RankingsPrimerArbiterPhysicistMathematicianContrarianQESites logClaimsSourcesResearch notes

Kelly-Patrick: Science Contrarian review (Step 2)

Reviewer: Science Contrarian (PhD standard; judges by experimental evidence; Mallove is a model for standards, not a persona). Tags: [read] = from the project's documents (deep reads, digests, research notes, or a page I read myself), [web] = Internet source with URL, [memory] = recalled, with confidence, [computed] = my arithmetic.

Source gap note: the project source is one 2,481-page compilation (V32.9, Feb 2018). Kelly says it is a report of what others say and "you are welcome to consider this a work of fiction" (p.11) [read]. It is the primary source for what Kelly and the texts he reproduces claim, not for any independent measurement. About 60 percent of pages were deep-read; the figure pages (circuits, graphs) were mostly not rendered. Where a verdict below rests on [memory] it is said.

What I read myself: all of deep/PJKbook-9cc3a5.md was consulted by section (Sweet, gravity, Tesla/Don Smith, Chapter 5, fuel-less engines, pulse charging, aerial/Moray, MEG and doubtful devices, water-splitting, suppression); the Investigator's notes in research/ and the Leads in sites.md; and one primary document the Investigator did not open, the NBS report on Newman (see claim 2). I did not open the Sweet video, the 1939 newsreel or any of the Investigator's PDFs beyond what its notes quote.

How I judge (the contrarian's rules, applied to both sides)

  1. A claim is "tested" only if input and output were measured together, by instruments and a method that are stated, and preferably by someone without a stake. Bulb brightness, nameplate ratings, terminal voltage after rest and "I saw it" are observations, not tests.
  2. "Dismissed without testing" and "never properly tested" are not the same as "tested and failed". I say which.
  3. Anomalous reports are kept and weighed. But the anomaly must be stated as the measured thing, not as the inventor's explanation of it. In this book the measured thing is rarely given.
  4. The same rigour applies to the critics: where a negative test is thin (a student project, a Wikipedia summary, a court's secondary report) I say so.

Corrections to Step 1 (and to earlier documents)

These are plain statements of what Step 1 got wrong or left incomplete, with evidence, for the orchestrator.

  1. claims.md claim 3 (and facts.json where it repeats it) says Kelly knows of no replication of the VTA and gives "COP about 1.6 million". Complete it: the 0.31 mW / 500 W pair is one filmed 1987 demonstration read by Bearden (10 V x 31.2 uA = 0.312 mW; 112.5 V x 4.7 A = 528.75 W) [web: https://www.aboveunity.com/thread/my-notes-on-floyd-sweet/, via sites.md]. Sweet's own 1987 letter says "one milliwatt" in and "a gain of some 200,000" at 200 W, "over 500" in recent runs [read: sites.md, from rex-sweet-sparky1987.pdf p.1]. So the 1.6 million figure is a ratio of two readings from one event, not a rate measured over a run. Also, 528.75/0.000312 is 1.69 million, not 1.61 million [computed]; Kelly's 1.612 million is 500/0.00031.
  2. claims.md claim 3 lists "Kawai COP 3.18" and "Annis and Eberly COP 200-400 for the switches" as results. Both are patentee figures. Annis/Eberly define COP as output over drive power "without the coupling circuit" [read: deep, E, p.219], so it is not a COP in the sense of claim 1. Say so in the claim text.
  3. claims.md claim 4 states Skinner's "1200 percent". Kelly reads it as 1200 times (COP 1200), because on the newsreel's 3,500-homes-per-horsepower boast a factor of 12 gives only 2.56 W per home [read: deep Chapter 4, section 14; computed 746 x 12 / 3,500]. The only primary text found says "about 1,200 per cent" as Skinner's own estimate, i.e. a factor of 12 on the usual reading, for a 1/8 hp motor [web: https://rexresearch.com/skinner/skinner.htm, via research/2026-10-08-moray-skinner.md]. Step 1 should record both readings and that COP 1200 is Kelly's choice, made to rescue the newsreel's arithmetic.
  4. claims.md claim 5 says Cramton gives "2,340 percent" without a basis. The deep read (section H) fixes the basis: Kelly's "Faraday" figure is energy-based, 2.34 Wh per litre (the thermoneutral 1.47 V figure), not the charge-based law [read: deep Chapter 10]. Reviewers must read "% of Faraday" as litres per watt-hour against 2.34 Wh/L. [computed: 360 L/h x 2.34 Wh/L = 842 W at the floor; the 36 W claim is 23 times lower].
  5. claims.md claim 6 says Moray's witnessed demonstrations were "635-675 W of lamps and appliances". Those are nameplate ratings; Hayes, the signing witness, says he had "no instrument in my laboratory able to measure" the current or voltage [read: deep Ch.7, C3 item 6]. Also "635 watts ... 0.878 of a horsepower" is internally inconsistent: 0.878 hp is 655 W [computed]. Step 1 should flag both.
  6. claims.md claim 7 gives "more than 10,000 US builders". Check the source: the deep read of the suppression essay finds "40,000 free-energy patents have been confiscated" by "reliably informed" [read: deep, E]. The 10,000 figure is not in the passages I read; I could not find it. Treat it as unsourced until located.
  7. claims.md says Kelly's source for the VTA conditioning procedure is a Sweet document. It is Kelly's reading of a YouTube video supplied by Horst Weyrich plus a lab photo [read: deep, A2-A3]; Ashley Gray's own build used unconditioned magnets and ran at about 400 Hz, not 60 Hz [read: deep, A4]. Kelly's own text therefore contains two different devices under the one name.
  8. Investigator's correction (sites.md): "Patrick J. Kelly" was a pen name; the real name per the FESIG memorial was Edward Dickson Williams; he died 11 July 2020; V32.9 is not the final edition (V34.5 June 2019; V34.8 Feb 2020) [web: https://www.youtube.com/watch?v=MnrNmV-vYO4; https://archive.org/details/pkj_free_energy]. I accept these as reported; the death date rests on a forum snippet plus the memorial video. Also the Benitez family has four GB documents, not three [web: Google Patents, via sites.md].
  9. Earlier digest error already noted by the deep reader: digest/PJKbook-9cc3a5.24.md (p.596) misreads "equipment examined and tested many times" as the germanium diode being tested [read: deep Ch.7 Corrections]. Also: the book says Moray's equipment "is said to have contained a germanium diode"; the Marshall passage presents the diode as Moray's illustration, not a confirmed component [read: deep Ch.7, C4]. And Moray's own words, "The valves are not rectifiers" (p.610), conflict with Kelly's "rectify" (project memory: claim 4 'rectifier' wording to fix) [read].
  10. Newman: sites.md describes the NBS result only through the court opinion (27 to 77 percent efficiency). I read the primary report's Statement of Findings myself: "At all conditions tested, the input power exceeded the output power. That is, the device did not deliver more energy than it used", and "in no case did the device's efficiency approach 100 percent". NBS tested for the Patent Office at the request of court orders; it was not asked about theory [read: NBSIR 86-3405 p.i, https://nvlpubs.nist.gov/nistpubs/Legacy/IR/nbsir86-3405.pdf]. I could not read the sections on how back spikes were handled (the PDF is image-only; I rendered the summary page only), so whether NBS answered Hastings' count of MHz spikes as output is [read: unknown from my pass]; the court opinion says NBS measured with resistors at the coil terminals [web via sites.md].

Investigator findings I weigh in every claim

  • Later editions exist (V34.5, V34.8); the preface caveats are unchanged [read: sites.md].
  • The Sweet numbers trace to one 1987 video, a 1987 fund-raising letter, Gray's 1994 notes and a 1988 typescript; Sweet's earlier 1986 lab-note outputs were 6-19 W then 100-110 W [web: hyiq.org, via sites.md]. Goldes funded him (stake).
  • Papp 1968: Ansley (San Jose Mercury News, 1989) and Feynman (via LASER) disagree on injuries (eight wounded vs three) and Feynman was a party to a suit that "was settled out of court, and Mr. Papp was paid something" [web: https://hoaxes.org/comments/papparticle.html, papparticle2.html]. A Medium essay says no explosives were found: untraced.
  • Moray: witness excerpts and the missing application contents are from the family's own publication; patent application 550,611 (13 July 1931) is not granted [web: rexresearch, via notes].
  • Meyer: 1996 Ohio finding of "gross and egregious fraud" is from Wikipedia (secondary); coroner ruled aneurysm; police found no foul play (Dispatch snippet) [web via sites.md].
  • Patent statuses: Annis/Eberly abandoned; MEG lapsed for an unpaid fee; Flynn lapsed for an unpaid fee. The reasons are not stated [web: Google Patents, via sites.md].
  • No independent test found for Campbell, Boyce, Kapanadze [web: Google, via research/2026-10-08-google-other.md]. Absence of evidence is recorded; it is not a refutation.

Verdict scale used

Established; Plausible, untested; Never properly tested; Tested, not supported; Refuted by calculation; Internally inconsistent; Not testable as stated; Historically superseded but instructive.

Claim 1: The "closed system" argument (COP above 1 does not break conservation, because real devices are open to an ambient or zero-point field)

a. Mathematics. Internally consistent as a definition. COP = useful output / operator's input is a valid ratio and is already above 1 for heat pumps, sailing boats and solar cells; Kelly's own examples say so (p.620) [read]. Where it fails is the next step: the ratio is applied to devices with no stated external source, and "zero-point field" is not given an energy density, a coupling or a measurable flux. Sweet's version states the logic exactly: gains are impossible "for a closed system", but the triode "provides a link between ... the Dirac Sea", so it is "open-ended" [read: deep, D, p.266]. That is a hypothesis about a source, not a calculation. Kelly's own Faraday table is consistent (the 2.34 Wh/L baseline matches the textbook thermoneutral energy to about 1 percent) [computed], which is to his credit.

b. Experiments. The weak form (an open system can have COP>1 for an operator who does not count the free input) agrees with every heat-pump and solar measurement. The strong form (a zero-point or ambient-field source delivering net useful work through the book's devices) is not supported by any measurement I can name. Where a neutral measurement exists in this book's territory it gave no excess: the NBS tests on Newman (input exceeded output at every condition) [read: NBSIR 86-3405 p.i]; Kelly's own lux test, where a Joule Thief was less efficient than direct LED drive [read: deep Ch.5, 11]; one student prototype of the MEG that gave "the same output as the input" [web: https://erepo.usm.my/handle/123456789/20255, summary via search; I did not read the thesis]. These are few and some are weak (see claim 3).

c. Tested. The general claim cannot be tested at the level of "the field"; it is tested device by device (claims 2-6). Tseung's formal statement, "X units supplied plus Y units led out", has no test data in the forum text [web: Energetic Forum thread 7530, via sites.md]. Not replicated because not independently tested.

d. If true. A usable, local, directional energy flux would have to exist that standard measurements (calorimetry of closed-loop devices, Casimir-type force measurements, gravitational and electromagnetic background) have not found; thermodynamics would need a loophole for non-equilibrium extraction; "negative energy" without a conserved-quantity bookkeeping would need new theory. At minimum, someone would have to state the energy flux in W/m^2.

Honest case for the claim: the logic is not circular. No law forbids an open system gaining energy; the entire dispute is empirical. The best framing the inventors have is Tseung's, and it is cleanly falsifiable: run a self-looped device for hours in a calorimeter and count the batteries.

Verdict: split. Weak form (COP>1 possible for open systems when the source is real and identified): Established. Strong form (ambient or zero-point energy is the source in these devices): Never properly tested at device level, and Not testable as stated while the source has no stated magnitude or coupling. Not dismissed without testing; the field has simply not been asked to produce a number.

Claim 2: Sharp pulses, resonance and a tiny drive tap the field (spark gaps, Joule Thief, Benitez, Bedini, Tesla coils, Don Smith, Kapanadze, Tseung FLEET)

a. Mathematics. Mixed. Don Smith's rule (power rises as V squared and f squared; "double both and the power goes up sixteen-fold") is relayed by Kelly [read: deep Ch.5, 8] and Mohamed's text "E = (Volts x Amperes) x (Cycles per Second) 2" has a garbled exponent [read: deep, I]. For a resonant tank, voltage multiplication by Q is real; power is not multiplied. Kelly's own "160 kW from 8,000 V and 20 A" is arithmetic (8,000 x 20); the 20 A is Smith's stated figure, not a reading [read].

Benitez GB 121,561 is the one case where an inventor's own worked example can be checked, and it fails. The patent's example: capacitor 0.222 uF charged to 1,200 V, inductance "0.0222 mH", circuit impedance 15 ohm, "72 oscillations per spark", 1,000 sparks per second, 40 A in the oscillating circuit [read: deep Ch.5, 6c]. My check: the characteristic impedance is sqrt(L/C) = sqrt(22.2e-6 / 0.222e-6) = 10 ohm; with R = 15 ohm the damping ratio is R/(2 x 10) = 0.75, Q about 0.67 [computed]. A circuit that damped cannot ring 72 times; the energy decays in a fraction of a cycle (time constant 2L/R about 3 microseconds against a period of 14 microseconds). The energy in each spark is 0.5 C V^2 = 0.16 J, so at 1,000 sparks per second the supply is 160 W [computed]. The claimed 40 A through 15 ohm would dissipate I^2 R = 24,000 W in the circuit, which is 150 times what 160 W of sparks can supply. The patent's own numbers therefore do not support its own claim that "400 watts taken, 2,000 W of continuous work available". This is an error inside the inventor's own derivation: not a mainstream-versus-inventor dispute.

Ace_Propulsion Joule Thief "COP 1.8": 12.5 mA x 2.35 V in vs 8 mA x 6.60 V out, from DC meter readings; Kelly's reader note: how the "output" was measured (average vs peak, load, waveform) is not stated [read: deep Ch.5, 12]. A pulsed 50 kHz circuit read with a DC meter can mis-state power in either direction.

b. Experiments. Consistent with established results: the large voltages (a coil's back-EMF is L di/dt, standard). Not consistent: any device that outputs more electrical energy than the sources supply over a stated time. Measured data in the book are rarely energy balances. The only ones: bulbs brightness (Giannopoulos 24 W driving two "very brightly lit" bulbs; Kelly's own caveat that bulbs light brightly at low power) [read: deep Tesla section, 8]; a nameplate comparison (758 W input to an induction plate vs 2,800 W of nameplate halogen lamps; no lamp power measured) [read: deep, 2]; battery terminal voltages (UFOpolitics: 37.00 V to 37.07 V and 34.94 V to 38.32 V, no amp-hours; Kelly's own FLEET swap, no numbers) [read: deep Tesla section, 5; Ch.6, K]. Kelly himself lists the measurement problem: "no instrument which can directly measure the flow of radiant energy"; the only reliable assessment is a timed discharge through a known load, which he never reports for any device [read: deep Ch.6, K1]. The surface-charge warning (Sucahyo) means a voltage rise can be a voltage artifact [read]. The best first-person run is Callanan's 33 Ah UPS batteries, 10 V to 14 V in about 6 h at 0.6 A with a plain-charger control, which "will take as long to charge as a normal charger"; no energy balance [read: deep Ch.6, E]. That control is useful, but it does not measure the energy stored.

Newman (listed as Hastings' COP 8): Hastings counted MHz back spikes as output; NBS measured input and output with several conventional instrument sets and found input greater than output at all conditions [read: NBSIR 86-3405 p.i; court opinion via sites.md]. This is a real controlled test that did not support the claim. The test was by a federal laboratory acting for one party in the suit (the Patent Office); Newman's side disputed the method. I could not read the report sections on the spikes.

c. Tested? Mostly no. Controlled test with reported result: Newman (NBS 1986, not supported). Kelly's own lux test (not supported). One null MEG-type test by Lawton (below, claim 3). Everything else: demonstrations, forum builds, patents. Not replicated independently under controls: Kapanadze (verbal "40 W start, 5 kW"; inventor withholds the design; anonymous translation) [read: deep Tesla section, 3]; Don Smith (160 kW stated, no measurement in the book); Campbell/Boyce/Kapanadze (no independent reports found by the Investigator) [web via sites.md]. Kelly's own repeated statement that results are "temperamental" and "many people never manage COP>1" [read: deep Ch.6] is itself a datum about replicability. For Kapanadze, the absence of documents (the inventor "refuses to share the details", the resonator "is secret until patented") is a gap to analyse, not a verdict: competing explanations are trade secrecy, commercial failure, or a result that does not hold up under measurement; the record that would decide is a recorded input/output measurement or a patent disclosure, and the two WO patents (2008/103129, /103130) [web via sites.md] are the place to look. I did not read them.

d. If true. A passive resonant coil system with no external source would break energy conservation unless it has an identified source. If the claim is "environmental static or RF energy", a field energy density can be stated and the output power bounded by collection area; the book does not do this. If true in the strong form, the energy-momentum bookkeeping of electrodynamics (Poynting's theorem) would need a new source term.

Honest case: Benitez-type charge-pumps, spark-gap and Tesla-coil circuits are dense in unusual transient behaviour, and standard measurement of short high-frequency pulses is hard; a DC voltmeter and a bulb cannot settle it. The strongest honest case is that nobody neutral has put a Tesla-coil or pulse-charging system in a calorimeter with a wideband power analyser (the sampling-wattmeter method NBS used on Newman is the right model). That is a legitimate "never properly tested". It is no argument that the claim is true.

Verdict: split. (i) Benitez GB 121,561 worked example (COP 6): Refuted by calculation, and Internally inconsistent (overdamped circuit cannot ring 72 times; 24 kW dissipation from 160 W). (ii) Newman: Tested, not supported (NBS 1986), with the caveat that the report's treatment of the spikes was not read by me. (iii) Joule Thief COP>1, Bedini, Kapanadze, Don Smith, FLEET, Tesla-coil devices: Never properly tested (no controlled energy-balance on record; some measurements mislead by instrument choice). It was not dismissed without testing; it was mostly not tested because the proponents never ran a calorimetric or wattmeter measurement and sceptics did not run one either.

Claim 3: Permanent magnets switched or "conditioned" give large gains (Flynn, Annis/Eberly, Heins, Kawai, Asaoka, Kunel, MEG, Sweet's VTA at COP 1.6 million)

a. Mathematics. Asaoka's own algebra is clean (eff = Wp/Win2 + alpha; "possible to make Wp/Win2 > 1" is a possibility, and the measured fact is a 40-fold drop in drive power for equal flux) [read: deep, G]. A permanent magnet is not an energy source but it can shift flux with small switching energy; the reduction in drive power is not the same as a COP>1 of the whole machine. Annis/Eberly define COP "without the coupling circuit" [read]. Kunel's cascade ("power multiplier of 10 for each generator, the sixth gives 1,000 MW from 1,000 W") is an extrapolation without a derivation; Kelly himself says he has "difficulty accepting this" [read: deep, F].

Sweet. The explanation quoted at p.266 is a hypothesis in words (negative energy, negative time, "more than four dimensions", "Dirac Sea"); it contains no equations that could be checked, while the paper's equation (2), E = (B x V) + (-B x -V) = 2(B x V), is the mathematical content, for a "motional E-field" whose mainstream meaning is E = v x B and does not create energy. In the quote, "unity and over-unity power gains are not possible" so long as positive energy is present in positive time is a conditional, not an argument. I leave the full maths to the mathematician. Internal numbers do not agree: output "more than 500 W", "1 kW", "5 kW"; input 0.31 mW, 1 mW, 10 mW; so gains 1.6 million, 200,000-500,000 or 500,000 [read: deep, B, C2; computed 5,000 W / 10 mW = 500,000].

b. Experiments. Consistent with magnets-switch-flux: yes, a magnetic amplifier or saturable reactor is standard. Not consistent with energy conservation: the VTA as claimed. A simple calculation: a pair of ferrite magnets of the stated size (about 150 x 100 x 25 mm = 3.75e-4 m^3) has a maximum energy product of order 25-30 kJ/m^3 for good commercial ferrite [web: https://www.magnetexpert.com/magnetic-properties-of-ferrite-magnets; via search summary 6-30 kJ/m^3], so the whole magnetic energy available to be "spent" is of order 10 J [computed; order of magnitude only, BHmax is not the exact stored energy]. A 500 W output would use that up in a few hundredths of a second, and Maurice's demonstration, Gray's 10-12 hour run, and Sweet's alleged runs are minutes to hours [read]. So if the output was real, the energy came from somewhere that is not the magnet, and the quoted input was not the whole input. The question for the contrarian is then not "is energy conserved" but "was the input fully measured". Three points in the record bear on it, and I put them here because they are the weak points of the data, in both directions. - Gray's notes: the input was 7.2 V at 143 microamps, read on what instrument at 400 Hz is not stated; V x I is apparent power; power factor is not reported; and Gray writes that "the signal diode divides the current into the circuit, and being parallel, puts a small microamp current into the power coil as well as the excitation coil" [read: deep, A4]. That is a second current path whose contribution to the input is not stated. Gray's device also is not the 60 Hz VTA. His own figures: 111 W out, 1 mW in; he says it ran 10-12 hours and was witnessed by three people (unnamed) [read]. - The 1987 video reading: "power factor measured to be 1" [web: AboveUnity notes, secondary, a builder's transcription]. For a magnetically coupled load the power factor of the output is easy to measure; the input side, 10 V at 31.2 microamps (impedance about 320 kilohm), is where an error would matter. If the input meter read only a fraction of the exciting power (for instance, because of harmonic content or a parallel path as Gray's wording suggests), the "COP" is not meaningful. This is a hypothesis, not a finding; the video was not watched in this run. - Heat: Maurice's account says the device cooled as load rose and lost weight ("about a pound"); frost on short circuit (Kelly, Chapter 10-ish passage) [read: deep, B, C1]. A 500 W output with a cold device would need the heat to come from somewhere; no temperature measurement is given. Both "support" and "against" are therefore stated: the 1987 demonstration has loads, meters and live witnesses ("120 V, five 100-watt globes, hair dryer, fan" per Maurice [read]); Maurice had no electronics training by his own account and the demonstration is a recollection [read]; the video can be rechecked frame by frame.

MEG: Kelly says on p.1145 that nobody achieved COP>1 on a replication and on p.1147 that it "has been replicated by others and verified as COP>1" [read: deep Ch.13, 9]; this is an inconsistency in the book. The only named test is Dave Lawton's: with a laminated-iron yoke (not nanocrystalline) removing the magnet made no difference [read: deep, 10]; and a Universiti Sains Malaysia prototype built from the Bearden and Naudin designs reported no COP above 1 [web: https://erepo.usm.my/handle/123456789/20255, search summary only]. Both tests are small and neither used the nanocrystalline core Bearden says is required; so "Tested, not supported" applies narrowly. The patent lapsed for an unpaid maintenance fee (2006) [web via sites.md]; that is a record fact with no stated reason (it could be cost or loss of belief in the device).

c. Tested? VTA: hyiq.org "have tried to replicate it for a number of years now, without any success" (relayed by Weyrich via Kelly) [read: deep, A1]. That is a failed replication by sympathisers; no protocol or data is given. Gray (1994) did get an output in a different build; no one else. So: no neutral test; not dismissed without testing in the strict sense, but the demonstration record is not enough to test. Sweet died July 1995 per Maurice; the record that would settle it is the 1987 video with instruments in view, Roberts' May-June 1986 lab notes, the Goldes/Aesop files, Bearden's 1987 paper and Michael Watson's coil data [read]. Flynn (US 5,455,474) lapsed 2007; Annis/Eberly application abandoned [web via sites.md]. Tseung's magnetic frame has an "independent replication" per Kelly with no name or data [read: deep, F].

d. If true. Magnets as gates for an unseen energy source would require a source with measurable power, and a mechanism not already in electromagnetism. Sweet's claim that the device can run self-powered and also be cooled by the load would need a heat flow from the environment that a thermocouple can detect.

Honest case for Sweet: the claim is not a vague "free energy" slogan; it names a number (500 W for 0.3 mW), a waveform (60 Hz), a demonstration with a Navy-style test set-up (wattmeter, power factor) and documents of lab notes with dates (19 May 1986). If a replication under measured conditions were built from the lab notes (as Gray's was) and failed in the way hyiq reports, that would be informative, and the details of those failures should be asked for. The VTA claim is testable in a day with a good power analyser.

Verdict: split. VTA at COP 1.6 million: Never properly tested (no neutral or instrumented test on record; replication attempts reported without data; the input measurement is where the claim would break). Not an unreasonable doubt but not a refutation. Energy for a 500 W run cannot be the magnet's stored energy [computed, order of magnitude], so the claim stands or falls on measurement. MEG: Tested, not supported, limited to two small tests (neither used a nanocrystalline core). Annis/Eberly COP 200-400 (switch drive only): Not testable as stated (the COP excludes the coupling circuit; no whole-machine measurement). Asaoka, Flynn, Tseung frame: Plausible, untested as whole-system COP>1; the measured effect (drive power reduction) is consistent with known magnetics.

Claim 4: Impulse, gravity and buoyancy devices give excess mechanical energy (Tseung/Campbell flywheels, Hardy jet, Milkovic, Dmitriev, Skinner 1939, Kwok "Hidro")

a. Mathematics. Gravity is conservative: over any closed cycle the net work of a static gravitational field on a body is zero. A pulsed or jerked flywheel returns to its starting state each cycle, so excess "from the gravitational field" is not available by this route; this is a calculation, not a consensus appeal [memory: high confidence, classical mechanics]. Tseung's "2mgr" excess per pulse (credited by Kelly) has no derivation in the text; the deep reader notes 2mgr equals the potential energy change of a mass lifted by one diameter, which is not the same thing as an energy gain per pulse [read: deep Ch.4, 2]. Milkovic's COP of 12 is presented as a force ratio ("12W thrust"), a lever advantage, not an energy ratio over a cycle [read: deep, 10]. Kelly's own buoyancy suggestion (the air column "does not have to push against the head of water") conflicts with hydrostatics: the pump must supply pressure equal to the head [read: deep, 12]. Skinner: Kelly's "spinning chair" analogy does not address that the weight circulates at 150-300 rpm and its kinetic energy comes from the driver, with losses supplying the output [read: deep, 14]; he gives no energy balance.

b. Experiments. A static gravitational field has been tested to great precision; nothing in those tests suggests path-dependent energy. Perpetual mechanical motion has not been observed in any controlled device. Against that: Bessler's 1712 wheel and the Skinner demonstration are old reports that remain unresolved because the device was not preserved (Bessler withheld the secret; Skinner never patented or disclosed it) [read].

c. Tested? Tseung: public demonstrations Oct 2009, "COP 3.3 (3.4 in heading)", no instrument readings reproduced [read]. Campbell: "confirmed the excess energy by switching the drive motor input to the output generator... self-powered"; replication video "750 W in, 2,340 W out", no instrument details [read]. This is cheap to test: close the loop and see whether it keeps running for an hour with a stated load; the book reports it as having happened, without a named witness. Skinner 1939: Pathé newsreel (film ID 1190.10, 31 July 1939) and a United Press story naming A. P. Michaels (letter: "a practical device") and George C. Estill (retired Florida Power and Light president; his "check was not complete enough to allow him to form a conclusion") [web: https://www.britishpathe.com/asset/68269/; https://rexresearch.com/skinner/skinner.htm, via notes]. These are 1939 expert statements by named men, one favourable, one inconclusive; no measurement is reproduced. I treat Estill's "not complete enough" as an incomplete test, not as a failed one. Kwok sells licences (stake) and the patent itself says solar supplies part of the energy [read: deep, 12]. Ribero's patent was never built, per Kelly, who also doubts it [read]. So: no neutral, instrumented test on record for any of them.

d. If true. Energy extracted from a conservative field in a cycle would require a non-conservative component in gravity or a hidden source; general relativity and all lab tests of gravity (torsion balance, lunar laser ranging) say the field is conservative at the needed precision [memory: high]. Skinner's "1200 percent" at 1/8 hp (about 93 W) would be about 1.1 kW at factor 12 and could not run a 12-foot lathe, drill press and hacksaw simultaneously unless the loads were light; the 1939 account does not state the loads' power draw [read].

Honest case: the key reports are old, were given by named engineers, and one (Michaels) wrote a favourable letter; the experimental question is easy: a belt-driven flywheel with a measured motor input and a measured generator output, run until the temperature is steady. If Campbell's loop closed, an independent person could show it in an afternoon, and no one in the record has. The book's own "gain grows with diameter, mass, rim concentration and pulse rate" is a falsifiable scaling statement.

Verdict: split. General principle (net energy from a conservative gravitational field in a cycle): Refuted by calculation (closed-cycle work is zero). Specific devices (Campbell, Tseung, Hardy, Dmitriev, Skinner, Kwok): Never properly tested (no neutral instrumented test on record; Skinner was examined by two named engineers in 1939, one inconclusive, one favourable, no figures). Milkovic "COP 12" as stated: Internally inconsistent (a force ratio presented as an energy ratio). Kelly's air-column buoyancy claim: Refuted by calculation (the pump must supply the hydrostatic head).

Claim 5: Water can be split, and engines run on it, for far less than the textbook energy (Boyce 211-1,200 percent, Meyer, Cramton 2,340 percent, Zigouras; generators on HHO)

a. Mathematics. Kelly's baseline is right: 2.34 Wh/L of HHO matches the textbook figure for splitting water (about 285.8 kJ/mol, thermoneutral 1.48 V) [computed, memory: high]. Against it, the claims: Boyce DC 998 W for 900 L/h means 3,992 J/L, equivalent to 0.70 V per cell, below the 1.229 V Gibbs minimum; Boyce pulsed 180 W for 900 L/h; Cramton 36 W for 6 L/min, 0.063 V per cell equivalent [read: deep Ch.10 section H, computed by the deep reader; I rechecked 360 L/h x 2.34 Wh/L = 842 W]. Cramton's own input is given three ways: "12 watts (1 amp at 12 volts)", "3 amps at 12 volts", "36 W = 24 W motor + 12 W rotor" [read: J]. Zigouras: the printed flow (17,500 L/min) implies about 2.5 MW at the 2.36 Wh/L floor against 140 W to 2.7 kW of input, a ratio of order 10,000 and 900 respectively, while the page says "COP 5-10" [computed: 17,500 x 2.36 x 60 = 2,478 kW]. Kelly says the COP derivation "is by no means at all clear" [read]. The "three-men" generator: 3 L/min HHO (180 L/h) carries at most about 0.43 kW of chemical energy [computed: 180 x 2.36 Wh], against a stated 5.5 kW electrical output [read: deep Ch.10 section 21]; cold water mist cannot add energy (it absorbs heat on vaporising).

b. Experiments. Faraday's laws of electrolysis are among the best-tested relations in chemistry: one electron per H atom. The cited exceeding of "Faraday" is not a violation of Faraday's charge law if the extra comes from the energy side (lower voltage); the 1.229 V figure is a thermodynamic minimum, not a law about charge, but it has no measured exception. An older claim is "monatomic" or "charged water clusters" gas worth four times hydrogen (Kelly pp.806, 821, 836) with no measurement [read]. Meyer: a 1996 Ohio court found "gross and egregious fraud" after three expert witnesses found "conventional electrolysis" [web: https://en.wikipedia.org/wiki/Water_fuel_cell, secondary; court file not reached]. Meyer had a stake in the suit as defendant; the experts were retained for a suit; both facts are stated. Meyer's death (1998): coroner ruled aneurysm; Grove City police found no evidence of foul play [web: Columbus Dispatch via sites.md, snippets]. His brother reported Meyer said "They poisoned me" [read: research note]. Competing readings stand side by side.

c. Tested? Boyce, Cramton, Zigouras, Lawton ("300-900 percent" with no named assessor): reported by designers or Kelly; no log, no method, no independent measurement in the book; the Investigator found none either [web via sites.md]. Honest reading: dismissed without testing in the sense that no one neutral has published a result; "never properly tested" in the sense that the claimed gas flows were never measured with a disclosed method. Meyer's cell: tested in court by experts and found conventional (secondary source; verdict-in-suit). Generators on water: nine claimed, none named; no measured output (the book itself says no measured electrical input or output for a running water-only generator appears on pp.847-871) [read].

d. If true. Splitting water below the thermodynamic floor would require an external energy source, not just a better electrode; the electrochemical potential of the cell is set by the free energy of the reaction, which is measured everywhere. Exceeding it with the surplus from "the environment" would need an identified source, as in claim 1.

Honest case: Kelly's calculations are right and his deep structure is sound: the Faraday table is correct; the generator claim does not require an efficient electrolyser at all. His statement that "with an electrolyser efficiency of only 50%... the gain is in running a generator as an internal combustion steam engine" is the real claim, and it is a chemical-energy claim that fails on the energy table above. The one aspect the contrarian can support: HHO additives (booster) fuel-economy anecdotes are an ordinary question of combustion efficiency and have never been controlled in this book; they are Never properly tested in the sense that Ewert, Maynard, Caggiano report mileage without protocol [read: deep, 20].

Verdict: split. Below-floor electrolysis (Boyce, Cramton, Zigouras as printed, with inputs as stated): Refuted by calculation if the stated input is the whole input; and given Cramton's three different input figures and Zigouras's printed flow vs COP, Internally inconsistent as presented. As a measurement claim: Never properly tested (no disclosed flow measurement for any figure in the range; I found no independent test). Generator on 3 L/min of HHO giving 5.5 kW: Refuted by calculation. Meyer's water fuel cell: Tested, not supported (court, expert witnesses; secondary source, party to a suit). Booster fuel-economy anecdotes: Never properly tested.

Claim 6: Aerial, passive and fuel-less devices deliver power (Moray, Tesla/Plauson aerials, Papp-style inert-gas engines, compressed-air "fuel-less" engines, Coler, pyramids, Joe Cell)

a. Mathematics. Moray's own theory contains the standard underdamped-RLC criterion R < 2 sqrt(L/C) and f = 1/(2 pi sqrt(LC)) (the text drops the brackets) [read: deep Ch.7, C7]; that is standard. The 635 W / 0.878 hp mismatch is an arithmetic slip somewhere (0.878 hp = 655 W) [computed]. Plauson: 100 kW needs "more than one hundred separate aerials" at under 1 kW each [read]. Papp: Kelly's phrase "Atomic Number below 19" for a mixture containing krypton and xenon is wrong in its own terms [read]; the claimed 300 hp "constant at 4,000 rpm" for a 90 hp engine (35 min test) and "over 100 hp" elsewhere are different engines or tests not reconciled [read].

The aerial-plate numbers can be bounded without a new theory. Fair-weather field is about 100-130 V/m and the air-earth conduction current density about 1-2 pA/m^2 [web: https://arxiv.org/pdf/0906.1280; https://space.fmi.fi/~makelaa/fairw.html]. The product, about 130 V/m x 2e-12 A/m^2, is about 3e-10 W/m^2 [computed]. A plate of 0.48 m^2 (Ascanius' 600 x 800 mm) intercepts about 1 pA, i.e. 1e-10 W; for Kljajic's 96 W the area would be about 4e11 m^2 [computed]. So the quasi-static atmospheric-electric source cannot give the reported power; any real output from such an aerial comes from other sources (local mains and broadcast fields, thunderstorm activity, earth-electrode potentials between a copper and a galvanised pipe) or is mismeasured. The book gives only voltages for Ascanius (2.6 V on an LED, about 20 V open, 30 V at a midpoint tap) and a single power figure for Kljajic (96 W, no method) [read: deep Ch.7, B]. Rayburn's "10 kW" has no current, load or method [read].

b. Experiments. Moray: 1920s-1930s demonstrations before named people (Judd, Eyring, Jensen, Hayes, Todd); Eyring "found no fault with the demonstration and the worst he could say was that it might be induction" and asked for a test 3-4 miles from power lines; Hayes states the device was taken "more than 26 miles from the nearest power line and over a hundred miles from the nearest radio station" [read: deep Ch.7, C3]. That is the right control, reported by the inventor's long-standing associate (stake); no measurement of V, I or frequency; the loads are described by their ratings (lamps and a flat iron); one bulb at about 75 percent brightness in Emigration Canyon [read]. The patent examiner demanded proof of a natural energy source ("No natural source of electric wave energy is known to the Examiner") [read: p.605, Kelly's quotation of a 1931 rejection]. The original application contents are missing from the Patent Office file, per the family's statement [web: rexresearch, via notes; read: p.605]. Competing explanations (suppression, loss of the Swedish-stone detector, induction or a battery in the box, change of interest to therapeutics) are set out in the Ch.7 deep read; the records that would decide are police or hospital records of the alleged shooting, the 1931 file, Moray's notebooks and the Hayes and Jensen original statements. "Mostly dismissed without testing" is the fair label for the patent-office stage: the examiner required proof; no test is recorded at the Patent Office. The one test that fits a modern standard (a neutral body measuring a full system) did not happen.

Papp: one documented dramatic event (the 1968 cylinder detonation in California, per Ansley and Feynman) [web: hoaxes.org]; "dozens of engineers saw the engine in closed rooms" and "certified 776 ft-lb at 726 rpm" and "a 1983 independent certification test" are in Kelly with no tester named [read: deep Ch.8, 6]. Feynman called it fake; he was a party to a suit that settled out of court with a payment to Papp [web: papparticle2.html]. A noble-gas explosion shows a rapid energy release only if energy was put in electrically; no net energy measurement is described. Papp's US 3,680,431 was granted (priority 1 Nov 1968) and US 4,428,193 (1984) [web via sites.md]; a granted patent is not a test.

Compressed-air "fuel-less" engines (Neal, Robertson, Rogers, Van Valkinburg): Kelly's argument mixes a real heat pump with an engine that needs its own compressor; no pressures, flows or powers are given [read: deep Ch.8, 0]. Rogers' patent US 4,693,669, as reproduced, is a supercharger that "supplements" tank recharge and makes no free-energy claim [read]. The Rogers report ("greater output than applied input... self-sustaining above 30 mph") is an interview remark. Clem's "350 hp measured over nine days" has no source; his death and funding timing is Kelly's insinuation without source [read].

c. Tested? Moray: witnessed demonstrations, not controlled measurements; not replicated, per Kelly. Ascanius: reported builds, voltages. Papp: demonstrated and litigated; no controlled measurement on record. For all, the neutral instrumented test is missing. Dismissed without testing at the institutional level (Moray 1931), tested only by sympathetic or party witnesses afterward.

d. If true. A self-sufficient 50 kW box without a fuel would require an energy source the whole Earth's sensors could see; the atmospheric-electric and cosmic-ray fluxes at the ground are many orders too small [computed; cosmic rays: memory, low confidence on exact flux; not stated numerically here].

Honest case for Moray: the demonstrations were repeated, in different rooms, before people with technical training, with the aerial and ground each shown to be necessary (output ceased on disconnecting either; the new ground rod made the light brighter as it was driven deeper). That behaviour pattern is what a real receiver of external electromagnetic energy does, and also what a rectifier of distant broadcast energy or a hidden source switched through the ground does; the demonstrations alone cannot separate these. The way to separate them (a spectrum analyser and a Faraday cage; an energy account over hours) was available only in later decades.

Verdict: split. Moray's device at kilowatt levels: Never properly tested (witness demonstrations, ratings not measurements; patent application dismissed without a test at the Patent Office, and the application file is missing); absence of the file is an anomaly to chase, not a verdict. Atmospheric-electric aerial plates delivering watts: Refuted by calculation for the source named (fair-weather field), leaving local-fields or earth-potential explanations Never properly tested. Papp: Never properly tested (one explosion, litigated; no net-energy measurement). Compressed-air fuel-less engines: Refuted by calculation if closed (the compressor must be driven by the engine); Plausible, untested as a modest heat-pump-assisted engine with an external heat source. Rayburn, Coler, pyramids, Joe Cell: Never properly tested; Kelly says nothing is replicated.

Claim 7: Suppression (inventors silenced, bought out or classified; "10,000 builders", "40,000 confiscated patents", Moray, Meyer, Sweet)

a. Internal consistency. The essay is assertion: "I am reliably informed that more than 40,000 free-energy patents have been confiscated", "not even Cal-Tech managed to get past that financial block", no case named [read: deep, E]. Kelly's own text undercuts the universal claim in places: the MEG patent was granted "even though it should not have been" [read], so some applications are granted; Annis/Eberly and Flynn patents lapsed or were abandoned with no reason recorded (fees are a plain alternative) [web via sites.md].

b. Evidence. Real, documented secrecy exists: US patent secrecy orders in effect numbered 5,445 (FY2013), 5,680 (FY2016) and about 5,976 (FY2021) [web: https://sgp.fas.org/news/secrecy/2016/10/103116.html; https://fas.org/publication/invention_secrecy_2011; FY2021 via a Statista citation, secondary]. Those are the whole of the national-security secrecy orders, not a count of energy patents; nothing I found links them to "free energy", and the number 40,000 is far from them. Specific records: Moray's missing application contents; the examiner's quoted rejection grounds; Papp's partner's and investors' suits; Meyer's coroner finding (aneurysm) and police finding of no foul play [read; web]; Maurice's allegations about Sweet (visitors, FBI removing equipment, his brother's death of a brain tumour, Glenn/DOE/GM) are recollections with no corroboration in the text [read: deep, C1], and one of the few the record shows is Sweet's own letter seeking financial backers, which makes funding the live stake [read]. Absence of evidence: Sweet's lab notes, Goldes's funding records, Moray's notebooks and the Papp suits have not been obtained; each is a record that would decide between suppression, commercial failure, secrecy and a failed device.

c. Tested? The general claim is not testable as stated. Sub-claims can be: each missing file can be searched (USPTO file wrappers, National Archives, court dockets). The Investigator's first run did not reach them (sites.md "not reached") [read]. So: no test and no systematic dismissal. A fair-minded statement: patents are sometimes classified; some inventors were defrauded or sued; some demonstrations failed; the record for "suppression of free-energy devices" as a pattern is mostly Kelly's assertion.

d. If true. A coordinated multi-decade suppression across countries would need a mechanism and a trail (orders, funding, personnel). The book offers neither.

Honest case: a thin record does not contradict suppression; inventors outside institutions leave thin records. The Moray patent file with empty jackets and the Papp suits are specific, checkable anomalies, and the Meyer case has two readings on the record. They need targeted record searches before any weighing. Applying the same rigour to Kelly: he is an interested party (runs the information site and an author of the claim) and the pen-name identity means his own biography is not in the review files beyond what the FESIG memorial says.

Verdict: split. General claim that inventors of working free-energy devices are suppressed: Not testable as stated. "40,000 confiscated patents": Never properly tested and unsourced (no source named). Moray's missing application file; Papp's and Meyer's litigation: Never properly tested (records not yet obtained; Meyer's death: coroner and police on one side, brother's report on the other). Real invention secrecy orders exist at the order of about 6,000, not 40,000, and are not tied to energy devices in any source I found [web, above].

Cross-cutting findings the arbiter should keep

  1. The book's numbers almost never meet the contrarian's own standard of a test. The best-measured items are not the over-unity claims: the Faraday table, Kelly's lux test (against), Callanan's charger control (mixed), Gray's 1994 tables (apparent power, instrument unstated), NBS on Newman (against), Ascanius's voltages (not power).
  2. The two cheapest decisive experiments are named by the book itself: a timed discharge through a known load (Kelly p.544) and a closed-loop Campbell/Hardy flywheel run on a stated load. Neither has a reported neutral result.
  3. Items where the inventor's own record is internally inconsistent and would not survive the inventor's own numbers: Benitez worked example; Cramton input; Zigouras flow vs COP; Sweet's input (0.31, 1, 10 mW) and output (0.5, 1, 5 kW); Moray 635 W vs 0.878 hp and aerial span 57/87/90 ft; Tseung 3.3 vs 3.4; MEG replicated and not replicated.
  4. Both-sided items for the reader: Moray's control (26 miles from power lines, per Hayes) and Hayes' own inability to measure; Newman's Hastings count vs NBS; Papp's Ansley and Feynman accounts; Meyer's court finding (secondary) and his brother's report.
  5. Records the next Investigator run should chase (from the not-reached list): NBSIR 86-3405 body (spike treatment), Moray application 550,611 file wrapper, the Roser v. Papp and Papp v. Caltech suits, 1996 Ohio judgment against Meyer, Sweet's 1987 video with meters visible, Roberts' 1986 lab notes, hyiq failed-replication data, the two Kapanadze WO patents, and the USM MEG thesis.

Web checks done (budget used: 8 searches/fetches)

  • NBS on Newman: read primary Statement of Findings (https://nvlpubs.nist.gov/nistpubs/Legacy/IR/nbsir86-3405.pdf).
  • MEG replication: search summary only, USM Malaysia prototype (https://erepo.usm.my/handle/123456789/20255); I did not open the thesis, so its details are unknown.
  • Ainslie COP 17 heater: the check failed (found only her 1999 patent application, EP 0932248; no independent test located). I did not read the book's Ainslie section (deep, 3540-3660) in detail, so no verdict on it beyond: patentee COP figures (WO 99/38247 at COP 1.086, 2.45, 0.976-1.176 per claims.md) include values below 1; these are not an over-unity result [read: claims.md].
  • Atmospheric electricity: confirmed 100-130 V/m, 1-2 pA/m^2 (arXiv 0906.1280; FMI page).
  • Secrecy orders: confirmed about 5,200-6,000 in effect (FAS).
  • Ferrite energy product: confirmed 6-30 kJ/m^3 typical (magnetexpert.com summary).
  • Not checked: Tesla US 685,957 grant date [memory: high; Kelly has two dates in the book]; Meyer court record (secondary only); cosmic-ray flux [memory: low, not used numerically].