Arbiter: Sykes-Chris, "Guidelines to Bucking Coils" (v2.7, 26.02.16)

Step 3. I read facts.json, claims.md, sites.md, the deep read, and the three reviews (physicist.md, mathematician.md, contrarian.md), and recomputed the numbers where they differed. Tags: [read] from the project files, [web] with URL, [memory] with confidence, [calc] my own calculation (venv Python, numbers in section 2).

Source gap note

There is no gap for the Guidelines themselves: the text layer is good and the deep read covers pp.14-19 with page images (the reviewers read the rest of the text layer, and the mathematician rendered pp.1, 8, 9, 46). The stale statement in claims.md and facts.json ("digest only, no deep read") should be corrected.

Gaps that matter for the verdicts: - The Guidelines contain no power-in/power-out table for any device the author built. Every quantitative power figure comes from outside the document: Sykes's later forum posts (read by the Investigator, page 1 of the threads only), a replicator's post, Sweet's lab notes (a transcription by Sykes of scans from "an anonymous source"), and Bearden. Where a verdict rests on those, it rests on the Investigator's notes and not on the primary pages. - Itsu's replication data (overunityresearch.com, login wall) are known only through Sykes's account of them. - The scans of Sweet's 1986 notes, the v2.5 copies of the Guidelines, the device videos, and Sykes's own scope data for the three-coil circuit ("some data off screen") were not seen. - Several verdicts below rest on [memory] at high confidence (reciprocity of mutual inductance, passivity, E = -grad phi - dA/dt). No verdict depends on a low-confidence [memory] fact.

Corrected claim list (Step 1 corrections the reviewers agreed on)

1. Claim 6 in claims.md is misworded. Page 23 says generators are "98-99% efficient", meaning 100% of shaft energy, 1-2% lost, and "the other 98-99% are the effects of Lenz Law creating Drag" [read: p.23]. Sykes is not describing a loss budget. He is calling the useful conversion work "drag". Corrected: "Generators are 98-99% efficient; the author labels the converted 98-99% as Lenz drag and argues that removing the drag is the route to higher efficiency." (All three reviewers agree.)
2. Claim 3: the "650 microseconds" and "75 ms" figures were not taken with cursors or a defined 63% point. 75 ms is about 3 tau of Experiment 1; 650 us is about the generator's pulse width at 650 Hz [read: deep read, page images pp.14-15]. They are not measured charge times. The word "proof" is the author's.
3. Claim 4: the alternator test is a motor-driven alternator with its field coil fed from its output. Sykes writes "Not Free Energy" and "ONLY because of Lenz Law, Resistive Mechanical Force is applied on the Shaft" [read: p.48]. It does not support "no extra input". The "input falls as output rises" mechanism he gives on p.46 is that the load is "another Power Source" returning power [read: p.46].
4. The p.17 Sweet "duty factor" quotation is not uncited. It comes from Sweet's "Letter to Mark from Sparky" on hyiq.org (the Guidelines text adds an "and") [web: https://www.hyiq.org/Downloads/Letter%20to%20Mark%20from%20Sparky.pdf, via the Investigator's saved copy].
5. Sweet's lab notes now have a transcription (hyiq.org, 15 pp., 17 scan pages numbered 5-48), made by an interested party from scans of unknown provenance. The scans themselves were not found.
6. The word "COP" does not occur in the Guidelines. "COP 1.7" comes from MileHigh's 2015 forum post attributing it to Sykes; Sykes's own statement of it was not found. Later steps should not attribute a COP to the Guidelines.
7. "Bloch wall" is used with a changed meaning: in mainstream magnetism it is the nanometre-scale transition layer between two domains inside a material (tens to a few hundred nm [web: https://www.sciencedirect.com/topics/materials-science/domain-wall, search-result summary read by the mathematician]); Sykes uses it for a macroscopic region between the coils.
8. Notation: the Guidelines use "Phi" for the scalar potential, for an undefined "magnetic component", and for magnetic flux; and emf = Bvl is given with "l = the Area", which is a unit error (l must be a length) [read: pp.4-9, 46; page image p.9].
9. The "5000 W" shoe-box figure (p.13, no source) and Bearden's "6 W" (pp.19-20) differ by a factor of about 800. Neither is a Sykes measurement.
10. (QE revision) The 1 mW input for 4 June 1986 is stated in Sweet's notes ("Oscillator put 143 uA, 7 v ... approximately 400 Hz 1 mW into the Excitation Coils"), not merely implied. The 43-fold difference with the 0.023 mW of 24 May is between two stated figures [read: lab-notes transcription pp.6 and 30]. Row in the numbers table amended.
11. (QE revision) Claim 4 "conditions": the circuit Sykes prescribes for beginners (p.38) is a sine drive with two bucking output coils and a 10 ohm load, with no diode. The diode and square-wave scheme is labelled "a work in progress" and "only appears to be valid under some configurations" (pp.40-41) [read: deep read pp.38-42]. A diode is therefore not a necessary condition of his own recipe.
12. (QE revision) Claim 1/4 gains a sub-claim from p.11: "Current, as you may think, is not stepped down as a result of the Voltage being stepped up" [read: p.11]. Claim 3 gets the loose-versus-tight comparison of p.32 as Sykes's one informal controlled comparison. Claim 5 gets the "Vector Oscillator" withheld part and the observation that the 1986 device in the notes has a single 12-turn output coil, not a bucking pair. See QE revision.
13. (QE revision) The p.21 equation "reflected power = -dPhi/dt" has the dimensions of a voltage (emf), not power [read: deep read p.21]; p.4 gives field decay as an "I^2 R factor", which is a power loss, not a decay law; p.9 defines velocity as "Speed * Distance Travelled". These are additions to the notation defects in item 8.

Numbers I recomputed

Web checks done by the arbiter (3 of 8)

• Generator efficiency. A search summary gave 95-98% for utility-scale synchronous machines and 90-98% for generators in general; the figure 99% was not found [web: https://eureka.patsnap.com/report-comparison-of-efficiency-between-synchronous-generator-and-asynchronous-generator, search summary only]. The previously [memory, high] "98-99%" is therefore slightly generous. The Claim 6 verdict does not depend on the exact figure.
• Sweet and Goldes. Search results state that Mark Goldes and Daryl Roberts were present at the 24 May 1986 demonstration, and that Goldes was Chairman and CEO of Magnetic Power Inc., which he founded [web: http://amasci.com/freenrg/sweetvta.html and https://wiki.naturalphilosophy.org/index.php?title=Mark_Goldes, search summaries; pages not opened]. This confirms the contrarian's [memory, medium] that Goldes was a promoter of Sweet. Sweet died 5 July 1995 of a heart attack, aged 83, per a search summary of a hyiq.org/naturalphilosophy-type source [web: https://www.hyiq.org/Reference/Profile?Name=Floyd+'Sweet' and https://www.wiki.naturalphilosophy.org/index.php?title=Floyd_A_Sweet, search summary only]. The contrarian's "ruled non-criminal" remark was not confirmed by me.
• Signal generator mains draw (relevant to the contrarian's metering point). Wavetek bench instruments of the type listed draw about 15 W (Model 188), about 20 W (2001) and about 25 VA (2002A) from the mains [web: https://pearl-hifi.com/06_Lit_Archive/15_Mfrs_Publications/38_Wavetek/Wavetek_188_Sweep_Fctn_Gen.pdf, https://manuals.plus/wavetek/2001-sweep-signal-generator-manual, https://chrisgrossman.com/manuals/wavetek/Wavetek-2002A.pdf, search summaries]. The model used in 1986 is not named in the notes I have, so this is the right order of magnitude, not a fact about Sweet's unit.

Not rechecked: the Aharonov-Bohm reference, Meyer's court case (not load-bearing), Koenig's patent claims, all Investigator material (taken as reported).

Claim-by-claim arbitration

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

Where they agree. All three find no derivation in the document, that the "A adds, B cancels" wording is not an equation, and that the weak form (a bucking pair reduces coupling and external field; loosely coupled gapped cores reflect less) is ordinary engineering. All three cite the same replication evidence: dewetw at most 0.82 COP, patrick1 "isnt OU", Itsu called "failed" by Sykes, and no neutral test.

Where they differ. For the strong form (load power delivered without reflection at the input, so gain), the physicist says "Refuted by calculation" plus "Tested, not supported"; the mathematician says "Refuted by calculation"; the contrarian says "Never properly tested", with the informal replications pointing against it.

Which argument is stronger. The calculation is stronger, with one stated scope. For any fully metered circuit built of resistances and inductors, linear or not, with no active element supplying power, average input power is at least average load power. The mathematician's loop model shows the reflected resistance (omega Mx)^2 Re(1/Z_loop) is positive whenever load current flows, and I reproduced his tables: output never exceeds input. A gapped core, a diode and a saturating core change waveforms and storage, not the sign of the energy balance (a hysteretic core stores or dissipates, it does not generate). The contrarian is right about the empirical record, though: nobody has published a calibrated, complete power-in/power-out measurement of a Sykes circuit, and the replications are volunteer bench builds on the inventor's forum, so "Tested, not supported" should be read as informal. The calculation closes the claim only if all power ports are metered; the hole in the evidence is that the metering has never been done to a standard that would test it.

Final verdict (split). - Weak form (bucking pair and loose coupling reduce reflected load and external field): Plausible, untested as a quantified claim. The mathematician's "Established" for cancelling external field is fair for the geometry alone; Sykes's actual weak claim, a measured reduction of reflection at fixed load, has no measurement. - Strong form (net real-power gain, or load power with no reaction): Refuted by calculation (passivity plus reciprocity, for any fully metered passive circuit). Informal replications (dewetw 0.82) are on the same side but are not a neutral test. - The "A is the electric component of B" vocabulary: Internally inconsistent (one symbol with three meanings; a units error in emf = Bvl).

What would settle it. A battery-powered (or separately metered at the DC supply) build of the three-coil circuit with a differential-probe, time-aligned average of v(t) i(t) over whole cycles at every port, and calorimetry of the whole assembly over several minutes, circuit and procedure signed off in advance by Sykes. Pass: energy in equals energy out within error. Cheap preliminary: recompute the 2019 "59.5 W RMS vs 17.1 W mean" shunt data using the sampled product.

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

Agree. All three agree that "no load current, no Lenz reaction" is textbook and Established, and that Sykes himself says energy cannot be created or destroyed (overunity.com reply #7, 2015 [web: Wayback copy, per research notes]).

Differ. The strong form: physicist and mathematician say Refuted by calculation; the contrarian says Not testable as stated.

Stronger argument. They are describing different statements. "Charge separation can occur without a reflected field" is, in the author's words, a hypothetical (p.21) and by itself has no measurable prediction: the contrarian is right on that. Once it is turned into the prediction a reader would test, namely that power can be delivered while the shaft or primary sees no additional load, the physicist and mathematician are right: P = e i = torque times angular speed, so no reaction means no delivered power. Sykes concedes "A Moving Charge constitutes a Magnetic Field" (p.10).

Final verdict (split). "Lenz opposition appears only when current is drawn": Established. "Charge separation without a reflected field, while delivering power": Refuted by calculation (in the sense of the energy balance; the bare wording, taken alone, is Not testable as stated).

What would settle it. A generator run at fixed speed with measured shaft torque (dynamometer) and electrical output, loaded and unloaded. The p.45 rotor experiment reports no numbers; a repeat with torque, speed and load would test it. The calculation already predicts the result.

Claim 3: the magnetic field slows charge flow (650 us vs 75 ms)

Agree. All three find Experiment 1 matches tau = L/R, Experiment 2 does not match it, the runs changed supply, drive and frequency at once, the quantity measured is current rise, not charge speed, and Sykes says this is "an expected behaviour" (p.16).

Differ. Only on labels: the physicist uses "internally inconsistent" for the time ratio, the mathematician "Tested, not supported" for 650 us as an intrinsic time, the contrarian "Not testable as stated" for the interpretation. They are compatible.

Stronger argument. All are right on their own sub-statement. I add that the ratio test (962 vs 115) and the effective inductance (4.0 mH vs 0.24 mH quoted) show that the "non-inductive" coil trace is limited by something other than the quoted inductance (drive circuit, the high-side MOSFET wiring as drawn, core effects at 0.7 A, or the pulse width itself). The 231 mH vs 222 mH ceiling is a second sign that the coil measurements were not made under controlled conditions. None of this makes the L/R physics wrong; it makes the quoted numbers unreliable as evidence for anything.

Final verdict (split). - Observation (aiding windings give a slow current rise, opposing windings a fast one): Established. - Interpretation ("proves" the field slows charge, i.e. something beyond inductance): Not testable as stated by this experiment, since ordinary inductance predicts the same curves; the specific numbers 650 us and 75 ms: not reproducible from the traces. - Sweet's duty-factor statement (10 times more current at the same temperature): Never properly tested. The plateau currents equal V/R for the same 7.8 ohm, so the document's data show no steady-state change.

What would settle it. Repeat both runs at the same supply, drive and frequency, with cursor-measured tau (63% point), then run a steady-state heating or DC-resistance comparison at equal current. L/R predicts about 26 ms and 27 us. A 10 minute bench test.

Claim 4: conditions (loose coupling, Bloch wall, diode and square wave at the BH knee, load), input falls as output rises, alternator test

Agree. Alternator self-excitation is standard and needs mechanical input; the requirements list has no derivation; the Bloch wall as used is not the mainstream object; informal replications give COP at most 0.82; the measurement basis is fragile (RMS vs mean, 59.5 W vs 17.1 W on one shunt; Partzman's diode-polarity objection).

Differ. For net gain the physicist says "Never properly tested ... informal COP 0.82", the mathematician says "Refuted by calculation", the contrarian "Never properly tested".

Stronger argument. Same as Claim 1: the energy-balance argument holds for a passive metered circuit, and the mathematician's resonant-loop model reproduces the qualitative effect (input current falls from 0.39 A to 0.02 A as a resonant loop is loaded) with P_out below P_in at every point, so falling input current is not evidence of gain. The contrarian's observation about structure is also fair: with "1 in 10 builds work", a minimum load, loose coupling, diode, frequency and winding direction all required, a failure can always be assigned to a missing condition, so the claim is hard to falsify as the author states it. Sykes's own 2020 test protocol (benchmark in transformer mode at about 80% efficiency and count anything above as gain) is a baseline that depends on identical metering, not a bound on what the circuit may deliver.

Final verdict (split). - Input current can fall as a resonant output loop is loaded: Plausible, untested as stated (reproduced in a linear model with P_out < P_in). - Net gain, or input falling to or below zero while output rises: Refuted by calculation (for a passive metered circuit); no neutral measurement exists, and the informal replications on record are 0.82 and "isnt OU". - Gap, diode, square wave, minimum load as necessary conditions: Never properly tested (no controlled comparison). - Macroscopic moving "Bloch wall" and "operation at the BH knee": Not testable as stated (no equation, no measurable prediction). - Alternator self-excitation: Established, with mechanical input, as the author says.

What would settle it. The same metered test as Claim 1, plus a B-H loop (hysteresis loop) measurement of the core at the operating drive, which tests the "knee, not hysteresis" statement independently. Also obtain Itsu's posts (overunityresearch.com 3951), Sykes's full scope captures for the 12 V 400 mA circuit and the original schematic, to decide whether the circuit drawn is the circuit measured (Partzman).

Claim 5: Sweet's VTA, Moray, Meyer, Jensen, Koenig, Frolov, Melnichenko and others share this principle

Agree. No neutral controlled test of any listed device is on record. Sweet's data are self-run, with interested witnesses (Goldes, a promoter of Sweet and founder of Magnetic Power Inc.; Roberts), and Bearden (co-author of the IECEC 1991 paper, promoter, an interested party). Sykes himself rejects Bearden's magnet conditioning ("simply wrong", p.17; "nonsense", 2025), so Sykes's account of the VTA differs from Bearden's, who worked with Sweet. Koenig's US4806834 shows a mirror-image two-coil geometry and a grant, not a measured gain [web: https://patents.google.com/patent/US4806834A/en].

Differ. The physicist stresses that the 1986 measurements are the kind where power is easily overstated; the contrarian gives them more credit and lists the specific gaps; the mathematician adds the arithmetic and flux checks. All three give Never properly tested.

Stronger argument. They complement each other, and none removes evidence. In favour of reading the notes seriously: they are dated, signed, witnessed, name the instruments (Beckman, Keithley, Wavetek generator) and the arithmetic is internally correct (4.6 A x 24.2 V = 111.32 W; 19.1544 W / 0.023 mW = 83.3 million %); nobody has re-run them. Against reading them as a power measurement: (a) a multimeter V x I product at 388-402 Hz into a highly reactive drive coil ("hundreds of thousands of ohms") is apparent power, not real power, unless phase is measured; (b) the two input figures in the notes differ by about 43 times (0.023 mW on 24 May; about 1 mW implied for 4 June) with no explanation I can find; (c) the output was read from meters on lamps, waveform and power factor unrecorded; (d) the flux needed (about 1 T swing through a 1.5 in coil at 400 Hz with 12 turns) is hard to reconcile with ceramic magnets of 0.38 T remanence [calc, assumes sinusoid and no flux concentration]; (e) a third-party forum analyst (Anor, 2026, one post) says the 1 kW and 5 kW claims do not sit comfortably with the stated wire gauges.

On the contrarian's point about the generator's mains draw: bench signal generators of that class draw about 15-25 W from the mains [web: see web checks]. That is small compared with 111 W, so unmetered generator mains draw alone cannot be the whole explanation of the 4 June output; it could account for 19 W on 24 May only if the generator model drew that much. This is an example where the point is real but weaker than first stated; the reactive-load (VA vs W) point and the unmetered waveform are stronger. I record both.

Final verdict (split). - Sweet VTA: Never properly tested (a documented 1986 informal measurement exists, with identified metering gaps; no independent laboratory test is on record; that Sweet declined certification is reported by Bearden, an interested party). Not dismissed without testing, and not tested neutrally. - Moray, Meyer, Jensen, Koenig, Frolov, Melnichenko and others each: Never properly tested in the project's material. That they share one principle: Not testable as stated until the principle is given a measurable prediction (the author himself notes Melnichenko's windings are not reversed but give results, p.29). - Bearden's wave/anti-wave and "negative time" language, as quoted: Not testable as stated. The Sweet algebra E = (B x V) + (-B x -V) = 2 (B x V) is correct but trivial (two sign flips in a cross product cancel), and the standing-wave passage (p.36) holds only at specific instants: cancelling amplitudes does not cancel energy, as Bearden's own quoted sentence on p.18 says.

What would settle it. Rebuild the 1986 device as the notes describe (two 4 x 6 x 1 in ceramic magnets, 250-turn #18 excitation coil, 12-turn #12 output coil), drive it from a battery-powered oscillator, and meter true mean power at the oscillator's DC supply, at the drive coil and at the load, with a wide-bandwidth sampled V x I, plus calorimetry. Obtain the scans of Sweet's notes and test whether the 0.023 mW and 1 mW input figures were different set-ups.

Claim 6: 98-99% of generator loss is Lenz drag; remove drag for efficiency; also "not new energy" and "open system connected to the Aether"

Agree, completely. All three find that the sentence mislabels the useful conversion as a loss; at 98% efficiency a 10 kW load costs 204 W of loss, and removing the "drag" removes the 10 kW; the maximum gain from removing true losses is 1-2 points. All note the document says both "no new form of Energy" (p.47), "Not Free Energy" (p.48) and "Free Energy Future is here NOW" (p.49).

Stronger argument. Nothing to arbitrate. One refinement from my web check: utility-scale efficiency is more safely stated as 95-98% [web: see web checks]; this makes the case for the conclusion slightly stronger, not weaker.

Final verdict (split). - Arithmetic as written (useful output counted as loss): Internally inconsistent. - Removing drag to raise efficiency: Refuted by calculation. - "Open system connected to the Aether" and "Free Energy Future": Not testable as stated, and contradicted by the author's own "not a new form of energy".

What would settle it. Nothing needed; the energy balance settles the efficiency argument. (A torque-and-speed test is the same as Claim 2.)

Final verdict table

What the record shows that the verdicts do not erase (kept in the record)

• Sykes's credits: full coil data for the p.14-16 experiment, a statement that the result is "expected behaviour", a rule that "Any COP > 1 Claims must be accompanied by Measurements", correct advice that scope mean, not RMS, gives power, publication of dewetw's 0.82 result on his own forum, no patents ("Can not be Patented"), and documentary work in finding Sweet's primary papers [read; web per Investigator].
• Sykes's own statements of failure: "one can have Asymmetrical Regauging and still be Below-Unity" and "we have not succeeded, in general" (aboveunity.com, as read by the Investigator).
• The measurement hazard Sykes himself demonstrated: 59.5 W from RMS values vs 17.1 W from the scope mean on the same 0.25 ohm shunt.
• Unfulfilled data requests: MileHigh (16 Jan 2015) asked for scope shots, load resistance and error bounds for the "alleged COP 1.7 device" on a thread of 514+ pages and 1.14 million reads by January 2017; the Investigator found no data response in the portion read (page 1 only).
• Single-source evidence stays: the 2014 hyiq.org line "Input is 1 Watt (5V 200ma) ... rated at 12V 300ma (3.6 Watt) ... slightly more Output" is one self-report with a rated, not measured, load and no method; Itsu's outcome is known only from Sykes's description; the Anor analysis and the Tinman quote ("a bucking coil arrangement dose indeed add energy to a system", mechanical devices, second-hand) are each one forum source.
• Tinman's and Sweet's "gain" statements and the 1991 IECEC paper are reported as the parties' own characterisations; Bearden, Goldes, Sweet and Sykes all have a stake in the VTA, and Bearden's accounts were challenged as "hearsay" on rexresearch.com and answered by Bearden [web: https://www.rexresearch.com/sweet/1nothing.htm, per research notes].
• Aharonov-Bohm: A is physical (it shifts the phase of an electron beam where B = 0 [web: https://www.hitachi.com/rd/research/materials/quantum/aharonov-bohm/index.html, per the contrarian]). That supports "A is not mere bookkeeping" but no one has derived load current without reaction from it.

Absence-of-evidence analysis

The record goes dark in five places where it would be expected to continue.

1. The overunity.com archive. The 514+ page thread (topic 15395, 16 Jan 2015 on; page 1772 reached by March 2021) now sits on a domain-for-sale redirect; the mirror overunityarchives.com says "Now Open for Paid Subscribed Members Only"; overunityresearch.com, which holds Itsu's build, requires a login; Wayback holds only partial snapshots (first pages of topics 15395 and 18464). - (a) Ordinary commercial or hosting decay (domain lapse, paywalled mirror). For: the domain sale and the paywall are documented in the Investigator's notes. Against: none found. - (b) Deliberate removal of the content. For: nothing found. Against: Sykes's thread was public and widely read, and Wayback snapshots survive. - (c) Forum decline in general. For: the forum scene shifted to other sites (Sykes's own forum aboveunity.com since 2019). Against: none. Evidence best fits (a) and (c). Records that would decide: a Wayback CDX listing for both topics (to count surviving pages), Itsu's own posts and videos, and any copy held by participants.

2. Itsu's replication. Known only from Sykes's account ("failed", a 6 V 0.6 W load, coil length unknown, "less than a basic Transformer Efficiency"). Explanations: a sound negative test; a flawed build the inventor correctly identified; a mismatch of conditions. Evidence for each is thin because the data are behind a login. Decisive record: Itsu's posts at overunityresearch.com topic 3951, his videos, or his schematic.

3. Sweet's 1986 notes and his own later records. The only copy is Sykes's transcription of scans from "an anonymous source" with pages missing, labelled randomly. Explanations: ordinary loss and mislabelling of a hobbyist's papers; selective release by an interested party; fabrication or alteration (no evidence of this found; the arithmetic is consistent, which is weakly against careless alteration, not against error). Decisive record: the scans themselves and the original lab books (Goldes's and Roberts's papers, Magnetic Power Inc. files).

4. Independent testing of the VTA and Sweet's death. Sweet reportedly never had the VTA independently certified and died 5 July 1995 [web: search summary, see web checks]. Bearden says Sweet feared his life would be "snuffed out" if tested. That is the account of an interested promoter; the cause of death in the summary I saw is a heart attack at 83, with no evidence of foul play found. Explanations: lack of funds and secrecy over the conditioning process, as defenders say; fear of an artefact being exposed, as critics say; suppression, as Bearden implies. Evidence for suppression is only Bearden's recollection. Decisive record: Magnetic Power Inc. and Goldes's correspondence, any funder's or lab's correspondence with Sweet, IECEC conference files for the 1991 paper.

5. The inventor's "deplatformed" account. aboveunity.com says "We have been DEPLATFORMED by everyone" and "PayPal De-Platformed me! They REFUSE to tell me why!" (interested party; no second source found). Explanations: payment-processor risk policy for the "free energy" category; a dispute with Sykes personally; suppression. The evidence is Sykes's statement only. Decisive record: the processor's notices to Sykes, which he could show.

Also absent: a power-in/out table for "COP 1.7", the YouTube device videos (not viewed), and any Sykes data after the 2016 Guidelines other than the forum snapshots. These are gaps in the evidence he would be expected to produce, and the missing metering data is the most telling absence in the project.

Open questions

1. Does any Sykes build show sustained P_out > P_in under full, time-aligned, calibrated metering plus calorimetry? (No such measurement is on record.)
2. What limited the rise in Experiment 2 (about 4 mH effective vs 0.24 mH quoted), and why does the aiding coil measure 231 mH when the ceiling is 222 mH?
3. What were the two different input figures in Sweet's May and June 1986 notes (0.023 mW vs about 1 mW), and what was the true power and power factor of the drive coil?
4. Are the original scans of Sweet's notes obtainable, and who was the "anonymous source"?
5. What did Itsu measure?
6. Does the geometry in Koenig US4806834 (claims unread) actually make the same energy-handling claim as Sykes, or only the same mirror-image coil arrangement?
7. Does the "diode, knee, loose coupling" recipe change measured reflected reactive power at fixed load in a way worth quantifying, even with COP below 1? (This is the part of the project that could be useful to a builder without any free-energy claim.)

Final verdicts in brief: C1 weak Plausible untested / strong Refuted by calculation; C2 weak Established / strong Refuted by calculation; C3 observation Established, interpretation Not testable as stated, duty factor Never properly tested; C4 alternator Established, net gain Refuted by calculation and never neutrally measured, Bloch wall and knee Not testable as stated; C5 Never properly tested, unification Not testable as stated; C6 Internally inconsistent and Refuted by calculation.

QE revision

Written after qe.md (verdict MINOR) and the new deep reads now appended to deep/Guidelines-to-Bucking-Coils-5ecaec.md: pp.21-24, 31-33, 40-42, 25-31, 4-9 and 34-39, and 10-13 (with a comparison against the Sweet lab-notes transcription). I re-read the lab-notes transcription lines I quote (pdftotext of research/files/hyiq-Lab-Notes-Transcribed-Floyd-Sweet.pdf). Result: no headline verdict changes. Four verdicts are reworded, one sub-claim is added, and a number of items that had dropped out of the record are put back. Tags as before.

Not done: the deep read for pp.45-49 (rotor experiment, "swipe" lamp experiment, alternator rig) is not in the deep file. The Claim 2 and Claim 4 statements about those pages still rest on the text layer and the earlier reviewers. This is stated, not hidden, and is the one open remediation item.

Web check added (4 of 8 used)

• Ceramic magnet remanence. Search results for the common grade Y30 give Br of 370-400 mT, i.e. 0.37-0.40 T [web: https://www.stanfordmagnets.com/magnetic-properties-of-sintered-ferrite-magnets.html and https://www.first4magnets.com/us/ceramic-magnet-information/grades-of-ceramic-magnets, search summary only; the summary itself mis-converted this to 0.037-0.040 T, which I corrected from the mT figure]. This retags my earlier [memory, medium] 0.38 T as [web]. Rosenthal gives Sweet's magnets as "grade 5 or grade 8" ferrite [read: research/2026-10-04-rexresearch.com.md]; I did not check those grade numbers. Point (d) under Claim 5 stands: the 1 T swing needed is well above what such a magnet supplies by itself, with the stated caveats (sinusoid assumed, no flux concentration, no core geometry in the notes).

Changes to the record, claim by claim

Claim 1 (and 4): the step-up rule (QE item M1). On p.10 Sykes states the ideal-transformer rule correctly (240 V x 1 A = 24 V x 10 A). On p.11 he writes: "Ideally, these devices work with a Step Up configuration. 1:3 is a common Ratio ... Current, as you may think, is not stepped down as a result of the Voltage being stepped up", and "Floyd Sweet used a Step Up configuration of 1:5" [read: pp.10-11]. The wording is ambiguous ("as you may think" can read either way). Taken plainly, it says secondary current does not fall with the turns ratio. Worked check [calc]: an ideal 1:5 step-up with 1 V in gives 5 V out; a 1 A load then needs 5 A at the input, 5 W in and 5 W out. Secondary current is stepped down and primary current is the larger. The statement can hold only if power is not conserved or the device is not behaving as a transformer. - Verdict: read as "output current is not reduced by the turns ratio, so output power exceeds input power": Refuted by calculation (energy balance). The wording as written: ambiguous, no measurement attached. - Two facts weaken its use as evidence about Sweet [read: deep read pp.10-13]. First, the "1:5" in the caption is FB1 (48 turns, #28, a feedback winding) to P1 (240 turns, #20 bifilar), not drive to output. The drawing is a 120 V, 60 Hz VTA with 2000-turn excitation coils and a 2-15 V battery. Second, the 1986 notes show a different device: 250-turn excitation coil, 12-turn output coil, about 400 Hz, coils at "90 Degrees Uncoupled". The Guidelines never reconcile the two. The source and generation of the p.11 schematic are not stated. - Also added to the notation defects: "reflected power = -dPhi/dt" (p.21) has the dimensions of a voltage, not a power; p.4 attributes field decay to an "I^2 R factor"; p.9 defines velocity as "Speed * Distance Travelled". "l = Area" in emf = Bvl is confirmed as written on p.9. The "Quantum Automobile" (p.5) is a thought experiment with no measurement; the lamination "Bloch wall" experiment (p.7) is qualitative ("extreme reluctance"), with no force reading. They stay in the record as the author's description beside the standard physics.

Claim 3 wording. I accept the QE's point. "Not testable as stated" for the interpretation was too broad. The weak reading (the traces are explained by inductance) is Established. The strong reading ("friction on charge" seen as heat, p.17) is testable by steady-state heating at equal current: Never properly tested. What can be said is that this experiment cannot distinguish it from ordinary L/R behaviour. Verdict table updated.

Claim 4: conditions. The deep read of pp.38-42 changes what the "conditions" are. - The circuit Sykes prescribes in "Follow the following Circuits" (p.38): sine function generator, power amplifier, input coil L1, two bucking output coils L2 and L3 in series with a 10 ohm load, no diode. He says to run most devices at about 20-600 Hz (p.39) [read: page image p.38; text p.39]. - The diode and square-wave scheme (pp.40-42) is "a work in progress", and "NOTE: This section only appears to be valid under some configurations!" [read: p.41]. The figures are Meyer's patent drawing (Fig. 3-23) with its switching diode boxed, a photo of Sweet's "stud screw diodes" and a photo of Sweet's bench with an HP 211A square-wave generator. No diode circuit drawn by Sykes exists in the document. His mechanism (p.42) is that the input field builds, the DC is switched off, the collapse reverses polarity, that "switches on the Diode" and the output coils then conduct. "Choked off" means the primary's build-up is held back by the output side's reaction when there is no diode. No part number, voltage or measurement is given. Melnichenko gets a single sentence and no drawing. - Partzman (2020): whether his objection was to the p.38 circuit or the p.42 scheme is not established. Nothing in the document mentions him, and the PDF metadata date quoted in the deep read (26 Jan 2017) is before 2020. His objection stays "one forum source, not checked against a schematic". - QE item M2 added: p.31 "All my experiments to date point to loosely coupled devices working better than tightly coupled devices ... My best results have been from Loosely Coupled devices", and p.32 "I re-built this device and tightly coupled the Laminations, like a transformer ... I was not able to make the device work the same." This is one self-reported before/after comparison, with no measurement, no definition of "work", one build each, and a video referred to but not in the PDF. The device also had a split core "gapped by small plastic spacers" (p.32). Sykes uses "loose coupling" for two geometries (an air gap, and a separate secondary core as in Alek's device, p.33). It stays "Never properly tested" because no controlled comparison with measurements exists, but it is no longer correct to say no comparison exists. - QE item M3 added: "I seem to get a better result if the Input is resonated for example, LC Tank ... There seems to be a time constant in the Magnetic Fields of the Partnered Secondary Coils" (p.24) [read]. The statement is hedged and has no circuit or numbers. A resonant input stores energy that swings between L and C and can lower the current drawn from the source; as with the mathematician's resonant-loop tables, that changes current without creating power. Verdict: Plausible, untested as stated (as for the output-loop effect), net gain Refuted by calculation as before. - QE item M8: Rosenthal's "The VTA likes to always see a minimum load of 25 watts" (p.46) is one of the author's listed conditions. Never properly tested. QE item M9: the "swipe" lamp prediction (p.47): with a big coil, a magnet and a light load the lamp lights only after a delay set by the coil's impedance, but "in this technology" the lamp lights nearly immediately. It is a qualitative, testable prediction with no numbers. Never properly tested, and I note that the only test described is by eye. Both rest on the text layer only (see "Not done").

Claim 5: Sweet, and the others. - Stated input (M7). The transcription says, for 4 June: "Oscillator put 143 uA, 7 v of power at approximately 400 Hz 1 mW into the Excitation Coils ... 4.6 A x 24.2 V = 111.32 Watts - measured twice, observed by 3 persons. It seems to be affected by the quality of the Sine Wave, negatively by harmonic distortion" [read: lab-notes transcription p.30]. The "105,000 x - 111,000 x" gain follows from 1 mW. The 24 May entry says "Input power as measured on a digital Beckman multimeter purchased 1hr prior: ? 7.5v (3.1 on 200 uA scale)" and 19.1544 W / 0.023 mW [read: transcription pp.5-6]. So the notes themselves carry two input figures about 43 times apart. 7 V / 143 uA is about 49 kilohm, which is of the order of the notes' "hundreds of thousands of ohm" (p.17). The favourable details (witnessed by three people, measured twice, sensitivity to waveform quality, a run of 10 hours "with no noticeable heating", p.31) are kept in the record beside the metering gaps already listed (meter V x I at about 400 Hz, no phase, no waveform). - Drive source (M6). The notes say plainly: "Power supply was from a Wavetek Signal Generator, with line power, producing a sine wave, vs. Tom's battery supplied square wave (20w?) oscillator" [read: transcription p.5]. This replaces my earlier statement that the generator's mains draw could not be settled: the generator was line-powered. Its mains draw (about 15-25 W for the models I checked, model not named) was not metered, which supports the contrarian's point, while still being smaller than 111 W. The notes also mention a separate battery square-wave oscillator of about 20 W ("20w?") that nobody has discussed; it is a question mark in the notes, not a measurement. - Withheld component (M5). The notes (p.46) record: "Checks still not in - 13 days late - Sparky explains that there is a little part in the Generator that he has withheld from disclosure - A $1.50 part from Radio Shack ... that is not a capacitor (no resonance effects) that divides the current somehow into the circuit being parallel, that puts a small uA current into the Power Coil as well as the Excitation Coil ... (he calls this unit a 'Vector Oscillator')" [read: transcription]. Consequences: (i) a rebuild "as the notes describe" (my earlier "What would settle it") may lack a part, so a failed rebuild would not refute Sweet and a working one would not isolate the principle; (ii) the line is evidence of money still owed ("Checks still not in"), which belongs with the interested-party discussion (Sweet, Goldes, Roberts, Bearden); (iii) it adds a documented reason, from the notes, why the notes alone cannot reproduce the device (absence-of-evidence item 3). It is also a reason a sceptic can cite, namely that an undisclosed component in a metered circuit is exactly where an unmetered power source could sit. I record both readings; the notes give no further description of the part. - The 1986 device is not a bucking pair (new). The transcription describes one excitation coil (#18, 250 turns, 1.5 in) and one output coil (#12, 12 turns, 1.5 in), mounted at 90 degrees "so that the output Currents do not couple with the excitation coil" (pp.30-32), with two 4 x 6 x 1 in ceramic plates [read: lab-notes, via the deep read pp.10-13]. The Guidelines' thesis needs two opposed output coils carrying current and coupled to each other. The p.11 schematic does show bifilar P1/P2 with opposed arrows, but it is a later 60 Hz design of unknown generation. So the Guidelines' claim that Sweet's VTA is an instance of the "same principle" is, for the 1986 documented device, not shown by the sources I have. This does not change "Never properly tested", but it weakens the "shares one principle" half of Claim 5 for the 1986 data and strengthens "Not testable as stated" for the unification. Open question 8 added below. - Self-oscillation (M4). Sykes quotes Manning ("On rare occasions, Sweet saw this effect, called self-oscillation, occur in electric transformers") and Bearden ("Sweet was also a transformer designer and expert, and he remarked that he had also observed specialized self-oscillation in certain transformers") on p.12 [read]. These are second-hand accounts by promoters; no measurement. Kept as the stated origin of Sweet's idea. - Weight-reduction test (R2). Bearden's account, preserved on rexresearch.com, describes a load box of parallel 100 W lamps fed with "120 volt, negative AC sine-wave power at 60 Hz" and weight recorded per 100 W step [read: research/2026-10-04-rexresearch.com.md; I did not open the page]. It is a claimed anomalous experimental result from an interested party, with no data tables seen. Verdict: Never properly tested; it is kept as a claim that a weighing test could check cheaply. - Survey of other devices (pp.25-31, M10). For Moray, Meyer, Jensen, Koenig, Markov, Frolov, Melnichenko, Mishail1971, Utkin and Alek, Sykes shows a drawing or photo and one remark, and gives no number except two captions: Frolov's title ("A Toroidal Over-Unity Generator", with the figure equation sum of H = H2 + H3 = 0) and Alek's "EFFICIENCY >> 100%". Performance evidence in the document: Moray's quotation (a lamp lit after about 10 minutes of tuning), Meyer's remark at 38:16 of a video ("add more turns on the Charging Chokes"), and nothing else. Mishail1971's sketch is stamped 2011.11.01 (NE555, IRF640, 56 and 240 turns). The survey is evidence of resemblance as Sykes reads it, not of performance. Melnichenko: Sykes says "the direction of turns is not reversed but results are still possible" and attributes the large gap to avoid a "short Circuit ... as a result of the A Vector Spins" (p.29), which is his hypothesis with no calculation. Alek's and Hooper's names also appear: W. J. Hooper, whom Bearden calls "the only physicist who ever methodically investigated EM stresses ... in the laboratory" (p.19), is related work that no review followed up.

Claim 6, Claim 2. Unchanged. The p.22 claim "if Lenz's Law could be by passed, then every Generator on the planet could be a perpetual motion machine of the first kind" is Sykes's own wording of what bypass would mean; I note it beside his "Not Free Energy" (p.48), as a further inconsistency inside the document.

Retention: material the QE found missing, now in the record (R1)

Positive and favourable claims, labelled as interested-party characterisations, none with data seen: - Sykes, aboveunity.com 2025: "Marcel [FringeIdeas] has succeeded, but now has the refinement phase", and "I have shown you how to get an output, how to Generate Energy in excess of the Input"; in the same post, "we have not succeeded, in general" [read: research/2026-10-04-aboveunity.com.md]. - The forum's own "Our Above Unity Machines" list names Baerndorfer's Experiment, Captainloz's A-R-E and Aetherholic's COP>4.0 among others, and the 2019 thread is tagged ABOVE-UNITY (355 comments, 41K views) [read: same file]. No power figures from these were found. - These sit beside the negative replications (dewetw at most 0.82 COP, patrick1 "isnt OU", Itsu "failed" in Sykes's account). The asymmetry the QE noted is corrected: positive and negative informal reports both exist, and neither set contains calibrated power-in/power-out data. - Related-work leads (R3), one line each: the Wiegand-wire field-cooling analogue (US3820090) and other industrial analogues in Anor's post; LordVeber's US4210859A; George I. Cohn on induction, which Sykes cites on p.9; and the later lineage Sykes himself cites (Kapanadze, Akula, Don Smith, Gunderson, MEG "asymmetrical regauging"). The patents and people were not checked by me. - Feynman: the quotation is on p.20, not p.21, and the "Vol. II ch.15" reference comes from Sykes's hyiq.org page, not the Guidelines [read: QE M11, research/2026-10-04-hyiq.org.md]. The Guidelines write "Sir Richard Feynman"; Feynman was not knighted [memory, high]. - Sykes's p.34 statement "For every one device that works, expect 10 devices that don't" and "I have never seen an experiment that has been done several times show incorrect information" are the author's own, kept; the first is the basis for the unfalsifiability point under Claim 4.

Additional settling experiments from the deep reads

• Claim 4 conditions: a bench A/B of the p.38 circuit with and without a diode in the Meyer/Sweet arrangement, same coils, equal metered input power. The loose-versus-tight comparison could be repeated with a gap spacer thickness and measured k.
• Claim 1: the step-up rule is settled by measuring primary and secondary current of any two-winding transformer built to 1:5; the ideal result (primary 5 times secondary) is the prediction.
• Claim 5: obtain the scans of the lab notes (to resolve the two input figures and the missing pages) and Sweet's "Vector Oscillator" description before attempting a rebuild.

Open questions added

8. Does the p.11 schematic (EX 2000 T, P 240 T bifilar, FB 48 T, 120 V 60 Hz) belong to a documented Sweet generation, and where did it come from? Which device did the 1986 witnesses actually see (single output coil at 90 degrees, or a bucking pair)?
9. What did the "Vector Oscillator" component do, and was it metered?
10. What exactly did Sykes observe when the loosely coupled device stopped "working the same" when tightly coupled, and is the video available?
11. pp.45-49 (rotor, swipe lamp, alternator rig) still need page-image deep reads before the Claim 2 and Claim 4 statements about them are treated as fully checked.

Net effect on the headline verdicts

C1 weak Plausible, untested / strong Refuted by calculation (unchanged; the p.11 step-up reading added, Refuted by calculation). C2 unchanged. C3 observation Established; strong reading of the interpretation Never properly tested (reworded from Not testable as stated); duty factor Never properly tested. C4 unchanged in rank: alternator Established, net gain Refuted by calculation and never neutrally measured, conditions Never properly tested (diode not part of the beginner circuit), Bloch wall and knee Not testable as stated. C5 Never properly tested; unification Not testable as stated, and weaker for the 1986 data. C6 unchanged.