RankingsPrimerArbiterPhysicistMathematicianContrarianQEClaimsSources

Science Contrarian review: Ainslie-Rosemary-Ann

Reviewer role: Science Contrarian (PhD standard, judges by experiment, same rigour for researcher and critics). Inputs: facts.json, claims.md, and my own reading of the sources.

Source gap (repeated from Step 1). The 1999 patent body (43 scanned pages), the raw oscilloscope data, the ABB / Cape Hope / BP reports and any neutral replication are not in the folder. Conclusions about the patent examples rest on Patrick Kelly's digest (third party). Conclusions about "independent tests" rest on the author's own list of names. Where I rely on [memory] or on forum reports I say so.

What I read myself

  • electric_heater_experiment.pdf (2009 reprint), full text layer, including Table 2 and the "Results" and "Error margins" sections [read].
  • quantum_october_2002.pdf pages 2, 3 and 5 as page images (circuit diagram, parts list, measurement text, conclusions) [read: page image p.2, p.3, p.5].
  • Ainslie_heater_circuit.pdf (Kelly): grepped for test, replication and tuning statements [read].
  • magnetic_field_model.pdf: keyword passages only [read].
  • Web: forum and press reports on replication (see "Web checks").

Corrections to Step 1

  1. The 1.22 MJ figure is not an integral of power. Step 1 (facts.json and claims.md) repeats "1.22 MJ" next to "about 17.5 W average". These are inconsistent. In Table 2 the "Joules" column equals (control watts at that row) x (elapsed seconds), not a time integral. I checked every row: e.g. 29 min: 30,333 J / 1,740 s = 17.43 W (the row's own control-watts value); 960 min: 956,978 / 57,600 = 16.61 W; the final row 997 min: 1,222,429 / 59,820 = 20.44 W, which is just the last, rising reading. The paper's own stated average is 17.53 W; 17.53 W x 59,820 s = 1.05 MJ, not 1.22 MJ. The headline energy ratio is therefore about 15.5 (1.05 MJ / 67.6 kJ), not 18.1. This is the author's arithmetic slip, not a Step 1 slip, but Step 1 should have flagged it.
  2. The 0.039 A control-battery current is inferred, not measured. Quantum p.5 says the voltage reduction (12.45 V to 12.04 V open-terminal) "was consistent with the small current flow (0.039 A) required". Step 1 reports it as a measured drain. It is a stated estimate [read: page image p.5].
  3. Step 1's "about 0.49 W control battery, about 43% of 1.13 W" is right as arithmetic (12.45 x 0.039 = 0.486 W), and the corrected input is about 1.61 W, making the ratio about 10.9 on the author's own numbers (using 17.53 W). Step 1 did not state the resulting ratio.
  4. The "17 times" ratio is power-based, but the stated "coefficient in excess of 16 after discounting" is not shown. The paper gives no calculation for it. Step 1's phrase "survives discounting negative samples" is the author's assertion only.
  5. Claim 6 in claims.md mixes a self-reported refusal (Stellenbosch) with "tested or supported by ABB, BP, Sasol". The paper only says ABB "conducted independent tests" of instantaneous power and BP "applied tests to measure the effect ... on battery duration"; no results from either are given. Sasol offered a bursary (Pretoria News), which is not a test. Step 1 treats these as "supported"; they are unsupported mentions.
  6. Step 1 correctly noticed the 2009 reprint drops the separate 12 V gate-supply battery and the 143-200 kHz oscillation. I confirm both: the 2009 text says only "connected ... to a 555 switching circuit" and "random oscillation" [read]; the 2002 article names both [read: page image p.3, p.5].

Web checks (budget used: about 5 calls)

  • Replication claims exist only on hobbyist forums and in an advocacy article (OpEdNews, Steve Windisch, 2009), which says Glen Lettenmaier ("Fuzzytomcat") reproduced it in 2009 with a Tektronix scope and reported COP greater than 4; it says five companies "fully verified" the claims but names none. URL: https://www.opednews.com/articles/The-Strange-Case-of-the-Ro-by-Steve-Windisch-ji-091219-425.html [web].
  • A forum participant, Poynt99, is reported in search results as saying the paper "does not contain convincing evidence to prove any claims of overunity" and as documenting measurement artifacts; a video titled "Rosemary Ainslie replication Cop 17. NO. Cop = 1" also exists. Search result pages: https://overunity.com/13743/rosemary-ainslie-quantum-magazine-circuit-cop-17-claims/30/ and https://www.youtube.com/playlist?list=PLml9VdOeqKa_5QWbelpO5Wp8MctLt5EdQ [web, snippets only; I could not open the threads (one returned blank, one 403), so I do not know the detailed critique or numbers].
  • Check failed: I found no web source for the ABB test report, the Cape Hope Metrology statement beyond what the paper says, or the Stellenbosch refusal other than the Pretoria News clipping in the folder. Treated as unverified self-report.
  • Check failed: no peer-reviewed or neutral-laboratory replication found.

Strongest honest case for the experiment

  • The design is not silly. It is a real, measurable question: does a switched inductor dumping through a MOSFET body diode return charge to the battery, so that the battery's net current is lower than a naive V/R estimate? Back-EMF recovery is real physics (every switch-mode converter uses it); the open question is only whether the accounting is right.
  • Load heating was measured with a method that does not depend on waveform analysis: a platinum probe on the resistor wall, calibrated with DC against the same resistor, run to thermal steady state in a draught-free room [read]. That is a sound way to avoid inductive phase errors, and the author documented the reason.
  • The 2002 article names its parts with supplier codes, says the results are "easily repeatable", gives calibration certificate numbers and openly invites duplication [read: page image p.2-3]. That is the right attitude and it makes the claim falsifiable.
  • The disclosure of the separate 12 V battery in 2002 is honest, and the author tried to bound its contribution.
  • The 16.5 h run was intended to rule out a transient effect. Battery voltage stayed between 24.3 and 25.1 V in Table 2 [read].
  • Institutions did, by the author's report, decline to look at it (Stellenbosch). If true, that is a documented failure to test, not a failed test. See claim 6.

Where the author's own work is weak (same rigour applied to her)

  • Battery input is a tiny signal in a hostile environment. 1.13 W / 24 V = 47 mA average, which across the 0.25 ohm shunt is about 12 mV mean (my calculation). The signal is a spiky, RF-radiating waveform (Kelly: it interferes with television) measured single-ended on a 200 MHz scope with DC coupling, with a shunt made of four 2 W carbon resistors (inductive at hundreds of kHz). A systematic offset of just a few millivolts from ground-loop pickup, probe ground-lead inductance or scope DC offset would change the answer by tens of percent or more. The stated 1.5% and 10% error figures are the scope's datasheet accuracy, not an uncertainty for this setup. The data are about 1,200 samples per stored screen at 1 us/div (a 10 us window against a 997-minute run) [read]. No averaged DC meter, no coulomb counter and no amp-hour test back this number.
  • Battery voltage is not a gauge of energy delivered. The paper itself says the capacity evaluation "was omitted" because lead-acid ratings "vary" [read]. A load really dissipating 17.5 W from 24 V needs about 0.73 A, about 12 Ah over 16.5 h from a 20 Ah pair, and a lead-acid battery under light load shows only a modest voltage fall for that. So Table 2's "no voltage fall" does not discriminate between 1.13 W and 17 W. (My arithmetic; the discharge behaviour of lead-acid is [memory] high confidence.)
  • Energy-budget check on the stated mechanism (my calculation, accepted laws only). With L = 8.64 uH [read], the largest inductor energy per cycle is at most 1/2 L I^2 with I no more than V/R = 2.4 A: 25 uJ. At 143-200 kHz that is 3.6-5.0 W absolute maximum if the current ran fully to 2.4 A every cycle. In the stated 1.3% duty mode the on-time is about 76 ns (0.013 / 170 kHz); the current rise is V t / L = 24 x 76e-9 / 8.64e-6 = about 0.21 A, giving 0.2 uJ per cycle, about 0.04 W at 170 kHz. At the stated 2.4 kHz, 3.7% setting, the first-principles input is V^2/R x duty = 57.6 W x 0.037 = 2.1 W, already above the reported 1.13 W. None of these routes gives 17.5 W of heat unless the battery supplies it. The inductor can only return energy it first took from the battery. So either the battery power is under-measured, or the load heat has another origin (for example the oscillation drawing far more current than recorded), or the heat measurement is inflated.
  • Unaccounted power: the 12 V control supply (about 0.49 W), the MOSFET dissipation and the radiated RF are not in the load-side or battery-side accounting.
  • Thermal calibration is crude. The author admits the "caloric test conditions were crude" and allows 10% error [read]. Heat loss is not linear in temperature rise (convection and radiation), yet the 0.34 W/K factor is applied across 16 to 60 K rises. RF heating (skin effect, dielectric losses in the ceramic former) would also put heat in different places than DC does; probe-on-wall readings could differ.
  • The instability. The circuit is described as "random oscillation" that the user tunes by gate resistor until the result appears. Kelly says only NE555N chips do it and tuning is by watching the battery voltage rise. That is a parasitic oscillator, which makes the operating state undefined and the measurement non-repeatable by design. 2002 says 143-200 kHz and 1.3% duty; the 2009 reprint keeps quoting 2.4 kHz and 3.7%; Kelly says over 500 kHz and 55% duty. These are three different circuits' behaviours described as one.
  • Kelly's chapter is advocacy, citing Aspden, Boyce, Don Smith and others without data. The "external energy inflow" reading is an interpretation, not an observation [read].
  • The 2002 article's own abstract contradicts the assumption that positive and negative shunt samples can simply be netted using a constant battery voltage; during oscillation the battery voltage itself is not constant at these timescales (internal resistance, wiring inductance), yet V is held fixed in the formula [read, equation on 2009 p.6].

Per-claim review

Claim 1: about 17.5 W of load heat for 1.13 W of battery power, a gain of 16-18, sustained 16.5 h with no voltage fall

a. Maths. The arithmetic of 17.74 W (13.32^2/10) and 67.6 kJ (1.13 x 997 x 60) is right. The 1.22 MJ figure is wrong (see Correction 1; about 1.05 MJ). The ratio on the paper's own corrected-for-12V numbers is about 10.9, or 15.5 before adding the control battery. No internal contradiction in the equations. b. Experiment. No conflict with a measurement, but conflict with the energy budget above: the inductor cannot supply the difference. Consistent with known measurement hazards of RF, sub-100 mV shunt signals. c. Tested. By the author and by named companies whose results are not published. Forum replications (Lettenmaier, COP greater than 4, per OpEdNews, also by amateur experimenters) are not neutral or reviewed. Poynt99 and others report artifacts or COP = 1 (snippets only, details unverified). No neutral controlled test is on record, and none was run with an amp-hour (coulomb) measurement. d. If true: energy conservation would fail in a closed circuit, or a hidden energy source exists. The weaker alternative (an accounting error) changes nothing. Verdict: Never properly tested (no neutral controlled test; not dismissed untested, since it was examined on forums and by claimed labs). Split: the strong form (true energy gain) is also Refuted by calculation on the paper's own parameters unless the battery power was under-measured; the weak form (the measured numbers as reported) stands as an unreplicated, probably artefact-driven measurement. I rate the artefact explanation far more likely than new physics.

Claim 2: gain comes from back EMF, with returning energy flowing back to the battery, growing with frequency

a. Maths. Back-EMF from a switched inductor is standard (V = L di/dt). The claim that it yields more energy out than in is not derived anywhere; the paper says no part of the model is developed mathematically [read]. b. Consistent with: flyback recovery is real and measured everywhere. Inconsistent with: any net gain, per the budget above (at most about 0.04 W of stored inductor energy per cycle stream at 170 kHz in the stated duty). c. Tested by the author only. The patent examples (COP 1.086 and 2.45 per Kelly's digest, not verified against the scan) are in the "5 kHz, 68 ohm" and "47 ohm" regime and use a thermal comparison with a resistor-only circuit; their margins are small (8.6% with a stated 5% error) [read, Kelly digest]. d. If true: Lenz-law energy accounting and the Poynting theorem would not close for a lumped circuit. Verdict: Split. Weak form (some energy returns to the battery during the off period): Established. Strong form (net energy delivered to the load exceeds battery energy): Refuted by calculation (energy stored in L cannot exceed what the supply put in; stated parameters cap it at a few percent of the claimed heat). The frequency scaling is Never properly tested.

Claim 3: tuned gate resistor causes self-oscillation far above the 555 setting, producing the gain

a. Maths. Not a mathematical claim; the three reported frequencies and duty cycles (143-200 kHz 1.3%; 2.4 kHz 3.7%; over 500 kHz 55%) are mutually inconsistent descriptions of what should be one state. b. Parasitic oscillation of a MOSFET stage with a long gate lead and a mis-tuned gate resistor is well known (memory, high confidence) and not mysterious. The chip-specific claim (only NE555N) is plausible as a device difference but unverified. c. Tested by Kelly (anecdotally). Forum replicators report similar waveforms (OpEdNews, Lettenmaier). No controlled study of what the oscillator does to measurement accuracy. d. Nothing in mainstream physics would change. This is an ordinary instability whose effect is to make power measurement hard. Verdict: The oscillation itself: Plausible, untested in the sense that it is expected from known physics, though no neutral spectrum measurement is on record. That it "makes the gain appear": Never properly tested (and see Claim 1; the oscillation is the best candidate source of the measurement artefact).

Claim 4: breach of unity; second law does not constrain electromagnetic energy; extra energy from an unseen environment

a. Maths. No derivation. The paper says it hopes to avoid references; it cites no engineering literature [read, p.2]. The inference from "magnetic fields comprise particles" to "the second law need not apply" is a non sequitur. b. Contradicts every closed-circuit energy balance ever measured (Poynting theorem, Joule heating). There are no experiments in support other than this one. The "environmental energy" reading (Kelly) is not tied to any measurement. c. No neutral test. Replications claimed only on hobby forums. No replication with a calorimeter plus a coulomb counter. d. Conservation of energy would need an open-system loophole identified; the author's own conclusion says "nowhere is there a claim to have a closed system" [read, 2009 p.13], which is in tension with the unity-breach wording. Verdict: Strong form (second law and energy conservation do not constrain the circuit): Refuted by calculation (no contrary measurement; the bound used is conservation of energy applied to a circuit with a stated battery) and, as worded, Not testable as stated for the "unseen medium" part (no measurable difference specified).

Claim 5: current is the motion of magnetic dipole particles (zipons, tachyons at about 2c); proton = three electrons; unifies non-locality, dark matter, mass ratio

a. Maths. Not developed mathematically, by the author's own account ("no part of that model has been developed mathematically", 2009 appendix [read]; "in defiance of the prescriptive use of mathematics" [read]). The mass-ratio calculation is numerology from assigned masses (zipon 0.5, electron 1.5) [read, summarised in facts.json]; I did not re-derive it in detail and mark it unverified. b. Conflicts with deep inelastic scattering (proton quark structure, [memory] high), with charge and lepton-number conservation, and with relativity (superluminal particles). It also does not predict the experiment: the paper says the experiment "may prove" it and that it "does not conclusively prove" it, which is a post-hoc fit. c. Not tested. No prediction in the model is stated that could discriminate it from standard theory other than the over-unity result itself, which is Claim 1. d. If true, special relativity, QED, the Standard Model and charge conservation would all fail. Verdict: Not testable as stated (no quantitative prediction), with the specific sub-claim "proton = three electrons" Refuted by calculation / by accepted experiment (a proton's charge and spin cannot come from three electrons: charge would be -3e, not +e, and spin-1/2 by three spin-1/2 gives the wrong baryon structure; [memory] high). I do not count it as a tested-and-failed claim in the experimental sense.

Claim 6: institutions refused to test (Stellenbosch refused a demonstration) while companies (ABB, BP, Sasol, Fluke distributors) supported it

a. Maths: not applicable. b. Not applicable. c. Evidence. Pretoria News (14 Nov 2002) reports Ainslie's statement that Stellenbosch refused to allow a demonstration; this is self-report with no response from the university in the clipping [read: page image]. The MTN ScienCentre displayed it and its director said it "is taken seriously" and there is a case for debate [read]. The same clipping notes she "couldn't give us the oscilloscope" or computer used for the displays [read], which weakens the "display" evidence. The 2002 article's list of organisations says they "approved the inclusion of their name", not that they endorsed the result [read: page image p.5]. Sasol offered a bursary, not a test. ABB and BP results are unpublished. I found no web record of any of them (check failed). d. If true: a pattern of institutional refusal would be a sociological, not physical, finding. Verdict: Refusal-to-test: Never properly tested (plausibly dismissed without testing at Stellenbosch, one-sided source; I cannot confirm). Corporate support: Not supported by the record (names listed, no reports); I use Never properly tested for the experimental claim and flag the independent-testing assertions as unverified.

Overall assessment (contrarian view)

The honest strong case for Ainslie is that she published a replicable circuit, with parts codes, and asked to be disproved; the mainstream response in the folder is mostly silence. But the contrarian standard also demands that the researcher's data be good, and it is not: the central quantity (47 mA mean, about 12 mV across a shunt in an RF-rich, self-oscillating circuit) is measured with a method whose error bars are understated, the energy ratio uses an arithmetic slip (1.22 MJ vs about 1.05 MJ), a control battery is excluded from input, and no amp-hour or calorimetric-coulometric closure is given. An energy budget on the stated components shows the inductor cannot supply the claimed heat. The result is best classed as a never neutrally tested measurement most likely explained by under-measured battery input, not as a physical anomaly.

Single decisive test. Replace the scope-shunt method: run the circuit from a fixed battery (or a large capacitor bank) and measure total energy drawn by a coulomb counter or by the capacitor's voltage drop (E = 1/2 C (V1^2 - V2^2)), include the 12 V control supply, and measure load heat calorimetrically (water bath or a proper calorimeter). If load heat exceeds net energy drawn, that is a first-class anomaly; if not, the artefact explanation stands.