Bedini-John: Arbiter (Step 3)

Inputs: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md. Tags: [read] folder document; [web] URL; [memory] with confidence. Numbers below were recomputed by the arbiter.

Source gap (repeated from claims.md)

The folder holds no Bedini book, no patent text, no lab notes and no measured data (no meter logs, scope traces or COP tables). The only substantial readable works are the 2008 BM3 Group build manual (volunteer-written, no results in it) and a saved web page about the science-fair motor; most other files are diagrams read as page images, several never opened. Anything below about whether the effect exists therefore rests on [web] sources (the patent page, one 2018 journal paper, an Energizer capacity figure) and on [memory] where marked, not on data in the folder. Obtain: Bedini/Bearden, Free Energy Generation: Circuits and Schematics (Cheniere Press, 2006), the full patent text, and the BM3 Yahoo group posts (largely gone).

Corrections the reviewers made to Step 1 (all accepted)

• COP is inverted in the manual: p.10 defines COP as work out / energy paid in; the p.11 table says "Input Joules / Output Joules" [read]. Read literally, a value above 1 means LESS energy out than in. The published replication below uses out/in, so the table is most likely a typo, but it is uncorrected in the manual.
• The test excludes the primary battery on purpose ("just try and forget it", p.21) and p.5 says the system boundary is moved to the charging battery [read]. Step 1 cited p.5 for the "forget it" text; it is p.20-22.
• Step 1's "Free Energy Generation (1984 book, Bedini and Bearden, 2006)" is garbled. The book is Bedini, Bearden and Bedini, 2006, Cheniere Press [web: reference [5] of the Sriphan paper, http://gratisenergi.se/mgbedini.pdf]; "1984" belongs to the N-pole motor drawing [read: page image].
• US 6,545,444 is a back-EMF recovery patent; "Radiant Energy charging" is Friedrich's annotation on the schematic, not patent text [read: page image; web below]. But the patent does use the language "open dissipative system ... not in thermodynamic equilibrium with the active vacuum" and "environmental source", and claims thermodynamic compliance [web: https://patents.google.com/patent/US6545444B2/en]. So it is not purely mundane in rhetoric, though it states a non-unity COP; the contrarian was right that the manual's rhetoric goes beyond the patent, and the physicist's "describes a conventional mechanism" is only half the picture.
• Claim 3 merged two different devices: the SSG self-charging variant ("backpopping", Friedrich schematic) and the 9 V all-North-pole motor with the 5-day run. Split below as 3a and 3b.
• A reviewer (contrarian) noted the manual's footer says version 1.00 while the file is V1.01 [read]; minor.

Web checks done (arbiter budget 8; 3 used plus a local text extraction)

1. Patent US 6,545,444 B2: title, filed 13 Mar 2001, granted 8 Apr 2003, lapsed 2015 for fees, non-unity COP stated, "open dissipative system" language [web: https://patents.google.com/patent/US6545444B2/en]. Upgrades the medium-confidence patent facts.
2. Energizer 522 9 V alkaline capacity: about 600 mAh at 25 mA continuous drain (other listings 625 mAh) [web: https://data.energizer.com/pdfs/la522_ap.pdf and https://www.newark.com/energizer/522bp/alkaline-zn-mno2-battery-9v/dp/50F120, via search summary; datasheet PDF not opened by the arbiter]. Replaces the reviewers' "roughly 0.5 Ah [memory]".
3. Sriphan et al. 2018 (J. Electrochem. Energy Convers. Storage 15(4), 041002): downloaded and text-extracted the PDF copy [web: http://gratisenergi.se/mgbedini.pdf]; Tables 1-3 read directly (see Claim 2). Resolves the contrarian's "not certain of the table labelling" partially. Not web-checked, left as [memory] high (standard textbook definitions, no verdict turns on a contested detail): Bloch wall = domain wall; zero-point energy = ground-state energy; flyback energy E = 1/2 L I^2; Yahoo Groups content removed 2019-2020. Left [memory] medium and not checked: lead-acid conditioning improving capacity over the first 5-10 cycles (the manual itself says this, p.12 [read], so the claim is [read] as the manual's statement, not independently verified); that a few hundred mA pulse cannot move domains in a ceramic magnet (reviewers' reasoning, [memory] low-medium; no verdict depends on it).

Recomputed numbers (where reviewers differed or rounded)

5-day motor [read inputs: 22 mA, 120 h, 8.9 V start, page's own 400 mAh figure]: - Charge: 0.022 A x 120 h = 2.64 Ah. With generator: 0.0132 A x 120 h = 1.584 Ah. - Energy at 9 V: 85,536 J; at 8.9 V: 84,586 J. (Physicist 85.5 kJ and mathematician 84.6 kJ both correct; they differ only in the voltage used.) - Ratio to the page's 400 mAh: 6.6 (22 mA), 3.96 (13.2 mA). Ratio to Energizer 600 mAh at 25 mA [web]: 4.4 and 2.6. The reviewers' "4-7" range stands for 22 mA; for the 13.2 mA case it is 2.6-4. - At 600 mAh, 22 mA lasts 27 h, 13.2 mA lasts 45 h, so even generously the run would be 27-45 h, not 120 h.

Sriphan et al. Table 3, original design, row 1 [web]: input 0.485 A x 39,600 s = 19,206 A s = 5.335 Ah; output 1.01 A x 25,200 s = 25,452 A s = 7.07 Ah; energy 241.6 kJ in, 321.2 kJ out; COP = 1.33; charge out/in = 1.325 (agrees with the contrarian). The paper's own rows reproduce (e.g. 270,000/138,240 = 1.95). The battery is 12 V 17 Ah sealed lead-acid; starting state of charge is not stated in the table [web: Table 1], so 7.07 Ah out of a battery that took 5.34 Ah in is possible only if it began partly charged, or if current was mis-measured.

System-level check, from the paper's own Table 2 [web]. WITHDRAWN: see QE revision below; the 12.71 W / 4.94 W figures are charge-battery means, so the 0.63 ratio is not a system efficiency. Original text, kept for the record: Table 2 labels two power figures over an average run of 19,350 s: "Running battery average power consume" 12.71 W (original), 23.68 W (replication), and "Charging battery average power consume" 4.94 W and 8.02 W. Reading "running battery" as the battery that powers the machine (the label fits; the paper never defines it in what I extracted, so this is an inference), then: original: 12.71 W x 19,350 s = 245.9 kJ drawn from the running battery; average COP 1.63 x 4.94 W = 8.05 W-equivalent delivered, so a system ratio of 8.05/12.71 = 0.63. Replication: 1.87 x 8.02 = 15.0 W against 23.68 W = 0.63. Both come out about 0.63, so IF the labelling is as read, the paper's own numbers show the whole system returned about 63% of what it drew, the usual lossy result, while the charge-battery-only "COP" read above 1. (Withdrawn claim:) this was described as the single most useful finding of the arbitration: the physicist's boundary argument (COP>1 on the charge battery is compatible with system efficiency well below 1) is supported by the one published data set, not just illustrated with invented numbers. Caveat: inferred labelling; the paper does not itself compute a system COP, and the conditions (input current logged in a different way from running power) are not guaranteed to be comparable.

Claim-by-claim arbitration

Claim 1. The SSG charges a second battery by "radiant energy", not conventional current

• Agree: all three give the weak form (flyback pulses from the coil charge the second battery through a diode; E = 1/2 L I^2 per pulse supplied by the primary battery) as Established, and the "radiant energy" term as having no measurable definition. Disagree: only on how to label the strong form. Physicist: Never properly tested plus Not testable. Mathematician: both. Contrarian: Not testable as stated.
• Stronger argument: the contrarian's. A charging channel that is not conventional current predicts nothing a terminal-pair measurement (P = v i) would show differently, so it cannot be tested; the group's own protocol (conventional meters) cannot detect it. "Never properly tested" is a correct statement of history but it is a consequence of being untestable.
• Tested / dismissed: not dismissed without testing; never given a testable definition. The patent itself (not the manual) invokes an environmental source [web], which makes the strong form a stated claim by the inventor, yet with no measurable prediction.
• Settles it: a whole-system integrated-power test (see "What would settle it").
• Final verdict: weak form Established; strong form Not testable as stated (and never tested in any neutral way).

Claim 2. Charge-battery COP above 1; cycling improves it

• Agree: the formula is inverted, the primary battery is excluded by design, the manual's meters and resting-voltage energy accounting are weak, and conditioning explains rising capacity. Disagree: the physicist labels it Refuted by calculation; the mathematician Internally inconsistent; the contrarian Never properly tested.
• Arbitration: the physicist's calculation used invented illustrative numbers (a 0.5 A primary draw), so it demonstrates compatibility, not refutation; the folder holds no measured numbers to refute. The accurate statement is that the INFERENCE (charge-battery COP above 1 shows free energy) is wrong by construction, because the system boundary drops the primary battery, and the manual's formula contradicts its own definition. That is the mathematician's verdict. The contrarian's evidence adds that the one published data set exists (Sriphan, 4 runs per design, 12 V 17 Ah lead-acid, peer reviewed) and that its charge ratios are real arithmetic (1.33 on charge, 1.325 on ampere-hours), so the reported numbers are not made up by the reviewers' critique; but the 0.63 whole-system ratio I earlier derived from its Table 2 is withdrawn (QE revision), so the paper gives no system-level figure at all. Not enough to say "tested, not supported": no matched conventional-charger control, no recorded start state of charge, no instrument-level integration.
• Contrarian's distinction kept visible: this claim is NEVER PROPERLY TESTED (a neutral controlled, instrumented, whole-system test is not on record), NOT dismissed without testing (the community and a university group did run non-neutral tests), and NOT tested-and-failed (no controlled experiment designed to falsify it). The mathematician's Internally inconsistent verdict applies to the over-unity reading of the manual's own accounting.
• Settles it: log v(t) and i(t) on both batteries, matched starting state of charge, with a plain-charger control; see below.
• Final verdict (split): as evidence of over-unity, Internally inconsistent (inverted COP, boundary excludes the primary); whole-system COP above 1, Never properly tested (the one published data set measures the charge battery only and gives no system figure); cycling improvement, Plausible, untested (ordinary conditioning, stated in the manual itself).

Claim 3a. Backpopping: the primary battery is recharged while it powers the machine

• Agree: no data exist (a single annotated schematic with "Not to be posted ... without permission" [read: page image]); partial recycling is ordinary; a net gain would violate conservation. Physicist splits Never properly tested / Plausible, untested; mathematician Established (partial) / Never properly tested (strong); contrarian Never properly tested.
• Stronger: the mathematician's labelling, since "partial recovery reduces net draw" is established circuit behaviour, whereas the claim as made ("recharges while powering everything") is the strong form. A rising terminal voltage is not state of charge.
• Note on the rising-voltage observation: the schematic's note that a capacitor "reduces the effect" is itself a testable statement, but no readings accompany it.
• Final verdict (split): partial energy recycling Established; net self-recharge (self-running) Never properly tested (not dismissed through a bad test; there is simply no measurement).

Claim 3b. The 5-day, 22 mA, 9 V motor run

• Agree: unaudited, second-hand, no end voltage, no witness, the page's own editor doubts it and suspects a slot-car motor [read]. Disagree: physicist offers "Refuted by calculation" as a conditional; mathematician and contrarian refuse it.
• Arbitration: the verdict scale's "Refuted by calculation" requires accepted laws applied to a quantified claim. Here the claim is quantified (22 mA, 120 h), and the energy budget is unambiguous: 2.64 Ah needed vs 0.4 Ah (page) or 0.6 Ah (Energizer datasheet, [web]) available, a factor 4.4-6.6. But the input is an anecdote whose measurement, duration and battery replacement are unknown, so what is refuted is the reported figure as a literal physical description, not a controlled experiment. Verdict per the contrarian's distinction: never tested, never dismissed, not tested-and-failed. The 22 mA to 13.2 mA fall when a load was added is an unexplained internal oddity pointing to a measurement artifact [read].
• Settles it: one logged 120 h run with a calibrated current integrator and battery voltage at start and end.
• Final verdict: Never properly tested; as reported the arithmetic does not close (4.4-6.6x the battery rating), so a measurement or reporting error is the likelier explanation; arbiter does not assign Refuted by calculation because the underlying measurement is not in evidence.

Claim 4. The coil pulse "opens the Bloch wall" and pumps in zero-point energy

• Agree (all three): no mathematics, no measurable prediction, and both technical terms are misapplied: a Bloch wall is a boundary between magnetic domains, costing energy; zero-point energy is the ground state and cannot be drawn down [memory: high, standard textbook]. Physicist adds Internally inconsistent for the net-extraction reading.
• Arbitration: the extraction reading conflicts with the terms' definitions, but a vague slogan is better classed as unfalsifiable than as a self-contradicting equation; there is no equation to be inconsistent with. The contrarian's refusal to rest the verdict on it is the sounder discipline.
• Final verdict: Not testable as stated.

Claim 5. Works only if built exactly as specified

• Agree: it functions as an escape clause; the manual restricts invitations to "advanced groups" to those who follow directions (p.5) [read]. Contrarian adds a contradiction: the manual says neodymium magnets are NOT suitable (p.15-16), yet the Sriphan paper used eight neodymium magnets and reports COP above 1 [web]. Either the exactness rule is not real or the two groups mean different things by COP.
• The narrow sensitivity (tuning depends on coil, timing, magnet strength) is ordinary pulse-circuit behaviour.
• Final verdict: Not testable as stated (no numerical pass criterion that survives the Claim 2 defects; failures can always be blamed on the builder); the narrow technical sensitivity is Established.

Claim 6. Collapse spikes harvested into a second battery (1984 N-pole lineage)

• Agree: standard flyback recovery. Examples: 330 uF at 200 V stores 1/2 C V^2 = 6.6 J [read: page image dualbatt.jpg values; arithmetic recomputed] (a bound on one capacitor charge, not a measured output). The 1984 drawing says "energy recovery", 25-300 mA, not gain [read: page image].
• Final verdict: Established as energy recovery; any Bedini-specific recovery efficiency above ordinary flyback circuits: Never properly tested.

Corrected claim list (what the evidence lets us say)

1. A pulsed coil with a diode and a second battery charges that battery by flyback current drawn from the primary battery (Established). A distinct "radiant" channel is undefined and cannot be tested as stated.
2. The manual's "battery COP" counts only the charge battery, is defined inconsistently, and cannot show over-unity. One 2018 journal paper reports charge-battery COP of 1.33-2.13 using the same method; its replication-design COP column does not match its own joule columns, and it gives no whole-system figure. A neutral whole-system test has not been recorded. The only primary-side numbers in the folder (a builder's 15 V x 2.2 A in vs 0.3 A charging out) point the conventional way. 3a. Partial recovery of coil energy into the supply is ordinary; a self-recharging primary battery has been asserted, not measured. 3b. A second-hand report of a 5-day, 22 mA run on one 9 V battery implies 2.64 Ah from a battery rated 0.4-0.6 Ah; undocumented.
3. "Bloch wall" and "zero-point energy" are used in ways their textbook definitions do not support; the statement makes no measurable prediction.
4. The "build exactly" condition is a reporting rule, not a physical claim, and it is contradicted by a published neodymium replication.
5. Spike recovery is standard and Established as recovery, not generation.

What would settle it

One bench test, cheap and decisive. Two identical 12 V batteries of the same age, each brought to the same state of charge (same rest time and rest voltage, or an Ah count from a CBA II discharge, which is the instrument in the folder). Log v(t) and i(t) of the primary battery and of the charging battery with a two-channel scope or logger sampled well above the pulse rate, with the meter left in place for the whole run, and integrate v x i on both. Run a matched constant-current charger control. Compare the system's total energy drawn (primary) with the energy gained by the charge battery plus the primary's recovery, over at least 20 cycles. Conventional prediction: whole-system ratio below 1 (the builder sheet SCH-illust gives about 33 W in against roughly 4 W out, an uncontrolled hint only; the withdrawn Sriphan 0.63 is not evidence). The manual's own p.23-24 charger comparison is not this test: it records only load run-time of the charge battery. A ratio above 1 beyond the instrument error, with state of charge matched, would be the first real anomaly. For 3b: one 120 h log with a coulomb counter and start and end voltages; for the single published table, the authors' raw data and the definition of "running battery".

Open questions

• (Resolved by QE: Table 2's 12.71 W and 4.94 W are charge-battery means from Table 3, not primary-battery power. Still open: what the paper's "running battery" label means, and why the replication COP column disagrees with its joule columns.)
• What was the charge battery's starting state of charge in Table 3 (7.07 Ah out vs 5.34 Ah in on a 17 Ah battery)?
• Is the manual's table formula (Input/Output) a typo, and does the online calculator it links to use the correct ratio? Not checked.
• Does the 22 mA to 13.2 mA drop under added load reflect a pulse-timing change, a meter artifact, or a different circuit? No data.
• The eight drawings previously unopened have now been read (QE revision). Still unopened: the three Trevor bench photos (no captions; no verdict depends on them).
• Is there any neutral laboratory test of the SSG? None found by any reviewer; a reviewer mentioned a 2004 television test but could not confirm it, and the arbiter did not use it.

Final verdicts (one line each)

QE revision (after qe.md, verdict MINOR)

Inputs added: deep/-5058f1.md (page images of all drawings, Self-runner notes) and deep/Bedini-Monopole-3-Group-Experiment-V1-01-c67ca7.md (manual pp.19-24). No new web checks were needed. No final verdict changes. Changes are to supporting evidence and to one withdrawn inference.

1. Withdrawn: the "whole-system ratio 0.63"

The system-level check above read Sriphan Table 2's 12.71 W and 4.94 W as "running" (primary) and "charging" battery powers. The primer and example.py show these are the means of Table 3's discharge and charge powers on the charge battery [read by QE: example.py reproduces them], so 0.63 is roughly T_out/T_in, not a system efficiency. It is withdrawn everywhere (Claim 2, Corrected claim list, What would settle it, Open questions, verdict table edited above). The paper gives no system-level figure. Claim 2's whole-system part was already "Never properly tested", so the verdict stands. I did not re-extract the paper in this revision; I rely on QE's check, which ran example.py [read: qe.md].

2. Sriphan replication table is internally inconsistent

Per QE and example.py, the replication design's COP column (2.13, 2.19, 1.73, 1.44) does not match its own joule columns (which give 2.90, 3.82, 3.35, 3.34), and its Table 2 powers do not match Table 3. The contrarian's "real arithmetic" point (1.33 on charge, 1.325 on ampere-hours) holds for the original-design rows only, which the arbiter did recompute; it must not be extended to the replication rows.

3. New primary-side numbers (Claim 2 / 3a)

• SCH-illust. 2coil 4inch (Badertscher redraw, 30 Dec 2005) [read: page image p.10, deep read]: "15V IN AT 2.2 AMP", "CHARGING 0.3 AMP OUT", 2 coils "sweet spot". QE's guess that this was a primary battery drawing 2.2 A is corrected by the deep read: the drawing shows a "PS" power supply, not a battery. Input is about 15 x 2.2 = 33 W; if the charge battery sat at 12-13 V (assumed, not on the page) output is about 4 W. No voltage, duty cycle or averaging method is recorded and the sheet makes no over-unity claim in numbers (only the redrawer's "SCALAR ENERGY" label). It is the only builder-recorded input/output pair in the folder, uncontrolled, and it points the conventional way. It is an anecdote; it does not move any verdict.
• N-Pole-Motor-6b (p.8) "CHARGING RESULT": SCR output 13.51 V, 0.8 A to 0.2 A charging; 5 x 1N4007 output 12.96 V, 1.0 A to 0.25 A [read: page image p.8, deep read]. The arrows are unexplained (input vs initial charge current unclear), so I do not interpret it beyond noting that both rows show output at one quarter of the first figure.
• North Pole Motor 2 batts (p.9): earliest drawing shows a pulse of about 7 V with a spike labelled 200 V, plus a points/bridge/330 uF path into battery 2 [read: page image p.9]. Ordinary flyback; supports Claim 6.

4. Claim 6 lineage and dates (corrected)

• SSG.jpg is "SSG North Pole Motor, 'Charging off collector'" with copper-bus output for small batteries and an SCR path for large ones [read: p.12]. This adds an SCR/collector-spike variant to the lineage; it is still flyback recovery.
• The "1979-1984 / 9-19-1984, John Bedini and Ron Cole" dates belong to the original N-pole motor, which p.6 calls "a test motor for energy recovery" (9 to 12 V). The 2005 dates (July 10, Aug 20, Sept 30, Oct 28, Dec 30) are Badertscher's redraw "copy-claim" dates, not circuit dates [read: deep read]. Verdict for Claim 6 unchanged: Established as recovery; Bedini-specific efficiency Never properly tested.

5. Claim 3a: the backpopping mechanism statement

Self-runner diagram (Rick Friedrich, May 2007) [read: page images p.1, p.13]: unnumbered box "Backpopping of primary takes place once every rotation or more to keep it in recharing [sic] mode always"; note 2 says the primary "is recharging between powering everything"; note 7 says an S2 that disconnects the primary while pulsing "improves rotation and charging significantly"; note 9 says S1 and S2 are 180 degrees out of phase. The diagram says S2 is "Not used in video", so notes 7 and 9 describe a variant not in the filmed run. The only measured numbers on the sheet are "40v max" and "over 2000v" (instrument not stated); no currents, battery voltage trend or power. The claim is asserted, not measured. A primary that is disconnected each cycle while its "recharge" is attributed to the pulses is a switched-battery arrangement; it needs a whole-system energy log like any other. Verdict unchanged: partial recycling Established; net self-recharge Never properly tested.

6. Claim 1: the 1-ohm resistor test (manual pp.19-20)

The manual says the test "helps us see that there are two kinds of energy going to the charging battery" and that conventional energy "is not enough to account for the observed charging" [read: p.20]. The procedure is a tuning check: charge battery removed, 1 ohm resistor across the terminals, pass if V is 1 V or less and the resistor stays cool [read: p.19]. The only number is that criterion. By Ohm's law it implies at most about 1 A and 1 W in the resistor (my arithmetic, assuming the meter reads an effective value). The manual never links this to the charge current. The test does not support the "second kind of energy" conclusion: (a) the output is pulsed and a DC meter reads roughly the average, so heating and peak current are not bounded except by the "not hot" check; (b) a resistor has no back-EMF and is not a 12-14 V battery, so the operating point differs; (c) the operator tunes until the reading is 1 V or less, so the bound comes from the procedure, not from a measurement of the real charging; (d) 1 A into a battery at 12-14 V is already 12-14 W. Effect on Claim 1: the strong "radiant" form stays Not testable as stated; the manual's own evidence for it is an assertion with no numbers. The weak form stays Established.

7. Claim 2: the manual already prescribes tests; they do not settle the question

• Matched-bulb power method (p.21): two bulbs matched by eye, power read as volts x amps on the reference bulb [read: p.21]. Brightness matching is a visual, nonlinear comparison; a bulb responds to heating, which tracks effective (RMS) current, not the average current that charges a battery. It replaces the battery with a bulb, so no stored energy is measured, and the manual says to "forget" the input battery.
• Conventional-charger comparison (pp.23-24): two same-batch batteries, one charged by the SSG and one by a conventional charger, compare load run-times over cycles, 8-24 h charges, SSG stopped at the conventional charger's shutoff voltage [read: pp.23-24]. So the earlier "What would settle it" and the primer's plain-charger control are not new: the community's protocol already has one. It still cannot test over-unity, because it compares chargers by battery run-time, explicitly ignores front-end energy ("not presently concerned with ... what amount of energy is consumed on the front end"), uses one battery per arm and no blinding, and never records the primary battery's energy. My "What would settle it" differs in logging v(t) and i(t) on both batteries and integrating energy over the whole system.
• COP protocol (pp.22-23): the "input joules" is the conventional power measured once at the charging battery, so COP above 1 compares battery output with the SSG's output, not total energy consumed. This agrees with the existing Claim 2 verdict (Internally inconsistent as evidence of over-unity).
• Cycling: "repeated cycles should improve the COP figure; do at least 20 cycles" attributes early low COP to sulphation [read: p.22-23]; consistent with the ordinary-conditioning reading.

8. Housekeeping

Patent number: the manual prints 6,545,444 on the p.19 figure and 6,454,444 on p.1 and p.24; the Google Patents number checked above is 6,545,444 [web, see Web checks]. The manual's footer says "Version 1.00" on a V1.01 file. The CBA manual (29 pp.) is a commercial battery-analyser manual with no Bedini content (digest confirms). The digest's "Terry Bedini" is an OCR error for John Bedini; no file used it.

Verdict changes

None. Claim 2's whole-system line loses its "about 0.63" inference and now reads "Never properly tested; no system figure exists". All other table rows are unchanged.