Smith-Don: Science Contrarian review (Step 2)

Reviewer: Science Contrarian (judges by experimental evidence, not consensus; strongest honest case first, then the same rigour applied to the inventor). Date 2026-10-08. Tags: [read] = from the project documents (page images or text I read myself), [web] = Internet source with URL, [memory] = recall with confidence.

Source gap (repeated from claims.md)

The folder holds no measurement, test report, witness statement, instrument reading or patent grant. The only physical evidence is seven uncaptioned photographs (binder pp.3-9) and Smith's statements that devices were demonstrated. Lagace's don-smith.pdf is a secondary source (his reconstruction, not Smith's words). No verdict below on Smith's hardware can rest on the folder alone. Where I say "never properly tested", I mean no controlled, neutral test is on record; recordings that I did not watch (below) may hold more.

What I checked myself

• Binder p.44 (page image): "Capacitor formula: W = .5 X CE X C.P.S." with "W = energy in Joules", "Both one henry and one farad equal one volt" [read: page image p.44].
• Binder p.49 (page image): example B ends "[ 50,000 x One Million squared = ]", the last line blank, then "greater than 50 Mega K-Joules [Watts]"; benefit 7: "A small energy source is used to start the device and remains fully charged at all times from the System" [read: page image p.49].
• ExtraOrdinary Science, Apr/May/Jun 1996, p.37 (page image, rendered at 200 dpi): example B reads "Joules = 0.5 x .000,010 x 100,000^2 x CPS^2 = 500 x One Million squared", answer "more than 500 MegaJoules [Watts] plus" [read: page image, research/files/1996-ExtraOrdinary-Science-Electrical-Energy-System.pdf p.7].
• Lagace pp.40-45 (text layer): the 1 F, 100 V scaling, 4,750 J out for 500 J in, "Efficiency: 950%" [read: p.42-45].
• I accept the deep reads for the rest after spot-checking the quoted passages against the extracted text.

Corrections to Step 1 (and to the Investigator)

1. Example B has three printed answers, and the magazine's intermediate step is itself wrong. The deep read and sites.md compare the binder's "greater than 50 Mega K-Joules" (5 x 10^10) with the magazine's "more than 500 MegaJoules" (5 x 10^8). Both are checked below against the printed formula, p.47: 0.5 x 10^-5 F x (10^5 V)^2 = 50,000 J, then x (10^6)^2 = 5 x 10^16. The magazine writes "500 x One Million squared": its 500 should be 50,000 (a factor-100 slip), and even 500 x 10^12 = 5 x 10^14 is not 500 MJ (5 x 10^8). So: binder 5 x 10^10, magazine 5 x 10^8, formula 5 x 10^16 (binder numbers) or 5 x 10^14 (magazine's own intermediate). Four numbers, no two equal. Neither document computes the last step; each leaves a figure unreconciled with the formula. Evidence: page images above. Change to make: facts.json and claims.md claim 1 should say "three printed results" (not two versions), and sites.md's correction should add that the magazine's own "500" disagrees with its formula.
2. Claim 6, "the printed balances do not close", can be put more strongly. The deep read tests the 5 mJ scale. The 1 F, 100 V scale is Lagace's headline (the "1 kW" and "950%"), and it fails in a simpler way. Energy in: Cap A 5,000 J at the start plus 500 J from the battery = 5,500 J. At the end of a cycle Cap A is back at 100 V (5,000 J), so at most 500 J has left the system. Lagace prints outputs of 2,500 J (Cap C) + 2,250 J (auxiliary) = 4,750 J (pp.42-44). Unaccounted: 4,250 J, i.e. 8.5 times everything the system received and 4,750/500 = 9.5, his "950%". It is also not necessary to use the deep read's 2x slips: this accounting needs no assumption about the loss per pulse. [read: p.42-45; arithmetic mine]. The one statement in this section that does hold is that the redistribution of two equal capacitors from 100 V to 50 V leaves 2,500 J in them, with 2,500 J lost (pp.42): charge sharing alone gives exactly that.
3. claims.md claim 7 and "Source gap" stand, but "patent position" needs the Investigator's addition: Smith also filed MX NL02000035 A (15 Nov 2002, published 20 May 2004 as a solicitud, i.e. application; status "Application Discontinuation" [web: https://patents.google.com/patent/MXNL02000035A/en, as recorded by the Investigator]). Kelly's "granted" and Perreault's "approved" are the two statements the record does not confirm. I did not re-fetch that page.
4. The Lagace statement "peaked over 35 years ago" (p.47) is contradicted by the dated record (Smith's public activity 1994-2005, Investigator). Add to facts.json for don-smith.pdf, as the Investigator proposes. I confirm the sentence is on p.47 [read: p.47].
5. Smith's 1996 suitcase claim is in no claim in claims.md. It is the only quantitative performance statement by Smith himself (12 V x 0.3 A in = 3.6 W; 120 V x 80 A out = 9.6 kW, "measured by an Electrical Engineer, picked by the audience ... as Documented in the VCR Tape", Investigator [web: https://energyevo.org/2015/01/18/updated-compilation-of-important-info-don-smith-shared-to-nuenergy-yahoo-group/]). claims.md claim 3 should carry it as the performance claim under test; I treat it under claim 3 below.
6. Minor: the deep read of Resonance pp.54-75 says (section 1) that 3.9 EJ/yr x 4.5 x 10^9 yr = 1.76 x 10^10 EJ, which differs from the printed 17.6 x 10^18 EJ by a factor of 10^9. I confirm the arithmetic: 3.9 x 4.5e9 = 1.755e10. That inconsistency belongs in claim 3's evidence.

Verdict summary

Claim 1: energy by resonant induction scales with the square of frequency

a. Mathematics. Arithmetic of every printed step is right (A: 0.072 x 3,600 = 259.2) [read: page images pp.47-49]. The problem is the formula. On p.44 the energy is 1/2 C E^2 x f (first power); on p.47 it is 1/2 C V^2 x f^2. Example A uses f^2, example B's binder answer (5 x 10^10) is the first-power result (5 x 10^4 J x 10^6 = 5 x 10^10), and the formula as printed gives 5 x 10^16 (correction 1 above). Units: 1/2 CV^2 is joules; times f it is watts; times f^2 it is J/s^2, which is neither energy nor power. Smith labels it "Joules [Watt Seconds]" and, in example B, "[Watts]". C is defined "in microfarades" but entered in farads. "Both one henry and one farad equal one volt" is dimensionally false: 1 H = 1 V s / A and 1 F = 1 A s / V [memory, high]. The f^2 factor comes from an analogy (frequency in the place of c in E = mc^2, p.38; or of velocity in 1/2 mv^2, p.43), not from a derivation. Inconsistent with itself, so it cannot be tested as written.

b. Established experiments. The strongest case for Smith: an honest f^2 and f dependences do exist. Induced EMF in a pickup coil is proportional to f, so power into a fixed resistive load rises as f^2 for a given field amplitude; radiated power from a small dipole at fixed current rises as f^2 (f^4 at fixed dipole moment); a capacitor driven by a sine wave circulates reactive power 2 pi f C V^2 [memory, high]. Smith's "1/2 C V^2 x f" is, in effect, the rate at which a capacitor's stored energy is exchanged with the source each cycle: reactive power, which returns to the source and does no net work. Reactive power is measured daily on power systems, so the quantity is real and large (his example B, 10 uF at 100 kV and 1 MHz, would circulate about 6 x 10^6 A, deep read). Real; not available output. The "f^2" version has no measured counterpart. Nothing in the folder is a measurement. The only text that comes close is the 1996 suitcase statement (claim 3), which is a power figure from a device and does not test the formula.

c. Tested? No test of the formula is on record, by Smith or anyone else. Not dismissed after a test; the formula was simply never accompanied by a measurement. Smith's own stated "establishment" figure ("less than 10 Watt Seconds", p.49) is unsupported arithmetic (his p.44 formula gives 4.32 W).

d. If true. Energy conservation would fail: a 50 kJ capacitor at 1 MHz would yield 5 x 10^16 J per second of "useful" energy by the printed formula, an absurd rate. The plain statement is enough: a device storing 50 kJ cannot deliver more than 50 kJ per full discharge, and the only way to deliver it 10^6 times a second is to refill it 10^6 times a second from a 5 x 10^10 W source. Nothing in the folder names one. Maxwell's equations and energy conservation (Poynting's theorem) would have to change.

Verdict. Internally inconsistent (f vs f^2; three printed results; units). Strong form (net energy proportional to f^2): refuted by calculation. Weak form (reactive power circulating at rate f): Established, no net energy.

Claim 2: zero impedance at resonance, room-temperature superconductor, Ohm's law inapplicable

a. Mathematics. Z = R + j(X_L - X_C). At series resonance X_L = X_C, the reactive parts cancel and Z = R; at parallel resonance |Z| is a maximum (about Q^2 R), not zero [memory, high]. Smith's "impedance is the sum of the system resistance, which becomes zero at resonance" (p.39) confuses cancellation of the reactances with vanishing of R. "Volts and amperes are equal" has no meaning without units. "Infinite Q at resonance" (p.50): Q = omega L / R is finite for any real R. Internally the claim treats Z as "resistance", then says Ohm's law does not apply because Z is zero; but V = I Z is exactly Ohm's law for AC, so with Z = 0 the voltage across the loop would be zero, not "volts and amperes equal".

b. Experiments. The strongest case: resonance is real and its effects are large. A high-Q coil rings, the current in a series-resonant loop is limited only by R, and voltage across L or C is Q times the drive. Tesla coil output voltages in the hundreds of kV from a few-kV drive are measured routinely. That is likely what Smith saw and what the "no ohms" talk describes. But Q is measured on every RF coil (Q-meters, vector network analysers) and comes out at tens to thousands, never infinite [memory, high]. At RF the current crowds into the skin depth, about 66 um in copper at 1 MHz, so effective resistance rises roughly as the square root of f; Smith's "electrons encircle the conductor and are free of the conductor's resistance" (p.48) describes skin effect but with the opposite consequence [memory, high]. Real superconductors are defined by DC resistance below a critical temperature; none exists at room temperature for these coils.

c. Tested. Not by Smith; the folder prints no Q, no resistance, no heating measurement. By everyone else, yes: every resonant coil ever measured. Tested, not supported, for zero resistance.

d. If true. Every Q measurement ever made on a coil would be wrong, and the definition of superconductivity (a thermodynamic phase with zero DC resistance and a Meissner effect) would not apply. Standard circuit theory would not need to change for the resonance effects themselves.

Verdict. Zero impedance / superconductor form: refuted by calculation and by routine measurement (Tested, not supported, by the ordinary record of Q). Weak form (resonance cancels the reactive part, giving large circulating current and voltage magnification): Established.

Claim 3: the extra energy comes from the Earth's and air's ambient electrical and magnetic background

a. Mathematics. The scale numbers conflict with themselves. 3.9 EJ/yr over 4.5 billion years is 1.76 x 10^10 EJ (1.76 x 10^28 J), yet the text prints 17.6 x 10^18 EJ (1.76 x 10^37 J), a factor 10^9 too large [read: p.54, 67; arithmetic mine]. The gamma example (1,000 ft squared x 1,600 gammas x 10 V) is 1.6 x 10^9 gamma-V-units, not "1.6 trillion", and "16 trillion volts" does not follow [read: p.55; deep read]. A magnetometer reading in gamma (a magnetic field, 10^-9 T) has no conversion to volts: "one gamma equals 10 volts" and "Gauss = 100,000 volts" are unit conversions between unlike quantities (a magnetic flux density and a potential difference), and the "squared distance times reading" correction is the wrong way round for a source field (a dipole field falls as 1/r^3, a monopole-like field as 1/r^2, so the source is larger than the remote reading by r^2 or r^3, but that correction gives a field in the same units, not volts) [memory, high]. The ambient figures (200,000 V, 500,000 V at the poles) have no instrument, place or date (p.39).

b. Experiments, strongest honest case. (i) There is a real, measured ambient electrical potential: the global atmospheric electric circuit has a ground-to-electrosphere potential of order 400 kV, a fair-weather field of about 100 V/m near the surface and a vertical current density of about 2 pA/m^2 [web: https://en.wikipedia.org/wiki/Atmospheric_electricity]. Smith's 200,000 V and 500,000 V are the same order as that potential, so he was not inventing a magnitude. (ii) Energy can be extracted from natural fields: the Kelvin water dropper, atmospheric-electricity antennas (Franklin bells), earth batteries and geomagnetically induced currents in power lines all do it, at small power. (iii) Smith's own account of the 1996 demonstration is a direct, specific, first-hand experimental claim: the briefcase device ran "continuous 120 V at 80 Amperes" from "12 V DC at 300 milliamperes", "measured by an Electrical Engineer, picked by the audience ... as Documented in the VCR Tape" [web: energyevo URL above, Investigator]. It is a claim of a public measurement with a recording that exists; the workshop itself is independently listed in the 1996 Symposium programme [read: research/files/1996-ExtraOrdinary-Science..._djvu.txt, per Investigator]. I do not dismiss it with a label. It is one account, written by Smith about five years after the event; the recording (Friedrich's list: a 2 h 39 min video of 1996, https://youtu.be/INPIfDGfMGs) is the deciding record and I have not watched it. (iv) The Investigator also records bench claims by others: tswift (2016) reports about 1 W continuous output and battery "self-charging" in a flyback, battery and earth-ground setup, and himself flags a possible meter artefact ("probably just a measurement error from a cheap meter"; scope probes do not show it) [web: energeticforum thread 13932, Investigator]. Anonymous forum reports, uncontrolled; they are evidence of what experimenters claim to see, not of excess energy. Lagace's 2026 videos claim a cold input inverter and an output that melts power supplies, with no power figures in the part the Investigator read [web, Investigator].

Against. (1) Power budget. If the ambient field were the source, the global circuit sets the ceiling. Using the page's values, the circuit power per unit area is the current density times the full potential: 2 x 10^-12 A/m^2 x 4 x 10^5 V = 8 x 10^-7 W/m^2 [arithmetic mine, from the [web] values above]. Over the Earth's surface (5.1 x 10^14 m^2 [memory, high]) that is about 4 x 10^8 W for the whole planet, and a 1 kW device would need the whole circuit's flow over about 1.25 x 10^9 m^2 (1,250 km^2). Even if one counted the full 400 kV across the collector (which a grounded coil cannot), a table-top device cannot draw kilowatts from this field. Ambient RF is smaller still away from a transmitter [memory, medium; not checked]. Smith's own 17.6 x 10^18 EJ is not a deliverable power; stored energy in a reservoir says nothing about the rate at which a coil can draw from it. (2) Energy "conservation laws are excluded" for open systems (p.31) is wrong as stated: open systems exchange energy with surroundings and conservation applies to the whole [memory, high]; the attribution to Einstein's biography has no source. (3) The suitcase figures have the shape of a product of RMS voltage and current: 120 V x 80 A is 9.6 kVA, which is real power only for a resistive load with sinusoidal waveform; an RF-fed, rectified or switching output read with an ordinary meter can show V x A without the same real power. Smith's own p.46 triangle separates "Available Power" from "Used Power", but the 1996 post does not say what the load was (a heater? lamps?) or how power was measured. I flag this as a question the recording would answer, not as a conclusion. (4) 80 A at 120 V into a resistive load in the room would be a visible 9.6 kW heater; the report names no load. Kelly's reader notes the choke wire in a Smith photograph looks too thin for the stated currents [web: Investigator, Kelly ch.31; reported by Kelly]. (5) The 1994 demonstration, as described by an interested party (Rick Friedrich, who sells a Smith-derived kit): the setup was run "to the point of the neon bulbs", the four-secondary coil was not loaded, parts failed, the audience reacted negatively; the "battery stays charged for a month, then a year, then 2 years" statements got longer with the telling [web: https://www.youtube.com/watch?v=BHr3eDELyHk, Investigator]. A battery that holds its charge for a month while also driving a high-voltage module is something a simple log could show; none is recorded. (6) The 1997 schematic note says the battery will recharge itself "if the plus connector wire's length is a wavelength component of the L-6 or L-7's frequency" (p.36) and the parts list contains no L-6 or L-7: the design had not been fixed ("will have to experiment for correct number"). The p.36 diode next to the battery is the obvious path by which a high-voltage, high-frequency source charges the battery by rectification, which is input energy, not ambient energy [reasoning, not tested]. Whether the "self-charging" is ordinary rectification of the module's own output is a measurement to make: log battery charge and the module's input power over hours.

c. Tested? Not by a neutral party. Smith's claim of a public measurement (1996) is on a tape that this review has not seen. A validation planned for Inventor's Weekend 2002 ("some good test equipment to validate his claims") did not happen because Smith had a stroke (Perreault, an interested party: organiser and video seller [web, Investigator]). The 2001 tape was edited by the organiser at the point where Monroe Fischer questioned Smith; the reason given is to protect a speaker weakened by a stroke, the critic site's reading is that an unfavourable exchange was removed; the uncut tape would decide (Investigator). Competing explanations for the absence of later test data (Investigator): illness (documented), contractual or commercial secrecy (Smith's own mention of "contract obligations", Kelly archive), no working device. No document I have decides between them. The bench replications are anonymous forum posts with a self-flagged artefact; none reports a measured input and output power over a stated interval with a calibrated instrument. This is "never properly tested" (no neutral controlled test on record), not "tested and failed" and not "dismissed without testing". It was not rejected by an institution; there is no record in the folder that any institution tested it at all.

d. If true. A collector drawing kilowatts from atmospheric or earth electrical fields would require a power source thousands of times larger than the global circuit, or a different, unmeasured reservoir; that would change atmospheric electricity, not circuit theory. If the energy came from somewhere else (stored in the battery, from rectified driver output, from a nearby transmitter), nothing in physics would change.

Verdict. Split: (i) a kilowatt-scale draw from the ambient atmospheric/earth electric field: refuted by calculation (budget above, pending retrieval of the Wikipedia-sourced numbers' primary sources); (ii) Smith's specific performance claim (3.6 W in, 9.6 kW out in 1996): never properly tested; the recording and the engineer's identity are the deciding records.

Claim 4: turn ratio and "squared flux" give a net energy gain in an air-core two-coil system

a. Mathematics. The arithmetic is right (8,000 V / 4 turns = 2,000 V/turn, p.45). Volts per turn is a real concept: V = N dPhi/dt. If every L-2 turn sees 2,000 V per turn, an N-turn secondary gives 2,000 N volts: linear in turns, not squared. The "squares" in the text apply to energy (1/2 C V^2), not to flux or field: field between plates is proportional to V (p.43 figure: "Four Volts = Sixteen" squares). In p.61 the quantity divided among the turns is volts; in the speech (p.72) it is "the energy input in L-1 is divided by the number of turns" and "if more turns are present in L-2 than L-1, there is a net gain in the Energy present, as demonstrated by the dominos". Volts and energy are different things; the two statements are different claims. No L-2 output is computed in any of them. The dominoes analogy confuses a stored-energy release (each domino's own stored potential energy) with energy created by the first flip, which is the same accounting as a transformer (primary energy triggers many secondary stored energies), and even for dominoes the total released energy was put there earlier by whoever stood them up.

b. Experiments. Voltage step-up by turn ratio is among the best-established facts in electrical engineering. A resonant (Tesla) transformer gives a voltage rise roughly V2/V1 = sqrt(L2/L1) at near-lossless transfer, or sqrt(C1/C2), with energy conserved [memory, high]. Strongest honest case: a high-Q secondary accumulates energy over many drive cycles, so the peak output power can be far above the average input power. That is real; it is storage and release, with output energy never above input energy plus what was already stored. Tesla coil operators measure this every time. Nothing in the folder reports an energy or power measurement of a Smith coil.

c. Tested. Transformers and Tesla coils are tested without end; their energy balance has never shown gain. No Smith-specific test (coil dimensions, coupling, input and output power) is on record.

d. If true. Energy conservation (Poynting's theorem for the coil pair) and the standard coupled-circuit model would fail.

Verdict. Voltage step-up by turns: Established. "Net gain in the energy" by turn-division or "squaring flux": refuted by calculation (power conservation in any linear coupled circuit; the voltage gain is paid for in current) and internally inconsistent (volts divided in p.61, energy in p.72).

Claim 5: dipole and capacitor plates convert a resonating dipole's radiated flux into useful energy; "mirror" plate effect

a. Mathematics. The binder's "Dipole Transformer Generator" (pp.11-19) and the Mexican filing are qualitative: no formulas, no dimensions, no numbers; the prototype is said to be "fully tested for utility by the Inventor" with no results [read: digest pp.11-14; deep read of these pages was not done]. Cannot be checked for consistency. The words "Heaviside current component" are used without definition. [memory, medium]: the term sits in the Bearden literature as a name for the curl-free part of the Poynting vector (the part that Heaviside's energy-flow argument allows and that is usually dropped); Smith's own meaning is not stated, so I do not read it into him. Lagace's Figure A shows two plates with a plastic sheet between them, not an air gap, which differs from "charging a plate at a distance" [read: page image p.2, deep read].

b. Experiments. Strongest honest case: near-field coupling between a resonating antenna or coil and a capacitor plate or second resonator transfers real power without a wire. Electrostatic induction through air is textbook (a charged plate induces charge on a nearby grounded plate; grounding the far plate lets current flow until the potentials equalise). Strongly coupled resonators have been measured: 60 W to a lamp over more than 2 m at about 40% efficiency [web: https://pubmed.ncbi.nlm.nih.gov/17556549/ (Kurs et al., Science 317, 83, 2007; figures from the summary at https://sciencedaily.com/releases/2007/06/070607171130.htm)]. Tesla's own resonant wireless-power demonstrations are the historical forerunner. So "capacitor plates at right angles to the dipole pick up energy" is plausible, and Smith's "mirror" demonstration, if run as a plate-to-plate induction, should work in the same way. Efficiency in any such system is below 100% of the energy the driver delivers [memory, high], and the energy captured cannot exceed what the dipole stores and radiates: an electrically small dipole stores far more reactive energy than it radiates, so "radiated and wasted energy" is a small fraction of the driver's input [memory, high].

c. Tested. The MX application is a filing, not a test. The Investigator found it was discontinued (status "Application Discontinuation"); Kelly's "granted in May 2004" is the likely origin of later "patented" statements, and the document is headed "SOLICITUD de PATENTE" [web: Investigator]. No measurement of a dipole-and-plate device by anyone is in the folder. The tswift replication used a flyback, a battery and a ground, not a dipole and plates. The 1994 and 1996 demonstrations (claim 3) used devices of unspecified type.

d. If true. For the weak form, nothing changes: this is near-field power transfer. For output above the driver's input, energy conservation would fail.

Verdict. Weak form (plates collect part of the energy a driven resonator puts into its near field): plausible, untested as built. Strong form (net gain; energy "radiated and wasted" recovered and exceeding input): never properly tested. The folder holds no number to test either.

Claim 6: capacitor displacement-current scheme (Lagace's reconstruction)

This claim is Lagace's, not Smith's. The 1 F / 1 kW statement reaches us only through Lagace, with no source, date or occasion (p.2). Related prior statement: Utkin (c.2012) quotes Smith: "the capacitor is charged by sparks and powerful displacement current is produced. The transformer with the ferromagnetic core is collecting this current" [orchestrator fix 2026-10-08: the source reads "is produced", not "is around"; the sentence is Utkin's caption for a 2006 rendering titled "Don Smith Device" (credited to Zero Point Films), not Smith's words; see deep/further-investigator-files.md] [read: Utkin PDF text layer, per Investigator]. That is the same idea in Smith's name, about a decade before Lagace, which supports Lagace's claim that it is Smith's, but there is no earlier, primary Smith source in the folder for it.

a. Mathematics. Chapter 3 is right: sharing charge between two equal capacitors, 10 V to 5 V each, loses half the energy whatever the resistance (1/2 (C/2) (delta V)^2) [memory, high]. The remaining chapters do not close: figure and prose differ by 1000 (10 mF in Figure C, "10 microfarads" in the text); 1/2 CV^2 for 10 uF at 10 V is 0.5 mJ, not 5 mJ (the deep read's ratio test still holds, on the document's own 10x scale); after step one of chapter 6 the printed state already totals 6.25 mJ for a 5 mJ input; chapter 4 gives 1.25 mJ per pulse and then "2.5 + 2.5 = 5" (a factor 2); the polarity-flip step claims capacitor A "holds all the energy" at 0 V while E = 1/2 C V^2 = 0 [read: deep read, spot-checked against p.186-540 of the extracted text]. At 1 F and 100 V the scheme prints 4,750 J out for 500 J in; the energy ceiling is 5,500 J in total and 500 J of output (correction 2). "Primary does not feel the secondary load" is contradicted by the document's own account: energy given to Cap C is energy the C1-C2 transient no longer dissipates, which is what a primary that feels the load looks like, and Figure D draws the primary in the series loop of C1, T1, C2 and SW2; "no closed current loop" (p.32) is not what the figure shows. Maxwell's displacement current in the capacitor gap equals the conduction current in the leads, so the current through the primary is the conduction current [memory, high].

b. Experiments, strongest honest case. The core engineering idea is sound: the half of the energy lost in charge sharing between two capacitors is only lost if the transfer goes through a resistor. Put an inductor in the path and the transfer rings (LC oscillation), and with a rectifier or a second winding extracting the oscillation into a third capacitor, much of that half can be recovered. That is adiabatic or resonant charge transfer, standard in switched-capacitor and resonant converters [memory, high; not checked this run]. So a transformer primary in series with the capacitor-to-capacitor path capturing "the half that is normally lost" is plausible and in fact the design the document draws. What it recovers is energy the first transfer would have dissipated, up to 100% of the input at best, never above. The step that is not defensible is the claim to 200%, 950%, or "over-unity". Lagace's 2026 bench videos claim a working modified replication; the Investigator found no input or output power figures in the part read, so there is no number to weigh. The videos' author is the author of the document (interested: he is promoting his reconstruction).

c. Tested. No measurement of the circuit in the document: "every number in chapters 3-7 is a worked example, not data" (deep read, confirmed by me for chapters 5-7). A simulator screenshot of two capacitors at 5.0 V is the only simulation output. Never properly tested in the document; the ordinary recovery effect is already known engineering, the over-unity form is a bookkeeping error, so no experiment is needed to refute that form (see calculation).

d. If true. For the over-unity form, energy conservation would fail. For the recovery form, nothing.

Verdict. Weak form (a series transformer or inductor can recover part of the energy normally lost in charge sharing, with total output at most the input): plausible, untested, and consistent with known physics. Printed form (4.75 kJ out for 0.5 kJ in, "950%"): refuted by calculation (own figures, correction 2) and internally inconsistent (the 1.25 mJ vs 2.5 mJ per pulse and factor-2 slips).

Claim 7: Smith's account of suppression

a. Mathematics. Not applicable. The inference chain on pp.50-51 is a logical argument and the contrarian reading of it: (a) over-unity applications are sent to a "metering group"; (b) so the Office accepts logarithmic measuring devices; (c) so it accepts the square law for kinetic energy; (d) so it knows over-unity exists. Step (a) is Smith's report of his own treatment (art unit, examiner, date and application number are not given). Steps (b)-(d) do not follow: assignment to a group that deals with measurement says nothing about whether logarithmic scales or square laws are accepted, and logarithmic scales are used for convenience over wide ranges (decibels), not because AC measurements are nonlinear [memory, high]. Smith's reasoning here is an unfalsifiable form: any treatment by the Office would confirm it.

b. Experiments and records. The record of how it was rejected is thin, and I say so plainly: (1) The meeting. "A meeting between J.P. Morgan, Edison and Tesla" where Tesla proposed a metered-free system (p.37). No date, place or source. A web search for it found the "where do we put the meter" line passed on secondhand in popular accounts with no primary document, and with the same line attributed to other people [web: search, e.g. https://archive.longislandpress.com/2010/09/16/teslas-last-stand-on-long-island/2/ among the hits; no primary record found]. What is documented is that Morgan financed Wardenclyffe (c. $150,000 for 51% of Tesla's patent rights, reported) and withdrew, with differing accounts of why [same search; secondary]. So Smith's story is a version of a popular anecdote; the three-person meeting is not in the sources I could find. [memory, medium]: no such Morgan-Edison-Tesla meeting is part of the standard record. (2) The Patent Office. MPEP 2107 treats a "perpetual motion machine" among its examples of "incredible" utility (citing Newman v. Quigg), and says the claimed device "must be totally incapable of achieving a useful result" for a utility rejection; the examiner must not call a utility incredible unless clearly proper [web: https://www.uspto.gov/web/offices/pac/mpep/s2107.html]. So the Office does have a documented, published path for refusing claims on utility grounds, and it is a rule the applicant can answer with evidence (a working demonstration). That bears on Smith's complaint and on the owner's interest in how claims are rejected: this is a refusal on a stated standard, not a secret group; I found no evidence for or against a "metering group" and Smith's account is a single, interested source. Smith's stake: a pending application (08/100,074) and a business selling books and devices. His statement that "no prior art" came "based on the Patent Office response" (p.22) is a second claim about the same file; the file wrapper would settle both. (3) The Mexican application was filed 2002, published 2004 and discontinued; no reason is public (the IMPI file was not reached by the Investigator). (4) The US serial: the Investigator notes that serial numbers around 08/100,000 date from 1993, not 4 February 1992 [memory, low-medium; not checked by me]. If 1992 is the earlier filing of a parent case, there may be two files. Unknown.

How the claims were rejected, by the sources: there is no record of any neutral lab testing Smith's devices and failing. There is no record of any funding or publication being blocked (Smith published in ExtraOrdinary Science in 1996 and spoke at the International Tesla Society convention in 1994 and 1996 and at TeslaTech in 2005 or 2006, per the Investigator). The mainstream's silence is the silence of not being asked: no test was offered with measured input and output, and the only validation announced (2002) did not happen for the reason given (illness). That is "dismissed without testing" in the sense that no one tested, but there was also no test requested or published by the inventor with numbers. The Texas company record: no TransWorld Energy entity is shown active 1992-2002 (Investigator, partial), so the "TransWorld Energy, CEO" letterhead of 2002 (p.1, p.21) is not matched by a registered company found so far; an assumed name is possible.

c. Tested. The claim of suppression cannot be tested as a physical claim. The records that would test it: the USPTO file wrapper of 08/100,074, IMPI's file for NL/a/2002/000035, the uncut 2001 Inventor's Weekend tape, county assumed-name records.

d. If true. Nothing in physics changes. A documented pattern of the Patent Office treating a class of devices differently from its published rule would be a legal and administrative finding.

Verdict. Not testable as stated (the "metering group" inference and the meeting); the underlying records (US file wrapper, IMPI file, assumed-name records) are not obtained. Smith's account is reported as an interested party's characterisation, not as the neutral fact.

Investigator findings: what I keep in the record, and how I weigh each

• 1996 suitcase claim (12 V 0.3 A in, 120 V 80 A out, audience-chosen engineer, VCR tape): kept; the strongest primary evidence of an anomalous claim; unverified; recording not watched; load and meter type unknown. Bears on claim 3.
• 1994 demonstration (Friedrich, interested party; video exists): partial failure, unloaded secondary, shifting battery-hold claims. Kept; evidence for and against; the tape decides.
• Planned 2002 validation not held; 2001 tape edited by the organiser; critic's report not found: kept as an absence-of-evidence anomaly with competing explanations (illness, secrecy, no working device); none is decided by what is on record.
• Utkin (c.2012) quoting Smith on displacement current captured by a transformer: supports that the claim-6 idea predates Lagace; Utkin is a secondary compiler.
• Lagace's 2026 videos and Friedrich's 170 pages of Smith originals: more material exists than the 96-page binder; not read. A review that rests on the binder alone should say so.
• tswift 2016 bench report (about 1 W, self-flagged meter artefact): kept; anonymous, uncontrolled.
• Kelly's chapter: "granted" is not borne out by the IMPI record; "forty-eight devices ... kilowatts of excess energy" is Kelly's statement, no test cited; "taken over by Big Oil" is his characterisation (a lapsed domain bought by a reseller is the competing explanation).
• The MX application and the dating issues (US serial 08/100,074): kept.
• Death date 22 Nov 2010: one forum snippet; no obituary found.

Where Smith and Lagace are held to the same standard as the critics

Bad data and unit errors that I confirm: the two capacitor energy scales (Lagace, 10x), "10 mF" vs "10 microfarads" (Lagace), f vs f^2 (Smith), "microfarades" entered in farads (Smith), a henry and a farad equal to a volt (Smith), "Joules [Watts]" and "50 Mega K-Joules" (Smith), the 10^9 error in the Earth-energy figure (Smith), the p.46 power triangle with an energy as a side (Smith). Unfalsifiable moves: "conservation laws are excluded for open systems"; "Ohm's law does not apply"; "the primary does not feel the load" with no number; the patent-office inference. Equally, the critics' side needs care: "tested and failed" is not the state of this record. No controlled test of any Smith device exists in the folder, and most mainstream rejection here rests on calculation, which is a legitimate basis (energy budgets) but different from an experiment. Where the calculation is strongest (claim 6's own numbers, claim 1's own formula), no experiment is needed. Where it depends on a mechanism Smith does not specify (claim 3's 1996 demonstration), the experiment is not yet done.

What would settle it

1. Watch the 1996 recording (https://youtu.be/INPIfDGfMGs): what load, what meters, who the engineer was, whether the battery was disconnected.
2. A neutral bench test: log input energy (battery and module) and output energy (calorimetric load) over hours, with the power-supply input measured at the source with a true-RMS power meter. If output energy exceeds the input energy over a long run, the claim has its first real datum; if not, it is closed.
3. The USPTO file wrapper of 08/100,074 and IMPI's NL/a/2002/000035 file.

Source tags and web checks

[web] checks done this run (5): atmospheric-electricity numbers (Wikipedia), MPEP 2107 (USPTO), Kurs et al. 2007 (PubMed/ScienceDaily summary; the Science page itself returned 403), Wardenclyffe/Morgan meter anecdote (search; no primary source found). Not checked: skin-depth number, adiabatic charge-transfer statement, the power-triangle conventions, the "1 gamma" unit statements (all [memory, high]); the USPTO serial-number dating (low-medium, not checked); ambient RF power (medium, not checked). The 8 x 10^-7 W/m^2 and 4 x 10^8 W figures are my arithmetic from the [web] current density and potential, and depend on Wikipedia's 2 pA/m^2 and 400 kV.