REVIEW 3 major objections 5 minor 5 references
Recent Advances in Frequency-Multiplexed TES Readout: Vastly Reduced Parasitics and an Increase in Multiplexing Factor with sub-Kelvin SQUIDs
T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read A cold readout redesign cuts parasitic inductance tenfold for TES arrays.
desk verdict A useful engineering demo of an integrated cold fMux module with a direct residual-impedance win, but the 132x multiplexing claim is only a passive resonator-yield test, not demonstrated readout. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the Cold Integrated fMux Module (CIMM): a single sub-kelvin package that holds the LC resonator comb, the TES detectors, and a series-array SQUID amplifier on one temperature stage, with a dissipationless inductive bias divider replacing the conventional bias resistor. The load-bearing quantity is the equivalent series impedance presented to each TES. Inductances common to the whole comb cannot be fully tuned out at the bias frequency and therefore look like a real resistance, so cutting stray inductance is the mechanism that improves detector stiffness, linearity, and crosstalk while raising the maximum number of channels per SQUID. The performance numbers come from a network-analysis fit to a circuit model with one capacitance and resistance per resonator plus five global parameters.
What would settle it
Refit the same network-analysis data with a circuit model that includes distributed effects, or measure the stray series inductance directly with a calibrated two-port S21 measurement that does not rely on the 85-parameter fit; if the inferred common series inductance is not below about $20\,\mathrm{nH}$, the claimed order-of-magnitude reduction would not be supported.
Extended reading notes
Core claim
The central claim is that a frequency-multiplexed TES readout can be made both simpler and better by eliminating all intermediate-temperature electronics: a sub-kelvin SQUID reads out the current, and the TES bias is applied through a nearly lossless inductive divider rather than through the usual series resistor. With this cold integrated architecture, the parasitic inductance in series with the LC comb drops by roughly an order of magnitude to about $10\,\mathrm{nH}$, and the measured residual impedance of a latched superconducting TES is about $0.02\,\Omega$, an order of magnitude below the previous system. The authors further claim that this low-parasitic environment—not any change in the SQUID itself—is what allows the multiplexing factor to grow from 68 to 132 channels, and they report a 94% resonator yield (122 of 130 connected resonances) as a lower limit. Meeting the white-noise requirement of $12\,\mathrm{pA}/\sqrt{\mathrm{Hz}}$ with this chain rounds out the demonstration.
Load-bearing premise
The headline parasitic numbers ($10\,\mathrm{nH}$ stray inductance, $200\,\mathrm{pF}$ comb capacitance, $4.2\,\mathrm{nH}$ bias inductor) come from an 85-parameter fit to an analytic circuit model the authors themselves call imperfect, and they give no quantitative systematic uncertainty; if that model is biased, the order-of-magnitude improvement could be smaller than claimed.
Editorial extensions
If this is right
- The low-inductance, low-thermal-conductance cables between the sub-kelvin and 4 K stages become unnecessary, simplifying cryostat assembly and reducing heat load.
- Lower-impedance TES detectors, around $200\,\mathrm{m\Omega}$ normal resistance, become usable; the authors expect this to improve robustness against excess current noise.
- The same architecture can scale toward the roughly $500{,}000$-detector class of next-generation CMB experiments, since cost and complexity scale with the number of readout channels.
- A 132x multiplexing factor, demonstrated with 94% yield on a prototype LC chip, implies that per-SQUID channel count can roughly double without sacrificing detector noise.
- The reduced parasitic impedance improves detector stability and crosstalk even at previously used multiplexing factors.
Reading between the lines
- The paper leaves open the systematic uncertainty in the 85-parameter circuit fit; a direct measurement of stray inductance with a two-port calibration would settle whether the order-of-magnitude claim holds outside that model.
- If the parasitic reduction is as large as reported, the same architecture could plausibly be pushed beyond 132 channels or used with lower bias frequencies, where the effective series resistance from common inductance is more severe.
- The reported 94% yield is a lower limit, since two channels were not connected; the true fabrication yield may be higher, which matters for large arrays where every dead channel costs observing time.
- The sub-kelvin SQUID's roughly $20\,\mathrm{nW}$ dissipation raises a question the paper does not address: whether this heat load is acceptable for the most demanding ultra-low-power focal planes.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This manuscript reports on a Cold Integrated fMux Module (CIMM) for frequency-domain multiplexed readout of transition-edge sensor (TES) arrays. The module places a SQUID amplifier and a reactive (inductive-divider) bias circuit on the sub-Kelvin stage, eliminating 4 K front-end electronics and the associated low-inductance cabling. The authors present network analysis of a 40-channel comb, from which they extract a parasitic series inductance of roughly 10 nH, no measurable real series impedance, a comb capacitance of roughly 200 pF, and a bias inductance of 4.2 nH. They show a single detector transition with a residual impedance of approximately 0.02 Ω when the TES is superconducting, and they report a noise equivalent current below the 12 pA/√Hz requirement, albeit with a frequency-dependent transfer function. They also present a network-analysis scan of a 132-channel resonator chip with superconducting shorts in place of TESs, observing 122 of 130 connected resonances (94% yield). The paper concludes that these advances reduce parasitic inductance by an order of magnitude and enable an increase in multiplexing factor for CMB-S4-scale arrays.
Significance. If the parasitics and yield results are borne out, the CIMM architecture is a credible path toward scaling fMux readout to the ~500,000 detectors planned for CMB-S4. The direct measurement of 0.02 Ω residual series impedance when the TES is superconducting is a genuinely informative result, and the white noise level near or below 12 pA/√Hz in a first integrated module is encouraging. The paper is a compact technology demonstration rather than a full system characterization, and its main value is in showing that sub-Kelvin SQUID operation with reactive biasing is practical. The strengths are the end-to-end assembly, the direct residual-impedance measurement, and the explicit list of the analytic model's limitations. The principal weakness is that the headline 'increase in multiplexing factor' is inferred from a passive resonator-yield test, not demonstrated by simultaneous readout of many TES channels.
major comments (3)
- [§3.5 and Fig. 6] The claim of an increased multiplexing factor, which appears in the title and abstract, is not supported by the experiment presented in §3.5. The test uses superconducting shorts in place of TES chips and counts resonant peaks in a network-analyzer scan; it does not bias, read out, or stabilize a single SQUID channel while 132 TESs are operated simultaneously. A multiplexing factor in fMux requires simultaneous operation with electrothermal feedback, controlled crosstalk at the designed frequency spacing, and acceptable noise across the full comb. The paper itself carefully describes the measurement as an 'LC resonator yield,' but the title and abstract state a stronger conclusion. Please either add a demonstration of simultaneous multiplexed readout of a large fraction of the 132 channels, or explicitly reword the claims to say that high resonator yield has been shown while full multiplexed operation remains to be demonstrated.
- [§3.2] The order-of-magnitude reduction in parasitic series inductance to ~10 nH rests on an 85-parameter fit to network-analysis data for which the authors state that quantitative uncertainties are not presented because the analytic model is known to be imperfect. Because this number is a headline result and is used in the conclusions to argue for improved linearity, stability, and crosstalk, the lack of any uncertainty budget or independent cross-check leaves the magnitude of the improvement undetermined. The direct ~0.02 Ω residual-impedance measurement is less model-dependent and is a genuine result, but it does not anchor the ~10 nH inductance value. Please provide at least a model-systematics estimate, a comparison with a simpler extraction method, or a direct reactance measurement to support the claimed factor-of-ten reduction.
- [§3.4 and Fig. 5] The statement that the noise requirement is met is stronger than the data shown. The measured white noise is below 12 pA/√Hz in a limited sense, but the noise falls with frequency because of the bias-circuit transfer function, and the authors state that they expect to correct this dependence to achieve approximately 10 pA/√Hz 'across the full readout bandwidth.' As presented, compliance has been demonstrated only at some frequencies, not over the full band. Please qualify the claim in the abstract and §3.4 by stating that the frequency dependence is a known property of the present bias circuit and that full-bandwidth compliance is planned rather than measured.
minor comments (5)
- [§3.2] The sentence 'the effects of which do not study in this work' is grammatically incomplete; it should read 'the effects of which are not studied in this work.'
- [§3.5] The text states that two of 132 designed channels were not connected and that 122 peaks were observed, 'representing a lower limit on the LC resonator yield of 94%.' Since 122/130 = 93.8%, the rounding to 94% is fine, but the eight missing connected resonances are never explained; please clarify whether they are attributed to fabrication defects, measurement artifacts, or an as-yet-unidentified cause.
- [§3.1] The quoted SQUID input coupling (25 µA/Φ0) and forward gain (up to 1500 V/A) are given without specifying the bias point; adding the relevant bias settings would improve reproducibility of the measurement.
- [Fig. 3] The caption for Fig. 3 does not specify which panel corresponds to 800 mK and which to 250 mK, nor does it define the plotted quantity; the text says 'left' and 'right,' but the caption should be self-contained.
- [§3.3] Figure 4 shows a single detector transition; it would be helpful to state whether this transition and the ~0.02 Ω residual impedance are representative of the full 40-channel array, since the network analysis indicates 100% yield.
Circularity Check
No significant circularity: the paper is an experimental characterization whose headline numbers are extracted from measurements, not assumed as inputs.
full rationale
This paper reports laboratory measurements of a Cold Integrated fMux Module (CIMM). The headline parasitics (stray inductance ~10 nH, comb capacitance ~200 pF, bias inductor 4.2 nH) are outputs of an 85-parameter chi-squared fit to network analyzer data, not inputs assumed in order to derive the claimed improvement; the authors explicitly note the model's systematic uncertainties rather than presenting the fit as a prediction. The residual impedance of ~0.02 ohm is a direct measurement from a resistance-voltage curve taken while the TES is superconducting. The 132x claim is supported only by a resonator-yield test with superconducting shorts, which is a limitation in the strength of the multiplexing demonstration, but this is a sufficiency-of-evidence concern, not circularity: no fitted parameter is renamed as a prediction and no self-citation is used to force the result. The self-citations (refs. 6 and 7) provide prior context and the circuit diagram, but the central results are new measurements that do not reduce to those references. Therefore the derivation chain is self-contained and no circular step is present.
Assumptions & free parameters
free parameters (5)
- Stray series inductance of LCR comb =
~10 nH
- Capacitance across LCR comb =
~200 pF
- Bias inductor value =
~4.2 nH (design target 5 nH)
- SQUID input coupling =
~25 µA/Φ0
- Residual series impedance (TES superconducting) =
~0.02 Ω
assumptions (4)
- domain assumption The analytic circuit model used for the network analysis fit accurately represents the physical CIMM circuit.
- domain assumption When the TES is fully superconducting, the measured residual impedance is dominated by the readout circuit rather than by contact or wiring resistances.
- domain assumption The SQUID gain calibration used to refer noise back to the TES is accurate.
- domain assumption Standard TES electrothermal feedback and voltage-bias models apply to these detectors.
Cite this review
Pith. "Pith review of Recent Advances in Frequency-Multiplexed TES Readout: Vastly Reduced Parasitics and an Increase in Multiplexing Factor with sub-Kelvin SQUIDs." pith.science (2026). https://pith.science/paper/IXMXXAXB
@misc{pith2026190807642,
author = {Pith},
title = {Pith review of: Recent Advances in Frequency-Multiplexed TES Readout: Vastly Reduced Parasitics and an Increase in Multiplexing Factor with sub-Kelvin SQUIDs},
year = {2026},
howpublished = {\url{https://pith.science/paper/IXMXXAXB}},
note = {Machine review of arXiv:1908.07642}
}
read the original abstract
Cosmic microwave background (CMB) measurements are fundamentally limited by photon statistics. Therefore, ground-based CMB observatories have been increasing the number of detectors that are simultaneously observing the sky. Thanks to the advent of monolithically fabricated transition edge sensor (TES) arrays, the number of on-sky detectors has been increasing exponentially for over a decade. The next-generation experiment CMB-S4 will increase this detector count by more than an order of magnitude from the current state-of-the-art to ~500,000. The readout of such a huge number of exquisitely precise sub-Kelvin sensors is feasible using an existing technology: frequency-domain multiplexing (fMux). To further optimize this system and reduce complexity and cost, we have recently made significant advances including the elimination of 4 K electronics, a massive decrease of parasitic in-series impedances, and a significant increase in multiplexing factor.
Reference graph
Works this paper leans on
-
[1]
Abitbol, M. H. et al. (2017), arXiv:1706.02464
arXiv 2017
-
[2]
Gottardi, L. et al. (2016) DOI:10.1016/Nuc. Instr. 824
work page doi:10.1016/nuc 2016
-
[3]
Dobbs, M. A.; Lueker, M. et al. (2012), DOI:10.1063/Rev. Sci. Instr. 83/7
work page doi:10.1063/rev 2012
-
[5]
Bender, A. N. et al. (2019), arXiv:1907.10947
work page Pith review arXiv 2019
-
[7]
de Haan, T. et al. (2012), DOI:10.1117/Proc. SPIE 8452
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.