Project 33

Discrepancy register

Two parts of the project disagreeing is information. Each entry names both sides, the claims it affects, and the next action. Resolved entries stay, with what resolved them; a record correction is not an engineering review.

Status	Effect	Count
OPEN	Contradicts a claim or requirement	5
OPEN	Informational	4
OPEN	Weakens a claim’s support	8
RESOLVED	Contradicts a claim or requirement	2
RESOLVED	Informational	3
RESOLVED	Weakens a claim’s support	2

D-001

Historical C1 note contradicts its own sweep

Field	Value
Status	RESOLVED
Effect	Contradicts a claim or requirement
Kind	claim vs artifact — What a claim or note says differs from what its artifact contains.
Claims	C1

• evidence/archive/notes-v1/C1_four_bar.md: Baseline is a change-point linkage with a 78.46-101.54 deg transmission angle and confirmed over-center locking.
• evidence/archive/C1_four_bar_sweep.v1.json: Baseline row: non-Grashof, 90.0/90.0 deg, envelope 0.0; every one of 81 rows is identical.

Next action. None for the record; the design question continues as D-002.

Resolution. The note is archived as superseded and the claim no longer repeats its figures. (Record correction in the evidence-discipline change; not an engineering review.)

Found by: Reading the committed JSON against the note during the evidence review of 2026-09-28.

D-002

The specified four-bar cannot be assembled

Field	Value
Status	OPEN
Effect	Contradicts a claim or requirement
Kind	requirement vs model — A model result does not meet a stated requirement.
Claims	C1

• evidence/archive/notes-v1/C1_four_bar.md: Ground 25, coupler 60, input 15, output 15 mm deploys the fin 0-92 deg.
• docs/EVIDENCE/C1_four_bar_sweep.json: Over the 0-92 deg drive range the B-D distance never exceeds 29.6 mm (40 mm at any angle), but the coupler and output circles only meet at 45 mm or more; no pose closes. None of the 81 variants closes at any pose.

Next action. A human confirms the intended topology and link dimensions; then the model is re-run and reviewed.

Found by: Loop-closure check (circle intersection) added in model revision 2.

D-003

C1 model revision 1 reported placeholders as results, and its tests passed on them

Field	Value
Status	RESOLVED
Effect	Contradicts a claim or requirement
Kind	implementation defect — Software produced values that do not follow from its inputs.
Claims	C1

• evidence/archive/C1_four_bar_sweep.v1.json: transmission_angle() returned 90.0 for poses that cannot close; an empty trajectory produced an envelope of 0.0; the Freudenstein constants were misassigned.
• evidence/archive/test_four_bar.v1.py.txt: Tests asserted 40-140 deg and a bounded envelope, and passed because the placeholders sat inside those bounds.

Next action. None; keep revision 1 archived as a worked example of a vacuous pass.

Resolution. Revision 2 returns null for unassemblable poses and solves closure by circle intersection; tests check geometry and null handling. (Model and test correction in the evidence-discipline change; software-tested, not engineering-reviewed.)

Found by: Comparing compute_theta4 with direct circle intersection for a linkage that does close.

D-004

Over-center locking is claimed but not modeled

Field	Value
Status	OPEN
Effect	Contradicts a claim or requirement
Kind	claim vs artifact — What a claim or note says differs from what its artifact contains.
Claims	C1

• evidence/archive/notes-v1/C1_four_bar.md: Over-center detection: confirmed (moment sign flips at toggle).
• mechanism/four_bar.py: over_center_moment_sign() compares the drive angle with 92 deg and ignores every link length.

Next action. Model the locking moment for an assemblable linkage, or state that locking is a physical-test question only.

Found by: Reading the helper’s implementation.

D-005

C1 CAD artifacts are illustrative bars, not the modeled mechanism

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	simulation vs cad — Simulation or model geometry disagrees with CAD geometry.
Claims	C1

• CAD Files/FusionScripts/Project33FourBar/Project33FourBar.py: The Fusion sketch draws the coupler from (0, 15) to (25, 15) mm, 25 mm long, and dimensions it as drawn; the CadQuery export places four separate bars and calls the assembly purely illustrative.
• docs/EVIDENCE/C1_four_bar_sweep.json: The modeled coupler is 60 mm and the loop cannot close.

Next action. Regenerate CAD only after D-002 is resolved; never cite the STEP as evidence the mechanism assembles.

Found by: Reading the CAD script coordinates against the model inputs.

D-006

Reference and textbook fin-term formulations disagree about the static-margin window

Field	Value
Status	OPEN
Effect	Contradicts a claim or requirement
Kind	model vs model — Two models of the same quantity disagree.
Claims	C5

• docs/EVIDENCE/C5_static_margin_inputs.json: Fin CP sweep term uses the root chord; margin 1.55-1.57 cal (inside 1.5-2.0).
• docs/EVIDENCE/C5_barrowman_crosscheck.json: Textbook Barrowman sweep term (tip leading-edge sweep X_R) and mid-chord line length give 0.74-0.76 cal (outside 1.5-2.0).

Next action. A human reviews both formulations against Barrowman (1967) or OpenRocket’s component CP output.

Found by: Independent re-derivation of the fin terms during the 2026-09-28 review.

D-007

The static-margin script burns a different motor than the OpenRocket model

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	model vs input — A model’s inputs differ from the design source it claims to use.
Claims	C5

• Simulation/Folding Stabilized Rocket.ork: Motor configuration: AeroTech G64W (29 mm reload).
• Simulation/static_margin.py: Hard-coded Estes C6-5: 24 g wet, 11 g dry, 1.8 s burn, CG at 0.520 m.

Next action. Use the OpenRocket motor (or justify the substitute) and record the motor mass data source.

Found by: Parsing the motor designation from the .ork file.

D-008

Static-margin inputs omit or misread parts of the OpenRocket model

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	model vs input — A model’s inputs differ from the design source it claims to use.
Claims	C5

• Simulation/Folding Stabilized Rocket.ork: Ellipsoid nose cone with a 0.067 kg override mass.
• Simulation/static_margin.py: Uses the ogive nose CP coefficient (0.466) and sums only body-tube override masses (1.185 kg).

Next action. Reviewer decides whether the omissions matter at the required precision.

Found by: Comparing parse_ork() with the .ork XML.

D-009

Model inputs were revised until paper targets were met

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	model tuned to target — Model inputs were revised until a target was met.
Claims	C5, C6, C8

• docs/EVIDENCE/CHANGELOG.md: C5 fins, C6 design point, and C8 fin mass were changed so results match the paper; entries are marked RESOLVED.
• evidence/archive/C5_openrocket_pre_retune.ork.xml: Before the change the OpenRocket fins had a 20 mm root chord; the reference script then gave 0.08-0.10 cal.

Next action. Record an independent source for each revised input (drawing, datasheet, measurement), or keep the claims open.

Found by: Reading the divergence log and the preserved pre-change model.

D-010

No shared design point across the mass, spring, reliability, and structural models

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	model vs model — Two models of the same quantity disagree.
Claims	C4, C6, C7, C8

• materials/parts.yaml: Fin volume 3.4 cm^3 each: about 5.3 g in CFRP; four listed parts total 98.1 g.
• docs/EVIDENCE/C6_design_point.json: 154 g fin, 120 mm chord, 180 mm span, 205 m/s egress.
• docs/EVIDENCE/C7_run_record.json: 15 g fin, 60 mm, 70 m/s egress.
• docs/EVIDENCE/C5_static_margin_inputs.json: Body-tube dry mass 1.185 kg.

Next action. Define one design point in a single reviewed input file that every model reads.

Found by: Cross-reading model inputs during the 2026-09-28 review.

D-011

Calculated spring margin is below the stated requirement

Field	Value
Status	OPEN
Effect	Contradicts a claim or requirement
Kind	requirement vs model — A model result does not meet a stated requirement.
Claims	C6

• evidence/archive/notes-v1/C6_spring_sizing.md: Requirement: at least 20% margin.
• docs/EVIDENCE/C6_design_point.json: Margin 0.189936 (19.0%).

Next action. A human changes the spring, the requirement, or the design point, and records why.

Found by: Requirement R-C6-MARGIN evaluated from the design-point record.

D-012

Test named for a 20% margin did not enforce it

Field	Value
Status	RESOLVED
Effect	Weakens a claim’s support
Kind	test vs requirement — A test does not check what its name or claim says.
Claims	C6

• evidence/archive/test_spring_sizing.v1.py.txt: test_design_point_has_20_percent_margin asserted k_required <= 1.2 * k_supplied, which accepts negative margins.
• tests/test_spring_sizing.py: Margin arithmetic is tested directly; the requirement is evaluated by the evidence record.

Next action. None.

Resolution. Test replaced; revision 1 archived. (Test correction in the evidence-discipline change.)

Found by: Reading the assertion against the test name.

D-013

The reliability simulation cannot produce the failures it counts

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	implementation defect — Software produced values that do not follow from its inputs.
Claims	C7

• Simulation/reliability_sweep.py: omega_final is forced to 0 when the toggle event fires; k (Nm/rad) is applied as a constant torque (Nm); latency shifts start time only.
• docs/EVIDENCE/C7_run_record.json: 10,000 of 10,000 trials succeed.

Next action. Rebuild the simulation so each failure mode it counts can occur, at the shared design point (D-010).

Found by: Reading the success criterion against the integrator’s event handling.

D-014

The historical 10,000-trial summary could not be reconstructed from its CSV

Field	Value
Status	RESOLVED
Effect	Weakens a claim’s support
Kind	claim vs artifact — What a claim or note says differs from what its artifact contains.
Claims	C7

• evidence/archive/notes-v1/C7_reliability.md: Trials 10000, successes 10000.
• docs/EVIDENCE/C7_reliability.csv: Header only: a failure-only export cannot state how many trials ran.

Next action. None.

Resolution. The simulation now writes a run record with trials, successes, failures, and seed; consistency checks K-C7-DENOMINATOR and K-C7-FAILURE-ROWS tie it to the CSV. (Record correction; reproduction is checked by CI.)

Found by: Evidence Observatory row count (0 data rows).

D-015

Laminate modulus is below the aluminium comparison

Field	Value
Status	OPEN
Effect	Contradicts a claim or requirement
Kind	requirement vs model — A model result does not meet a stated requirement.
Claims	C8

• evidence/archive/notes-v1/C8_clt.md: Claim: match or exceed aluminium stiffness; relies on increased section thickness.
• docs/EVIDENCE/C8_clt.csv: Ex = 56.68 GPa, below 69.0 GPa.

Next action. Calculate equal-mass section bending stiffness (R-C8-COMPONENT-STIFFNESS) or restate the claim.

Found by: Requirement R-C8-MODULUS evaluated from the CLT output.

D-016

Architecture diagram omitted the rocket’s fins-deployed ignition gate

Field	Value
Status	RESOLVED
Effect	Informational
Kind	documentation vs implementation — Documentation disagrees with firmware or software behavior.
Claims	C2

• docs/ARCHITECTURE.md: ARMED –> IGNITING on the IGNITE command.
• Firmware/Rocket/src/main.cpp: IGNITE is refused unless is_fins_deployed(); the refusal was printed to USB serial only and never reached the telemetry record.

Next action. Bench-verify that the refusal now reaches the dashboard CSV.

Resolution. Diagram updated; the rocket now reports CMD_REJECT:ignite_fins_not_deployed and CMD_REJECT:ignite_not_armed over UART, which the launcher already forwards. Compile-verified only. (Documentation and firmware logging change; not bench-verified.)

Found by: Reading the IGNITE branch in processSerialCommands().

D-017

Ignition requires deployed fins, but deployment requires a boost-level acceleration

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	design logic — Two implemented rules cannot both be satisfied in the intended sequence.
Claims	C2

• Firmware/Rocket/src/main.cpp: Deploy fires only after a >15 g x-axis reading while not IDLE; IGNITE is refused unless the deploy flag is set.
• Firmware/Launcher/src/main.cpp: After sending IGNITE the launcher waits 5 s for IGNITED, then aborts.

Next action. A human decides the intended ordering. On the bench this ordering acts as an extra interlock; do not remove it without a safety review.

Found by: Tracing the nominal command sequence through both firmwares.

D-018

Live telemetry timestamps are launcher relay times, not rocket sample times

Field	Value
Status	OPEN
Effect	Informational
Kind	telemetry vs expected — Recorded telemetry disagrees with its documented meaning or rate.
Claims	C2

• Firmware/Launcher/src/main.cpp: T packets are built as T,<launcher millis()>,… when a DATA line is relayed.
• docs/WIRING.md: T,,... without saying whose clock.

Next action. Proposed protocol change: carry rocket millis, a sequence number, and a boot identifier in DATA. Needs a bench-tested firmware revision.

Progress. The audit labels T as launcher_relay_millis and LOG as rocket_millis, and the docs state each clock domain.

Found by: Reading processSerial2() in the launcher.

D-019

Launcher reported 0.0 m altitude when the barometer was missing

Field	Value
Status	RESOLVED
Effect	Informational
Kind	documentation vs implementation — Documentation disagrees with firmware or software behavior.
Claims	none

• Firmware/Launcher/src/main.cpp: filteredAlt was initialized to 0.0 and sent unchanged when bmpHealthy was false.
• docs/WIRING.md: ENV altitude field documented as altitude, with no missing-value convention.

Next action. Bench-verify that ENV carries nan with the barometer disconnected.

Resolution. The launcher sends nan for altitude when the barometer is unavailable; the dashboard shows ‘not reported’. Compile-verified only. (Firmware change; not bench-verified.)

Found by: Unknown-is-not-zero audit of telemetry fields.

D-020

PID comparison fragmented windows, merged streams, and zero-filled bad samples

Field	Value
Status	RESOLVED
Effect	Informational
Kind	implementation defect — Software produced values that do not follow from its inputs.
Claims	none

• docs/PID_TUNING.md: Samples are grouped by PID setting.
• Firmware/analyze_pid.py: A new window started at every STATUS packet (one per telemetry packet); LOG and T rows shared windows; unparseable values became 0.0.

Next action. None.

Resolution. Windows change only when gains change; T and LOG are summarized separately; invalid samples are counted and excluded. (Software change with regression tests.)

Found by: Tracing the launcher’s T-then-STATUS packet order through the analyzer.

D-021

Bill of materials does not match the actuators in firmware and CAD

Field	Value
Status	OPEN
Effect	Informational
Kind	bom vs design — The bill of materials disagrees with the design or firmware.
Claims	none

• materials/bom.csv: Five SG90-class servos.
• Firmware/Rocket/src/main.cpp: Rocket drives five servos (four canards plus ignition); the launcher drives a sixth. The folding-rocket CAD archive references an MG996R.

Next action. Reconcile the BOM with an as-built inventory; record purchased prices where known.

Found by: Counting Servo.attach() calls against the BOM.

D-022

The C4 baseline is not the printed prototype

Field	Value
Status	OPEN
Effect	Weakens a claim’s support
Kind	model vs input — A model’s inputs differ from the design source it claims to use.
Claims	C4

• materials/parts.yaml: Baseline body tube and fins in 7075-T6.
• docs/CAD_ASSEMBLIES.md: Prototype airframe and folding-fin structure in PLA+ or PETG.

Next action. State which assembly the 30-40% claim compares, and make the baseline match it.

Found by: Comparing material tables.

D-023

OpenRocket screenshots have no recorded model revision

Field	Value
Status	OPEN
Effect	Informational
Kind	provenance gap — An artifact cannot be traced to the model revision that produced it.
Claims	C5

• Simulation/Stability_Graph.png: An OpenRocket stability plot with no settings, date, or model hash.
• evidence/archive/C5_openrocket_pre_retune.ork.xml: The fins changed substantially in a later revision; the plot may predate it.

Next action. Re-export from the current .ork with its hash and settings recorded, or label the images historical.

Found by: Provenance review of committed images.

D-024

The deploy bench sketch measures a timer with the same timer

Field	Value
Status	OPEN
Effect	Informational
Kind	measurement method — A measurement procedure cannot establish what it is meant to measure.
Claims	C2

• Firmware/Rocket/tests/bench_deploy_isr.ino: Elapsed time from esp_timer_get_time(), polled every 10 us until the callback’s flag changes.
• docs/EVIDENCE/C2_latency.csv: Same column layout, synthetic values.

Next action. Use an external instrument on GPIO 33 (P-001). Sketch output can only describe callback-to-poll latency.

Found by: Reading the sketch against requirement R-C2-JITTER.

Failure archive

Superseded models, rejected tests, and historical claims, kept byte for byte. The digest in evidence/archive/index.json makes any edit detectable.

File	Kind	Originally	Why archived	Discrepancies
evidence/archive/C1_four_bar_sweep.v1.json	superseded model output	docs/EVIDENCE/C1_four_bar_sweep.json	Placeholder values (90 deg, 0.0 envelope) reported for linkages that cannot close.	D-001, D-003
evidence/archive/C5_openrocket_pre_retune.ork.xml	prior model	scratch/ork_unzip/rocket.ork	OpenRocket model before the fins were revised to meet the paper’s margin target.	D-009, D-023
evidence/archive/notes-v1/C1_four_bar.md	historical claim	docs/EVIDENCE/C1_four_bar.md	Statements contradicted by its own sweep.	D-001, D-004
evidence/archive/notes-v1/C2_latency.md	historical claim	docs/EVIDENCE/C2_latency.md	Replaced by the generated claim passport.	—
evidence/archive/notes-v1/C3_naca_sweep.md	historical claim	docs/EVIDENCE/C3_naca_sweep.md	Replaced by the generated claim passport.	—
evidence/archive/notes-v1/C4_delta.md	historical claim	docs/EVIDENCE/C4_delta.md	Replaced by the generated claim passport.	D-010, D-022
evidence/archive/notes-v1/C5_static_margin.md	historical claim	docs/EVIDENCE/C5_static_margin.md	Generated by the model script with a ‘confirms’ verdict; replaced by the passport.	D-006, D-007
evidence/archive/notes-v1/C6_spring_sizing.md	historical claim	docs/EVIDENCE/C6_spring_sizing.md	Replaced by the generated claim passport.	D-011
evidence/archive/notes-v1/C7_reliability.md	historical claim	docs/EVIDENCE/C7_reliability.md	Summary not reconstructable from its CSV; replaced by the run record and passport.	D-014
evidence/archive/notes-v1/C8_clt.md	historical claim	docs/EVIDENCE/C8_clt.md	Replaced by the generated claim passport.	D-015
evidence/archive/notes-v1/C8_hinge_fos.md	historical claim	docs/EVIDENCE/C8_hinge_fos.md	Model output is now structured JSON; replaced by the passport.	—
evidence/archive/test_four_bar.v1.py.txt	rejected test	tests/test_four_bar.py	Passed vacuously on placeholder values.	D-003
evidence/archive/test_spring_sizing.v1.py.txt	rejected test	tests/test_spring_sizing.py	Named for a 20% margin but accepted negative margins.	D-012

This site is open source. Improve this page.