Performance Variance of Low Noise Resonant Capacitance Bridges While Replacing their Ungapped MnZn Ferrite Cores

C.Y. Lui; J. Lipa; S. Buchman; S. Saraf; S. Wang

REVIEW 4 major objections 4 minor 3 references

Performance Variance of Low Noise Resonant Capacitance Bridges While Replacing their Ungapped MnZn Ferrite Cores

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Ten ferrite cores from one manufacturing batch, swapped into the same resonant capacitance bridge, leave the noise floor nearly unchanged: weighted-mean variance below 0.3% at room temperature and below 0.6% at 140 K.

desk verdict Solid batch-consistency dataset, but the <0.3% figure is the standard error of the mean, not the core-to-core spread; the conclusion overstates what the data show. read the letter →

arxiv 2501.00681 v1 pith:4BB2WJEJ submitted 2024-12-31 physics.space-ph physics.ins-det

S. Saraf , S. Buchman , C.Y. Lui , S. Wang , J. Lipa This is my paper · ORCID

classification physics.space-phphysics.ins-det

keywords resonantcapacitancebridgeMnZnferritecorecore-to-corenoisevarianceSIFERRITN41sub-attofaradfloorlow-temperaturemulti-sensorreproducibilityfrequencyscalingof

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether the ferrite transformer core in a precision resonant capacitance bridge is a meaningful source of sensor-to-sensor variability. It reports that ten ungapped MnZn ferrite cores from the same TDK N41 batch, installed one at a time in the same bridge board, produce noise floors that cluster tightly: a weighted-mean variance below 0.3% at 293 K and below 0.6% at 140 K, with all ten cores inside roughly a ±1% band at room temperature. If true, this means multi-sensor systems such as gravity and gravitational-wave instruments can be built from batch-matched cores without per-core noise selection. The same batch also agrees with a previously measured core from a different batch to about 9% at room temperature and 6% at 140 K after frequency and temperature scaling.

What carries the argument

The central object is the resonant capacitance bridge (RCB): a precision AC bridge whose transformer is a planar printed-circuit-board winding on an ungapped MnZn ferrite core, operated at sub-attofarad noise levels. The argument is carried by the circuit-noise relation $S^{1/2} = (1/U_i)\sqrt{8 k_B T / [(2\pi f_0)^3 L(T) Q_{\rm RCB}]}$, which combines with $L \propto f_0^{-2}$ at fixed equivalent capacitance to give $S^{1/2} \propto f_0^{-1/2}$ at fixed temperature. That scaling lets the authors shift each core's measured noise minimum to a common frequency and compare the ten cores on equal footing. The statistical machinery is a weighted mean with a corrected variance that accounts for over- or under-dispersion, applied to the parabolic fits of each spectrum.

What would settle it

Measure the same core repeatedly, removing and reinstalling it and thermally cycling it between 293 K and 140 K each time; if the run-to-run spread of a single core is comparable to the reported <0.3% and <0.6% inter-core spreads, then core-to-core differences are not separately established.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the noise performance of a resonant capacitance bridge is effectively independent of which ungapped MnZn ferrite core from the same manufacturing batch is installed. Fitting each core's noise spectrum to a parabola around its minimum out to the +3 dB points, the ten cores A through J give a weighted-mean minimum noise of 0.3027 aF/√Hz at 293 K with a corrected variance of 0.0012 aF/√Hz, and 0.1815 aF/√Hz with 0.0016 aF/√Hz at 140 K; after scaling every core to a common frequency, the corrected variances become 0.0008 aF/√Hz and 0.0011 aF/√Hz. The authors take these numbers to be a <0.3% and <0.6% variance of the weighted mean, and they point out that the noise minima scale with resonant frequency as $S^{1/2} \propto f_0^{-1/2}$, consistent with what the circuit model predicts for fixed temperature and nearly constant quality factor.

Load-bearing premise

The conclusion rests on each of the ten cores being measured only once, so the tiny spread across cores could in principle include the effect of removing, reinstalling, and cooling the core rather than the core itself.

Editorial extensions

If this is right

A multi-sensor array assembled from cores of one manufacturing batch can be expected to share a common noise floor without per-core selection, within about 1% at 293 K and 1.5% at 140 K for 90% of cores.
The observed $S^{1/2} \propto f_0^{-1/2}$ behavior gives a simple rule for predicting a bridge's noise at other resonant frequencies once one core is characterized.
The combination of frequency and temperature scaling, using the core permeability, provides a way to compare noise measurements made with different bridges, temperatures, and batches.
The cross-batch agreement of about 9% at 293 K and 6% at 140 K suggests that batch-level reproducibility may be adequate for many applications without per-core grading.
For instruments requiring tighter matching than 0.3-0.6%, the paper's own conclusion is that cores would need to be screened from a larger sample.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Because each core in this study was measured only once, the reported <0.3% and <0.6% spreads include any noise from removing, reinstalling, and thermally cycling the core; a repeated-measurement control on a single core would separate core-to-core differences from setup noise.
If the batch-level stability holds broadly, procurement for large sensor arrays could specify material and manufacturing batch rather than individual noise sorting, an operational consequence the paper leaves implicit.
A natural extension is to apply the same swap-test protocol to other core materials or to gapped cores; that would show whether this uniformity is specific to ungapped MnZn N41 or a more general property of ferrite cores.
The paper's single cross-batch comparison is a sample of one; a multi-batch study would tell whether the 6-9% agreement is typical or just a favorable pair of batches.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a useful engineering dataset, and the measurement work is careful, but the headline claim is built on the wrong statistic. The paper reports a weighted mean variance of <0.3% and concludes that RCB noise performance is independent of the ferrite core. That variance is the standard error of the weighted mean, not the spread between cores. Looking at the scaled values in Table 1, the unweighted sample standard deviation is about 0.9% of the mean. So the cores do not match to 0.3%; they match to roughly 1%. That is still useful, but it is not the stated result.

What is new and good: a ten-core same-batch characterization, each core installed in the same board and measured at 293 K and 140 K, with parabolic fits, overdispersion-corrected errors, and a consistency check against their earlier single-core result. No new technique, but a clean extension of an established method.

The main soft spot is the missing repeatability baseline. Each core was measured once, so the across-core variance includes reinstallation and drift. Measuring one core several times would have separated setup noise from core variability. Second, the frequency-scaling assumption: they use f^{-1/2} to normalize to 100 kHz, but their own room-temperature fit gives f^{-0.32±0.03}. That mismatch is large enough to shift individual points by a few percent, comparable to the spread they are trying to quantify. Third, the '±1% band' statement is only true if you include the error bars; the central values themselves are not all inside ±1% of the mean.

Overall, the data are probably honest and the practical takeaway — same-batch cores are consistent to about a percent in noise floor — is plausible. But the central quantitative claim as written is not supported. With a repeatability measurement and a corrected statistical interpretation, this would be a solid contribution.

Recommendation: send to peer review. A referee should ask for the single-core repeat run and for the conclusion to be rewritten around the actual spread, not the standard error of the mean.

Referee Report

4 major / 4 minor

Summary. The manuscript reports sequential noise-floor measurements of ten nominally identical SIFERRIT N41 ferrite cores (A–J) installed one at a time in the same resonant capacitance bridge circuit, at 293 K and at 140 K. The authors fit each noise spectrum with a second-order polynomial around the minimum, extract the minimum noise value and its uncertainty, compute weighted means and corrected variances of the weighted mean, and scale the data to common frequencies using an assumed f^{-1/2} dependence. They compare the results with a previous measurement of a core from a different manufacturing batch. The central conclusion, stated in the Abstract and Conclusions, is that within a manufacturing batch the noise performance of the RCB sensors is independent of the ferrite cores, with a variance of the weighted mean below 0.3% at 293 K and below 0.6% at 140 K.

Significance. If the batch-insensitivity claim were established, it would be practically useful for multi-sensor arrays, allowing arrays to be built without per-core noise matching. The paper has clear strengths: ten cores tested at two temperatures with the same board and instrumentation, documented conversion from dBm to aF/√Hz, use of parabolic fits with stated R^2 values, an overdispersion-corrected weighted mean, and an explicit consistency check against Ref. 1. However, the headline statistic is the standard error of the ensemble mean, not a measure of core-to-core dispersion, and the absence of repeated measurements of a single core leaves setup variability uncharacterized. As written, the result supports a statement about the precision of the average noise of the batch, not a demonstration that individual cores are interchangeable at the 0.3% level.

major comments (4)

[Section II, Eqs. (5)–(6), and Conclusions] The quantity reported as “variance of weighted mean” is the standard error of the ensemble mean, not the variance of the individual core noise values. With n = 10, a sample standard deviation of roughly 1% of the mean yields a standard error of about 0.3%, so the quoted “<0.3%” does not establish that individual cores match to that level. For example, the scaled-to-100 kHz values in Table 1 have an unweighted sample standard deviation of approximately 0.9–1.2% of the weighted mean, depending on whether one uses the parabolic-fit or scaled values. The abstract and conclusions should therefore report a dispersion statistic (sample standard deviation or an inclusion band) as the primary metric for core-to-core variability, and should not present the standard error of the mean as a measure of independence.
[Sections II and III, experimental procedure] Each core A–J was measured only once, with the core removed and reinstalled between runs. Any variability from connector contact, thermal cycling, or slow instrument drift is therefore included in the observed spread. The paper does not provide a control measurement in which a single core is remounted and remeasured multiple times, so the reported spread cannot be attributed entirely to the cores themselves. A repeatability baseline is needed to support the claim that the cores, rather than the measurement setup, match to the quoted level.
[Section II, Eq. (3), and scaling to 100 kHz] The room-temperature power-law fit reported in the text, S_min^{1/2}(f) ∝ f^{-0.32±0.03}, is inconsistent with the assumed f^{-1/2} scaling used to normalize the data. Over the observed 81–90 kHz range, the difference between scaling with the empirical exponent and with −1/2 changes the normalized values by roughly 3%, which is comparable to the spread being characterized. The authors should either use the empirical exponent for the frequency normalization or provide a more detailed justification for the theoretical scaling, and they should show that the main conclusions are robust to this choice.
[Section II, Table 1, and Conclusions] The statement that “all ten cores are contained in a ±1% band” is not supported by the scaled values in Table 1: cores B (0.2824 aF/√Hz) and G (0.2832 aF/√Hz) lie more than 1% above the weighted mean of 0.2788 aF/√Hz, and core E (0.2758 aF/√Hz) lies more than 1% below. The Figure 6 caption correctly states that the band “include[s] all data in within their error bars,” but the abstract and conclusions omit this qualification. The inclusion band and the weighted-mean precision should be reported separately and with explicit reference to the error bars.

minor comments (4)

[Section II, after Figure 4] The cross-reference “Error! Reference source not found.” appears in the sentence referring to Table 1; the reference is broken and should be fixed.
[References] Reference 3 and Reference 6 are identical (Armano et al., Phys. Rev. D 96, 062004 (2017)); please deduplicate and renumber.
[Table 1, Raw Data column] The ‘Raw Data’ minima (e.g., 0.289 aF/√Hz for core A) differ noticeably from the parabolic-fit minima (0.3019 aF/√Hz for core A); the text should clarify that the raw value is the minimum of the averaged spectrum before fitting and explain the origin of the difference, such as frequency resolution or fit bias.
[Abstract and Eqs. (5)–(8)] The reported quantities such as 0.0012 aF/√Hz are standard deviations of the weighted mean, not variances as defined in the text; please use consistent terminology throughout to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the batch-variance result is self-contained, and the f^{-1/2} normalization is derived rather than fitted.

full rationale

The paper's derivation chain is not circular. The frequency-scaling relation S^{1/2} is proportional to f0^{-1/2} is obtained from equations (1)-(3) with stated assumptions (constant C_eq, approximately constant Q at fixed temperature, and L proportional to f^{-2}); it is used only to normalize the measured noise minima to 100 or 150 kHz, not to construct the variance claim. The reported variances are computed directly from ten independently measured core minima and their parabolic-fit errors (Section II, equations (5)-(6); Section III, equations (7)-(8)). The comparison with reference 1 is a same-author consistency check against a different core from a different batch; it is not an input to the batch-variance result, and its temperature and frequency scaling uses equation (2) and the manufacturer's permeability data rather than the cores under test. No fitted parameter is relabeled as a prediction, no uniqueness theorem is invoked, and no known result is merely renamed. The strongest statistical caveat, that the quoted 'variance of weighted mean' is the standard error of the weighted mean rather than the core-to-core population spread, is a matter of statistical interpretation rather than circular reasoning, and it does not raise the circularity score.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper relies on standard RLC circuit noise physics, the assumption of constant Q across cores, the assumption that only L changes when cores are swapped, and the claim that the cores are from a single batch. No new entities are posited. The parabolic fit procedure is a data-analysis choice rather than a fundamental axiom.

assumptions (5)

domain assumption The RCB noise floor follows the thermal noise formula S^{1/2} = (1/U_i) sqrt(8 k_B T / ((2 pi f0)^3 L Q)).
Equation (2) is the basis for all scaling; it assumes a simple resonant RLC circuit with a quality factor Q_RCB.
domain assumption Q_RCB is approximately constant across cores at fixed temperature.
The authors state in Section I that 'Data shows that Q_RCB is approximately constant for a given temperature'; Table 1 shows Q between 6.2 and 6.5 at 293 K.
domain assumption Replacing the core changes only the inductance L, with C_eq constant.
Section I: 'Replacing the transformer cores modifies only the inductance L of the circuit, with all other parameters, and in particular C_eq, remaining constant.'
domain assumption The ten cores A-J are from the same manufacturing batch.
They were 'purchased as a batch from TDK' (Section I); no batch code verification is reported.
ad hoc to paper The noise minimum is reliably estimated by a second-order polynomial fit over the +3 dB region.
Sections II and III use this procedure; R^2 varies from 0.87 to >0.92, and the fit window is defined by the +3 dB points of the measured spectra.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Performance Variance of Low Noise Resonant Capacitance Bridges While Replacing their Ungapped MnZn Ferrite Cores." pith.science (2026). https://pith.science/paper/4BB2WJEJ

@misc{pith2026250100681,
  author       = {Pith},
  title        = {Pith review of: Performance Variance of Low Noise Resonant Capacitance Bridges While Replacing their Ungapped MnZn Ferrite Cores},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4BB2WJEJ}},
  note         = {Machine review of arXiv:2501.00681}
}

read the original abstract

Precision AC resonant capacitance bridges, with a planar printed circuit board transformer using an ungapped MnZn ferrite core, have shown excellent noise performance in high-precision measurements. As many applications use an ensemble of bridges, consistency of performance is critical to the functionality of the systems. If part of the same manufacturing batch and scaled to the same frequency, the noise performance of the ferrite cores at 293 K has a weighted mean variance of <0.3%, with all the cores within a +/- 1% band. At 140 K the weighted mean variance is <0.6% and the 90% inclusion band is +/- 1.5%. Ten cores of SIFERRIT Material N41, from TDK Ferrites Accessories, were tested at room temperature, 293 K, and in a liquid nitrogen dewar at 140 K. Fitted to a parabolic function, to + 3dB on both frequency sides, the weighted mean of the noise minima at 293 K was 0.3027 aF/rtHz, with a variance of 0.0012 aF/rtHz spanning a frequency range of 85.0 +/- 2.7 kHz. Scaled to 100 kHz, the weighted mean and variance were 0.2788 aF/rtHz and 0.0008 aF/rtHz. Corresponding noise values at 140 K were 0.1815 aF/rtHz with a variance of 0.0016 aF/rtHz for the range of 152.7 +/- 6.3 kHz, and 0.1835 aF/rtHz with a variance of 0.0011 aF/rtHz when scaled to 150 kHz. Scaled for resonant frequency and temperature these results are consistent with a previous measurement of another ungapped core (same supplier and type, different manufacturing batch) to within 9% and 6% at 293 K and 140 K respectively.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 2 canonical work pages

[1]

Noise and thermal performance of a sub-attofarad capacitance sensor for precision measurements, with applications in gravitational wave detectors

1 S Saraf, S Buchman, CY Lui, S Wang, J Lipa, “Noise and thermal performance of a sub-attofarad capacitance sensor for precision measurements, with applications in gravitational wave detectors” Review of Scientific Instruments 95(3) 034503 (2024) 2 Davor Mance, “Development of Electronic System for Sensing and Actuation of Test Mass of the Inertial Sensor...

work page 2024
[2]

Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors

Brandt, M. Caleno et. al., “Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors ” Phys. Rev. D, 96, 062004 (2017). 4 GRACE-FO, Gravity Recovery and Climate Experiment Follow-On, (2018-continuing) 5 MICROSCOPE, MICROSatellite with Compensated drag for the Observat...

work page 2017
[3]

Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors

Brandt, M. Caleno et. al., “Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors ” Phys. Rev. D, 96, 062004 (2017). 7 LISA team, “LISA Laser Interferometer Space Antenna” (2024) arxiv.org/pdf/2402.07571v1. 8 Z. Luo, Y Wang, Y Wu, W Hu, G Jin , “The Taiji program: ...

arXiv 2017

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Noise and thermal performance of a sub-attofarad capacitance sensor for precision measurements, with applications in gravitational wave detectors

1 S Saraf, S Buchman, CY Lui, S Wang, J Lipa, “Noise and thermal performance of a sub-attofarad capacitance sensor for precision measurements, with applications in gravitational wave detectors” Review of Scientific Instruments 95(3) 034503 (2024) 2 Davor Mance, “Development of Electronic System for Sensing and Actuation of Test Mass of the Inertial Sensor...

work page 2024

[2] [2]

Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors

Brandt, M. Caleno et. al., “Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors ” Phys. Rev. D, 96, 062004 (2017). 4 GRACE-FO, Gravity Recovery and Climate Experiment Follow-On, (2018-continuing) 5 MICROSCOPE, MICROSatellite with Compensated drag for the Observat...

work page 2017

[3] [3]

Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors

Brandt, M. Caleno et. al., “Capacitive sensing of test mass motion with nanometer precision over millimeter-wide sensing gaps for space-borne gravitational reference sensors ” Phys. Rev. D, 96, 062004 (2017). 7 LISA team, “LISA Laser Interferometer Space Antenna” (2024) arxiv.org/pdf/2402.07571v1. 8 Z. Luo, Y Wang, Y Wu, W Hu, G Jin , “The Taiji program: ...

arXiv 2017