Pith. sign in

REVIEW 3 major objections 4 minor 11 references

Automated Signal Integrity Analysis Framework for High-Speed Interconnects in the PPCB-1347-MuPix11 Probe Card

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read No single route descriptor—insertion loss, mode conversion, or eye size—ranks the four 1.25 Gbps MuPix11 probe-card links; physical length alone is not a proxy for link quality.

desk verdict Honest, reusable SI pipeline for one probe card, but the central comparative claim rests on differences smaller than the stated numerical cross-check; treat the route ranking as preliminary. read the letter →

arxiv 2608.09462 v2 pith:HRIAMJX7 submitted 2026-08-10 hep-ex cs.SYeess.SY

D M S Sultan , R. Plackett , A.E. McDougall , A.J.A. Knight , A.S. Rotelli , J. Vossebeld , D. Bortoletto This is my paper · ORCID
classification hep-excs.SYeess.SY
keywords signalintegrityS-parametersmixed-modeconversionhigh-speedinterconnectprobecardeyediagrambiterrorrate8b10bencoding
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 argues that a reusable signal-integrity framework can convert four-port S-parameter files into traceable link-level evidence, and that applying it to the four 1.25 Gbps differential routes of a MuPix11 probe card shows no single descriptor—insertion loss, mode conversion, or eye size—ranks the links. The practical stakes are that design reviews often use physical route length or one loss number as a proxy for link quality; the paper finds those proxies disagree. It reports that DP3 has the lowest Nyquist insertion loss, DP1 the strongest differential-to-common isolation, and DP4 the largest modeled eye, while the 23.9 mm length span changes simulated insertion loss by only 0.05 dB. All analytical bit-error-rate projections fall below the stated reporting floor, so the paper declines to rank the routes by BER. If the framework is right, the same automated chain can be reused on any compatible four-port S-parameters, from simulation or after fabrication.

What carries the argument

The load-bearing machinery is the automated pipeline's chain of transforms: a power-normalized single-ended-to-mixed-mode conversion (sum and difference waves) that separates differential transfer from modal conversion; a non-negative route-length-aware loss decomposition into conductor $\sqrt{f}$, dielectric $f$, and higher-order $f^2,f^4$ bases; a causal loaded-channel rational model that includes the stated source and load termination and the 0.35 pF receiver input; a full PRBS-31 convolution to the receiver waveform; and a conditional BER model that keeps deterministic ISI in the sampled voltage level and adds only receiver noise and slope-scaled aperture jitter as Gaussian terms. The 8b10b encoder/decoder and exhaustive 1024-word enumeration turn those waveform statistics into separate coded-link event projections.

What would settle it

Measure the fabricated probe-card routes with a calibrated VNA at the same reference planes and run the same mixed-mode and eye pipeline; if the measured differential-to-common isolation of DP1 is not clearly the strongest, or if measured Nyquist $S_{DD21}$ does not reproduce the simulated −0.30 to −0.35 dB spread, the claimed metric-dependent ranking is a property of the simulation, not the physical card.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central claim is that the four probe-card routes do not admit a single ranking: at 1.25 Gbps, DP3 minimizes differential insertion loss ($S_{DD21}=-0.300$ dB), DP1 provides the strongest differential-to-common isolation ($-31.038$ dBc), and the shortest route DP4 has the largest modeled FEB-input eye ($0.588$ V), while DP4 is neither the lowest-loss nor the most mode-isolated route. The paper frames this as evidence that physical path length is not a sufficient proxy for link quality: the 99.42–123.33 mm path span produces only a 0.05 dB range in $S_{DD21}$ and a 2 mV range in channel-only eye height. It also establishes a traceability boundary: all analytical BER and 8b10b event-rate values lie below declared floors, so those columns cannot support a route ranking, and only the weaker finite-record confidence bounds are defensible.

Load-bearing premise

The simulated four-port S-parameters used for every route-level result faithfully represent the real probe-card channels, including the localized return-current discontinuities at anti-pads and vias; the paper itself states that final verification requires hardware validation.

Editorial extensions

If this is right

  • All four routes retain modeled FEB-input eyes within 0.586–0.588 V at 1.25 Gbps, so under the stated transmitter and receiver assumptions none of the short probe-card links is a bottleneck from vertical margin alone.
  • Because each route wins on a different metric, acceptance for the probe card must be written as a set of per-metric thresholds rather than a single best-route choice.
  • A 23.9 mm length difference yields only 0.05 dB in Nyquist $S_{DD21}$ and 2 mV in eye height, so choosing by length alone can misorder links in this class of interconnect.
  • The sub-$10^{-15}$ BER and sub-$1.25\times 10^{-6}$ events/s 8b10b projections are reporting-floor bounds, not measurements; longer simulations or timed hardware tests are needed before any BER-based ranking is claimed.
  • The same checked pipeline applies to any compatible four-port Touchstone file, making the comparison reproducible for VNA-measured data after fabrication.

Reading between the lines

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

  • The metric dependence the paper finds suggests that the practical acceptance document for such a probe card should be a small vector—loss budget, mode conversion, eye height, and a BER bound—rather than a scalar figure of merit; the framework already outputs all four columns.
  • The strongest route discriminator here is differential-to-common conversion, whose 2.8 dB spread dominates the eye and loss spreads; a natural next check is whether this spread survives on fabricated hardware, since it is driven by local layout asymmetries rather than length.
  • Because the loss decomposition uses correlated basis functions, the apparent cross-route differences in conductor versus dielectric contributions should not be read as material differences; de-embedded coupon measurements could test whether those basis assignments are physically meaningful.
  • If the framework is applied to VNA-measured data after fabrication, the eye and mode-conversion columns could be compared directly with the simulation; the paper's own 0.06 dB agreement on $S_{DD21}$ suggests the simulated ranking is testable at that precision.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces a MATLAB-based signal-integrity analysis framework that converts four-port S-parameter data into mixed-mode parameters, a causal loaded-channel model, propagated waveforms, eye metrics, conditional BER, and 8b10b coded-link projections. The framework is demonstrated on four differential routes (DP1-4) of the PPCB-1347-MuPix11 probe card using pre-production CST simulation data. The main comparative claim is that physical route length is not a sufficient ranking variable because different metrics favor different routes: DP3 has the lowest Nyquist insertion loss, DP1 the strongest differential-to-common isolation, and DP4 the largest modeled eye. The authors repeatedly and explicitly distinguish model projections from finite-record observations and report sub-floor BER values as bounds rather than as measured values.

Significance. If the comparative claims are supported, the paper offers a reusable, transparent pipeline that could be valuable for pre- and post-fabrication SI assessment of detector interconnects, and its insistence on separating raw S-parameter evidence, model assumptions, and finite-record observations is methodologically sound. The work also honestly documents its limitations, including the non-uniqueness of the loss decomposition and the need for hardware validation. However, the central comparative claim rests on very small numerical differences, and the paper does not currently establish that those differences exceed the numerical uncertainty of the supplied simulation data and the re-processing pipeline; this is the load-bearing issue for the conclusion that length is not a sufficient proxy.

major comments (3)
  1. [§3, Table 2, Table 4] The central claim that physical length is not a sufficient ranking variable relies on route-to-route differences in SDD21 of only 0.05 dB and in RX eye height of only 2 mV. However, Section 3 reports that a direct re-evaluation of the supplied Touchstone files reproduces the PTSL SDD21 values only within 0.06 dB, and SDD11/SDD22 within 4.94/4.79 dB. The paper does not report a CST mesh-convergence study, a tolerance for the rational-fit approximation, or residual bounds for the eye pipeline. As a result, the observed SDD21 span and the eye differences are comparable to or smaller than the stated re-processing uncertainty, so the route-ranking distinctions (for example, DP3 versus DP4, and the equal SDD21 of DP1 and DP4) are not currently supported. Please provide an uncertainty budget for the simulated S-parameters and the analysis chain, or explicitly weaken the comparative conclusion to differences that exceed the numerical noise.
  2. [§2.2, Eq. (2.4), Table 2] Equation (2.4) is acknowledged in the text to be non-unique because the bases are correlated over the fit band, yet Table 2 and Section 5 report dielectric-proxy values of 0.183 dB for DP4 and 0.000 dB for DP1-DP3, and then present path-normalized values in dB/m. These numbers carry physical connotations (dielectric loss) that the non-identifiability does not support. The paper should either omit the per-route decomposition from the comparative evidence or demonstrate that the coefficients are identifiable, for example by reporting the fit residual under alternative basis choices or by using a regularized decomposition. As written, the decomposition may mislead readers into interpreting a mathematical artifact as a physical difference between adjacent traces.
  3. [§4, §5, Table 2] There is an internal inconsistency in the reported eye heights. Section 4 states that at 1.25 Gbps 'the eye height remains 0.593V' for DP4, while Table 2 lists the RX eye for DP4 as 0.588 V and Table 4 lists channel-only eye differences of 1-2 mV among routes. The manuscript defines both 'channel-only eye' and 'FEB-input eye' but does not clearly state which quantity is reported in each location. Please reconcile the 0.593 V value with Table 2, and define precisely whether the reported eye includes the 2 mV noise and 2 ps jitter or not. This matters because the 2 mV eye difference among routes is only marginally above the declared receiver noise.
minor comments (4)
  1. [§4, text after Eq. (4.1)] The sentence 'a zero dielectric proxy does not prove' is incomplete; it should state what a zero dielectric proxy does not prove (e.g., that the dielectric is lossless).
  2. [Table 1] Table 1 lists all seven differential pairs including CLK, SIN, and SYNC RES, but the paper's analysis focuses on DP1-4. Please define the exact mapping from DATAOUT1-4 to DP1-4 in the table caption or in the text.
  3. [§5] The phrase 'spTAB bonded alone for the Mu3e outer pixel detector HDI-flex' is unclear; please spell out 'spTAB' and clarify whether this is a bonding technology or a route-specific feature.
  4. [§2.5] The eye metric uses the 1st/99th percentiles of the received waveform, but the finite record length of the PRBS-31 sequence is not stated in the eye-metrics section. Please report the number of decisions contributing to the percentiles and any finite-record confidence bound on the eye height, since the reported inter-route differences are 1-2 mV.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the route-ranking metrics are direct transformations of the supplied S-parameters, and the only fitted decomposition is explicitly disclaimed and not used for prediction.

full rationale

The paper's derivation chain is self-contained relative to the PTSL-supplied four-port S-parameters. The headline quantities — SDD21, differential-to-common conversion (D-to-C), SCC21, and the modeled eye — are computed directly from those S-parameters via the algebraic mixed-mode transformation (Eqs. 2.1–2.2) and the causal channel convolution (Eq. 2.6); none of them is produced by fitting a parameter to the quantity it is then said to predict. The only fitted model is the loss decomposition of Eq. (2.4). That fit is used solely for a descriptive split into conductor/dielectric/higher-order basis terms, and the paper explicitly disclaims its physical interpretation: the bases are correlated over a finite band, the decomposition is non-unique, a zero dielectric proxy does not prove a lossless dielectric, and a non-zero higher-order coefficient does not by itself locate radiation or discontinuities. Thus Eq. (2.4) is not a load-bearing prediction and cannot make the argument circular. The eye, BER, and 8b10b results are model projections under explicitly declared transmitter and receiver assumptions; the 10^-15 reporting floor is transparent, and the paper explicitly states that the equal sub-floor bounds support no BER or coded-event ranking. The comparative claim that physical length is not a sufficient ranking variable is simply an observation about the computed S-parameter-derived metrics (Table 2) and does not reduce to a fitted input. The PTSL report [5] is an external source of raw simulation data, not a self-citation; and no uniqueness theorem or prior-work assertion is used to forbid alternative interpretations. The skeptic's concern about unquantified CST and MATLAB re-processing uncertainty is a correctness or uncertainty limitation, not circularity: invalidating the input simulation would invalidate all results, but that is external-input dependence, not definitional self-reference. No pattern from the enumerated circularity classes is present.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The central results depend on two sets of inputs: the proprietary CST S-parameters and the explicit transmitter/receiver model assumptions. The only fitted quantities are the loss-decomposition coefficients, which the paper itself warns are non-unique and not physically separable. No new physical entities are introduced.

free parameters (8)
  • A_fixture (insertion loss intercept) = not reported
    Whole-path intercept in Eq. (2.4), fitted per route to the S-parameter insertion loss over 0.1-0.8 GHz.
  • a_c (conductor loss coefficient) = contributions 0.073-0.303 dB (Table 2)
    Fitted sqrt(f) basis coefficient in Eq. (2.4); non-unique due to correlated bases.
  • a_d (dielectric loss coefficient) = contributions 0.000-0.183 dB (Table 2)
    Fitted linear f basis coefficient; zero value does not imply lossless dielectric per paper.
  • a_2, a_4 (higher-order coefficients) = contributions 0.000-0.021 dB (Table 2)
    Fitted quadratic and quartic residual-curvature bases; paper cautions they do not measure radiation.
  • Receiver input capacitance = 0.35 pF
    Assumed Arria V input capacitance, stated as not a measured specification (Section 1).
  • Receiver voltage noise sigma_v = 2 mV RMS
    Assumed receiver noise used in Eq. (2.8); not measured.
  • Aperture jitter sigma_t = 2 ps RMS
    Assumed sampling clock jitter in Eq. (2.8); not measured.
  • Transmitter launch parameters = 0.7 Vpp, 180 ps edge, -0.08 post-cursor
    Assumed MuPix-like transmitter model stated in Section 4.
assumptions (4)
  • domain assumption CST-simulated S-parameters accurately model the physical channel as a linear time-invariant system.
    All results derive from these S-parameters; fidelity not experimentally verified.
  • domain assumption The rational-function approximation preserves the causal channel response over the band of interest.
    Used in Section 2.4 to build impulse response; non-causal pre-cursor energy is discarded.
  • domain assumption The 8b10b model with running disparity captures the coding properties relevant to the link analysis.
    Section 2.7 states the model does not reproduce FEB firmware or CDR loop.
  • domain assumption Receiver noise and jitter are independent Gaussian sources.
    Eq. (2.8) combines them in RSS; the paper acknowledges this is a model, not a measured distribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Automated Signal Integrity Analysis Framework for High-Speed Interconnects in the PPCB-1347-MuPix11 Probe Card." pith.science (2026). https://pith.science/paper/HRIAMJX7

@misc{pith2026260809462,
  author       = {Pith},
  title        = {Pith review of: Automated Signal Integrity Analysis Framework for High-Speed Interconnects in the PPCB-1347-MuPix11 Probe Card},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HRIAMJX7}},
  note         = {Machine review of arXiv:2608.09462}
}
read the original abstract

A reusable MATLAB signal-integrity (SI) framework is presented that converts compatible four-port S-parameter data, measured by VNA or obtained from electromagnetic simulation, into traceable link-level evidence rather than a single loss metric. The framework is demonstrated on the four 1.25 Gbps differential routes (DP1-DP4) of the PPCB-1347/MuPix11 probe card using PTSL CST Microwave 3D-Solver-derived four-port S-parameters and a virtual time-domain solver. The automated pipeline preflights file structures, performs a power-normalized mixed-mode transformation, applies route-length-aware loss decomposition, constructs a causally loaded channel model, and propagates full PRBS-31 sequences into eye-diagram, conditional-BER, and 8b10b-coded-link analyses. At the 1.25 Gbps data rate (Nyquist 0.625 GHz), DP1-DP4 exhibit differential insertion loss (SDD21) from -0.350 to -0.300 dB, differential-to-common conversion from -31.038 to -28.236 dBc, and modeled FEB-input eye openings from 0.586 to 0.588 V. The comparison shows that path length alone is not an adequate SI ranking variable: DP3 has the lowest Nyquist insertion loss, DP1 the strongest differential-to-common isolation, and DP4 the largest modeled eye. All analytical BER values remain below the reporting floor and therefore do not support a BER ranking. By preserving the distinction between route-dependent waveform behavior, model projections, and finite-record observations, the framework provides an extensible basis for comparative high-speed-interconnect SI analysis from design review through calibrated VNA measurement interpretation.

Figures

Figures reproduced from arXiv: 2608.09462 by the authors.

Figure 1
Figure 1. summarizes the per-route MATLAB implementation. Each CST Touchstone file is first checked for frequency coverage, storage convention, and physical port pairing; the CST header assignment is cross-checked numerically before a power-normalized single-ended-to￾mixed-mode transformation is applied. The response-mode-first matrices provide differential transfer, differential-to-common conversion, common-to-differential c… view at source ↗
Figure 2
Figure 2. PTSL probe-card overview [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 2
Figure 2. PTSL probe-card overview [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: CST time-domain reflectometry profiles for all probe-card differential pairs. 5. Signal-path results The PTSL-exported ODB++ layout was imported into the CST Studio 3D electromagnetic field solver with port configurations of excitation (1,2) → response (3,4). Single-en…
Figure 4
Figure 4. Figure 4: (a) Differential S-parameters and operating-band loss decomposition; (b) mixed-mode conversion and loss-residual correlation [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 4
Figure 4. Figure 4: (a) Differential S-parameters and operating-band loss decomposition; (b) mixed-mode conversion and loss-residual correlation [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: (a) Modelled MuPix-side transmitter eye; (b) FEB receiver-input eye with nominal noise; (c) 2.50 Gbps stress eye at 1.25 GHz Nyquist. A MuPix-like differential signal is launched with a deterministic PRBS-31 generator that maps the generated bits to NRZ levels of plus …
Figure 6
Figure 6. Figure 6: (a) Receiver BER bathtub; (b) receiver BER versus data rate; (c) projected receiver 8b10b [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

11 extracted references · 4 canonical work pages

  1. [1]

    Augustin et al., The MuPix system-on-chip for the Mu3e experiment, Nucl

    H. Augustin et al., The MuPix system-on-chip for the Mu3e experiment, Nucl. Instrum. Methods Phys. Res. A 845 (2017) 194-198, doi:10.1016/j.nima.2016.06.095

    doi:10.1016/j.nima.2016.06.095 2017

  2. [2]

    Weber, H

    A. Weber, H. Augustin and I. Perić, MuPix10 Documentation, Mu3e Internal Note 0052, version 0.7 (2023)

    2023

  3. [3]

    Berger and M

    N. Berger and M. Müller, Front -end board v2.0 Manual, Mu3e Internal Not e 0057, version 1.2 (2024)

    work page 2024

  4. [4]

    Köppel and M

    M. Köppel and M. Müller, Protocol for the data transmission between Frontend - and Switching-Board, Mu3e Internal Note 0047, version 0.2 (2018)

    2018

  5. [5]

    PTSL, PPCB-1347 - MuPix11 - Final SI Results, revision R02, internal technical report (2026), not publicly available

    work page 2026

  6. [6]

    IBIS Open Forum, Touchstone File Format Specification, version 2.0, technical specification (2009)

    2009

  7. [7]

    Bockelman and W.R

    D.E. Bockelman and W.R. Eisenstadt, Combined differential and common -mode scattering parameters: theory an d simulation, IEEE Trans. Microwave Theory Tech. 43 (1995) 1530. - 16 -

    work page 1995

  8. [8]

    Pozar, Microwave Engineering, 4th edition, Wiley, Hoboken U.S.A

    D.M. Pozar, Microwave Engineering, 4th edition, Wiley, Hoboken U.S.A. (2012)

    2012

Show all 11 references
  1. [9]

    Arndt et al., Technical design of the phase I Mu3e experiment, Nuclear Instruments and Methods in Physics Research Section A, Vol

    K. Arndt et al., Technical design of the phase I Mu3e experiment, Nuclear Instruments and Methods in Physics Research Section A, Vol. 1014 (2021)

    2021

  2. [10]

    Hall and H.L

    S.H. Hall and H.L. Heck, Advanced Signal Integrity for High -Speed Digital Designs, Wiley, Hoboken U.S.A. (2009)

    2009

  3. [11]

    Widmer and P.A

    A.X. Widmer and P.A. Franaszek, A DC -balanced, partitioned -block, 8B/10B transmission code, IBM J. Res. Dev. 27 (1983) 440

    1983

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.