QE (Step 4b): Charles Proteus Steinmetz, project "Steinmentz-Charles"

QE agent, 2026-10-08. Inputs read: _text/manifest.json, all 35 digests (Discharges 01-05, Alternating-Current 01-13, Transient 01-17), all three deep reads, facts.json, claims.md, sites.md and the research notes, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Spot checks were made against the extracted text in _text/ and against research/pages/ia-electric-circuits-1917.txt. Recomputations were done in the project venv (scratch script in /tmp/qe-stz). Tags: [read] = extracted text (PDF page number given), [calc] = QE arithmetic, [memory] = recalled, with confidence.

Verdict: MINOR

The seven central verdicts survive QE. The arithmetic behind them reproduces, and the review is careful and honest about what it did not read. But three parts of the Transient book that bear on claims 3, 4 and 7 rest on digests alone: Section V, Section IV chapters VIII-XI, and Section III chapters VIII-IX. These parts contain text that contradicts two statements in the arbiter and the primer:

• On the transformer "danger frequencies" (claim 4), the review says Steinmetz gives no measurement in the 1920 book. In fact the 1920 book says "experimental investigations have shown" that winding natural frequencies fall in the 20-100 kHz band. It also derives that band for transmission lines from attenuation against frequency [read: Transient PDF pp.674, 676].
• On the energy-transfer constant and transitions (claim 3), the arbiter takes over the Contrarian's criticism that the √(z2/z1) rule ignores abrupt junctions. Steinmetz's own 1920 treatment gives the abrupt-junction coefficient 2c1/(c1+c2), with at most twice the incoming voltage [read: PDF pp.631-634].

The fixes are deep reads, corrections to facts.json and claims.md, a "QE revision" section in arbiter.md, and edits to primer.md. Steps 2-4 do not need to be redone. One verdict wording, claim 3 ("energy transfer": Not testable as stated), may change after deep read 2.

1. Coverage

• Every manifest document was digested and none was skipped. The manifest lists no Word, image or audio files.
• facts.json has one entry per document, and its how_read fields describe accurately what was done.
• No source text was left unread through a skip. Coverage: PASS.

2. Shortcuts

Digest-only coverage is acceptable for the machine, polyphase and rectifier chapters, which are peripheral. It is a defect where a central claim depends on digest-only pages:

1. Transient Section V, PDF pp.652-703 (claims 3, 4, 7): defect. The preface of the 3rd edition singles out this section as the new answer to the constant-r, L, g, C assumption [read: p.15]. It contains: - frequency-dependent constants in four frequency ranges (p.659-661); - wave decay tables IV-VI (p.663-675); - the "danger frequencies of transmission systems, 20,000 to 100,000" cycles, plus the statement on transformer windings [read: p.674, 676, quoted in section 4]; - the law of wave-front flattening, gradient G proportional to E/√(distance) (eqs. 27-28, p.698-702).

The arbiter's correction 8 calls the constant-parameter assumption a hidden tension with claim 7 "unless r is read as r(f)". Steinmetz does exactly that in this section, so the tension is his stated limitation and his own fix, not a hidden assumption. 2. Transient Section IV chapters VIII-XI, PDF pp.602-648 (claim 3): defect. These chapters cover the compound circuit and its power and energy, including: - the worked example in which the transformers transfer 1,110 kW to a line that dissipates 1,800 kW [read: p.628]; - the approximate common decrement u0 = Σλu/Λ, of which Steinmetz writes "Whether this expression (46) is more general is still unknown" [read: p.626]; - reflection and refraction, with a transmitted voltage of at most twice the incoming voltage and a "law of refraction" based on a ratio of tangents [read: p.633-634, 638]; - inductive discharge into a line.

The verdict on the "energy-transfer constant" rests on the Discharges lectures and on Section IV chapter III only. 3. Transient Section III chapters VIII-IX, PDF pp.431-482 (claim 7): partial defect. The Mathematician and the arbiter read page images of p.462 and p.467, and the appendix reader read pp.436-445. Two parts of this range were read by no one: - §69 (p.432-433), Steinmetz's own phase-lag argument that radiation resistance at 1 MHz is "630 times the ohmic resistance" (6 ft spacing) and "12,600 r0" (ground return) [read]; - Tables I-III of chapter IX, with the iron-against-copper comparison (p.467-482). 4. Transient Section I chapters VI and VIII, PDF pp.104-157 (claims 1 and 4; experimental record): minor defect. These chapters contain the few test results in the folder (see section 3). The folder statement "no measurement by Steinmetz beyond oscillograms" was never checked against them. 5. Lightning-chain equations (claim 5). Steinmetz's eqs. (17)-(19) were not re-derived; the review recomputed from the constants instead. That shortcut is acceptable, because the conclusion does not depend on the printed closed form.

3. Missed material (from the digests; not in facts.json, claims.md, the reviews, the arbiter or the primer, or only in passing)

Bearing on the claims - The 1916 book in the folder already has the falling arc characteristic. It gives "voltage consumed in the arc flame decreases with increase of current, approximately inversely proportional to the square root of the current", e = e0 + c/√i, with a residual-vapour branch e = r i. There is a worked hard-carbon case: e = 30 + 40/√i, r = 70 Ω, arc power factor 77%. Metal arcs give voltage peaks "frequently several thousand volts" [read: AC PDF pp.382-386]. - This is the same functional form as the 1917 e = 35 + 200/√i that the primer uses. - The reviews and the arbiter date Steinmetz's falling-characteristic account to 1917 and 1923 only. In 1916 it was used for wave distortion, not yet as the energy source of an oscillation. That use is the point of claim 2, and the timeline should say so. - Danger frequencies, 1920. Steinmetz derives the 20-100 kHz band from the frequencies at which a transmission-line oscillation lasts the most cycles: 65 cycles at 100 kHz against 9 at 1 kHz [read: p.674, 676]. He then says "experimental investigations have shown, that the natural frequency of oscillation of the high voltage windings of large power transformers usually is within this range" and links "the not infrequent disastrous experience with such transformers" to it [read: p.674-676]. - This is a claim of measurement with no reference given. It is not "no measurement cited". - It is single-source in the folder and should be quoted as such, with the missing reference flagged. - Wave-front flattening law (Section V, ch. IV). The steepest gradient of a rectangular front falls roughly as 1/√(distance) out to about 30-100 km, and approaches 1/distance² at great distance. - Practical conclusions: a close return conductor carries a steep front far, and a short iron wire at the station gives "practically no protection" against nearby lightning fronts [read: p.685-702]. - This is a quantitative, testable prediction that the review never examined. It is the natural partner of claim 4 and of the reader's "danger" question. - Transition points (Section IV ch. X). The transmitted voltage is 2c1/(c1+c2) times the incoming voltage, at most twice it. Steinmetz says this "explains the frequent increase to destructive voltages, when entering a station from the transmission line" [read: p.633-634]. The Contrarian's abrupt-junction criticism should be set against this text. - Compound-circuit power transfer (Section IV ch. IX). In the worked example (r1 = 52, L1 = 0.21 on the line and so on) the transformers dissipate 22 kW and transfer 1,110 kW to the line. The line dissipates 1,800 kW, of which its own field supplies 690 kW [read: p.627-628]. - This makes "energy transfer" a predicted, measurable flow between sections, as in the 28-mile oscillogram of Discharges Fig. 41, not only a vocabulary choice. - Steinmetz marks the averaging formula for u0 as approximate and of unknown generality [read: p.626]. - Radiation, §69. At 1 MHz Steinmetz states r_rad = 630 r0, which is 162 Ω/km with r0 = 0.257 Ω/km [calc]. His own table gives 48 Ω/km [read: p.444], and the standard antiphase-pair value is 1.43 Ω/km [calc, method as arbiter B]. - So the chapter's own numbers disagree by a factor of about 3.4, and both are far above the standard value. - This supports the existing verdict "Refuted by calculation" and is an instructive example of the crude phase-lag (cos 92.2°) argument. - Recorded tests and observations in the folder, which contradict "no measurement by Steinmetz beyond oscillograms": - Oscillating-current generator: a hot-wire ammeter read 29 A against a calculated 28.88 A [read: Transient p.123]. - Constant-current mercury-arc rectifier: performance curves "from tests" (Fig. 66) and oscillograms Figs. 67-77, including a high-frequency oscillation at the end of overlap that "may" cause a disruptive discharge [read: p.306-309]. - Destructive surge: an "experimental 10,000 V, 40 cycle line" in which a short-circuiting arc caused destructive effects, with the generator voltage observed to drop to about 5,000 V. The calculated peak was 316,000 V at about 1,800 Hz [read: p.157; calc 707·√(20,000·10) = 316,180 V, 44.7 × 40 = 1,788 Hz]. - Synchronous motor: V-curves observed for a synchronous motor (Fig. 163) [read: AC p.355-356]. - Transformer inrush: starting current calculated at 510 A against a normal 4.5 A, with an oscillogram of a 25-cycle transformer [read: Transient p.226-227].

Other items of interest to the reader (not central) - Telephone loading. A loaded line is far more sensitive to wet-weather leakage than an unloaded one. In Steinmetz's example 0.11% of the wave arrives on the loaded line against 0.35% on the unloaded one [read: p.567-568; calc: the attenuation constants 0.00341 and 0.00282 per mile reproduce]. - Upper frequency limit of condenser discharge. About 2 × 10⁹ Hz with a wavelength of about 6 cm. Light "would need spheres ... approaching molecular dimensions" [read: p.110-111]. - Iron at extreme frequency. "almost as good as copper" at very high frequency, and better per unit cost; iron conductors suppress stationary oscillations [read: p.478-479, 676]. - Mutual inductance between outside conductors. In a flat three-phase layout, "a power transfer by mutual inductance occurs between the outside conductors" (C. M. Davis, 1911) [read: Discharges p.170]. - Viscous hysteresis. Steinmetz rejects "viscous hysteresis", citing magnetic tests at 100-200 kHz [read: Transient p.398].

4. Accuracy spot checks (against the extracted text)

example.py runs cleanly (exit 0). Its energy ledger closes (error 1.2 × 10⁻¹⁴ W, then −3 × 10⁻⁸ mJ), and the table matches primer section 5.

5. Retention check

Nothing pertinent was dropped from the arbiter or the primer because a reviewer judged it negatively. In particular these are all kept beside the verdicts: the Contrarian's case for the cloud premise and the magnitude agreement with modern estimates of intracloud potential, the 1922 lightning generator, Dollard and "Steinmetz Decoded" (marked as interested or secondary), the published winding-resonance literature, and the 1912 address.

Losses through condensation, not through negative judgement, to restore in the arbiter's "Investigator's finds kept" list and in primer section 6 or 9: - Commonwealth Edison 1919 trouble report. Alternator groups oscillated against each other, with the lag of field flux treated as damping. It bears directly on the open question of whether the 1914 "lag" wording came from machine hunting. It is in mathematician.md and contrarian.md but absent from the arbiter and the primer. - Slepian and Peters, "Surges on Transmission Lines", 1923. This is Westinghouse's contemporary treatment, presented in the same AIEE session; Westinghouse was a competitor, so it is an interested party. It is absent from the arbiter and the primer. - Creighton's arcing-ground suppressors (J. AIEE 1923). These are reduced to "1923 J. AIEE discussion" in the arbiter and are absent from the primer.

The test records in the folder listed in section 3 are not a retention defect, since no one judged them. They are missed material.

6. Process

• Source tags. Tags are present and honest throughout. Two problems:
• Primer section 3.4 tags "ambient fields ... of order 100-200 kV/m" as [web], citing arXiv 0906.0429 and physicist.md. The Physicist took the figure from an informal search summary, and the Contrarian says the 100-200 kV/m figure "was not confirmed". Retag the figure as [memory: medium] or give a confirmed source. (Balloon soundings of roughly 100-400 kV/m maxima [memory: medium] would support the order of magnitude.)
• facts.json and claims.md still carry the pre-review statements on claim 2 ("does not use the words negative resistance") and on claim 4 ("asserted, no measurement cited"). The corrections listed in the arbiter have not yet been applied to these source-of-truth files.
• Web checks. Done for the load-bearing facts: Peek's 10⁻⁵ factor, the dipole formula, Maxwell-Wagner-Sillars polarisation, intracloud potentials and winding resonances (the last from search snippets only, which the arbiter says).
• Verdict scale. Applied consistently, and split verdicts are clearly labelled. One wording issue: on claim 7 the arbiter writes "correct for its idealisation; inapplicable to real conductors: Refuted by calculation". The scale name is right, but the evidence table in the primer should keep the qualifier "for real wires".
• Primer against STEP 4 and the reader profile.
• The nine sections are in the required order, the answers are at the end, symbols are defined, and the worked example has real numbers and a script.
• The "Being fair to the starting points" paragraph is good practice.
• The primer is about 39 KB, long but justified.
• example.py runs.

7. Corrections to earlier documents

1. arbiter.md claim 4 ("Steinmetz cites no measurement in the 1914/1920 pages") and arbiter correction 2 ("the danger frequencies carry no measurement in 1914/1917/1920 pages"): wrong for the 1920 book. - Section V says "experimental investigations have shown, that the natural frequency of oscillation of the high voltage windings of large power transformers usually is within this range of danger frequencies", and attributes "not infrequent disastrous experience" to it [read: Transient PDF pp.674, 676]. - It also gives a derivation: the line-oscillation duration in cycles peaks between 20 and 100 kHz. - The same error is in claims.md claim 4 ("asserted, no measurement cited"), in facts.json (Transient key claims and summary), in the primer evidence table row 4b ("No measurement cited") and in all three reviews. - The verdict "Plausible, untested" can stand, because there is no neutral, controlled test of his numbers on record. The evidence must now read: "claimed as experimentally observed in 1920, reference not given; single source in the folder".
2. arbiter.md claim 3, the Contrarian's point that the √(z2/z1) rule is the gradual-transition result: the 1920 book itself gives the abrupt-junction transmission factor 2c1/(c1+c2), with a maximum of twice the incoming voltage at an open end [read: p.633-634]. The criticism applies to the 1914 Discharges lecture only. Say so.
3. arbiter.md correction 8 and primer, the constant-parameter "tension": Steinmetz states the limitation himself in the 3rd-edition preface and replaces constant constants with frequency-dependent r, L, g and C in Section V [read: p.15, pp.652-661]. Describe it as the inventor's own stated scope, not as a hidden assumption.
4. arbiter.md claim 2 and primer section 4 row "How the arc feeds it": the falling characteristic e = e0 + c/√i is already in the 1916 Alternating-Current book in the folder [read: AC p.383-385]. It is used there for wave distortion, not yet as the energy source of an oscillation. Add this to the 1914 → 1916 → 1917 → 1923 timeline.
5. claims.md source-gap paragraph, arbiter and primer source gap ("no measurement by Steinmetz beyond oscillograms described in the text"): incomplete. Add: - the hot-wire ammeter reading of 29 A against 28.88 A calculated [read: Transient p.123]; - the rectifier test curves [read: p.306-307]; - the 10,000 V, 40-cycle experimental line [read: p.157]; - the observed synchronous-motor curves [read: AC p.355-356].

None is a test of a contested claim. They show that he did measure and compare. 6. facts.json (Transient summary): "Section V rest[s] on digests only" is accurate, but its key_claims miss the wave-front flattening law and the Section V danger frequencies. Add them after deep read 1.

8. Remediation actions

Deep reads (at most 8, most important first; see the structured list): 1. Transient PDF pp.652-703, Section V (Opus). Claims 3, 4, 7. 2. Transient PDF pp.602-648, Section IV ch. VIII-XI (Opus). Claim 3. 3. Transient PDF pp.431-482, Section III ch. VIII-IX (Sonnet). Claim 7. 4. Transient PDF pp.142-158, Section I ch. VIII, low-frequency surges (Sonnet). Claim 4 and the experimental record. 5. Transient PDF pp.104-124, Section I ch. VI, oscillating currents and generator (Sonnet). Claim 1 and the experimental record. 6. Alternating-Current PDF pp.379-386 and 402-404, the alternating arc and resonance by harmonics (Sonnet). Claims 2 and 4. 7. Transient PDF pp.292-316, the mercury-arc rectifier with tests and oscillograms (Sonnet). Arc behaviour, claims 1 and 2.

What each file must change (MINOR route) - facts.json: - Transient entry: add the Section V claims (danger frequencies of transmission systems 20-100 kHz with the "experimental investigations" quote; wave-front gradient roughly E/√distance) and the transition-point factor 2c1/(c1+c2) with its maximum of 2. - Add the test records in the folder to the Transient summary. - Update how_read once the deep reads are done. - Alternating-Current entry: add the arc characteristic e = e0 + c/√i (p.383-385). - Apply arbiter corrections 1-10, which have not yet been applied. - claims.md: - Claim 2: add the 1916 arc characteristic and the 1917/1923 negative-resistance account (arbiter correction 1). - Claim 4: replace "asserted, no measurement cited" with the 1920 quote and its single-source status, and add the flattening law as a sub-claim (or a new claim 8, "Variation of circuit constants: steep fronts flatten roughly as 1/√distance; danger frequencies of systems 20-100 kHz"). - Claim 3: note the abrupt-junction factor in the 1920 book. - Revise the source-gap paragraph (correction 5). - arbiter.md: add a "QE revision" section that - restates the claim 4 evidence and keeps or justifies "Plausible, untested"; - reconsiders claim 3, "energy transfer: Not testable as stated", in the light of the compound-circuit power-transfer prediction. A split is likely: uniform line = vocabulary; compound circuit = testable by exact calculation or measurement of the inter-section power flow; u0 = Σλu/Λ = an approximation Steinmetz flagged himself; - gives a verdict on the flattening law. It is testable against later surge-propagation measurements [memory: medium that such data exist (corona-damped surge-front distortion)], to be checked on the web; - adds §69's 630 r0 figure to claim 7 as further support for "Refuted by calculation", with the internal 3.4× inconsistency; - adds the Commonwealth Edison report, Slepian and Peters (interested party: Westinghouse) and Creighton to the Investigator's finds kept; - adds the 1916 arc characteristic to the claim 2 timeline; - records corrections 1-5 above under "Corrections to earlier documents". - primer.md: - Section 4 table: add the 1916 row. - Section 6, row 4b: replace "No measurement cited" with the 1920 quote and its status. Add a row for the flattening law and for transition-point doubling. - Section 7 "What is genuinely open": add "What experimental investigations stood behind the 1920 statement on winding frequencies?" and "Does the 1/√distance front-flattening law match later measurements?" - Section 3.4: fix the field-strength tag. - Section 5 step 1: either save the 1917 page image (archive.org PDF p.217) to research/files and cite it, or state that e = 35 + 200/√i was read by the primer author from an image not kept in the project. - Add the Commonwealth Edison report to further reading or to the open question on hunting. - example.py: no change required. Optionally, add a line that reproduces Steinmetz's 1916 hard-carbon arc (e = 30 + 40/√i) beside the 1917 arc. - Recheck (Sonnet): confirm items 1-5 of section 7 are applied in facts.json, claims.md, arbiter.md and primer.md. Confirm that the deep-read files exist for items 1-3 at minimum. Re-run example.py.

Recheck

Recheck agent (STEP 4b), 2026-10-08. Read: qe.md, facts.json, claims.md, arbiter.md ("QE revision" section, lines 125-245), primer.md, deep/ (Transient headings at lines 631, 711, 772, 1087, 1269), example.py. example.py was run with the project .venv/bin/python.

Residual notes (not defects): - arbiter.md line 115 (the original "Corrections to earlier documents" list, item 2) still carries the old "carry no measurement in 1914/1917/1920 pages" wording. The QE revision's corrections 11 supersedes it, and the revision section says so. Cosmetic; a "see QE revision" tag on that line would remove any chance of misreading. - The reviewer files (physicist.md, mathematician.md, contrarian.md) keep their original wording by design; the arbiter records the correction.

Final: PASS