{"id":"c3779099-fa5b-4f2c-9514-e066f9433474","arxiv_id":"1910.01983","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"FEM simulations of four transformer winding styles at 20 MHz and 20 kHz confirm that skin effect raises AC resistance and lowers leakage inductance, with foil windings cutting resistance at the cost of higher leakage.","lead":"This paper uses finite element simulations to compare circular, square, and foil windings in a high-frequency transformer, and shows how skin effect changes AC resistance and leakage inductance between 20 kHz and 20 MHz. The results give designers a quantitative trade-off map for very-high-frequency power converters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table IV's type-c AC resistance at 20 kHz (0.103 mΩ) is below the DC resistance floor for an AWG18-equivalent 5-turn foil winding, so the 400x ratio is a modeling artifact.","rationale":"The reader correctly identified missing validation, absent mesh-convergence study, and the unsupported 'except type c' exception as weaknesses. However, the more load-bearing concern is independent of simulation settings: Table IV itself contains a physically impossible resistance value for type (c) that violates the DC-resistance lower bound for a winding with the stated cross-section and turn count. This is a decisive internal inconsistency, not simply a missing benchmark. The paper's qualitative message—skin effect increases AC resistance and decreases leakage inductance—is well established and consistent with the general trends in Tables III and IV, so that part may survive. But the paper's headline quantitative contribution, the 'up to 400 times' AC-resistance reduction and the accompanying 13x leakage-inductance variation, rests on a value that cannot be correct. Because the quantitative conclusion is a central claim of the abstract and conclusion, and because the error is not something a mesh refinement or added Dowell validation would fix without first correcting the model, the appropriate verdict for the current manuscript is REJECT. The concern is concrete, testable, and not ad hominem: it points to a specific numerical inconsistency in the reported data.","tokens_in":6960,"tokens_out":7065,"duration_ms":73164,"concrete_test":"Compute the analytic DC resistance of the type-c foil winding from the stated geometry: R_DC = ρ L / (N A), with ρ = 1.68×10⁻⁸ Ω·m, N = 5 turns, A = 0.823 mm² (AWG18), and mean turn length taken from Table II dimensions. Compare this value to Table IV's type-c entry at 20 kHz (0.10297 mΩ). If R_DC exceeds 0.103 mΩ by the expected factor of ~18, the FEM extraction is defective. A complementary check: rerun the same FEM model at f = 1 Hz (or DC) for all four types; all should converge to the same DC resistance. If type (c) still yields ~0.1 mΩ, the model or post-processing is invalid.","verdict_should_be":"REJECT","load_bearing_attack":"The central quantitative claim—that foil type c lowers AC winding resistance by up to 400x relative to AWG18 conductors—depends entirely on Table III's type-c value of 0.19381 mΩ at 20 MHz and Table IV's 0.10297 mΩ at 20 kHz. These values are physically impossible for the stated geometry. Section II states that all conductor types have cross-sectional area equivalent to AWG18 (0.823 mm²) and that primary and secondary each have 5 turns. Type (a) at 20 kHz has Rac = 1.8264 mΩ, which is an excellent proxy for the DC resistance of the 5-turn AWG18 winding: R_DC = ρL/A gives about 1.8 mΩ for a mean turn length near 18 mm from Table II. Since all four winding types share the same conductor area and the same turns, their DC resistances must be nearly identical. Type (c) reporting 0.10297 mΩ at 20 kHz is roughly 18x below this DC floor. Even at 20 MHz, type (c)'s 0.19381 mΩ remains below DC resistance, which is impossible for any series-connected winding regardless of skin or proximity effects. No mesh density, solver setting, or missing convergence study can explain a resistance below the DC value. This indicates an error in the FEM model definition (e.g., foil turns inadvertently shorted in parallel, an over-wide foil cross-section, or incorrect post-processing of Rac). Consequently, the headline 'up to 400 times lower AC resistance' and the related 13x leakage-inductance spread cannot be accepted. The qualitative statement that skin effect raises Rac and lowers leakage inductance is not undermined, but the paper's distinctive quantitative conclusions are not credible as reported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a finite-element method (FEM) study of skin effect in a 1:1 high-frequency isolation transformer with four winding configurations: AWG18 circular wires, square wires of equivalent cross-section, overlaid foil windings (type c), and non-overlaid foil windings (type d). For each configuration, the authors present current-density and magnetic-field distributions at 20 kHz and 20 MHz, and tabulate magnetizing inductance, leakage inductance, and winding AC resistance at both frequencies, along with frequency sweeps. The stated central claims are that skin effect increases AC winding resistance and decreases leakage inductance with frequency, and that for a nearly constant magnetizing inductance the winding topology can vary leakage inductance by up to 13 times and AC resistance by up to 400 times.","tokens_in":7303,"tokens_out":5792,"duration_ms":57836,"significance":"If the quantitative results were reliable, the comparison of winding topologies at very high frequency would be practically useful for designers of SiC/GaN-based converters, where parasitic inductance and winding loss are critical. The study is not circular: no parameters are fitted to the reported outputs, and the inductance and resistance values are taken directly from FEM solutions of Maxwell's equations. The qualitative observation that conductor shape and winding arrangement affect leakage inductance and AC resistance is consistent with textbook physics. However, the quantitative findings, especially the headline '400 times lower AC resistance,' are compromised by a physically impossible value in the tables and by the complete absence of mesh-convergence or validation evidence.","major_comments":[{"comment":"The AC resistance reported for type (c) violates the DC floor for the stated geometry. All winding types are described as having cross-sectional area equivalent to AWG18 (0.823 mm²) with 5 turns per winding. Using the mean turn length implied by Table II (about 18 mm), copper resistivity, and series-connected turns gives R_DC ≈ 1.84 mΩ, which matches type (a) at 20 kHz (1.8264 mΩ). Type (c) reports 0.10297 mΩ at 20 kHz and 0.19381 mΩ at 20 MHz, both well below the DC resistance of the same conductor material and cross-section. No skin effect, eddy-current redistribution, or FEM post-processing error can produce a resistance below the DC value for a series-connected winding of fixed total conductor volume. This indicates an error in the FEM model definition for type (c), such as inadvertently shorting foil turns in parallel or misinterpreting the post-processed quantity. Consequently, the abstract and conclusion claims of 'up to 400 times' lower AC resistance, and any comparison involving type (c), are unsupported and must be re-derived after correcting the model.","section":"III-B, Tables III and IV"},{"comment":"The claim that the skin-effect trend is observable in all transformer types 'except type c' is contradicted by the paper's own tables. For type (c), the leakage inductance decreases from 13.451 nH at 20 kHz to 13.2631 nH at 20 MHz, and the AC resistance increases from 0.10297 mΩ to 0.19381 mΩ. Both changes are in the same direction as the stated skin-effect trend, albeit smaller in magnitude than for the wire conductors. The accompanying explanation that the structure of type (c) 'cancel[s] the skin effect and prevent[s] the eddy current to flow' is also inconsistent with a resistance that nearly doubles over the frequency range. Please reconcile the text with the data or provide a corrected interpretation.","section":"III-B, paragraph after Table IV"},{"comment":"The manuscript provides no information on the FEM solver, mesh density, element order, boundary conditions, or convergence criteria, and it includes no validation of the simulation setup against an analytical benchmark (e.g., Dowell's formula for a foil winding) or against measurement. Given that Table IV contains a physically impossible resistance for type (c), the accuracy of all reported numbers, including the leakage-inductance ratios and the frequency-response curves in Fig. 11, is called into question. The authors should add a mesh-convergence study and at least one validation case demonstrating that the FEM model reproduces a known AC resistance and leakage inductance before the quantitative comparisons can be accepted.","section":"III-A and III-B, FEM methodology"}],"minor_comments":[{"comment":"Equation (3) is garbled: the symbols '𝜄𝜔' and '𝜔𝜔' are unclear, and the expression for AC resistance is dimensionally inconsistent as printed. It should be rewritten, for example as R_ac = ρ l / (w δ) with δ = sqrt(ρ/(π μ0 f)), to make the skin-depth dependence explicit.","section":"II, Eq. (3)"},{"comment":"The text says 'Equation (4) shows the relation between magnetic energy and leakage inductance,' but the displayed equation is numbered (5). Please renumber or fix the cross-reference.","section":"III, before Eq. (5)"},{"comment":"Table I is difficult to parse because the three flux-density values and the three core-loss values are not clearly aligned in separate columns. Please reformat so each pair is unambiguous.","section":"Table I"},{"comment":"The line breaks in the manuscript text make the table entries hard to associate with their rows. Please ensure the actual published tables have clear row and column separation.","section":"Tables III and IV"},{"comment":"The conclusion states that the comparisons are made 'for the same value of the magnetizing inductance,' but Table III shows magnetizing inductances ranging from 397.08 to 422.14 nH, a spread of about 6%. Please qualify this statement.","section":"Conclusion"},{"comment":"Minor language issues include 'at presence of eddy currents' (should be 'in the presence of'), 'trans former ration' (should be 'ratio'), and inconsistent capitalization of 'KHz'/'kHz' and 'MHZ'/'MHz'.","section":"Abstract and throughout"},{"comment":"The FEM software and solver type are not identified. Please state the tool (e.g., ANSYS Maxwell, COMSOL, or another package), the element type, and the mesh settings used for the 20 MHz and 20 kHz simulations.","section":"Section II"}],"recommendation":"major_revision","confidential_remarks":"The physically impossible DC-floor violation for type (c) is the key risk in this manuscript. If the corrected simulations do not restore a physically plausible AC resistance, the quantitative claims will not stand. I would also ask the editor to request the FEM model or a detailed setup description, because the current manuscript does not contain enough detail to audit the simulation independently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: the paper's qualitative conclusion is textbook and its quantitative headline is not just unvalidated—it's physically impossible. The type-c foil winding reports 0.103 mΩ at 20 kHz and 0.194 mΩ at 20 MHz. All conductors are specified as AWG18-equivalent cross-section with 5 turns per winding, so the DC resistance of type c must be close to the 1.83 mΩ seen for the round-wire type (a) at 20 kHz. Reporting a 20 kHz resistance 18x below that floor, and a 20 MHz resistance still below DC, means the FEM model has an error—most likely the foil turns being shorted in parallel or an incorrect cross-section in post-processing. No mesh refinement or validation study can fix a resistance below the DC value.\n\nThat said, the paper is not without merit. The FEM sweep of circular, square, and two foil winding layouts at 20 kHz and 20 MHz gives a concrete picture of current density and field distribution, and the direction of changes—skin effect raises AC resistance and lowers leakage inductance—is correctly identified. The plots in Figs. 5-10 are informative. The problem is that the authors over-interpret their simulation output without any sanity check. Their own tables contradict their \"except type c\" claim: type c's leakage drops from 13.451 to 13.263 nH and its AC resistance rises from 0.103 to 0.194 mΩ between 20 kHz and 20 MHz, exactly the same direction as the other types. So the stated exception is simply wrong.\n\nThere are smaller issues: equation (3) is garbled, the Table III formatting isolates the type (b) row, and there is no mesh convergence study, no analytical benchmark (Dowell would have been easy for the foil), and no data or code. The references are thin but not misleading.\n\nWho is this for? A designer looking for rough FEM-based guidance on winding shapes at HF might find the trends useful, but not the numbers. The 400x and 13x ratios in the abstract and conclusion should not be cited. The paper needs a corrected model, a DC-resistance sanity check, and external validation before it is publishable. Right now it does not deserve referee time; it should be returned to the authors with the DC-floor problem made explicit.\n\nIn short: the qualitative story is fine, the quantitative claim is broken, and the paper is not ready for review.","headline":"The 400x claim is physically impossible (type-c resistance sits below the DC floor for AWG18-equivalent foil), so the paper's headline numbers are not credible, even though the qualitative skin-effect trends are standard.","tokens_in":7844,"tokens_out":2461,"would_cite":false,"duration_ms":23431,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Skin effect raises AC winding resistance and lowers leakage inductance as frequency rises, and the paper reports that winding shape and layout alone can vary these parasitics by up to 400-fold and 13-fold at the same magnetizing inductance.","keywords":["skin effect","eddy currents","leakage inductance","AC winding resistance","high-frequency transformer","finite element method","foil windings","very high frequency power conversion"],"falsifier":"Measure AC resistance and leakage inductance of the four winding arrangements on the same EE core at 20 MHz with an impedance analyzer and compare them with the simulated values; the claim fails if the reported ordering and rough magnitude—especially the roughly 400-fold lower resistance of the overlaid foil winding—are not reproduced.","tokens_in":1559,"feed_emoji":"⚡","tokens_out":2336,"duration_ms":91104,"temperature":0.7,"pith_summary":"The paper simulates a 1:1 high-frequency isolation transformer in four winding styles—round wire, square wire, overlaid foil, and spaced foil—at 20 kHz and 20 MHz to see how the skin effect changes parasitic behavior. It argues that skin effect crowds current toward conductor surfaces, so as frequency rises the winding AC resistance increases and the leakage inductance decreases. It also reports that, at the same magnetizing inductance, the winding arrangement and conductor shape can move leakage inductance by roughly 13 times and AC resistance by 400 times, giving designers a wide lever for tuning loss and resonance. This matters for very-high-frequency converters using fast SiC and GaN switches, where knowing leakage inductance and winding resistance is needed to maintain efficiency and avoid unwanted oscillation.","feed_headline":"Foil windings can cut high-frequency winding loss 400-fold","feed_subtitle":"Simulations show winding shape and layout move leakage inductance up to 13-fold while magnetizing inductance stays nearly fixed.","key_machinery":"The central object is the skin depth, $\\delta=\\sqrt{\\rho/(\\pi \\mu_0 f)}$, which sets how thin the effective current-carrying shell becomes at 20 MHz. The finite-element simulations produce current-density and magnetic-field distributions for each winding arrangement, and leakage inductance is obtained from the volume integral $L_{\\text{leakage}}=\\iiint H\\cdot B\\,dv\\,/\\,I_{\\text{primary}}^2$ taken over the window with the secondary short-circuited. Foil windings work because a thin, wide conductor oriented along the field carries current with less surface crowding, and the overlaid pattern (type c) prevents eddy-current loops from forming.","core_discovery":"On the paper's own terms, the discovery is a quantitative finite-element comparison: at 20 MHz, the round-wire design has AC resistance 80.637 mΩ, the square-wire design 73.377 mΩ, the overlaid foil design 0.19381 mΩ, and the spaced foil design 6.7345 mΩ, while leakage inductance ranges from 3.8030 nH to 51.765 nH. The authors read this as showing that geometry, not just frequency, controls skin-effect losses: overlaid foil conductors keep the current path spread out and suppress eddy-current circulation, whereas round and square wires confine current to thin surface shells. Because the magnetizing inductance stays nearly the same across designs (about 397–422 nH), the parasitic differences are attributed to winding topology. The paper's general conclusion is that the skin effect increases AC resistance and decreases leakage inductance with frequency, with the overlaid-foil winding showing that trend only weakly because its structure largely cancels the eddy-current mechanism.","pith_inferences":["An immediate testable extension would be to wind the four geometries on a physical EE core and measure impedance from 20 kHz to 20 MHz; the paper's 400-fold resistance claim would survive only if the foil winding's advantage persists after terminal and lead resistances are included.","The same skin-effect logic suggests litz-wire designers could reduce loss by using square or shaped strands rather than round strands, since the simulations show a weaker skin effect in square conductors; the paper notes the possibility but does not test a litz bundle.","If leakage inductance can be varied 13-fold at fixed magnetizing inductance, one could co-design the transformer as the resonant inductor of a ZVS or ZCS tank, shrinking the converter by removing a separate inductor—an implication the paper points toward but does not build.","Because core loss at 20 MHz grows steeply with flux density under the reported Steinmetz fit, the 400-fold winding-resistance improvement may matter less than core loss in a full converter, so foil-winding choice should be weighed against total loss rather than winding loss alone."],"forward_implications":["A converter designer can reduce winding AC loss by roughly two orders of magnitude at 20 MHz by switching from AWG18 round wire to overlaid foil windings, at the cost of higher leakage inductance.","Leakage inductance can be tuned over a 13-fold range without changing magnetizing inductance, so the same core can be wound to place the parasitic where a resonant tank needs it.","Square conductors of the same cross-sectional area as round wire show a slightly weaker skin effect, so conductor shape alone is a small but usable loss lever.","The frequency-response curves imply that wire-wound transformers change parasitic values markedly between 20 kHz and 20 MHz, while foil-wound transformers remain comparatively stable.","Because magnetizing inductance stays near 400 nH across all four designs, the observed parasitic differences are attributable to winding geometry rather than to changes in the core."],"supporting_citations":[{"why":"Supplies the very-high-frequency design context and the core-sizing procedure used to fix the transformer dimensions.","marker":"[1]"},{"why":"Establishes the high-frequency converter loss and EMI challenges that make accurate parasitic estimates necessary.","marker":"[2]"},{"why":"Shows how known leakage inductance can be absorbed into ZVS/ZCS resonant tank design, motivating the comparison.","marker":"[6]"},{"why":"Provides the Steinmetz core-loss constants and the design handbook basis for the transformer dimensions.","marker":"[7]"},{"why":"Supplies the 67-material B-H curve used as the core model input for the FEM simulations.","marker":"[8]"}],"fun_headline_variants":["Foil windings slash high-frequency transformer losses","Skin effect tamed by foil winding geometry","Overlaid foil cuts AC resistance 400-fold at 20 MHz","Winding shape, not just frequency, drives skin-effect losses","Simulations show foil windings beat round wires at 20 MHz"],"cache_read_input_tokens":9728,"weakest_assumption_plain":"The quantitative comparisons rest on the finite-element simulation being accurate at 20 MHz; the paper does not provide a mesh-convergence check, an experimental measurement, or a comparison with a standard closed-form resistance formula, so a wrong field solution would shift every reported ratio.","fun_headline_variants_meta":{"raw":{"variants":["Foil windings slash high-frequency transformer losses","Skin effect tamed by foil winding geometry","Overlaid foil cuts AC resistance 400-fold at 20 MHz","Winding shape, not just frequency, drives skin-effect losses","Simulations show foil windings beat round wires at 20 MHz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000376,"raw_usage":{"total_tokens":2000,"prompt_tokens":938,"completion_tokens":1062,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":554,"completion_tokens_details":{"reasoning_tokens":981}},"tokens_in":554,"tokens_out":1062,"duration_ms":8131,"temperature":1.0,"reasoning_tokens":981,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:14:45.310960+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure AC resistance and leakage inductance of the four winding arrangements on the same EE core at 20 MHz with an impedance analyzer and compare them with the simulated values; the claim fails if the reported ordering and rough magnitude—especially the roughly 400-fold lower resistance of the overlaid foil winding—are not reproduced.","supporting_citations":[{"cited_title":"Opportunities and Challenges in Very High Frequency Power Conversion","cited_arxiv_id":null,"evidence_quote":"Supplies the very-high-frequency design context and the core-sizing procedure used to fix the transformer dimensions."},{"cited_title":"Design and Development of Very High Frequency Resonant DC –DC Boost Converters","cited_arxiv_id":null,"evidence_quote":"Establishes the high-frequency converter loss and EMI challenges that make accurate parasitic estimates necessary."},{"cited_title":"Design of Le akage Inductance in Resonant DC -DC Converter for Electric Vehicle Charger,","cited_arxiv_id":null,"evidence_quote":"Shows how known leakage inductance can be absorbed into ZVS/ZCS resonant tank design, motivating the comparison."},{"cited_title":"Available: https://www.fair-rite.com/67-material-data-sheet/","cited_arxiv_id":null,"evidence_quote":"Supplies the 67-material B-H curve used as the core model input for the FEM simulations."}],"review_version":1}