Flynn-C: Quality Engineering (Step 4b)

Project: US 6,246,561 B1, Charles J. Flynn, "Methods for controlling the path of magnetic flux from a permanent magnet and devices incorporating the same" (filed 31 Jul 1998, granted 12 Jun 2001). QE date: 2026-10-03. Inputs checked: _text/manifest.json, all 3 digests, facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py, and the full extracted text _text/US6246561B1-...-7ef49f.txt (49 pages, all OCR).

Verdict: MINOR. The central verdicts stand. The fixes are: correct some factual and page-number errors, take one wrong "correction" back out of the reviews and arbiter, reconcile the claim-4 ("twice the energy") verdict with the primer's own calculation, and add one physics topic that every file missed (series magnets, "saturation" of a permanent magnet, bypass paths and "excess flux").

---

1. Coverage

• All 49 pages were extracted and every chunk has a digest. No files were skipped, so no skipped file holds text that should have been read.
• Digest defect (pagination and omission). Digest 02 says it covers pp.40-47. It puts the end of the description (power conversion, embodiments 1000 and 1030) on pp.45-47. In the extracted text the description ends on p.45 and "What is claimed is:" starts on p.45 [read: p.45]. Claims 1-13 are on p.45, claims 14-21 on p.46 and claims 22-26 on p.47 [read]. So digest 02 page-shifted the description and left out claims 1-25 entirely. Digest 01 also garbles the force formula as "F = BA²/μ₀". This does little harm, because Step 1 and all three reviewers read pp.34-49 directly from OCR rather than relying on the digest. It is recorded here so that later steps do not cite digest page numbers.
• The drawing sheets (pp.2-33) have almost no text, as expected. The primer studied page images p.3, p.7, p.10 and p.12. No one studied the others. For this project that is acceptable, except for the sheets behind the topic in section 3, item 1.

2. Shortcuts

• Step 1 ran before digests existed, but it did not sample. facts.json how_read says the description and claims were read in full from OCR. The reviewers say they re-OCR'd pp.34-49 themselves. The quotations in the reviews match the extracted text (section 4). The central derivations (pp.36-39) and the power-conversion text (pp.43-44) were read in full. No central claim rests on a digest alone.
• Shortcut that matters: no one read the description of FIGS. 3-4, 11-12 and 15-16C (series and parallel sources, bypass paths, "excess flux", the critical point where flux reverses) as physics. These passages were transcribed, not reviewed. See section 3, item 1.
• The arbiter wrote its "twice the energy" discussion without seeing FIG. 14 ("drawings ... nobody examined"). The primer later looked at FIG. 14 [read: page image p.7] and found that the comparison device is one U-shaped leg with coils. That removes the arbiter's escape route (a "longer circuit"). The arbiter verdict was not updated to match (section 5).
• Web checks: the shared search quota was exhausted for all reviewers, for the arbiter and for this QE pass. My own WebSearch for QM Power test data was refused ("used its web search budget (200 of 200)"). So no source was checked on whether independent tests of Flynn/QM Power devices exist. This is disclosed honestly in every file. It still leaves the "Never properly tested" verdicts on claims 1, 5 and 6 resting on "none in the folder" only.

3. Missed material

1. Series magnets, "saturation" of a permanent magnet, and bypass/excess flux (pp.36-39). Missed by every file. This is the most important omission. The patent states [read: p.36]: - two permanent magnets (PMs) in series give "double" H but the same B as one magnet; - with a PM and an electromagnet in series, "since the permanent magnet is near flux density saturation B, the flux density can only be increased ... to B_sat of the permanent magnet", and further coil flux "does not contribute ... unless a bypass path around the permanent magnet-type member is provided".

The whole "bypass" and "excess flux" family of embodiments (FIGS. 11A-12E, 15-16; claims 12-14 and 37; the bypass members in the 740 motor) is built on this picture. In the standard model it is wrong, or at least nonstandard. A sintered NdFeB or ferrite magnet has recoil permeability μ_r ≈ 1.05 and behaves like an air gap with an MMF source in series, not like a flux limiter. A series-aiding coil drives extra flux through the magnet along its recoil line. Two series magnets in a gapped circuit also raise B rather than leaving it unchanged. This is a candidate new claim of type "Internally inconsistent" or "Refuted by calculation" (to be decided by the reviewers or arbiter), and it is a good teaching point for the primer's "where Flynn departs" table. The same physics bears on FIG. 16C (flux on the opposing side passes through zero at a "critical point", after which reverse flux appears). It also bears on Hall-switch current limiting [read: p.39], which the primer mentions only as a sketch. 2. The parallel-source statement (p.36): two PMs in parallel double B "while H remains unchanged". Not discussed. It is roughly right for a closed, low-reluctance circuit. 3. Embodiments 1000 and 1030 (pp.44-45, claims 29-31): a two-magnet bridge with coils between the magnets, and a PM-stator/PM-rotor motor. These are in the digest but absent from facts.json, the reviews and the primer. They are peripheral, but they are the configurations most often reproduced in later "Flynn parallel-path" builds [memory: low]. 4. Motor embodiments 500, 580, 650, 740 (pp.40-44) and their position-sensing drive circuits (LED/phototransistor, Hall switch, commutator brushes; p.43 [read]). Claim 5 is judged only in the abstract. No reviewer notes that these are ordinary switched-reluctance or brushless-PM commutation schemes. That point would strengthen the "nothing new to physics" reading. 5. Front-page prior art: the front page cites six earlier Flynn patents: 5,254,925; 5,304,881; 5,455,474; 5,463,263; 5,710,493; 5,753,990 [read: p.1]. facts.json and the primer list only the first three. The front page also cites Sturman (3,683,239; 3,743,898) and Monroe (3,670,189), which are latching and PM actuator prior art. That is useful evidence for the "established engineering" verdicts and was not used. 6. QM Power (current assignee, Q-Sync motor) test data: this is the most likely source of independent efficiency measurements for claims 4-5. It was never checked, because all web calls failed. It is listed as an open item, not as a defect of the reasoning.

4. Accuracy spot checks (against the extracted text)

5. Process

• Source tags: present and honest throughout. [memory] items carry confidence levels. Failed web checks are disclosed every time. Minor problem: page tags mix PDF pages and printed column numbers (contrarian, physicist "p.7"). The primer states its convention (PDF pages), which is good.
• Web checks: required, but not achievable (quota exhausted). The arbiter used WebFetch on Wikipedia pages (latching relay, stepper motor) to retag the medium-confidence bias-effect claim, which is acceptable. Still unverified: the Rauch-Johnson 1955 flux-switch alternator (no verdict depends on it) and any independent tests of Flynn or QM Power devices (the claim 5 and claim 6 verdicts lean on this).
• Verdict scale: applied consistently and with split verdicts where they fit. One inconsistency:
• (a) The arbiter's "Reconciling" item 4 repeats, without correcting it, the physicist's and mathematician's claim that I²R coil loss could fall by 4x. The mathematician claims the electromagnet "needs 2x MMF" in the ideal limit. That contradicts the arbiter's own item 5 and the primer's SymPy result: the blocking coil's MMF F₁ = F_mR/(R+R_m) equals exactly the coil-only MMF Φ₂R, for any values. I reran example.py and it prints "ratio blocking / coil-only = 1". The physicist's and mathematician's factor-of-2 (MMF) / factor-of-4 (I²R) statements are an error that the arbiter left standing.
• (b) Once FIG. 14 is seen, the "twice" figure fails within the standard linear magnetic-circuit model, and the patent offers no calculation or data of its own. The arbiter verdict "Plausible, untested" for the "twice" figure should be revisited. One option is a split: "Refuted by calculation within the linear model; real nonlinear or leaky hardware untested". The primer currently argues against its own evidence table and keeps the old verdict.
• Primer: follows STEP 4 order 1-9, with answers at the end. Symbols and acronyms are defined (MMF, EMF, OCR, NdFeB, μ_r). The worked example has real numbers. The level suits the reader profile. Errors to fix: "claims (pp.46-49)" should read pp.45-49; the Flynn prior-patent list is incomplete; the section-4 table lacks the series-magnet/bypass departure.
• example.py: runs cleanly with .venv/bin/python. Its output matches the numbers quoted in the primer (Φ₀ = 5.969×10⁻⁵ Wb, gain 1.990, 28.3 N to 112.2 N, 94.5 A-t for both coils, ratio 1, Part D table).
• claims.md: the Source-gap note is present. It says "31 drawing sheets (pp.3-33)", which should be 32 (pp.2-33).

6. Remediation actions

Deep reads (most important first): 1. Opus, pp.36-39 plus drawing page images pp.3, 8-9, 11-12: the series/parallel-source passages (FIGS. 3-4), the bypass and "excess flux" embodiments (FIGS. 11A-12E, 15-16C) and the FIG. 16C critical point. Question: what exactly does Flynn claim about permanent-magnet "saturation", series magnets and bypass paths? Is it consistent with the linear recoil-line model (μ_r ≈ 1.05)? Give a worked check of a coil in series with an NdFeB magnet. 2. Opus, pp.37-39 plus page images p.3, p.7, p.10: FIG. 13A versus FIG. 14 energy comparison. Question: with FIG. 14 now seen, can any reading of the patent's geometry (including the armature air gap 324, unequal legs, or counting both coils 122 and 124) reproduce "twice", or is it refuted within the linear model? Confirm or overturn the primer's ratio-1 result. 3. Sonnet, pp.45-49 (claims 1-42): the digest omitted claims 1-25. Question: list the independent claims and say which embodiments they protect (bypass claims 12-14 and 37, power conversion claims 13-14 and 38-42). Do any of them assert an efficiency or energy result? 4. Sonnet, pp.40-45: motor embodiments 500, 580, 650, 740, 1000, 1030 and the drive circuits. Question: are these standard switched-reluctance or brushless-PM commutation schemes? Does the text anywhere give torque, speed or efficiency values?

Other actions: - Web (when quota allows): search for published efficiency tests of QM Power Q-Sync / "parallel path" motors and of Flynn devices. Update the claim 5 and claim 6 verdicts and the primer evidence table to match. - facts.json: claims run pp.45-49; drawing sheets pp.2-33 (32, as on the front page); add the six cited Flynn patents and the Sturman/Monroe prior art; add key claims for the series-magnet/saturation/bypass assertion and for embodiments 1000 and 1030. - claims.md: fix "31 drawing sheets (pp.3-33)" to 32 (pp.2-33). Add a claim 7: "A permanent magnet limits added coil flux to its saturation value unless a bypass path is provided; two series magnets double H but not B." - arbiter.md: add a "QE revision" section that: - (i) withdraws the 31-sheet correction; - (ii) marks the physicist's and mathematician's "electromagnet needs 2x MMF / I²R 4x" as an error, citing the example.py ratio-1 result; - (iii) re-decides the claim-4 "twice" verdict now that FIG. 14 has been seen; - (iv) gives a verdict for the new claim 7. - primer.md: fix the claims page range; extend the Flynn patent list; add a row to the section-4 table and a sentence to section 7 on PM "saturation" versus the recoil line; align the evidence table with the revised arbiter verdict on "twice". - Reviews (physicist, mathematician, contrarian): no rewrite is needed (MINOR). The errors above are to be carried as corrections in the arbiter's QE revision.

---

Recheck

Date: 2026-10-03. Checked against the current facts.json, claims.md, arbiter.md, primer.md, the reviews, example.py (re-run; prints ratio 1, gain 1.990) and deep/.

Defects

Remediation actions

Open issues

1. primer.md not updated (D3, D5, D6, R9).
2. Claims deep read R3 not done.
3. Motor-embodiment deep read R4 not done.
4. Independent test data R5 not obtained (cannot be done without web quota).

Final: OPEN (4 issues)

---

Recheck 2

Date: 2026-10-03. Re-checked the four open issues from Recheck against the current primer.md, arbiter.md, facts.json, claims.md and deep/.

Final: OPEN (1 issues)