Gunderson: Science Contrarian review (Step 2)

Reviewer role: judge by experimental evidence, build the strongest honest case, and apply the same rigour to the inventor and the critics. Tags: [read] document; [web] URL; [memory] with confidence. Web check budget: one fetch attempted (Wikipedia "Permanent magnet", for energy-product values); it contained no numbers, so the energy-product figures below stay [memory, medium] and are used only as order-of-magnitude scales.

Inputs read: claims.md, facts.json, both deep reads, the two patents' extracted text (spot checks), sites.md "Leads for review" and all eight research notes. No files were downloaded to research/files/ (the Investigator read archived PDFs in the browser only).

Source gap and evidence base (state this first)

• Neither patent contains any measurement. US 2006/0163971 has one unquantified sentence ("Experiments show that high-quality permanent magnets can be magnetically 'pushed against' ... for very long periods", para 0040) [read: p.8; confirmed in extracted text]. US 8,093,869 has no data at all; its numbers are illustrations (5000 m/s, 125 kHz, 1 cm) [read: deep read p.11].
• The only numerical performance evidence is outside the patents and is of four kinds: (1) the inventor's own conference descriptions (ESTC 2015/2016) [web: emediapress.com]; (2) one named engineer's forum account of the 2016 demo [web: energeticforum thread 13945]; (3) assignee statements (Magnetic Power Inc 2008, Chava website) [web]; (4) an anonymous critic. All are second-hand to me; none is a controlled, documented test. Notably the 2016 device (ferrite C-cores, ~800 V H-bridge, switched output spike) is a different device from the 2006 toroid and the 2008 acoustic bar, so it is related evidence, not a test of either patent.

Corrections to Step 1

1. facts.json (and claims.md) treat US 2006/0163971 as an unpublished-outcome A1 ("grant status unknown"). The Investigator found it granted as US 7,830,065 B2 (9 Nov 2010), lapsed for the 12th-year fee (2022) [web: patents.google.com/patent/US7830065B2/en]. The granted claim 1 adds a resonant circuit with a square-wave source and a new claim 11. Claim 2 of claims.md (Lenz "aiding") is therefore in the description of the granted patent too (para 0044 kept), but the grant softened it ("is believed to"). I did not open the B2 myself; this rests on the Investigator's comparison [web, one source].
2. claims.md claim 2 correctly says no numbers; it should add that no patent claim recites gain or Lenz-aiding [read: deep read sec.1]. The energy language lives only in the description.
3. claims.md claims 3-6 are all from US 8,093,869; the patent's own text names a third mechanism family ("Lenz field addition", [A]) and two contradictory statements within it: the reaction fields "tend to cancel" and also "increase the total electrical energy delivered" [read: deep read sec.2]. Claim 4 as written is fair but should note the contradiction.
4. Minor: claims.md claim 5 says power "scales with the square of the summed harmonic amplitudes". The patent wording is "the square of the sum of the energy induced by each acoustic frequency" [read: deep read sec.4] (squares a sum of energies, not of currents). The paraphrase is fair but the unit slip is the patent's.
5. Claim 1's "virtual motion" has a second, separate meaning in the inventor's later talks ("motional field", barium ferrite "programmed to resonate", energy "from ambient energy") [web: Wayback grahamgunderson.com/ou]. Step 1 does not carry this; the energy source stated differs: magnets (2006) versus ambient energy (2018) versus assignee's zero-point energy (2008 site) [web]. Three stated sources for one effect is itself a fact to weigh.

No other factual errors found in Step 1; page citations I spot-checked (para 0040, 360-degree passage, "less acceleration input") match the extracted text.

---

Claim 1: Permanent-magnet flux steered ("virtual motion") by a modulated soft core induces EMF in holed-core wire

a. Mathematics: The patent has no equations. In flux-linkage terms the EMF is -dPhi/dt of flux linked by the wire loop through the holes; steering magnet flux between hole neighbourhoods changes that flux. Nothing inconsistent; "flux cutting a conductor" is a picture, not an error here. "Virtual motion" is a metaphor (the sun seen under agitated water). b. Experimental consistency: Consistent with well-known behaviour: variable-reluctance and flux-switching generators, saturable reactors, magnetic amplifiers and fluxgates all induce EMF by changing a core's permeability in a permanent-magnet bias field with no moving parts [memory, high]. The patent examiner cited Sony US4103197, US4639626 (PM variable reluctance generator) etc. [web: Google Patents via Investigator]. So the EMF is not anomalous. c. Tested: No test of the toroid in the record. The 2015 transformer (inventor-reported COP about 1.02-1.03, forum attendees) is a different build [web: energeticforum/rexresearch]. A COP near 1 with a "motional field" is what an ordinary transformer with a magnet bias would show; it is neither an anomaly nor a test of claim 1. d. If true: Nothing, as stated. The weak form (induced EMF from magnetically modulated PM flux) is standard. Verdict (split): Weak form (it makes an EMF with no moving parts): Established (as known flux-switching/parametric action; energy comes from the modulating coil, not the magnet). Strong form (the magnet's field "moves" so the magnet is the energy source): see claim 2. Strongest honest case: the geometry (hole in mid-plane, alternating magnet faces) is a real, buildable flux-steering design; the granted claim 11 recites it. Whether it beats other flux-switch designs in efficiency is untested.

Claim 2: Lenz reaction directed against the permanent magnets aids the input, so the input is unloaded or amplified, less input is needed as more output is drawn, magnets are the source until demagnetised

a. Mathematics: No equations. The argument is qualitative flux-arrow bookkeeping at one hole (paras 0036-0044). It never closes an energy budget. The proposed measurement (instantaneous v x i, para 0048) is instantaneous power; at 90 degrees quadrature its average is zero, so it measures circulating reactive power, not real power [read: deep read sec.6]. That is a methodological hole in the inventor's own test recipe. Also claims 8 and 9 allow the magnets to be replaced by electromagnets or a DC bias, which contradicts "magnets are the energy source". b. Experimental consistency: No known experiment shows a passive magnetic core system returning more than it takes. Counter-check by energy scale (order of magnitude, [memory, medium] for energy products: NdFeB about 200-400 kJ/m3, hard ferrite about 25-35 kJ/m3): - The 2016 witness account: 20 magnets of 2 x 3 x 0.5 in = about 980 cm3 total; 9.5 W for one hour = 34,200 J. Total magnetic energy product of that volume is about 30 J (ferrite) to 390 J (if NdFeB). The magnets would be "spent" in about 3 to 40 seconds, not an hour; the witness reports no demagnetisation. So if the reported readings are right, the magnet cannot be the source in the sense of the patent ("until demagnetized"), and the "very long periods" sentence has no number to test. This is a calculation, not a replication. - A cleaner explanation exists for the reported input going to zero: a high-Q resonant circuit is mostly reactive, and the real input power is a small difference between large products. Worked: apparent power 100 VA at 89 degrees phase gives P = 1.75 W; at 88 degrees, 3.5 W. A 1-degree error is only 55 ns of voltage/current probe skew at 50 kHz. Input readings near zero or negative (reported -0.25 W) are exactly what phase error on a nearly pure reactive load produces. Power analyzers can also mis-measure kilovolt narrow switching spikes (bandwidth). This is an instrumentation hypothesis, not a finding; the witness says two analyzers and a Tektronix scope agreed, which lowers but does not remove it [web: energeticforum; one witness]. - Parametric operation is an ordinary way for an "input" to look unloaded: the modulating (pump) coil supplies the power, the signal port may draw little. The patent does not distinguish pump from signal. c. Tested: Never properly tested on record (not "tested and failed"): no controlled, independent, documented measurement of the 2006 device exists, no replication of the 2016 device was completed as of Oct 2016 [web: energeticforum pages 1 and 5 only; pages 2-4 and thread 13946 unread], and the assignee's own claims (2008 report: "reaction-less behavior ... soundly demonstrated"; Chava: "more than 100 times the input" at "very low power") offered NDAs and no data [web: Wayback magneticpowerinc.com; rexresearch]. No evidence that it was dismissed without testing; equally, the inventor never published a test protocol. Gunderson himself in 2016: "still in the process of validating", "not quite sure why it works" [web: emediapress], a candid statement that counts for him as honest and against the claim as established. d. If true: A passive structure with a net energy output from a permanent magnet would break energy conservation for the magnet-as-source reading (the magnet is a static state; a cyclic modulated process returns it, net work zero apart from hysteresis loss) and Lenz's law's role as a conservation statement. The weaker reading (uses ambient energy, as the 2018 text says) would require an undetected energy bath, a bigger change. Verdict: Strong form (output exceeding input with magnets as the source): Refuted by calculation for the magnet-as-source reading (the energy-product scale above, using the witness's own numbers); for the broader "COP over 1 by any source" reading: Never properly tested (one witness, uncontrolled, instrumentation questions open; not dismissed untested, but not independently replicated). Weak form (input appears lightly loaded at resonance): Plausible, untested (reactive-power effects). Strongest honest case: an electrical engineer (PE) with a background to know better reported two instruments agreeing, a repeating zero-input condition, and phase-dependent behaviour (input from 4 W to -1 W as phase moves 110 to 140 degrees) that looks like a systematic physical dependence, not noise. The inventor reported failures too (2015 COP about 1.02) and does not claim to understand it. The decisive, cheap test: close the loop (feed the output back to power the drive, as McKay suggested), or measure battery energy over hours.

Claim 3: In a resonant magnetostrictive bar, the Lenz reaction returns an acoustic wave with 90 + 180 + 90 = 360 degrees, aiding the drive; less drive energy needed

a. Mathematics: The numbers check (5000 m/s, 125 kHz gives a quarter wavelength of 1 cm; EM equivalent about 600 m) [read: deep read]. The phase sum is bookkeeping, not derivation. (i) The 180 degree step is assumed: the patent itself says the current phase depends on inductance and a tuning capacitor [read: p.20, 22], and in a resistive circuit the induced current is 90 degrees from the flux. (ii) The 90 + 90 = 180 degrees of travel is the round trip over half a wavelength; adding 180 is just the condition for resonance of the bar itself, which exists without any coil. (iii) The conclusion of "less drive for the same output" is never reconciled with energy: the output power I^2 R is drawn from the acoustic wave, which the source must supply. A passive load always presents a non-negative real impedance (it absorbs power) to the drive; "aiding" the source in the sense of lowering required drive power below the delivered output requires an active element. A loaded high-Q resonator lowers its Q (as quartz or transducer motional impedance does) [memory, high]. Measured behaviour in transducers: an electrical termination shifts the resonant frequency and damping (short vs open circuit resonance), so with a fixed drive frequency, the drive current can fall or rise through resonance shifts without any gain [memory, high]. b. Experimental consistency: Consistent with magnetostrictive transducer and harvester practice; inconsistent with any known case of a load lowering the drive power needed per watt delivered. A body of independent patents cites US 8,093,869 for ordinary magnetostrictive harvesting (DOE US9960341B1, Oscilla Power WO2015116275, Sumitomo Riko WO2014119202, Mitsumi WO2015002069) [web: Google Patents via Investigator, not opened by me]; I found none that reports self-aiding drive. That is silence, not contradiction. c. Tested: Never properly tested on record. The assignee's 2008 report claims 2007 proof-of-concept prototypes and "generation without damping or counter-force" but "inefficiently" and "a few watts self-sustaining expected", with NDAs and no data [web: Wayback magneticpowerinc.com, interested party, does not name Gunderson]. A second, neutral line: the Tae Chang N.E.T. "Sound wave resonance generator" (priority same day, 3 Dec 2008) cites it [web], which says someone else worked on a related idea, not that it was tested. No replication. d. If true: Passive magnetoelastic transducers would have to have negative real damping in their motional impedance; electromechanical energy conversion bookkeeping (a lossless two-port) would have to allow output greater than input. Verdict: Refuted by calculation in its strong form (net saving of drive energy greater than the delivered output): power balance across a passive coupled resonator. Never properly tested for a weaker, measurable form (reaction wave changes resonance and drive current; a cheap bench test: drive impedance with the pickup coil open versus loaded). Phase argument: Internally inconsistent in that it asserts both that the reaction cancels and that it adds, and its 180 degree step is assumed.

Claim 4: Opposing currents in neighbouring coils at a harmonic node add magnetic opposition that increases delivered energy and cancels inductive reactance

a. Mathematics: Two parts. (i) Reactance: coils wound/connected to oppose, with mutual inductance M, have total inductance L1 + L2 - 2M, so reactance falls. This is textbook and correct [memory, high]. (ii) Energy: the patent's own transformer analogy says the secondary's opposing current "increases the current flow within the primary", which is the standard result that load raises the input. It then concludes the opposite (extra power without extra drive). The text says both that the reaction fields "annihilate to zero net flux" and that each "generates" in the other coil [read: p.21]. b. Experiments: Series-opposing coils at adjacent stress maxima of opposite sign adding their EMFs (FIG. 8) is the sound engineering part; it is a design of two pick-ups. No experiment shows added energy. c. Tested: Never tested on record. d. If true: Same as claim 3. Verdict: Split. Reactance reduction by opposed coils: Established (standard circuit theory). "Increases total delivered energy" beyond what the drive supplies: Internally inconsistent and Refuted by calculation (power balance); never tested.

Claim 5: Square-wave drive makes odd harmonics coincide at a central node; power I^2 R scales with the square of summed harmonic amplitudes and is "substantially increased"

a. Mathematics: Wrong as stated. For a linear system, average power is the sum of the powers in each harmonic (Fourier components are orthogonal, so cross terms average to zero), not the square of the summed amplitudes. Worked check (computed): for a square wave, harmonic powers go as 1/k^2 for odd k; the fundamental carries 81% of the total, the first three harmonics 93%. The coincident peak is just the square wave's own peak (amplitude A, reconstructing the input), and the average power equals the drive's own power (Parseval), so there is no gain from "adding" harmonics. The "square of a sum of energies" has a unit slip (energy is already quadratic) [read: p.13]. b. Experiments: Consistent with ordinary resonator excitation; inconsistent with any claimed gain. Peaks matter for peak stress and saturation, not mean power. c. Tested: Never tested on record. d. If true: Parseval's theorem would fail for a linear system; so would the definition of RMS. Verdict: Internally inconsistent (squares the sum rather than summing the squares) and, as an energy gain, Refuted by calculation. Harmonic content as a drive method: Plausible, untested as a bandwidth/efficiency choice, with no gain over the input.

Claim 6: 1/8 (preferably 1/4) acoustic-wavelength source-to-pickup spacing, bias magnet and separated source distinguish the device from prior gyrators and harvesters

a. Mathematics: Consistent: a quarter wavelength is a 90 degree propagation delay, the arithmetic is right. b. Experiments: Phase delay in acoustic delay lines and resonators is well established [memory, high]; whether a pickup at a node of a driven bar is novel is a prior-art and legal question, not a physics one. c. Tested: No test that the spacing produces any benefit beyond the delay itself. d. If true: Nothing; it is a design choice. Verdict: Plausible, untested as a geometric feature; Not testable as stated as a claimed distinction (it separates the patent from prior art, it does not show a performance difference). The patent also hedges: FIG. 7 is described as "an embodiment designed to harvest power from a source of vibratory energy" [read: p.22], blurring the claimed distinction from harvesters.

---

How the claims were treated (examination of rejection and record)

• Not dismissed without testing, and not tested: the record is silence. The US Patent Office classed the family under H02K53/00 ("alleged dynamo-electric perpetua mobilia") yet granted structural claims [web: Google Patents], and softened the energy language at grant ("believed to", deletion of "without consuming natural resources", of "persists until the magnets are no longer magnetized", of "reduction of required input power as loads are applied"), while keeping para 0044 (input unloading) and the "Experiments show" sentence [web, Investigator's comparison; one source]. The prosecution file wrappers (why claims narrowed, examiner's rejections) were not reached; that is the key unread record.
• Funding/publication: no sign of blocked publication; Gunderson published through a commercial conference seller (A & P Electronic Media, with a stake in sales) and a private site. The anonymous "Physics Review Board" fraud allegation about Chava's solicitations is a critic's characterization; no court or regulator finding was found [web: rexresearch]. Chava's own "more than 100 times the input at a very low power level" is an interested-party claim, with no data. The 2009 FOIA entry on Magnetic Power Inc is a snippet only [web].
• Inventor's rigour: three stated energy sources over time (magnets, 2006; zero-point/quantum vacuum, assignee 2008; ambient energy, 2018), "not quite sure why it works", and claims of 570% over-accounting with "still validating". The inventor's own v x i test recipe measures reactive power. I call these out; they cut against the claims.
• Absence-of-evidence anomaly: Chava promised "several prototypes", a "very broad" application and 30 patents in 3 years; no later families exist. Competing explanations the Investigator lists (did not survive validation; lapsed for cost; trade secrecy/NDA; funding or internal conflict; suppression). Evidence I can add: the failed-validation explanation fits the soft ("believed to") grant text and the 2016 "still validating"; the cost/secrecy explanations fit that a provisional was filed in 2016 but nothing published (provisionals are never published). No evidence for suppression was found. Decisive record: USPTO file wrappers and status of the 2016 provisional's conversion, Chava filings (FL/NV), Gunderson's 2019 video.

Findings kept from the Investigator (pertinent; not removed whatever the verdict)

• 7,830,065 grant, lapse 2022; 8,093,869 active to 2029; two US-only families.
• Inventor's ESTC 2015/2016 descriptions (570%, input dialled to zero, COP about 1.02-1.03 for 2015).
• Mark McKay PE witness account: 1.53 W in, 9.43 W out (COP 6.16); an hour later input 0.0000 W, output 9.5 W; Gunderson reports negative input about -0.25 W; "Reiyuki" phase-dependent table and about 3.5 W in/8.5 W out; no completed replication seen.
• Related work: Floyd Sweet VTA, MEG US6362718 (Patrick et al.), E.V. Gray, TK/Kapanadze, Tae Chang N.E.T. sound-wave generator; mainstream harvesters citing 8,093,869.
• Assignee chain and MPI history (Goldes; GENIE = title of US 8,093,869); FOIA log; items unread (about 10 archived MPI PDFs, forum pages 2-4, thread 13946, 2019 YouTube).

Summary table

What would settle it: (1) close the loop (output powers the drive) in the 2016 device, or log battery energy over hours; (2) measure real input power with a calorimetric or battery-energy method, not v x i; (3) bench test of drive impedance with the claim-3 pickup open versus loaded.