain detector operation before those same design choices are multiplied inside Prime-Cam. A plausible implication is that Mod-Cam is best understood not as a reduced standalone instrument, but as the qualification environment for the Prime-Cam module ecosystem.

2. Receiver and module architecture

The later CCAT design papers describe Mod-Cam as a $0.9\ \mathrm{m}$ diameter, $1.8\ \mathrm{m}$ long single-instrument-module cryogenic receiver with an off-axis dilution-refrigerator architecture and cryogenic stages at 40 K, 4 K, 1 K, and 100 mK. The vacuum shell is 6061-T6 aluminum, the 40 K shell is mainly 6063-T5 aluminum, and the 4 K shell is 6061-T6 aluminum. Cooling of the 40 K and 4 K shells is provided by a Cryomech PT-420 pulse-tube cooler, while the sub-kelvin stages are provided by a Bluefors LD-400 dilution refrigerator (Vavagiakis et al., 2022, Lin et al., 29 Sep 2025).

A defining mechanical feature is the off-axis or “side-car” dilution refrigerator. This leaves the rear of the receiver accessible, so instrument modules can be installed from the back and cantilevered from the 4 K plate. The 4 K plate is the sole structural mount for the instrument module, while the module’s 1 K and 100 mK stages are coupled to the dilution refrigerator through OFHC copper cold fingers, clamps, and braided OFHC straps (Vavagiakis et al., 2022, Lin et al., 29 Sep 2025).

The optical and structural content of the 280 GHz module is Prime-Cam-compatible. The module is about 41 cm in diameter and 130 cm long, with three metamaterial anti-reflection-coated silicon lenses, staged infrared and low-pass filtering, 4 K magnetic shielding in A4K material, a 1 K carbon-fiber tube, and a 100 mK carbon-fiber truss. Stray-light control uses about 240 wedge-shaped injection-molded carbon-loaded plastic metamaterial tiles in the upper 4 K section, a blackened 1 K Lyot stop, and additional 1 K ring baffles and shield blackened with Stycast 2850 FT loaded with carbon powder (Vavagiakis et al., 2022).

The thermal design margins reported for the 2022 receiver configuration were substantial. Estimated total loads were 28.4 W at 40 K, 0.51 W at 4 K, 0.907 mW at 1 K, and 50.6 $\mu$W at 100 mK, against available cooling powers of 110 W, 4 W, 24 mW, and 400 $\mu$W respectively (Vavagiakis et al., 2022).

3. The 280 GHz detector module and readout topology

The module currently installed in Mod-Cam is the first CCAT instrument module, a 280 GHz broadband camera. It contains more than 10,000 KIDs split across three detector arrays; each array is divided into six readout networks of about 570 detectors each, giving 18 RF chains total (Keller et al., 29 Sep 2025). The 2022 detector design paper gave the more explicit breakdown summarized below (Vavagiakis et al., 2022).

Component	Configuration
Detector arrays	3
Total detector count	10,352 total; 10,286 optically coupled
Array composition	1 TiN array, 2 Al arrays
Networks per array	6
Total RF networks	18
Resonators per network	572 or 576
Resonator band	roughly 500 MHz to 1 GHz

The array composition is technologically mixed. One array is a TiN/Ti/TiN trilayer KID array coupled to aluminum feedhorns and contains 3,456 total detectors, of which 3,450 are optically coupled. The two additional arrays use aluminum KID designs with silicon platelet feedhorn arrays; each contains 3,448 total detectors, of which 3,418 are optically coupled. The mixed deployment was motivated by dark-test evidence that the aluminum design reduced low-frequency spectral noise relative to the TiN design, while still allowing on-sky comparison of the two detector technologies (Vavagiakis et al., 2022).

The cold readout chain begins at the arrays at about 0.1 K. On the input side from 0.1 K to 4 K, the system uses 304 stainless-steel coax; on the output side it uses NbTi coax. Input attenuation is distributed among the 0.1 K, 1 K, and 4 K stages, and the paper notes that optimization is still ongoing because the design must balance focal-plane thermal stability against readout noise. At 4 K, cryogenic low-noise amplifiers made at Arizona State University are mounted on the back of the instrument module. Flexible coax then connects the 4 K chains to the stripline interface through custom SMP-to-SMA transition PCBs, and from 4 K to 300 K six stripline assemblies carry the signals through the cryostat wall to the warm electronics (Keller et al., 29 Sep 2025).

Warm simultaneous readout of all detector networks has already been demonstrated using CCAT’s RFSoC-based electronics. That integrated demonstration exercised the full chain from KID networks through cryogenic coax, 4 K LNAs, transition PCBs, striplines, and warm digital RF electronics (Keller et al., 29 Sep 2025).

4. Flexible stripline as thermal harness and electrical interconnect

A major engineering theme of Mod-Cam is the 4 K–300 K flexible stripline solution that Prime-Cam is intended to inherit. Each assembly uses a 46 cm long DuPont flexible circuit with a buried copper signal layer between polyimide dielectric layers and copper ground layers on the outer surfaces. At dedicated clamp points the ground plane widens across the full circuit width to improve thermal sinking at the 4 K and 40 K stages; elsewhere the ground traces are narrowed to reduce thermal conduction. Profilometry measured the ground traces to be about $22~\mu\mathrm{m}$ thick and $0.43~\mathrm{mm}$ wide, larger than the original $18~\mu\mathrm{m}$ by $0.38~\mathrm{mm}$ design, likely because of unexpectedly thick gold plating (Keller et al., 29 Sep 2025).

Thermal characterization began with a residual resistance ratio measurement on a $\sim 5$ cm segment. The stripline measured $0.9\ \mathrm{m}$0 at 300 K and about $0.9\ \mathrm{m}$1 at 4 K, implying $0.9\ \mathrm{m}$2, consistent with high-purity OFHC copper. The heat-flow model used the relation

$0.9\ \mathrm{m}$3

with literature $0.9\ \mathrm{m}$4 for OFHC copper integrated numerically over the relevant temperature range to obtain $0.9\ \mathrm{m}$5. Direct conductivity measurements in a Bluefors SD250 dilution refrigerator were consistent with OFHC-copper expectations for $0.9\ \mathrm{m}$6 roughly 50–150 and with the measured $0.9\ \mathrm{m}$7. Polyimide conduction was neglected because its cryogenic conductivity is more than 4000 times smaller than copper’s (Keller et al., 29 Sep 2025).

The resulting harness heat-load estimates quantify the design trade between Mod-Cam and Prime-Cam.

System	300 K $0.9\ \mathrm{m}$8 40 K per harness	40 K $0.9\ \mathrm{m}$9 4 K per harness
Mod-Cam	0.71 W	0.64 W
Prime-Cam	0.79 W	0.53 W

The paper attributes Prime-Cam’s lower critical 40 K–4 K load, despite a much larger stripline population, to two specific changes: halving the copper pour thickness in signal and ground traces to $1.8\ \mathrm{m}$0, and roughly doubling the circuit length between the 4 K and 40 K stages by about 20 cm compared with Mod-Cam (Keller et al., 29 Sep 2025).

Electrical characterization showed that the stripline concept itself was not the dominant source of inter-network coupling. Warm $1.8\ \mathrm{m}$1 measurements over 300 MHz–1.5 GHz showed that stripline-alone inter-channel crosstalk was at least 50 dB below the desired transmission and met project requirements. When transition PCBs were included on both ends, crosstalk rose substantially, especially among channels 2, 4, and 6 that shared a PCB; those combinations showed crosstalk levels up to 30 dB higher than combinations not sharing a PCB. Cold measurements showed no significant reflections, reinforcing the conclusion that the main issue was pickup between neighboring readout chains rather than mismatch-induced artifacts (Keller et al., 29 Sep 2025).

The proposed remedy was a redesigned transition PCB in which the signal is carried through an impedance-matched via and rapidly buried in an inner stripline layer between ground planes. Ansys HFSS was used to optimize the microstrip, stripline, and via dimensions for

$1.8\ \mathrm{m}$2

Simulation predicted crosstalk no greater than $1.8\ \mathrm{m}$3 dB for a single transition structure and $1.8\ \mathrm{m}$4 dB across the full PCB, about a 20 dB improvement over the original design (Keller et al., 29 Sep 2025).

5. Cryogenic performance and impact on 280 GHz KID noise

A separate 2025 characterization paper focused on cryogenic behavior and on whether focal-plane temperature fluctuations add appreciable noise to the 280 GHz arrays. Two cooldown configurations were emphasized: an “optically open” cooldown to validate overall cryostat performance and a “cold load” cooldown to measure focal-plane thermal stability and detector thermal responsivity (Lin et al., 29 Sep 2025).

During the optically open test, the abstract reports stable base temperatures of 1.5 K on the 1 K stage and 85 mK at the detector stage. The body also reports module plate temperatures near $1.8\ \mathrm{m}$5–$1.8\ \mathrm{m}$6 and $1.8\ \mathrm{m}$7–$1.8\ \mathrm{m}$8, but contains internal inconsistencies, including a “173 K” entry for the optically open 40 K front shell that the paper itself does not reconcile. The same section also contains a cooldown-figure caption inconsistent with the accompanying text. The low-temperature conclusion is nonetheless clear: Mod-Cam reached stable near-100 mK detector-stage operation in both dark and optically representative configurations (Lin et al., 29 Sep 2025).

The 40 K interface required a documented engineering change. Homemade clamped OFHC copper ribbon straps and a copper adapter had produced an upward thermal drift of about $1.8\ \mathrm{m}$9. Replacing that interface with three braided OFHC copper straps from TAI connected to welded bosses on the 40 K shell reduced the 40 K front plate base temperature from 72.8 K to 55 K and improved the long-term drift to $\mu$0 (Lin et al., 29 Sep 2025).

At the focal plane, the detector stage was servo-controlled at 100 mK with a PID loop. Over an 8-hour, 10 Hz temperature timestream, the detector focal plane RMS stability was

$\mu$1

Thermal responsivity was measured by stepping the bath from 96 mK to 105 mK in 2 mK increments and fitting resonator frequency shifts. The reported median responsivities were

$\mu$2

$\mu$3

The sign difference was interpreted physically: negative TiN responsivity followed standard Mattis-Bardeen expectations, while positive Al responsivity was attributed to a nonlinear two-level-system effect dominating the measured slope (Lin et al., 29 Sep 2025).

The paper then translated thermal fluctuations into equivalent optical loading using

$\mu$4

and

$\mu$5

Using imported optical responsivities at 5 pW loading, it obtained $\mu$6 equivalent power fluctuation for aluminum devices and $\mu$7 for titanium-nitride devices. Relative to a 5 pW incident photon power, these correspond to $\mu$8 for aluminum and $\mu$9 for TiN, which the paper judged negligible for CCAT science goals (Lin et al., 29 Sep 2025).

An important limitation is that optical responsivity was not measured directly in Mod-Cam during this campaign; it was imported from related detector characterization. The noise-impact argument is therefore partly inferential rather than a complete in-situ optical NEP measurement. Even so, the measured thermal stability and responsivity data support the conclusion that cryogenic bath fluctuations are not expected to limit the 280 GHz module (Lin et al., 29 Sep 2025).

6. RFSoC warm readout, software stack, and scaling to Prime-Cam

The warm readout architecture shared by Mod-Cam and Prime-Cam is based on Xilinx ZCU111 RFSoC boards and the primecam_readout software framework. In the configuration described for FYST instruments, each board can drive up to 4000 KIDs, the software runs on a centralized control computer connected to the boards via dedicated Ethernet, and computational demands are intended to increase linearly with detector count (Burgoyne et al., 2024). Control uses dynamically generated channels via Redis, while high-rate time-ordered data are streamed separately over UDP and converted into G3 format by rfsoc-streamer; higher-level tuning and acquisition logic is handled by ccatkidlib (Patel et al., 7 Oct 2025).

For the 280 GHz module in Mod-Cam, the 2025 RFSoC system paper reported a concrete deployment of five custom-packaged RFSoCs. Each RFSoC can simultaneously read out four RF channels with up to 1,000 detectors spanning a 512 MHz bandwidth per channel using the current firmware, so five boards provide sufficient capacity for all 18 readout networks in the installed module (Patel et al., 7 Oct 2025).

The readout procedure is sweep-based. A full-network VNA-style sweep uses 1000 equally spaced tones over 512 MHz around a numerically controlled local oscillator; resonators are identified from

$\mu$0

together with phase information from

$\mu$1

After candidate resonators are found, narrower target sweeps are used to place tones and optimize drive level, typically about 1–3 dB below bifurcation. Cable delay is corrected by multiplying the complex transmission by

$\mu$2

after which the resonator loop is fit in the complex plane, translated to the origin, and decomposed into frequency and dissipation quadratures (Patel et al., 7 Oct 2025).

A representative detector summary from an Al network under $\mu$3 K cryogenic blackbody loading showed stronger fluctuations in the frequency quadrature than in the dissipation quadrature, which the paper interpreted as detector-noise-dominated rather than readout-noise-dominated behavior. The same campaign also reported preliminary averaged spectral responses for one TiN network and one Al network measured through the full Mod-Cam optical stack with a Martin–Puplett interferometer; both showed a $\mu$4 GHz bandpass centered at $\mu$5 GHz (Patel et al., 7 Oct 2025).

These demonstrations matter because Mod-Cam’s readout architecture is the direct template for Prime-Cam scaling. The 2025 stripline paper notes that one Prime-Cam harness is planned to use 18 striplines and 48 of 54 available readout chains, with six spare chains (Keller et al., 29 Sep 2025). The RFSoC system paper frames the Mod-Cam achievement as the foundation for reading out $\mu$6 KIDs in Prime-Cam’s future full-capacity configuration (Patel et al., 7 Oct 2025). A plausible implication is that Mod-Cam’s significance lies less in one 280 GHz module by itself than in demonstrating that the complete CCAT cold-and-warm readout architecture is already operable at the $\mu$7-detector scale.