Flynn-C: Arbiter (Step 3)

Project: US 6,246,561 B1, C. J. Flynn, "Methods for controlling the path of magnetic flux from a permanent magnet" (filed 31 Jul 1998, granted 12 Jun 2001).

Source gap (repeated from Step 1)

The only document is the patent. The PDF has no text layer; the description (pp.34-45) and claims (pp.45-49) were read by the earlier agents from OCR of rendered page images. The 32 drawing sheets (pp.2-33) were largely unstudied when this was first written; the QE deep reads later viewed pp.3, 7, 8, 9, 10, 11 and 12. The patent has no measurements or test data. No test reports, motor data or later publications are in the folder. Therefore every statement below about whether Flynn-type devices work in practice, or were or were not independently tested, rests on [memory] or on the few web checks listed at the end, not on the folder. Where a conclusion depends on [memory] alone, it is marked.

One consequence of the gap (now resolved by the QE deep reads): the patent's "twice the energy" comparison (FIG. 13A vs FIG. 14) depends on the circuit topologies in the drawings, which were not examined at first. See the QE revision.

Reconciling the reviewers

Factual points where reviewers differed or Step 1 needed correcting:

1. Drawing sheets: facts.json says 32, claims.md says 31. [Superseded by the QE revision below: the front page says 32 sheets, pp.2-33; use 32.]
2. "Unexpected" quarter (claim 3). All three agree: per armature the force is 1/4, the total over both armatures is 1/2. I recomputed: B = 1 T, face area 1 cm^2: one face B^2 A/(2 mu0) = 1 x 1e-4 / (2 x 1.2566e-6) = 39.8 N; two faces (one armature) 79.6 N; at B = 0.5 T each armature is 19.9 N (a quarter), both together 39.8 N (a half). The contrarian's figures (A = 1e-3 m^2: 398 N, 99.5 N, 199 N) also check out. The three sets of numbers differ only in area chosen; all are right. The contrarian also notes the patent's sentence "reduces by one-fourth" literally means a 25 percent reduction while the formula gives "to one-fourth". This is a wording slip [read via OCR, per the contrarian]; I treat the formula as authoritative.
3. Flux gain table (mathematician). I recomputed gain = (R + 2 R_m)/(R + R_m) with R_m/R = 0.01, 1, 10, 100: 1.010, 1.500, 1.909, 1.990. Correct. The patent's "flux doubles, force x4" is the limit R_m >> R (magnet with large internal reluctance feeding a low-reluctance closed iron path). It is a hidden assumption, not a general result. The contrarian's "load line" point is the same fact seen from the magnet side: opening a second path moves the operating point, so total flux is not constant unless R_m dominates. Physicist treated constant flux as "a good first approximation only", which is consistent.
4. "Twice the energy" (claim 4). [Partly withdrawn by the QE revision below: the 2x MMF / 4x I^2R idea is an error.] Physicist and mathematician say coil I^2R loss at fixed turns could fall by 4x when MMF halves, so "twice" is a different, undefined metric; the patent also says "one half the power" (p.37) in a separate place. Contrarian says the energy argument is "not internally consistent". I judge: not self-contradictory, but underived and mixing two metrics (MMF, energy, power). "Underived and metric-undefined" is the accurate label, stronger than "fine" and weaker than "inconsistent".
5. A point none of the reviewers made, from the same ideal model (my own algebra, lossless linear iron, no leakage). With the opposing coil on path 1, path 2 carries Phi_2 = F_m/(R + R_m) and the blocking coil needs MMF F_1 = Phi_2 R. A plain electromagnet on path 2 that produces the same flux Phi_2 also needs MMF Phi_2 R. So in this simple model the blocking coil's MMF is not smaller than a coil-only device's MMF for the same path-2 force. The saving the patent claims must therefore come from how FIG. 14 is arranged (for example an electromagnet that has to build flux in both paths or in a longer circuit), which is in the unread drawings. This is a flag for the reader, not a verdict-changer: it makes the weak form of claim 4 less certain than the physicist's "physically reasonable" implies.
6. Web status facts (all three): current assignee QM Power Inc, patent expired about 31 Jul 2018, no test data in the document [web: https://patents.google.com/patent/US6246561B1/en, fetched by the physicist and contrarian; I did not refetch]. The 20-year term from filing gives 2018, consistent.
7. Over-unity: all three agree the patent text does not claim output above input [read]. Any over-unity use is from outside.

Contrarian's categories (kept visible)

Nothing was tested and failed on this record, and no source shows anything being dismissed without testing or blocked. "No test on record" means "none in the folder and none found", not "none exist": my own search for independent tests failed (budget exhausted, see below). The contrarian found no suppression story and I see none here; this is an ordinary patent with ordinary claims.

Per-claim arbitration

Claim 1. Pulsed coils on two paths switch essentially all of the magnet's flux between them.

• Agreement: all three say the mathematics is consistent and the principle is ordinary magnetic-circuit engineering; nothing in mainstream physics changes.
• Disagreement: physicist "Established"; mathematician "Established in mainstream form, specific device Plausible, untested"; contrarian split with the strong form "Never properly tested".
• Stronger argument: the split form. "Substantially all" needs the coil MMF to match the path MMF drop (residual flux = error/R), and leakage and saturation prevent exactly 100 percent. The principle is Established; the patent's quantitative claim for its own devices has no data.
• Settles it: build or simulate (finite-element magnetostatics) one two-path circuit with a measured magnet B-H curve and measure flux in each leg versus coil current.
• Final verdict: Established (principle, as a magnetic-circuit effect). Specific patent devices: Never properly tested on this record (not dismissed).

Claim 2. A small excitation controls a much larger magnet flux.

• Agreement: no conservation-of-energy issue; steering flux is not power gain.
• Disagreement: physicist says "Established (weak) / Plausible, untested (strong)" and notes the patent's own figure is a factor of two ("one half the power"), so "much smaller" is not supported; mathematician "Established with R_m >> R"; contrarian "Plausible, untested as a ratio".
• Stronger argument: physicist and mathematician together. In the ideal model F_1/F_m = R/(R + R_m), small only when R << R_m, but F_1 is also comparable to what a coil-only device needs (item 5 above). So "small relative to the magnet's MMF" is true, while "small relative to an electromagnet doing the same job" is the factor-of-two claim, undemonstrated.
• Settles it: measure coil ampere-turns needed to block one leg versus ampere-turns an electromagnet needs for the same leg flux, same iron and gap.
• Final verdict: Split. Qualitative (a pulse can redirect or release a large magnet force near a balance point, as in latching devices): Established. Quantitative ("much larger", with the patent's own factor of 2): Not testable as stated until a ratio and definition are fixed.

Claim 3. Splitting flux over two equal paths cuts the force on each armature to one quarter.

• Agreement: arithmetic correct; "unexpected" is rhetoric; total force halves; follows in one line from F proportional to B^2.
• Additional point (contrarian, confirmed by my gain table): holds only for constant total flux.
• Final verdict: Established (per armature, constant-flux idealisation). Description as "unexpected" is misleading. Real ratio departs from 1/4 when R_m is not dominant.
• Settles it: not in dispute; a single Gaussmeter and load cell on one real magnet would show the departure.

Claim 4. More force and work per unit of electrical input than an electromagnet-only design (twice the coil energy).

• Agreement: force at one stroke position is not work over a cycle; a conservative magnet adds no net energy per closed cycle; patent does not claim over-unity; no test in the record.
• Disagreement: physicist and mathematician say weak form Plausible, untested; contrarian says weak form Established [memory medium-high]; contrarian also calls the argument "not internally consistent", physicist/mathematician say underived/ill-defined.
• Stronger argument: physicist/mathematician. The general fact that polarised and hybrid actuators need less coil MMF than coil-only ones for a given holding force is supported by [web: https://en.wikipedia.org/wiki/Latching_relay]: "a permanent magnet that produces part of the force required... the coil supplies sufficient force... by aiding or opposing the field of the permanent magnet". That supports the general bias effect (and [web: https://en.wikipedia.org/wiki/Stepper_motor] describes hybrid steppers combining permanent magnet and variable reluctance "to maximize power in a small size"). It does not support the patent's specific "twice" figure, and item 5 shows the saving depends on the comparison topology. So Established for the general idea, untested for this patent's number.
• Settles it: a cycle energy balance: measure electrical energy in and mechanical work out over a full close-and-reopen stroke for the PM-switched device and for a coil-only device of equal force and stroke, including holding-phase coil power. A FIG. 13A vs FIG. 14 circuit calculation from the drawings would settle the "twice" figure on paper.
• Final verdict: Split. (i) General bias effect, less coil effort than coil-only for the same force: Established in general; for the patent's "twice" figure Plausible, untested. (ii) "More work per unit input energy" as written: Not testable as stated (no stroke, cycle or loss definition). (iii) Net energy gain over a cycle (not claimed): Refuted by calculation: for a conservative source, work out over a closed cycle = electrical energy in minus losses.

Claim 5. Motors built on this are more efficient and have better torque than many conventional motors.

• Agreement: no equations, numbers or baseline; nothing in mainstream physics at stake unless efficiency above 100 percent is meant, which it is not.
• Disagreement: physicist "Never properly tested"; mathematician "Not testable as stated"; contrarian "Not testable as stated, narrow form Never properly tested". These are compatible.
• Final verdict: Not testable as stated ("many currently used motors"). The narrow form (a named built motor versus a named conventional motor at stated load) is Never properly tested on this record; no dismissal is documented. Conclusion rests on absence of data in the folder and a failed search; it is not a finding that the motor fails.
• Settles it: dynamometer efficiency map of a built motor against a commercial motor of equal rating.

Claim 6. Fixed armatures plus secondary coils turn static flux into time-varying flux (transformer or inverter-like).

• Agreement: Faraday's law gives EMF = N dPhi/dt; the control coils are the real primary; the magnet only biases the flux; power out cannot exceed control-circuit power in minus losses. The mathematician's flux swing check (secondary flux Phi_m, twice the control coil's added flux Phi_m/2) matches the patent's "factor of twice".
• Disagreement: only wording of verdicts (physicist "Established" for mechanism; mathematician and contrarian "Plausible, untested" for a working converter).
• Final verdict: Split. Mechanism (flux modulation induces EMF; magnet-biased switching converter): Established [web support for flux-switching machines is weak, see below; Faraday induction itself is textbook]. Practical efficiency or advantage over an ordinary transformer or inverter: Never properly tested. The reading "the magnet's static flux supplies the primary power" (patent p.44 wording invites it): Refuted by calculation.

Summary table

Corrected claim list

1. A coil that opposes the magnet's MMF in one leg of a two-leg circuit can drive that leg's flux to about zero and push the flux into the other leg; in the ideal limit R_m >> R the other leg's flux doubles, otherwise it rises by (R + 2R_m)/(R + R_m) (1.5 at R_m = R). "Substantially all" needs a balanced coil and low leakage.
2. The blocking coil's required MMF is F_m R/(R + R_m), small compared with the magnet's MMF only if R << R_m; in the idealised model it is not smaller than the MMF a coil-only device needs for the same leg flux. Whether the patent's electromagnet comparison is fair depends on drawings not reviewed.
3. For constant total flux, splitting over two equal paths cuts force per armature to one quarter and total force to one half. Constant flux holds only if R_m >> R.
4. A permanent-magnet bias helps hybrid actuators in general (web-supported), but for the patent's FIG. 13A versus FIG. 14 comparison the coil MMF and steady I^2R are identical (ratio 1, any leg reluctances). The "twice" figure is reproduced only as the field-building energy (integral of i d(flux linkage)) with equal closed legs, ratio (R + 2R_m)/(R + R_m), 1.99 in the example; that energy is stored and returned, so "more work per unit input electrical energy" does not follow. No net energy gain is claimed or possible.
5. The patent asserts, without numbers, that flux-switching motors beat "many currently used motors" (p.34 only); no torque, speed, power or efficiency value appears anywhere on pp.40-45, and none of the 42 claims mentions efficiency [read: Round 2 deep reads]. No data exist in the folder.
6. With fixed armatures, the control coils act as the primary and the magnet as a DC bias; the secondary EMF is N dPhi/dt, and output power is paid for by the control circuit, not the magnet. The patent's own sentence "static flux ... is useless" with continuous paths and its "factor of twice" (p.44) describe flux swing in the secondary pole pieces, not energy [read: p.44, Round 2].

7. (Added by QE.) The patent says a permanent magnet is "near saturation B_r", can be raised only to a magnet B_max, and that series magnets double H but not B and excess coil flux needs a bypass (the symbols are B_r and B_max, not B_sat) [read: page image p.36]. In the linear recoil model (mu_r about 1.05) a magnet is an air-like reluctance in series with an MMF source and has no B ceiling; the real limit is iron saturation. The bypass advice is nevertheless sound, because the magnet is a poor path for coil flux.

8. (Added in Round 2.) The 42 claims recite structure, timing and drive modes only: 13 independent claims (1, 16, 18, 19, 20, 21, 22, 26, 29, 30, 31, 32, 38); no claim states efficiency, energy gain, output above input or "twice" [read: pp.45-49]. "Twice" and "increased efficiency" appear only in the description. The patent protects three drive modes separately (opposing-only, claim 16; aiding on one coil with opposing on the other, claim 18; simultaneous push-pull, claim 19). The motor drives (FIGS. 40-44: LED and phototransistor, Hall switch, brush commutator, MOSFET switching) are ordinary position-synchronised commutation, which the patent calls "circuitry common in the art" [read: p.41, p.43]; what is unusual is the magnetic circuit being switched.

What would settle it

• Cycle energy balance (electrical in, mechanical out, holding power included) for a PM-switched actuator versus a coil-only actuator of equal force and stroke. Settles claims 4 and the energy side of 2.
• (Done in the QE deep read from FIG. 7-8, 13A and 14; see the QE revision.) A bench measurement of coil current, flux and stored energy for one-leg versus two-leg devices would still be needed to confirm it in hardware.
• Finite-element or bench measurement of leg fluxes versus control current for one built circuit. Settles claims 1, 2, 3 quantitatively.
• Dynamometer efficiency map of a built flux-switching motor against a commercial motor. Settles claim 5.
• Secondary power versus control-coil power in a built fixed-armature converter. Settles claim 6.

Open questions

• (Answered in the QE revision: "twice" is the field-building energy, not MMF or I^2R.) Does the p.37 statement "one half the power" mean power or energy? The patent uses both words.
• Did Magnetic Revolutions LLC or QM Power (current assignee) ever measure and publish efficiency of these devices? Not found; search failed.
• How large is the real gap between the ideal "flux doubles, force x4" and a real device with armature gaps, leakage and saturation?
• How much holding-phase coil power does the blocking coil cost, and does it erase the claimed saving?
• Does the PM load-line shift in a real device make the "one quarter" figure noticeably different?

Web verification

Budget about 8 calls; the shared search quota was already exhausted (WebSearch refused all three attempts), so only WebFetch was usable. - [web: https://en.wikipedia.org/wiki/Latching_relay] confirms the permanent-magnet latching relay principle (coil aids or opposes the magnet's field). Retags the physicist's and contrarian's medium-confidence [memory] on polarised/latching devices and the bias effect, in the general sense only. - [web: https://en.wikipedia.org/wiki/Stepper_motor] confirms hybrid synchronous steppers combine permanent-magnet and variable-reluctance designs, supporting the hybrid-machine point at a general level (it gives no efficiency comparison). - [web: https://en.wikipedia.org/wiki/Permanent_magnet_synchronous_generator] mentions "flux switching alternator" only in a sidebar with no detail. The "1955 Rauch-Johnson flux-switch alternator" and "flux-switch alternator literature dates from the 1950s" claims remain [memory: medium, unverified]; no verdict depends on them. - The permanent-magnet load-line effect (operating point shifts when the circuit's reluctance changes) remains [memory: high] and is consistent with the standard magnetic-circuit algebra I recomputed; no web page was fetched for it. - Independent tests of Flynn, Magnetic Revolutions or QM Power devices: check failed (search quota exhausted). Statements about their absence or presence are unverified; verdicts that say "never properly tested" mean "no test on this record". - The Google Patents page [web: https://patents.google.com/patent/US6246561B1/en] (assignee QM Power Inc, expired, no test data) was fetched by two reviewers; I rely on their report.

QE revision (after qe.md, verdict MINOR)

Date: 2026-10-03. New evidence comes from two Opus deep reads in deep/ (pp.36-39 with page images p.3, 7, 8, 9, 10, 11, 12) and the QE's own checks of the extracted text. All numbers are idealised linear magnetic-circuit calculations [calc], not experiments. Web checks were still impossible (search quota exhausted), so no verdict here depends on a [web] source that was not already fetched. Remediation items 3 (claims 1-42) and 4 (motor embodiments pp.40-45) were not available as deep reads at this point; they are reported in Round 2 below.

Withdrawn corrections and errors

1. Drawing sheets: the front page reads "42 Claims, 32 Drawing Sheets" [read: p.1]; sheets run pp.2-33. My earlier "use 31" is withdrawn. facts.json was right.
2. Claims run pp.45-49, not pp.46-49 [read: p.45]. Digest 02 mis-paginates pp.40-47 and omits claims 1-25; do not cite digest page numbers.
3. The physicist's and mathematician's idea that an opposing-coil device needs half the MMF and a quarter of the I^2R is an error, and my "Reconciling" item 4 left it standing. The blocking MMF is F_block = F_m R_2/(R_2 + R_m) = Phi_2 R_2, identical to the MMF a coil-only device needs for the same leg flux. The deep read shows this holds for every leg reluctance (R_1 drops out), so neither the air gap 324 nor unequal legs change it. example.py prints ratio 1.
4. My earlier suggestion that the saving might lie in a "longer circuit" in FIG. 14 is removed. FIG. 14 is one U-shaped leg with coils on it (pieces 348 and 350 separated by non-magnetic spacers 352) [read: page image p.7].
5. The QE wrote "B_sat". The page image shows the patent's symbols are B_r and B_max [read: page image p.36].

New finding: claim 7 (permanent-magnet "saturation", series magnets, bypass)

• Patent text [read: page image p.36, p.39]: a PM near "saturation B_r" can only rise to B_max; two PMs in series double H but not B; coil flux through a PM "does not contribute" past B_max unless a bypass is provided; excess flux "cannot be through a permanent magnet member".
• Reviewers' question (a), math consistent? Parallel sources: yes. Series electromagnets: yes. PM in series: no. A PM in its working range is a straight line B = B_r + mu0 mu_r H with mu_r about 1.05 [memory: high], that is, an MMF source F_m in series with an air-like reluctance R_m. It has no B ceiling. Worked check [calc]: B_r = 1.30 T, 10 mm magnet, 1 cm^2, g = 1 mm: B rises 1.176 T to 1.236 T to 1.356 T to 1.774 T for 0, 500, 1500, 5000 A-t, past B_r. Two series magnets raise B by 1% (g = 0.2 mm) to 34% (g = 10 mm), so "B unchanged" is only a closed-circuit approximation.
• (b) experiments: none cited. FIG. 16C (flux against coil current, 300-3100 mA, zero crossing at point 393 near 1.0-1.1 A) has no flux scale and is not stated to be a measurement. Its shape (linear fall through zero, aiding side bending over) is what the linear model plus iron saturation predicts; it shows no magnet B_max. A two-path calculation reproduces the zero crossing at I* = 0.405 A for 500 turns and "flux doubles" to 1.98 [calc].
• (c) tested: not on this record. (d) If true, magnet physics would need a B ceiling for the recoil line; the data on every commercial magnet say otherwise.
• Why the bypass advice is still right: the magnet is about 10x the reluctance of a 1 mm gap here, so coil flux strongly prefers an iron path (Delta B 1.88 T against 0.18 T for 1,500 A-t). Beyond I*, 97.9 percent of the excess coil flux goes around the armature loop, as Flynn draws for flux 384 (FIG. 15). So "cannot go through a magnet" is about 98 percent true, for a reason the patent never gives.
• Side findings: bypasses short-circuit part of the magnet's flux when the coils are off (visible in FIGS. 11A, 11D, 12A; not discussed); the Hall switches (FIGS. 16D-E) are on/off flux-zero cutoffs, not current regulators; two-path devices keep the magnet operating point nearly fixed (1.286 to 1.273 T), avoiding the demagnetisation risk of fully blocking a single-path magnet [calc; coercivity data [memory: medium]].
• Reviewer positions: none discussed it, so there is no disagreement to reconcile. The stronger argument is the linear recoil model, which is standard magnetics [memory: high].
• Settles it: measure B-H on a magnet while driving an aiding coil through it, or the flux of a built FIG. 4 stack versus coil current (a gaussmeter and a variable supply).
• Verdict: split. As magnet physics ("saturation" cap on added flux, B unchanged in series): Refuted by calculation. The internal inconsistency is mild: Flynn treats coils as circuit elements and the PM as a flux-capped element. As design advice (give excess flux a low-reluctance bypass): Established (reluctance engineering). FIG. 16C as evidence: Never properly tested (no scale, not stated as measured, not dismissed).

Claim 4 "twice the energy": verdict changed

The earlier verdict "Plausible, untested" for the "twice" figure is replaced. The deep read of FIG. 13A, FIG. 14 and pp.37-39 gives a three-way split depending on what "electrical energy" means (the patent says "power" on p.37 and "energy" on p.39 and gives no formula) [read: p.37, p.39]:

• Where this changes the earlier text: the physicist's and mathematician's factor-of-2 MMF and factor-of-4 I^2R are wrong; the "one half the power" on p.37 is wrong if read as power; and the general bias effect (web-supported) is not evidence for the patent's number. The primer's "ratio 1" is right for MMF and power but must not say "twice" is unreproduced.
• FIG. 13A as drawn has unequal legs (armature 118 is far from the pole ends, gap 324) [read: page image p.10]. With gap 324 at 1 mm per face the energy ratio is 21, but armature 120 already gets 91 percent of its blocked-state force with the coils off (0.908), so the coil is nearly unnecessary there. The exact "twice" matches the equal-leg FIGS. 7-8, not the gapped FIG. 13A.
• What survives of a real benefit: the blocking coil sees about half the inductance, L = N^2/R_seen against N^2/R_2, so current rises faster and less reactive energy is stored per stroke [calc]. That is not less energy consumed.
• Settles it: a cycle energy balance, plus a coil-current and flux measurement of one-leg versus two-leg devices.
• Final claim 4 verdict: Split as in the table above and the updated summary table. Claim 1's "flux doubles" is confirmed in the model (1.98 at the critical current).

Other verdicts

Claims 1, 2, 3, 5 and 6 stand as written, with these notes: - Claim 2: the MMF result is now confirmed in general, so "small control" holds only against the magnet's MMF, not against a coil-only device (ratio 1). - Claim 3: the force-splitting formulas are standard Maxwell-pull arithmetic [read: page image p.36]. - Claims 5 and 6: not changed. The QE suggests the motor drive circuits (LED/phototransistor, Hall switch, brushes; pp.40-44) are ordinary brushless-PM or switched-reluctance commutation, which supports "Not testable as stated" but was not deep-read.

Round 2 (second QE pass)

New evidence: two further Opus/Sonnet deep reads appended in deep/, covering the claims 1-42 (pp.45-49) and the motor embodiments and drive circuits (pp.40-45). They were read from OCR text; no page images of FIGS. 19-50 were rendered, and no web checks were run. Statements about circuit omissions (no flyback diode, unipolar drives) rest on the text only and are not used in any verdict.

Verdict changes: none. Every verdict in the table above stands. What the new reads changed is the evidence behind them.

Corrected claim list: items 5, 6 and the new item 8 above are updated in place. Items 1-4 and 7 are unchanged.

Other Round 2 points for the reader: the patent's own drafting defects (claim 6 repeats "first pole piece"; claim 8 depends on claim 6 but refers to the second bypass; claim 26 and claim 27 contain garbled wording; claim 38's preamble and body disagree on "third path portion") do not bear on any verdict. The review's "claim 5" and "claim 6" mean claims.md claims, not patent claims 5 and 6; for example patent claim 5 recites a permanent-magnet rotor. Drive circuits in FIGS. 40-44 are unipolar, but motors 1000 and 1030 require coil polarity reversal (p.45); the drive for that is not shown [read: p.43, p.45, drawings not checked]. The p.45 sentence "the same principles ... as shown in FIG. 42" cites a figure that p.43 describes as the interrupter wheel; FIG. 49 is probably meant (possible OCR or drafting slip).

Still open

• Independent test data for Flynn, Magnetic Revolutions or QM Power devices: not checked (web quota exhausted).
• (Round 2: embodiments 1000 and 1030 and the claims 1-42 map are now covered by deep reads, from text only. Drawing sheets for FIGS. 19-50 were not rendered.) The six cited Flynn patents plus Sturman and Monroe prior art (front page p.1) are still not covered by a deep read.