Physicist review: Chris Sykes, "Guidelines to Bucking Coils" (v2.7, 26.02.16)

Reviewer: PhD physicist (Step 2). Tags: [read] from the document or Investigator files, [read: p.N] a page of the Guidelines, [web] with URL, [memory] with confidence, [calc] my own calculation (venv Python, numbers below).

Source gap: none for the Guidelines (text layer good). Gaps that matter: no data table, power-in/power-out measurement or COP calculation appears anywhere in the Guidelines; the only quantitative experiments are the two coil traces on pp.14-16. The post-2016 device data are forum posts read only by the Investigator, page 1 of the threads. Itsu's replication data are behind a login and are known only through Sykes's account.

What the project says (short)

Sykes proposes that two "partnered output coils", wound oppositely on a loosely coupled core so that their magnetic parts cancel, deliver current to a load without the load's reaction (Lenz drag) appearing on the input. He reads this as the working principle of Floyd Sweet's Vacuum Triode Amplifier (VTA) and of Moray, Meyer and others.

The mainstream view, stated fairly

1. Faraday and Lenz. The emf in a loop is -dPhi/dt (Phi = magnetic flux through the loop). Lenz's sign says the induced current's own field opposes the change. It is energy conservation written as a sign.
2. Mutual inductance is symmetric. For two coils M12 = M21 = M. The emf in coil 2 from coil 1 is M dI1/dt and the flux linkage on coil 1 from current I2 is M I2. Whatever geometry lets coil 1 drive coil 2 also lets coil 2 push back on coil 1. This is the key point against "output without reflection" (see claim 1).
3. Energy. Power into a linear passive network is d/dt(1/2 L1 I1^2 + 1/2 L2 I2^2 + M I1 I2) plus I^2 R losses. Delivered power is paid for by the source. A cancelling geometry changes how much is stored and returned, not how much a load can take out.
4. Potentials. B = curl A. E = -grad(phi) - dA/dt. A is the vector potential, phi the scalar potential. A is not "the electric component of the magnetic field"; Sykes's claim is a different thing from textbook usage (Griffiths ch.10, Feynman Lectures II ch.15 [memory, high]).
5. Mainstream can describe real subtleties here: bucking coils exist (humbucking pickups, non-inductive wirewound resistors, balanced transformers, common-mode chokes), loosely coupled gapped cores, flyback/diode circuits (Bedini-type) that return stored energy. None give COP above 1 [memory, high].

Claims

Claim 1: bucking output coils cancel the magnetic part, A adds, so no Lenz drag on the input

a. Mathematics. The document has no derivation, only words and drawings; the "A adds, B cancels" statement is not an equation. As stated it is not well defined (see mainstream point 4). Where it can be made precise it fails: take an E-core with the primary on the middle leg and the two outer legs carrying the output coils. If the coils are wound so the emfs from the primary flux add in series, the reaction mmfs (magnetomotive forces, amp-turns) also add in the same common-mode flux path and are reflected into the primary. If instead they are wound so the reaction fluxes circulate around the outer legs and cancel in the centre leg, then the primary flux, which splits through the outer legs, induces cancelling emfs, so there is no output either. This is the reciprocity M12 = M21 in geometric form. [calc/argument, not a theorem from the document.] Verdict on (a): not internally contradictory (it is too loose to be), but incompatible with reciprocity when made quantitative. b. Established experiments. Transformers, including multi-winding and "bucking" ones, obey P_in = P_out + losses to the precision of calorimetry (efficiency 95-99 percent at large size) [memory, high]. The Guidelines themselves quote the ideal-transformer relation [read: p.10-11]. c. Tests. By Sykes: the first (2010-11) and second device videos; a 2014 hyiq.org update "Input is 1 Watt (5V 200ma) ... Globe rated 12V 300ma (3.6 Watt) ... slightly more Output than this" [read, Investigator hyiq.org note]. The "rated" figure is a nameplate, not a measurement, and no meter or waveform is given. By others: dewetw (aboveunity.com, 30 Oct 2019), who built the three-coil circuit on an N87 ferrite core, reproduced the sawtooth waveform and "never saw ... more than about 0.82 COP" (documented in one forum post with photos; Sykes's reply addressed the waveform and did not dispute the number) [read, Investigator]; patrick1: replication "successful of course, but thats because it isnt OU"; Itsu (overunityresearch.com topic 3951, 2020), whom Sykes calls "failed" with a load of 0.6 W and "less than a basic Transformer Efficiency" (Sykes's account only; data not seen); MileHigh asked in 2015 for the missing power-in/out data and found none in the PDF. Measurement basis is itself in question: Sykes's own 2019 thread shows RMS across a 0.25 ohm shunt gave 59.5 W where the scope mean gave 17.1 W, and Partzman noted diodes reversed relative to the scope traces [read, Investigator]. Note the sources are Sykes's own forum and an interested party's characterisation of the failing replication; they are not neutral controlled tests, but they are the only tests on record and they point one way. Sykes himself writes one can have this "and still be Below-Unity" and that Sweet-style gain "we have not succeeded, in general" [read, Investigator]. d. If true: reciprocity of mutual inductance and the energy theorem (Poynting's theorem) would fail for linear media, and the first law with them. Sykes's own words differ here: he says "Energy can not be created or destroyed. You and I agree" and "Not Free Energy" [read: p.48; overunity.com reply #7], while elsewhere "Free Energy Future is here NOW" [read: p.49] and "Input can become Negative" (2020). The document is ambiguous between "lower reflected load on an input" (a loss-reduction, in principle harmless) and "output exceeds input" (conflicts). Verdict: split. Weak form (a bucking geometry can reduce coupling or reflected reactive power; a diode/flyback circuit can return energy): Plausible, untested as a quantified claim. Strong form (load current drawn with no reflection into the input, i.e. real power gain): Refuted by calculation (reciprocity plus energy conservation), and Tested, not supported by the only replications on record, which were informal and not neutral.

Claim 2: Lenz's law is the result of drawing current from already-separated charge, not the cause

a. Consistent in a trivial reading and wrong in the strong one. In a generator or transformer, no load current means no reaction torque or reflected current; Sykes states this correctly [read: p.21-22; hyiq.org Lenz's Law page]. "Charge separation" is just the induced emf, E = -dA/dt - grad(phi), which exists before current flows. The strong inference, that you can draw current and have no field, contradicts Ampere's law: current makes B. Sykes concedes "A moving charge constitutes a magnetic field" [read: p.10]. The step from "drag appears only when current flows" to "so it can be avoided while current flows" does not follow, because the drag is the mechanical-to-electrical energy conversion itself: the force on a wire carrying I in field B is F = I L B, and F x v = emf x I. b. Consistent with all induction experiments; the weak form is in textbooks. c. Tested: every generator ever measured. No test of the strong form is reported beyond the forum devices above. d. If strong form true: Lenz's law would become "optional", and with it energy conservation in electromechanics. Verdict: split. Weak form (drag exists only when current is drawn): Established. Strong form (drag avoidable while delivering power): Refuted by calculation.

Claim 3: the magnetic field itself slows charge flow (650 microseconds vs 75 ms at 7.8 ohm)

This is the only claim backed by data in the document, so I checked it closely, using the deep read (pp.14-19, with page images) and my own numbers. a. Mathematics. Consistent: the standard law is i(t) = (V/R)(1 - e^(-t/tau)), tau = L/R. [calc] For R = 7.8 + 1 = 8.8 ohm: Experiment 1, L = 0.231 H gives tau = 26 ms, 3 tau = 79 ms, final current 1.75 A; Experiment 2, L = 0.24 mH gives tau = 27 microseconds, 3 tau = 82 microseconds, final current 0.80 A. Sykes's "75 ms" matches about 3 tau of Experiment 1. The coil data fit two tightly coupled windings: series opposing L = 2L(1 - k) with 0.24 mH and 55.6 mH per winding gives k = 0.998 [calc]. Hence the "field slows charge" statement is the textbook statement that inductance resists change of current, in different words. Sykes admits this: "an expected behaviour for the Current in the Coil when the Inductance Changes" [read: p.16]. Problems: (i) the times were not measured with cursors or a defined 63 percent point; "650 us" is the pulse width at 647 Hz (about 640 microseconds read from the trace) and "75 ms" is about the Experiment 1 pulse width or 3 tau [read: deep read, page images pp.14-15]; (ii) the ratio of claimed times is 75 ms/0.65 ms = 115 against an inductance ratio 231/0.24 = 960 [calc], so the second trace does not follow L/R; the nearly linear rise to 0.7 A in about 0.4 ms implies an effective inductance near 4 mH [calc: L_eff = V/(di/dt) = 7 V / (0.7 A/0.4 ms)], about 16 times the quoted residual, so something else (drive circuit, MOSFET wiring as drawn, core effects at 0.7 A, leakage) limits the rise; (iii) supply voltage (15.4 V vs 7.0 V), drive and frequency (6.1 Hz vs 647 Hz) changed between runs; (iv) it measures the current rise di/dt, not any "speed of charge"; the drift velocity of conduction electrons does not change; (v) no repeat runs or error bars. The plateau current is V/R in both, so the DC resistance did not change. b. Consistent with L/R behaviour everywhere. It is inconsistent with the stronger reading, "friction on charge itself that shows as heat" [read: p.17]. Resistance heating is I^2 R; the back-emf of an ideal inductor stores energy (1/2 L I^2) and returns it, it does not dissipate it. c. Tested only by Sykes, once, with two scope photos (OWON PDS7102T, 100 MHz). Anyone repeating it would get the textbook curves. No replication needed or reported. d. If the strong reading were true: inductance would have to cost I^2 R-type loss and "duty factor" of copper would be variable; Sweet's passage "a conductor ... temperature rise ... factor of 10 would now operate at 1.0, the same gauge wire would carry 10 times more current" [read: p.17; source is the Sweet letter "Letter to Mark from Sparky", Investigator] would have to hold. Neither experiment tests heating or steady-state resistance. Superconductors and cryogenic copper do change resistivity, but through lattice scattering, not by fields cancelling. Verdict: split. Weak form (inductance, i.e. the field, slows the rise of current): Established, and the document's own traces show it. Strong form (field is friction on charge, wire carries 10x current at the same temperature, shown by this experiment): Not supported by this experiment, and the numbers claimed are not reproducible from its traces (internally inconsistent in the time ratio); the Sweet duty-factor statement is Never properly tested.

Claim 4: loose coupling, Bloch wall between coils, diode with square-wave drive at the BH-curve knee; input falls as output rises; alternator test

a. Mathematics. No equations. Statements are about qualitative behaviour. "Input power falls as output rises" [read: p.46] is attributed to output current feeding back to raise the input coil's impedance. In an ordinary loaded transformer a source in the load circuit in antiphase does exactly reduce input power, but then the "load" is a second power source delivering energy, so net power must be accounted for. The text's own example says this ("a source of power returning back", [read: p.46]). A "Bloch wall" is a boundary between magnetic domains in a ferromagnet, a few tens to hundreds of nanometres thick, set by exchange and anisotropy energy [memory, high]. Sykes treats it as a macroscopic "wall of force" that grows with a coil's current and sits between the coils [read: p.6-7, 23]. A domain wall does not behave that way; the lamination "jumping over it" is the ordinary field of a magnet. b. Alternator test [read: p.48]: connecting the field coil to the output gives output with no separate field supply. That is a self-excited alternator, standard since Siemens and Varley/Wheatstone (dynamo principle, 1860s) using remanent magnetism [memory, high]. The mechanical input still pays: "ONLY because of Lenz Law, Resistive Mechanical Force is applied". Sykes says "Not Free Energy". It is not a test of the claim. c. Tests: no measurement in the document. Later forum data are above (claim 1c). The Investigator's later quotes (benchmark in transformer mode at about 80 percent efficiency, count only what exceeds that; "Input can become Negative"; 12 V 400 mA input with output "some data off screen") are untabulated. A 10 percent duty, 3000 Hz pulsed circuit with RMS/mean ambiguity is the classic setting for overestimating power. The first-order test is cheap: average v(t) i(t) over whole cycles for input and output, with a calorimetric check. [Not done on record.] d. If true: either the first law fails, or the extra energy is drawn from a store (the Aether/vacuum), and the Sykes text does say the system is "directly connected to the Aether" [read: p.47] and "no new form of energy" [read: p.47] in the same breath. No measured excess would need no change in physics. Verdict: split. Weak form (loose coupling and diodes affect waveform, a flyback/Bedini-type circuit returns stored energy to the supply, and a self-excited alternator needs no external field): Established (ordinary circuit and machine behaviour). Strong form (input falls to or below zero while output rises, net gain): Never properly tested with a neutral controlled test; the informal replications report COP below 1 (0.82). The Bloch-wall mechanism as described: Not testable as stated.

Claim 5: Sweet's VTA, Moray, Meyer, Jensen, Koenig, Frolov, Melnichenko, Hendershot and others all work on this principle

a. Mathematics. The Sweet source Sykes depends on ("Nothing is Something", 1988; Investigator file) gives E = (BxV) + (-Bx-V) = 2(BxV) for opposing fields. Dimensionally E = V x B is the motional field, and the sum of a term with both signs flipped does not cancel, since (-B)x(-V) = BxV, so the "2" is algebraically right but the physical premise (what the opposed field B and velocity V are) is not defined. It is not derived from Maxwell's equations. I leave detailed notation to the mathematician review. b. Experiment. The record is of unreplicated claims. The 1986 lab notes (transcribed by Sykes from scans of an anonymous source, 15 pp., 17 scan pages numbered 5-48; Investigator) record 24 May 1986 witnessed by Mark Goldes and Darryl Roberts; 4.6 A x 24.2 V = 111.32 W out, input quoted at "0.023mw" or 7 V x 143 uA, gains of "105,000x-111,000x" and "200,000 rectified", with Beckman and Keithley digital multimeters at 388-402 Hz [read, Investigator]. A multimeter reading of a waveform near 400 Hz from a ferromagnetic reactive device is exactly where mean/RMS and true-power errors arise; "gain" computed from a microwatt DC input to a 100 W AC output is not a power measurement unless power factor, waveform and phase are measured. The numbers also conflict inside the sources: Guidelines p.13 says generation six gave "5000 Watts" in a shoebox; the 1991 Sweet-Bearden paper reports 60 Hz and over 1 kW at 6 lb; Bearden elsewhere says the device produced "about 6 watts" [read: p.19-20]. A third-party analyst on the forum (Anor, Sep 2026) lists "what does not add up": no replication of conditioning, the 12-turn flux-swing problem, and "The 1 kW and 5 kW claims do not sit comfortably with the stated wire gauges" (a forum post; one source) [read, Investigator]. Independent testing: Sweet "never had the VTA certified by independent testing" and kept secret his magnet-conditioning process [web: https://www.rexresearch.com/sweet/1nothing.htm and search summary of Sweet/Bearden accounts; I did not open the page, so this is a search-result summary only]. Sykes rejects Bearden's conditioning in the Guidelines (p.17, "simply wrong") and now as "nonsense" (2025). Since Bearden and Sweet say magnet conditioning was essential, the Sykes principle, bucking coils, is not what the inventor of the device said made it work. That is a point the document itself leaves open. c. Tests: Sweet's lab notes (self-run); a weight-reduction test account (rexresearch; Investigator, not read); no independent replication on record. Koenig US4806834A (filed 1987, granted 21 Feb 1989; two mirror-image coils on a magnetisable bar; Investigator [web] https://patents.google.com/patent/US4806834A/en) shows only the geometry, not a measured gain. Others, e.g. Melnichenko, Kapanadze, Meyer, Hendershot, are not tested in the project's material; I know of no independent controlled test of any of them [memory, medium]; Meyer lost a 1996 Ohio court case on his "water fuel cell" with expert testimony [memory, medium; not on this claim's evidence, mentioned only because Sykes cites him as a bucking-coil precedent]. d. If true: the claim that these devices share one principle would require a mechanism that works for devices as different as a ceramic-magnet plate pair with 12-turn output, a Meyer water-cell circuit and a motor-generator. The simpler hypothesis is that they have nothing in common but a bucking geometry and an unmeasured gain. Verdict: Never properly tested (no neutral, controlled test of Sweet's device on record; Sweet declined independent certification, a point Sweet's defenders would say reflects a lack of funds and secrecy, and critics say fits an artefact; not dismissed without testing but not tested either). The "same principle" unification: Not testable as stated.

Claim 6: 98-99 percent of generator loss budget is Lenz drag, 1-2 percent friction, so removing drag is the route to higher efficiency

a. Mathematics: an arithmetic and physical misreading. The text says generators are "98-99% efficient; this means that 100% of shaft energy, with a 1-2% loss ... and the other 98-99% are the effects of Lenz law creating drag" [read: p.23]. Efficiency 0.98-0.99 means 1-2 percent of shaft power is lost as heat and friction. The other 98-99 percent is the electrical output. Lenz "drag" is the mechanical force that carries that power across: torque x angular speed = emf x current. It is not a loss. [calc: loss budget 1 - 0.98 = 0.02; 1 - 0.99 = 0.01.] Removing the drag removes the conversion, so no electrical output remains. b. Consistent with measured generator efficiencies; large utility generators are 98-99 percent efficient [memory, high], so the maximum conceivable gain from "removing losses" is 1-2 percent, not a large factor. c. Tested: every dynamometer measurement of an alternator. d. If true: shaft power would not be convertible to electrical power by induction. Verdict: Internally inconsistent (the claim's own premise, 98-99 percent efficiency, contradicts its conclusion), and Refuted by calculation. The sentence is a plain misinterpretation and a good worked example for the reader.

Investigator's findings that bear on these claims (kept whatever the verdict)

Own work: Guidelines v2.7 original Word export on hyiq.org (better text layer); hyiq.org 2014 update (1 W in, 3.6 W rated lamp, "slightly more output"); 2015 overunity.com thread 15395 (reply #4 "Lenz's Law is ONLY apparent when Current is drawn ... Energy is not Generated at all! It was already there!"; reply #11 "Loose Coupling is a requirement!"; MileHigh reply #8 asking for data); 2019-2026 aboveunity.com "Non-Inductive Coil Experiment" (three coils, two diodes, 10 percent duty, tagged ABOVE-UNITY by the forum; Sykes: "and still be Below-Unity"); 2020 Builders Group (80 percent transformer benchmark, "Input can become Negative!"); 2025 rejection of Bearden magnet conditioning. Sykes states his work is public domain and "can not be patented", so no patents to read. Supporting or testing by others: dewetw COP at most 0.82; patrick1 "isnt OU"; Itsu "failed" (Sykes's account); Partzman diode-polarity and RMS-versus-mean critique; Tinman's mechanical-device comment (second-hand quote); Anor's analysis of the VTA numbers. Related work: Sweet's own texts (lab notes, "Letter to Mark from Sparky", "Nothing is Something" 24 Jun 1988), Sweet-Bearden IECEC 1991, Koenig US4806834A, Melnichenko, Jensen unidirectional transformer, Figuera 1908 patents, Wiegand wire analogues, Cohn on induction. My reading: Figuera's 1908 generator and Koenig's coils are variable-flux, mirrored-coil ideas; I did not see a measurement establishing gain for any [memory, medium]. Historical record: overunity.com is now a domain-for-sale redirect (GoDaddy) and its archive is paywalled; Wayback holds partial snapshots. This is an "absence of evidence" item. Competing explanations (Investigator's list): commercial decision, hosting lapse, deliberate removal. The only evidence found points to the first two; Sykes's own claim that aboveunity.com was "DEPLATFORMED by everyone" and PayPal refused service is from an interested party and I could not check it. It does not change my physics verdicts, but it means the 2015-2021 replication debate may be lost; Itsu's data at overunityresearch.com should be obtained before any final judgement on "Tested, not supported" for the 2020 circuit. Sweet's lab notes come from "an anonymous source" per Sykes: a provenance gap, since the scans were not found.

Corrections to Step 1

1. claims.md and facts.json describe the filar/bifilar result as the author's "proof" and treat "650 us vs 75 ms" as measured "charge times"; the deep read shows the times are not measured by cursor and match pulse width or 3 tau. facts.json summarises it as "charge flow ... took about 75 ms versus about 650 us" without that qualification. Added in claim 3 above.
2. claims.md item 6 repeats "98-99 percent of loss budget is Lenz drag". The text actually calls generators "98-99% efficient", so the claim as Sykes states it is a misreading of efficiency; the "loss budget" wording in claims.md already contains the author's error. Corrected in claim 6.
3. facts.json says the Sweet duty-factor quote has no source; the Investigator found it is from Sweet's "Letter to Mark from Sparky" (the Guidelines version differs by an added "and"). Treat the quote as sourced to Sweet, but the document does not cite it.
4. facts.json says no copy of the Sweet lab notes is in the digest: the Investigator found a transcription on hyiq.org (15 pp., 17 scanned pages numbered 5-48). The notes are now readable but still a transcription by an interested party.
5. Claim 4 in claims.md lists the alternator test as support ("output with feedback and no extra field input"). It is a self-excited dynamo and Sykes says himself "Not Free Energy"; it does not support net gain.
6. The string "COP" does not occur anywhere in the Guidelines text layer (grep, 0 hits). The "COP 1.7" figure comes from MileHigh's 2015 forum post attributing it to Sykes; Sykes's own statement of it was not found. Step 1 does not claim otherwise, but later steps should not attribute a COP to the Guidelines.
7. facts.json states the claim of Bearden barium ferrite magnets (p.20-22) and "6 W" as the author quoting Bearden; consistent with deep read p.19-20. No correction needed other than that Sykes's "5000 W" (p.13) and Bearden's "6 W" disagree by three orders of magnitude, which Step 1 did not flag.

Summary table

Web checks: one search made (Sweet/Bearden independent testing); retagged [web] with a search-result summary; the Koenig patent metadata is the Investigator's [web]. The remaining [memory] items (Bloch wall thickness, utility generator efficiency, self-excited dynamo history, Meyer case) are high or medium confidence and no verdict depends on a medium one alone.