Systematic Debugging & Pre-Compliance

01 / 10

Changing everything can erase the cause

The node loses range intermittently. Should you replace the antenna, update the firmware and change the supply together?

The fictional 2.450 GHz condition-monitoring node works on an open bench and drops reports when mounted. One engineer suspects antenna detuning. Another sees a low conducted-power reading. A third proposes new firmware. All three changes might make the symptom disappear, but the result would not explain which change mattered or whether a second fault remains.

Think about itIf all three changes restore the link, what evidence have you lost?

Answer

You lose the controlled contrasts: same firmware with a different supply; same supply with a different mount; same mechanical state with a verified RF command. You may also remove the only reproducible failure configuration. Keep the original specimen, firmware, raw records and a reproducible operating script before intervention.

Debugging starts with a question that can be falsified. “Bad RF” cannot be tested. “At fixed command 7, an independently corrected on-time mean at R1-TX remains below the internal minimum” can. It separates a conducted-output symptom from antenna efficiency, link geometry and the analyzer display.

Three statements with different evidence burdens
Statement	What is needed next
Symptom: reports disappear under mounting D	Freeze route, traffic, report deadline, receiver state, observation window and environment.
Requirement: conducted mean below 19.5 dBm	Qualified reference plane, correct gate/statistic, calibration and declared uncertainty rule.
Cause: supply droop reduces PA output	Direct supply/RF contrast at unchanged actual command, followed by safe mechanism-specific confirmation.

This capstone uses the methods from 08.1 uncertainty, 08.2 lab integrity, 08.3 spectrum, 08.4 VNA and 08.5 modulated TX/RX tests. All case evidence is Illustrative / Derived, frozen on 2026-09-09. “Verified” here identifies supplied fictional evidence, never a physical inspection or certification.

Common misconceptionThe last changed part must have caused the failure.

A change can alter supply impedance, shielding, coupling, firmware state or the measurement setup at once. Preserve the symptom and compare mechanisms before choosing a correction.

Engineering decision → record update

Open M08-06-HANDOFF-D. Freeze HW-D1/FW-D1, the mounting, source script and raw acquisition IDs. Keep the apparent failure as evidence even when its source may be the measurement system.

02 / 10

Freeze the symptom, configuration, and raw record

Which observations are usable before any fault is eliminated?

The audit bundle deliberately combines different method witnesses and defective setups. It is not one consistent instrument session. A corrected number needs its raw parent, exact configuration and transformation. A repeat needs a new acquisition. A different processing view of the same data gives a new question, not independent evidence.

Use four separate labels: the symptom, the requirement decision, the interpretation, and measurement eligibility. An invalid VNA comparison stays visible as a setup finding; it cannot eliminate an antenna mechanism. Calibration corrects a characterized model and leaves residual error, drift and random effects. [VNA-ERROR]

Initial flawed audit bundle and explicitly repaired evidence
Immutable record	Original observation / problem	Eligible inference / next action
PWR-CAL-A	SYN-IND-01: 18.050 dBm; n=16, s=.120 dB.	Upper endpoint 19.660416425819 < 19.700 dBm internal screen. With CAL/CABLE correlation .5, U=.407308237088 and endpoint 19.707308237088: inconclusive.
TX-DEBUG-RAW-D	M-PWR-D indication 16.000 dBm; supply 2.900 V in event window; command 7 unchanged.	Indication belongs at M-PWR-D, not R1-TX.
TX-DEBUG-D	16.000 dBm at M-PWR-D, retained.	Even upper endpoint 17.36 < 19.5 dBm minimum. Supports local deficit; source of deficit remains P or F.
SPEC-ARTIFACT-D	Input-referred internal product −110 dBm at 10 dB attenuation; −130 dBm at 20 dB.	Setup artifact is consistent with the controlled model. A DUT spur need not follow that slope.
VNA-BAD-D	Unused Γv=−.2; corrected estimate −.2191235059760957.	DUT trace is ineligible for E4. A smooth fit does not remove wrong model bias.
VNA-REPEAT-D	At 2.450 GHz reference Γ=.10+j0; mounted Γ=.35+j0. Three untouched-cable pairs: Δ\|Γ\|=.250,.248,.252.	Supplied local shift rule: every Δ\|Γ\|−.02 > .10. E4=true is a reviewed synthetic signature, not a radiation-efficiency result.
EVM-PROC-D	Four unit-energy QPSK symbols with exactly common +4° rotation: RMS EVM 6.979899340500%.	Two processing views are not two independent repeats and do not establish a physical repair.
PER-COUNT-D	0 errors / 1000 externally counted trials; no retries; fixed 100 ms cycle, 100 s horizon.	Cannot support .001 target. Count is descriptive if IID or frame identity is unknown; not E5 for a different mode.
LEAK-CONTROL-D	Source −40 dBm; wanted path 70 dB → −110 dBm. Bypass isolation 90 dB → −130 dBm.	Initial flawed audit cannot enter the clean Boolean baseline.
ISO-D2	Synthetic reviewed isolation 112 dB; source −40 dBm gives −152 dBm bypass; upper bound −150 dBm includes supplied 2 dB bound.	E2=false only under ISO-D2 local ≥110 dB criterion, not absence of visible tone.
STATE-TIME-D	RF 10.000 ms; supply logger 10.300 ms; firmware logger 9.800 ms. Supply 3.300→2.900 V and RF 20.0→17.0 dBm; command stays 7.	Same-time signatures support correlation; a 50 µs difference cannot establish precedence with overlapping ±40 µs bounds.

Read raw values, corrections, conditions and repairs for every record

PWR-CAL-A

PWR-CAL-A · 08.1 method/bench reference TX-PWR-A; not DEBUG-QPSK-D
Record field	Supplied evidence
Parents	M08-01-RAW-A · M08-01-DERIVED-A
Raw / observed	SYN-IND-01: 18.050 dBm; n=16, s=.120 dB.
Processing / correction	+1.200 dB cable + .050 dB sensor residual = 19.300 dBm. Internal sensor correction is already in indication.
Validity conditions	R1-TX on-time mean; 2.450 GHz; real 50 Ω; 2.0–23.6 ms gate. Independent variance .032475 dB²; uc=.180208212909 dB; k=2.
Supported decision	Upper endpoint 19.660416425819 < 19.700 dBm internal screen. With CAL/CABLE correlation .5, U=.407308237088 and endpoint 19.707308237088: inconclusive.
Repair / next evidence	Preserve the independent and correlated versions. Neither budget is silently transferred to debug-mode TX.

TX-DEBUG-RAW-D

TX-DEBUG-RAW-D · DEBUG-QPSK-D / CFG-AUDIT-D1
Record field	Supplied evidence
Parents	None; new raw acquisition
Raw / observed	M-PWR-D indication 16.000 dBm; supply 2.900 V in event window; command 7 unchanged.
Processing / correction	Raw gated mean only; no external plane correction yet.
Validity conditions	SYN-SENSOR-D/firmware SIM-2; linear-power gate 2.0–23.6 ms over 16 bursts at 1 s cadence; RF passband ±100 kHz. No added pad; reviewed fictional sensor load.
Supported decision	Indication belongs at M-PWR-D, not R1-TX.
Repair / next evidence	CAL-PWR-D2 and CABLE-D2 independently characterize +1.000 dB path and supply a separate U=.36 dB (k=2, approximate coverage).

TX-DEBUG-D

TX-DEBUG-D · DEBUG-QPSK-D / fixed command 7
Record field	Supplied evidence
Parents	TX-DEBUG-RAW-D
Raw / observed	16.000 dBm at M-PWR-D, retained.
Processing / correction	16.000 + 1.000 dB cable = 17.000 dBm at R1-TX. Sensor internal calibration is already included; no second correction.
Validity conditions	CAL-PWR-D2, F-D2 stable cable, 298.15 K. Supplied debug-mode U=.36 dB includes cable, sensor, repeatability and loading jointly; its component budget is not supplied.
Supported decision	Even upper endpoint 17.36 < 19.5 dBm minimum. Supports local deficit; source of deficit remains P or F.
Repair / next evidence	E3 is a supplied synthetic review of this exact requirement/statistic. Acquire STATE-TIME-D and controlled supply confirmation; real budget remains a handoff gap.

SPEC-ARTIFACT-D

SPEC-ARTIFACT-D · 08.3 cubic instrument fixture; continuous two tones, not debug QPSK
Record field	Supplied evidence
Parents	SYN-IM3-D-A10 · SYN-IM3-D-A20
Raw / observed	Input-referred internal product −110 dBm at 10 dB attenuation; −130 dBm at 20 dB.
Processing / correction	Internal input correction included once; −2 dB/dB attenuation slope in valid cubic region, ≥10 dB model reserve.
Validity conditions	Fixed external input tones and gain; no automatic attenuation/preamp coupling, clipping or floor masking.
Supported decision	Setup artifact is consistent with the controlled model. A DUT spur need not follow that slope.
Repair / next evidence	Independent source/DUT spectra and valid analyzer checks remain separate. Never remove a physical DUT fault from this unrelated trace.

VNA-BAD-D

VNA-BAD-D · 08.4 wrong-standard-definition audit / CFG-AUDIT-D1
Record field	Supplied evidence
Parents	SYN-VNA-WRONG-LOAD-RAW-v1
Raw / observed	Unused Γv=−.2; corrected estimate −.2191235059760957.
Processing / correction	Wrong load: actual Γ=.02, assumed zero. Local residual = .0191235059760957 > .01.
Validity conditions	M-CAL real 50 Ω; 2–3 GHz, 101 points, −10 dBm CW, IFBW 1 kHz; unused reference never fitted.
Supported decision	DUT trace is ineligible for E4. A smooth fit does not remove wrong model bias.
Repair / next evidence	Retain this raw ID. New CAL-VNA-D2 verifies Γv=−.2 with residual 0 in the exact fixture; then repeat mounted/reference pair.

VNA-REPEAT-D

VNA-REPEAT-D · New acquisition ACQ-VNA-D2 / CFG-DEBUG-CLEAN-D2
Record field	Supplied evidence
Parents	VNA-BAD-D
Raw / observed	At 2.450 GHz reference Γ=.10+j0; mounted Γ=.35+j0. Three untouched-cable pairs: Δ\|Γ\|=.250,.248,.252.
Processing / correction	Correct standard definitions; CAL-VNA-D2 at R2. Max unused-reference residual 0 in exact fixture; supplied comparison bound ±.02.
Validity conditions	Same antenna, calibrated cable F-D2, source −10 dBm, IFBW 1 kHz, bias off for CW comparison, 298.15 K; mounting is the sole intended contrast.
Supported decision	Supplied local shift rule: every Δ\|Γ\|−.02 > .10. E4=true is a reviewed synthetic signature, not a radiation-efficiency result.
Repair / next evidence	Separate S0 controlled comparison needed. This record does not retroactively make VNA-BAD-D eligible.

EVM-PROC-D

EVM-PROC-D · 08.5 four-symbol algebra; separate from TX-DEBUG-D
Record field	Supplied evidence
Parents	PHASE-RAW-D
Raw / observed	Four unit-energy QPSK symbols with exactly common +4° rotation: RMS EVM 6.979899340500%.
Processing / correction	Ideal known-phase inverse gives zero. PHASE-RAW-D and PHASE-CORRECTED-D share one acquisition.
Validity conditions	R3 symbol decisions, reference-energy normalization, no noise/estimator error. e^(+jωt).
Supported decision	Two processing views are not two independent repeats and do not establish a physical repair.
Repair / next evidence	Compare identical processing contracts before interpreting changes; preserve raw symbols.

PER-COUNT-D

PER-COUNT-D · 08.5 fixed-n packet method witness
Record field	Supplied evidence
Parents	RX-COUNT-C
Raw / observed	0 errors / 1000 externally counted trials; no retries; fixed 100 ms cycle, 100 s horizon.
Processing / correction	One-sided 95% exact zero-error upper bound pU=1−.05^(1/1000)=.002991249545095.
Validity conditions	IID Bernoulli assumptions supplied only for this fixed state and population. R1-RX calibrated level, R3 CRC/timeout decision; no optional stopping.
Supported decision	Cannot support .001 target. Count is descriptive if IID or frame identity is unknown; not E5 for a different mode.
Repair / next evidence	A predeclared 2995-trial zero-error plan could support .001. New frame/mode means a new count record.

LEAK-CONTROL-D

LEAK-CONTROL-D · Continuous-tone isolation witness / CFG-AUDIT-D1
Record field	Supplied evidence
Parents	None; new raw acquisition
Raw / observed	Source −40 dBm; wanted path 70 dB → −110 dBm. Bypass isolation 90 dB → −130 dBm.
Processing / correction	Coherent voltage ratio .1; phase-dependent total offset −.915149811214…+.827853703165 dB, not incoherent power sum.
Validity conditions	R1-RX real matched 50 Ω. Unknown search floor/orientation means no visible tone is not E2=false.
Supported decision	Initial flawed audit cannot enter the clean Boolean baseline.
Repair / next evidence	ISO-D2: re-route/shield, independently verify ≥110 dB isolation (≤−150 dBm bypass), search bound ≤−153 dBm and coherent error <.1 dB; recalibrate wanted path, sources on/off, terminated paths.

ISO-D2

ISO-D2 · ACQ-ISO-D2 / CFG-DEBUG-CLEAN-D2
Record field	Supplied evidence
Parents	LEAK-CONTROL-D
Raw / observed	Synthetic reviewed isolation 112 dB; source −40 dBm gives −152 dBm bypass; upper bound −150 dBm includes supplied 2 dB bound.
Processing / correction	Wanted −110 dBm remains calibrated. Worst allowed voltage ratio .01; magnitude error ≤.087296108050 dB.
Validity conditions	Calibrated reference receiver verifies floor ≤−153 dBm at 10 Hz measurement bandwidth; source off/on and all alternate path terminations recorded.
Supported decision	E2=false only under ISO-D2 local ≥110 dB criterion, not absence of visible tone.
Repair / next evidence	Repeat after a cable, shield, source, fixture or firmware-state change; old leaking raw record remains.

STATE-TIME-D

STATE-TIME-D · Deterministic timing witness, 30 ms horizon
Record field	Supplied evidence
Parents	LOG-SUPPLY-RAW-D · LOG-FW-RAW-D · RF-RAW-D
Raw / observed	RF 10.000 ms; supply logger 10.300 ms; firmware logger 9.800 ms. Supply 3.300→2.900 V and RF 20.0→17.0 dBm; command stays 7.
Processing / correction	Offsets +.300/−.200 ms removed; corrected events all 10.000 ms with ±.040 ms bounds including residual latency.
Validity conditions	Local pretrigger 2 ms/posttrigger 4 ms around event; independent timestamp offset calibration TIME-D2, 20 MHz supply capture bandwidth. Unaligned logs are ineligible.
Supported decision	Same-time signatures support correlation; a 50 µs difference cannot establish precedence with overlapping ±40 µs bounds.
Repair / next evidence	Safe mechanism-specific intervention supplies causal confirmation. An unchanged shared 10 MHz frequency reference alone does not calibrate offsets.

The clean exercise starts later, at CFG-DEBUG-CLEAN-D2. Independent analyzer/unused-standard verification supports E1=false. A new qualified isolation acquisition ISO-D2 supports E2=false. The earlier VNA-BAD-D and LEAK-CONTROL-D remain in the audit; their corrupt DUT interpretations are not quietly relabelled valid.

\begin{aligned}\text{eligible} &= \text{configuration} \land \text{plane} \land \text{method} \land \text{uncertainty} \\ &\land \text{calibration} \land \text{setup} \land \text{controls} \land \text{source} \land \text{timing}\end{aligned}

Definition: eligibility is a conjunction of relevant, matching source fields. Missing, invalid, clipped, stale or changed-state evidence cannot constrain the DUT question. Source status is not a confidence slider.

Think about itA new calibration makes the VNA pass its unused standard. Can the old DUT trace now supply E4?

Answer

No. Preserve the original trace and its failed verification. Create a new calibration ID and a repeated mounted/reference acquisition with unchanged cable and state controls. The new observation can become eligible; the old trace remains a record of the earlier invalid measurement.

Engineering decision → record update

Reject the audit bundle as a clean starting state. Record CAL-D2, ISO-D2 and new configuration/acquisition IDs before entering the Boolean exercise. Retain raw, corrected, reprocessed and repeated records separately.

03 / 10

Partition the chain and create competing mechanisms

Where could the failure enter, and what would each mechanism predict?

A conducted power result samples R1-TX; a feed comparison samples R2; a spatial test uses S0; a packet decision occurs at R3. R3 is never the receiver RF input. Clock, supply and firmware state can affect several boundaries. Leakage can bypass the intended attenuator, while an invalid analyzer can create a false apparent symptom.

Solid arrows: RF/decision chain. Dashed lines: causal or setup influences. Only declared RF connectors assume real positive 50 Ω; S0 is spatial and R3 is a decision boundary.

Use the existing RF-chain hypotheses and antenna evidence limits. In particular, corrected feed impedance is not antenna efficiency, and packet reception is not a full-sphere antenna measurement. Keep real positive 50 Ω only at declared RF connectors; the e^+jωt convention is unchanged.

Seven mechanisms · predictions must come with falsifiers and unchanged controls
ID / domain	Mechanism / signature	Falsifier	Unchanged controls / limit
P / bit 1 / Supply → R1-TX	Supply droop reduces PA output. E3 and E8; power follows the specified droop at fixed RF command.	The deficit persists after a separately verified supply correction.	RF command, gain, waveform, load, temperature and cable route. Time correlation alone is insufficient; no universal supply-to-dB law.
A / bit 2 / R2 / S0	Mounted antenna detuning. E4; a reproducible feed-impedance shift under controlled mounting.	An independent valid mounted/reference comparison has no shift.	Feed/calibration plane, cable route, antenna specimen, channel and spatial setup. Feed impedance does not determine efficiency or radiation pattern.
B / bit 4 / R1-RX → R3	External blocker degrades reception. E5; wanted-link degradation in the declared fixed ambient state.	Calibrated external source isolation leaves the same degradation.	Wanted power, packet population, local clocks/state, source noise and isolation. Quiet wanted-link means no deliberately added blocker; fixed ambient interference may remain.
C / bit 8 / Clock → R1-RX / R3	Local clock-related interference. E5 and E6; a controlled clock shift moves the interfering signature.	The signature stays fixed under a verified clock-only perturbation.	RF LO, supply, firmware duty, acquisition clock and wanted source. Changing a shared clock may alter several domains; declare that confound.
F / bit 16 / Digital control → R1 / R3	Incorrect firmware RF state or duty. E3, E5 and E7; forcing the verified intended state restores its symptom.	Independent state capture confirms correct operation while the symptom persists.	Supply, antenna, RF command, waveform and thermal state. A software log alone is not proof of actual hardware state.
L / bit 32 / Measurement path → R1-RX	Stimulus bypass leakage. E2; qualified terminated-path test detects the bypass.	A valid isolation test bounds leakage below the declared criterion.	All source outputs, shields, terminations, cables and detection floor. An invisible tone with an inadequate floor cannot establish isolation.
M / bit 64 / Instrument / calibration	Invalid analyzer or VNA configuration. E1; independent configuration/standard verification fails.	Independent verification clears with the exact DUT settings and connections.	Verification device not used in the fit; frequency, level and reference plane. A setup fault is investigated with a separate valid verification system.

Hypothesis × evidence matrix · singleton signatures (combine by OR)
Mechanism / mask	E0	E1	E2	E3	E4	E5	E6	E7	E8
{P} / 1	T	F	F	T	F	F	F	F	T
{A} / 2	T	F	F	F	T	F	F	F	F
{B} / 4	T	F	F	F	F	T	F	F	F
{C} / 8	T	F	F	F	F	T	T	F	F
{F} / 16	T	F	F	T	F	T	F	T	F
{L} / 32	T	F	T	F	F	F	F	F	F
{M} / 64	T	T	F	F	F	F	F	F	F

T/F are supplied Boolean predictions, valid only in this curated catalogue. E0 means a reproducible apparent RF failure, including L or M. E3 is P OR F; E5 is B OR C OR F. These are logical alternatives, not additive dB models. “Quiet wanted-link” retains the fixed ambient interferer state AMB-D; B is external to the DUT and need not vanish because a deliberately added blocker is off.

Complete no-fault, single-fault and two-fault signature catalogue

Complete 29-set catalogue · supplied synthetic signatures
Mechanism / mask	E0	E1	E2	E3	E4	E5	E6	E7	E8
∅ (no listed fault) / 0	F	F	F	F	F	F	F	F	F
{P} / 1	T	F	F	T	F	F	F	F	T
{A} / 2	T	F	F	F	T	F	F	F	F
{P,A} / 3	T	F	F	T	T	F	F	F	T
{B} / 4	T	F	F	F	F	T	F	F	F
{P,B} / 5	T	F	F	T	F	T	F	F	T
{A,B} / 6	T	F	F	F	T	T	F	F	F
{C} / 8	T	F	F	F	F	T	T	F	F
{P,C} / 9	T	F	F	T	F	T	T	F	T
{A,C} / 10	T	F	F	F	T	T	T	F	F
{B,C} / 12	T	F	F	F	F	T	T	F	F
{F} / 16	T	F	F	T	F	T	F	T	F
{P,F} / 17	T	F	F	T	F	T	F	T	T
{A,F} / 18	T	F	F	T	T	T	F	T	F
{B,F} / 20	T	F	F	T	F	T	F	T	F
{C,F} / 24	T	F	F	T	F	T	T	T	F
{L} / 32	T	F	T	F	F	F	F	F	F
{P,L} / 33	T	F	T	T	F	F	F	F	T
{A,L} / 34	T	F	T	F	T	F	F	F	F
{B,L} / 36	T	F	T	F	F	T	F	F	F
{C,L} / 40	T	F	T	F	F	T	T	F	F
{F,L} / 48	T	F	T	T	F	T	F	T	F
{M} / 64	T	T	F	F	F	F	F	F	F
{P,M} / 65	T	T	F	T	F	F	F	F	T
{A,M} / 66	T	T	F	F	T	F	F	F	F
{B,M} / 68	T	T	F	F	F	T	F	F	F
{C,M} / 72	T	T	F	F	F	T	T	F	F
{F,M} / 80	T	T	F	T	F	T	F	T	F
{L,M} / 96	T	T	T	F	F	F	F	F	F

Engineering decision → record update

Give every hypothesis a causal boundary, expected signature, falsifier and unchanged controls. Add a new mechanism when real evidence falls outside these seven; do not force it into the closest name.

04 / 10

Rank consistency without inventing certainty

Can a plausible explanation coexist with another fault?

With seven mechanisms there are 1 + 7 + 21 = 29 zero-, single- and two-fault sets. Before a symptom, even the no-fault set is possible. E0=true removes only the empty set. Clean E1=false excludes M, and E2=false excludes L, leaving 15 sets over P, A, B, C and F. There is no probability attached to these sets.

C=\{c:\operatorname{prediction}(c,o)=\operatorname{outcome}(o)\text{ for every eligible known observation }o\}

C is the currently compatible collection of candidate sets; o is an eligible observation. A set survives exactly when its supplied signature matches every eligible observed outcome. Unknown and ineligible observations add no condition.

Think about itPredict E3=true followed by E4=true. Must there be just one faulty block?

Answer

E3=true leaves nine masks [1,3,5,9,16,17,18,20,24]. E4=true leaves [3,18]: {P,A} or {A,F}. Both require A and one other mechanism. A single-fault-only assumption creates a contradiction because it forbids both remaining pairs.

Compatible means a set has not been excluded. A mechanism is required within the catalogue only if it occurs in every survivor; it is possible if present in some. With no survivors, all classifications become unresolved conflict, not “everything excluded, therefore healthy.”

Reveal the complete pinned guided sequence after predicting

Pinned guided stages · cumulative eligible observations
Added evidence	Remaining count	Exact masks / sets
None	29	All zero/single/pair sets
E0=true	28	1 {P} · 2 {A} · 3 {P,A} · 4 {B} · 5 {P,B} · 6 {A,B} · 8 {C} · 9 {P,C} · 10 {A,C} · 12 {B,C} · 16 {F} · 17 {P,F} · 18 {A,F} · 20 {B,F} · 24 {C,F} · 32 {L} · 33 {P,L} · 34 {A,L} · 36 {B,L} · 40 {C,L} · 48 {F,L} · 64 {M} · 65 {P,M} · 66 {A,M} · 68 {B,M} · 72 {C,M} · 80 {F,M} · 96 {L,M}
E1=false	21	1 {P} · 2 {A} · 3 {P,A} · 4 {B} · 5 {P,B} · 6 {A,B} · 8 {C} · 9 {P,C} · 10 {A,C} · 12 {B,C} · 16 {F} · 17 {P,F} · 18 {A,F} · 20 {B,F} · 24 {C,F} · 32 {L} · 33 {P,L} · 34 {A,L} · 36 {B,L} · 40 {C,L} · 48 {F,L}
E2=false	15	1 {P} · 2 {A} · 3 {P,A} · 4 {B} · 5 {P,B} · 6 {A,B} · 8 {C} · 9 {P,C} · 10 {A,C} · 12 {B,C} · 16 {F} · 17 {P,F} · 18 {A,F} · 20 {B,F} · 24 {C,F}
E3=true	9	1 {P} · 3 {P,A} · 5 {P,B} · 9 {P,C} · 16 {F} · 17 {P,F} · 18 {A,F} · 20 {B,F} · 24 {C,F}
E4=true	2	3 {P,A} · 18 {A,F}
E7=false	1	3 {P,A}
E8=true	1	3 {P,A}

The guided order illustrates controlled contrasts; the greedy cost heuristic initially prefers E7. E8=true corroborates P but adds no separation after only mask 3 remains. It does not close a root cause.

Honest alternatives when the model cannot decide
Case	Result / next decision
Equivalent-signature variant	Restrict explicitly to single faults P or F after E3=true/E4=false/E6=false; E5/E7/E8 predictions are unknown for this variant. Each possible-outcome set is {false,true}; no remaining eligible test separates P from F. Retain both.
Contradictory same-state E3=true and E3=false	Both eligible acquisitions stay in the record. The two form an inclusion-minimal conflict; investigate timing/state, setup or the model. No majority vote.
E4=true, E6=true, E8=true	Requires A,C,P. No ≤2-fault set fits. The exercise reports model/evidence conflict, not a least-bad winner or silent third fault.

Common misconceptionOne surviving mask proves the real root cause.

It establishes necessity only within a finite supplied model. Unlisted mechanisms, more faults, shared confounds or a wrong prediction model can still explain reality. Confirmation and regression need separate acquisitions.

Engineering decision → record update

At [3,18], mark A required, P/F possible and B/C/L/M excluded only within the model. Retain every elimination reason so a changed eligibility assessment recomputes the full set from the preserved record.

05 / 10

Choose a discriminating eligible test

Which test separates the most plausible sets for the effort—and is it eligible?

Write the predictions before obtaining the answer. A defined test names the stimulus, state, plane, estimator, observation interval, decision rule, uncertainty and unchanged controls. “Another screenshot” names none of them. Safety and measurement prerequisites are hard conditions, not penalties that can be outweighed by an attractive score.

\begin{aligned}\operatorname{separation}(t)&=\#\{\{c_i,c_j\}:O_t(c_i)\cap O_t(c_j)=\varnothing\}\\\operatorname{score}(t)&=\frac{\operatorname{separation}(t)}{\operatorname{cost\_units}(t)}\\\text{Binary deterministic test:}\quad\operatorname{separation}&=n_0n_1\end{aligned}

Oₜ(c) is the possible-outcome set for candidate c under test t. Count unordered pairs of candidate sets with disjoint outcome sets. Unknown maps to all allowed outcomes. Cost units and durations are synthetic budgeting inputs, not probabilities or promised bench times.

Clean-baseline scoring anchors · fixed synthetic costs and durations
Test	True / false	Cost / min	Pairs / score
E3 · Conducted TX power	9 / 6	4 / 4	54 / 13.5
E4 · Controlled feed impedance	5 / 10	3 / 3	50 / 16.666666666667
E5 · Quiet wanted-link packets	12 / 3	4 / 4	36 / 9
E6 · Clock-only perturbation	5 / 10	5 / 5	50 / 10
E7 · Verified firmware state	5 / 10	2 / 2	50 / 25
E8 · Synchronized supply capture	5 / 10	6 / 6	50 / 8.333333333333

Think about itE7 has score 25. Would cost .1 make it acceptable if the state-forcing safety review were missing?

Answer

No score can clear an unknown safety prerequisite. The engine withholds ranking for that test. With missing E1/E2 controls, it recommends independent setup/isolation verification before any dependent E3–E8 test.

At {P,A} versus {A,F}, both E7 and E8 separate one pair. E7 costs 2 and scores .5; E8 costs 6 and scores 1/6. That is a transparent local preference, not expected entropy or the probability of finding a cause. Retain equal scores as ties. Unknown cost is unranked; zero/negative cost is invalid.

Class 1 · synthetic reasoning exercise

RF Debug Evidence Graph

Audit the bundle first. The clean baseline has distinct calibration and isolation records. Compare predictions, reveal a supplied outcome, and keep every still-compatible explanation.

[data-debug-controls]{display:none!important}JavaScript is off. The complete canonical result, signature catalogue, guided stages, counterexamples and completed handoff remain below; native answers open normally.

Evidence caseDefault: clean initial record. Each preset replaces all observations, quality fields, costs, safety and closure records atomically. No result is saved.

Inspect test / edit observationDefault E3; choosing a different test loads its committed fields and discards that test editor's uncommitted draft. Selection changes no evidence.

Predict the supplied outcome's effectUngraded prediction; does not change physics. Required before a guided reveal. Default: no prediction.

E3 · Conducted TX power

Fixed command 7, on-time mean, gate 2.0–23.6 ms; 17.0 ± .36 dBm.

Predicted true / false: 9 / 6 sets. Separation: 54 pairs. Score: 13.5.

Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls.

Guided contrasts: E3 → E4 → E7 → E8. The calculated greedy alternative can differ. Reveal is unavailable for other tests, missing predictions, uncommitted edits, completed constraints, failed prerequisites, or a full 24-record history.

Edit a custom observation or test plan

Apply creates a new hypothetical acquisition with its own ID. Earlier entries stay in the ledger as replaced local comparisons; a repeated processing view never counts as an independent acquisition. This editor cannot authorize a physical test.

Outcome and test plan

Observed signatureDefault E3–E8: unknown. E0 true, E1/E2 false. Unknown is not zero or a pass.

Safety prerequisite evidenceDefault satisfied; unknown or failed makes this proposed test ineligible before scoring.

Cost (exercise units).1…20 units, step .1; or type unknown. Default 4. Duration is a synthetic budget, not acquisition time.

Nominal duration (min).1…120 min, step .1; or type unknown. Default 4. Duration is a synthetic budget, not acquisition time.

Prediction model availabilityDefault fixed. Unknown predictions separate no pair; they do not count as false.

Fault-count assumptionThe restricted view is a counterexample, not a preferred diagnosis.

Source fields used to derive eligibility

All fields default to valid for the clean synthetic fixture. Unknown withholds inference; invalid includes a mismatched state/plane, stale calibration, clipping, unaligned timestamps or incompatible processing. Each field is assessed separately; no force-pass control exists.

Configuration evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Plane evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Method evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Uncertainty evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Calibration evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Setup evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Controls evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Source evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Timing evidenceApplies to this new observation and proposed E3 test. Reset restores its canonical source fields.

Clean baseline: 15 compatible candidate sets. Predict before revealing E3.

Clean initial record · committed result15 compatible sets

E7: 50 separated candidate pairs / 2 cost units = 25. Predictions and controls first.

Within at most two of seven listed mechanisms. Counts and scores are not probabilities. A unique mask is not root-cause confirmation.

Compatible masks: 1 {P} · 2 {A} · 3 {P,A} · 4 {B} · 5 {P,B} · 6 {A,B} · 8 {C} · 9 {P,C} · 10 {A,C} · 12 {B,C} · 16 {F} · 17 {P,F} · 18 {A,F} · 20 {B,F} · 24 {C,F}

Required, possible and excluded mechanisms
ID	Current scope	Closure evidence
P	possible	Not independently confirmed
A	possible	Not independently confirmed
B	possible	Not independently confirmed
C	possible	Not independently confirmed
F	possible	Not independently confirmed
L	excluded within catalogue	Not independently confirmed
M	excluded within catalogue	Not independently confirmed

Next-test comparison · candidate-set pairs, not individual faults
Test	True / false candidates	Pairs separated	Cost / min	Score / status
E7 · Verified firmware state	5 / 10	50	2 / 2	25
E4 · Controlled feed impedance	5 / 10	50	3 / 3	16.666666666667
E3 · Conducted TX power	9 / 6	54	4 / 4	13.5
E6 · Clock-only perturbation	5 / 10	50	5 / 5	10
E5 · Quiet wanted-link packets	12 / 3	36	4 / 4	9
E8 · Synchronized supply capture	5 / 10	50	6 / 6	8.333333333333
E0 · Reproduce the reported symptom	15 / 0	Ineligible	1 / 1	Completed constraint; an independent repeat needs its own question.
E1 · Independent setup verification	0 / 15	Ineligible	1 / 1	Completed constraint; an independent repeat needs its own question.
E2 · Terminated-path leakage control	0 / 15	Ineligible	1 / 1	Completed constraint; an independent repeat needs its own question.

Inspect predicted partitions and unchanged controls

E7 · Verified firmware state

True: {F} · {P,F} · {A,F} · {B,F} · {C,F}. False: {P} · {A} · {P,A} · {B} · {P,B} · {A,B} · {C} · {P,C} · {A,C} · {B,C}.

Digital / R1 / R3 · Force intended state and verify command/log/physical duty independently. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E4 · Controlled feed impedance

True: {A} · {P,A} · {A,B} · {A,C} · {A,F}. False: {P} · {B} · {P,B} · {C} · {P,C} · {B,C} · {F} · {P,F} · {B,F} · {C,F}.

M-CAL = R2 · Stepped CW −10 dBm, IFBW 1 kHz, 2–3 GHz/101 points; untouched cable F-D2. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Unused verification ≤.01 residual; cable/fixture unchanged; reference/mounted pair at the same plane and bias.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E3 · Conducted TX power

True: {P} · {P,A} · {P,B} · {P,C} · {F} · {P,F} · {A,F} · {B,F} · {C,F}. False: {A} · {B} · {A,B} · {C} · {A,C} · {B,C}.

R1-TX · Fixed command 7, on-time mean, gate 2.0–23.6 ms; 17.0 ± .36 dBm. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E6 · Clock-only perturbation

True: {C} · {P,C} · {A,C} · {B,C} · {C,F}. False: {P} · {A} · {P,A} · {B} · {P,B} · {A,B} · {F} · {P,F} · {A,F} · {B,F}.

Clock / R1-RX / R3 · Shift clock by +1%; verify unchanged LO, supply and actual RF state. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E5 · Quiet wanted-link packets

True: {B} · {P,B} · {A,B} · {C} · {P,C} · {A,C} · {B,C} · {F} · {P,F} · {A,F} · {B,F} · {C,F}. False: {P} · {A} · {P,A}.

R1-RX → R3 · Fixed AMB-D ambient source; counted packet/control comparison under the declared state. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E8 · Synchronized supply capture

True: {P} · {P,A} · {P,B} · {P,C} · {P,F}. False: {A} · {B} · {A,B} · {C} · {A,C} · {B,C} · {F} · {A,F} · {B,F} · {C,F}.

Supply / R1-TX · Aligned supply and RF capture with bounded timing; fixed command/load. Eligible E1=false and E2=false in CFG-DEBUG-CLEAN-D2, plus the named controls. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E0 · Reproduce the reported symptom

True: {P} · {A} · {P,A} · {B} · {P,B} · {A,B} · {C} · {P,C} · {A,C} · {B,C} · {F} · {P,F} · {A,F} · {B,F} · {C,F}. False: none.

R1 / R2 / S0 / R3 · Repeat the fixed debug protocol; preserve apparent setup failures. Independent valid verification/reproduction method; no dependent DUT inference required. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E1 · Independent setup verification

True: none. False: {P} · {A} · {P,A} · {B} · {P,B} · {A,B} · {C} · {P,C} · {A,C} · {B,C} · {F} · {P,F} · {A,F} · {B,F} · {C,F}.

M-VERIFY → M-CAL · Unused reference and independent analyzer configuration verification. Independent valid verification/reproduction method; no dependent DUT inference required. Controls: Verified analyzer CAL-D2 and cable F-D2; source/terminated-path control ISO-D2; clock and RF command held unless the named test changes them.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

E2 · Terminated-path leakage control

True: none. False: {P} · {A} · {P,A} · {B} · {P,B} · {A,B} · {C} · {P,C} · {A,C} · {B,C} · {F} · {P,F} · {A,F} · {B,F} · {C,F}.

M-SOURCE → R1-RX · Source on/off and terminated wanted path; calibrated isolation/floor. Independent valid verification/reproduction method; no dependent DUT inference required. Controls: ISO-D2 ≥110 dB isolation; wanted −110 dBm; coherent leakage ≤−150 dBm; detection bound ≤−153 dBm. Source on/off, all alternate paths terminated.

Risks: Exercise-only prerequisite. Actual instruments, power/DC limits, terminations and approved low-energy procedures must be reviewed before physical work.

Immutable evidence and field-by-field eligibility
Record / observation	Value / acquisition	Eligibility reasons
OBS-D2-E0 / E0	true / ACQ-OBS-D2-E0	Eligible under all declared source fields.
OBS-D2-E1 / E1	false / ACQ-OBS-D2-E1	Eligible under all declared source fields.
OBS-D2-E2 / E2	false / ACQ-OBS-D2-E2	Eligible under all declared source fields.

Configuration, source metadata and record lineage

SYN-NODE-D / gateway G-D; illustrative only. DEBUG-QPSK-D/1.0; 2.450e9 Hz; generic QPSK 10e3 symbol/s, RRC .35, PRBS-9 0x1FF; no standardized PHY. 3.300 V supply nominal; 298.15 K; fixed mounting and ambient source AMB-D; exact method-specific snapshots below. Frozen synthetic 2026-09-09; baseline records acquired before interventions. No browser clock.

OBS-D2-E0
Field	Retained value
Configuration / method / plane	CFG-DEBUG-CLEAN-D2 / METHOD-D2-E0 / R1 / R2 / S0 / R3
Raw origin	Supplied, reviewed synthetic Boolean signature; not an automated inference from the inherited audit bundle.
Statistic / interval	Boolean signature under p08-m06-evidence-rules-v1; unknown is not false. 0…25600 µs; timestamp 10000 µs.
Uncertainty / rule	Supplied categorical review of method-specific decision and controls; quantitative uncertainty is not generated by the graph.
Source fields	configuration: valid; plane: valid; method: valid; uncertainty: valid; calibration: valid; setup: valid; controls: valid; source: valid; timing: valid
Parents / replaced local comparisons	none / none
Method note	Repeat the fixed debug protocol; preserve apparent setup failures.

OBS-D2-E1
Field	Retained value
Configuration / method / plane	CFG-DEBUG-CLEAN-D2 / METHOD-D2-E1 / M-VERIFY → M-CAL
Raw origin	Supplied, reviewed synthetic Boolean signature; not an automated inference from the inherited audit bundle.
Statistic / interval	Boolean signature under p08-m06-evidence-rules-v1; unknown is not false. 0…25600 µs; timestamp 10000 µs.
Uncertainty / rule	Supplied categorical review of method-specific decision and controls; quantitative uncertainty is not generated by the graph.
Source fields	configuration: valid; plane: valid; method: valid; uncertainty: valid; calibration: valid; setup: valid; controls: valid; source: valid; timing: valid
Parents / replaced local comparisons	none / none
Method note	Unused reference and independent analyzer configuration verification.

OBS-D2-E2
Field	Retained value
Configuration / method / plane	CFG-DEBUG-CLEAN-D2 / METHOD-D2-E2 / M-SOURCE → R1-RX
Raw origin	Supplied, reviewed synthetic Boolean signature; not an automated inference from the inherited audit bundle.
Statistic / interval	Boolean signature under p08-m06-evidence-rules-v1; unknown is not false. 0…25600 µs; timestamp 10000 µs.
Uncertainty / rule	Supplied categorical review of method-specific decision and controls; quantitative uncertainty is not generated by the graph.
Source fields	configuration: valid; plane: valid; method: valid; uncertainty: valid; calibration: valid; setup: valid; controls: valid; source: valid; timing: valid
Parents / replaced local comparisons	none / none
Method note	Source on/off and terminated wanted path; calibrated isolation/floor.

Inspect every eliminated candidate and its evidence

Compatibility exclusions · recomputed from all eligible records
Excluded set / mask	Contradicting eligible observations
∅ (no listed fault) / 0	OBS-D2-E0
{L} / 32	OBS-D2-E2
{P,L} / 33	OBS-D2-E2
{A,L} / 34	OBS-D2-E2
{B,L} / 36	OBS-D2-E2
{C,L} / 40	OBS-D2-E2
{F,L} / 48	OBS-D2-E2
{M} / 64	OBS-D2-E1
{P,M} / 65	OBS-D2-E1
{A,M} / 66	OBS-D2-E1
{B,M} / 68	OBS-D2-E1
{C,M} / 72	OBS-D2-E1
{F,M} / 80	OBS-D2-E1
{L,M} / 96	OBS-D2-E1 · OBS-D2-E2

Committed confirmation and regression records
Cause	Mechanism / corrective / regression status	Evidence links
P	Unconfirmed / correction open / regression open	No confirmation records selected.
A	Unconfirmed / correction open / regression open	No confirmation records selected.

rf-debug-evidence-graph/2.0 · p08-m06-evidence-rules-v1. Current configuration and observation history are local to this page visit. The separate completed review below is a worked example, not saved progress.

Go deeperConflict and ordering contract

The engine evaluates all eligibility fields, deduplicates equal views of the same acquisition, then filters a canonically ordered universe. Observation insertion order and candidate permutation cannot change the result. For a conflict, it removes observations in stable-ID order whenever the remaining eligible constraints still conflict, repeating until no deletion preserves inconsistency. This yields an inclusion-minimal subset, not necessarily a unique or globally smallest explanation.

Engineering decision → record update

Record the chosen test, both outcome partitions, fixed cost/time and controls. If the catalogue cannot split further, request confirmation, a new mechanism/test or a data-quality review; zero separation does not make those activities unnecessary.

06 / 10

Substitute, inject, and reverse one controlled change

What does a substitution isolate, and what else did it change?

A golden unit is useful only when its firmware, mounting, temperature, antenna and method match the question. A dummy load removes antenna radiation and may alter common-mode currents. Injection bypasses upstream blocks but can introduce a leakage path. Loopback excludes the external channel but may also exclude the fault you need to observe.

Controlled comparisons and their remaining confounds
Technique	Prediction / control	What it cannot prove alone
Golden-unit substitution	Run the identical source script through the same verified fixture and actual RF state.	A full board swap changes many physical variables. It localizes the problem to a boundary, not a component.
Dummy load / injection	Name the new R1/R2 boundary, termination and every bypass; reverify isolation.	A conducted pass does not establish radiated efficiency or original cable/common-mode behavior.
Safe A/B/A	One reversible mounting spacer change; retain cable, feed, calibration and command; compare a return to the original safe state.	A/B/A with a changed cable route or temperature still contains a confound.
Independent supply confirmation	Reviewed current-limited source substitution at fixed actual command; capture supply and conducted RF together.	Do not deliberately recreate an unsafe droop/overstress condition merely to complete an A/B/A pattern.

Think about itThe supply is replaced and firmware command changes from 7 to 8. Output returns. Has P been confirmed?

Answer

No. The intervention changes two variables that can each affect conducted output. The multi-variable preset exposes this in its record and withholds P confirmation. A record with a restored symptom cannot substitute for a controlled causal contrast.

For the supplied {P,A} resolution, the supply intervention restores R1-TX power while the mounted feed shift remains. The spacer intervention then changes the feed and same-direction OTA evidence while the repaired supply stays fixed. One successful correction need not resolve the other fault.

Separate confirmation contrasts and a two-variable counterexample
Record	Intended / actual change	Prediction and observation	Unchanged controls
CONF-P-D	Supply impedance / Supply impedance	At unchanged command 7, a reviewed current-limited source removes droop and restores conducted output; antenna shift remains. Supply minimum 3.28 V; three R1-TX on-time means 20.00,20.02,19.98 dBm, U=.36 dB. Each lower endpoint ≥19.5. Mounted Δ\|Γ\| still .25.	Same command, waveform, sensor, cable, load and temperature; independent low-energy supply substitution after approved inhibit/connection procedure. No unsafe fault recreation. Synthetic independent supply check replaces unsafe A/B/A; physical electrothermal model remains open.
CONF-A-D	Mount spacer / Mount spacer	A safe spacer-only A/B/A changes feed and directional OTA evidence, without changing the already repaired supply. Δ\|Γ\| .250→.010→.249, bound ±.02. S0 same-direction received levels −58.0→−54.0→−58.1 dBm, supplied difference bound ±.6 dB.	Same antenna/feed, CAL-VNA-D2, untouched cable, command, 2.450 GHz, supply D3; fixed distance/polarization/orientation and reference antenna before/after. Directional OTA witness is not efficiency, TRP or TIS. Full pattern and production mounting variation open.
CONF-PF-D	Supply impedance / Supply impedance, Firmware RF command	At unchanged command 7, a reviewed current-limited source removes droop and restores conducted output; antenna shift remains. Symptom disappears after simultaneous supply and firmware command 7→8 change. Single-mechanism contrast lost.	Waveform, sensor, cable, load and temperature held; actual command changes and is therefore a confound. P remains unconfirmed. This distinct acquisition cannot inherit CONF-P-D single-variable evidence.

Engineering decision → record update

Require a before/after ID, actual changed-variable list, predicted effect and independent verification for each mechanism. Keep unsafe reversals out of the plan; document the independent safe alternative.

07 / 10

Correlate domains without confusing correlation with cause

The supply log, RF trace and firmware event align. Does that tell you which came first?

Correlation needs the same specimen, state, plane and time interval. EVM-PROC-D is a four-symbol algebraic example; it is not a before/after measurement of TX-DEBUG-D. Likewise, an attenuation-dependent internal analyzer product is not evidence that a physical DUT spectrum improved.

Correct known timestamp offsets, then carry residual timing bounds and signal-path latency. A shared reference frequency controls frequency drift; it does not automatically align event timestamps. Manufacturer gated-FFT examples show how an identified time window can link spectral behavior to an event. The physical mechanism still needs a controlled contrast. [EMI-TIME]

\begin{aligned}t_{\mathrm{aligned}} &= t_{\mathrm{recorded}} - \mathrm{offset} - \mathrm{known}_{\mathrm{latency}} \\ &a \mathrm{precedes} b \mathrm{only} \mathrm{if} t_{a} + \mathrm{bound}_{a} < t_{b} - \mathrm{bound}_{b}\end{aligned}

Times are stored as integer µs. The offset is defined as logger time minus reference time; subtract it. Corrected event bounds include the supplied residual offset and latency uncertainty. An order claim requires strictly disjoint bounds.

STATE-TIME-D · fixed timestamp alignment
Domain	Raw timestamp / offset (µs)	Aligned (µs)	Residual bound / state
RF	10000 / 0	10000	±40 µs; 20.0→17.0 dBm
Supply	10300 / 300	10000	±40 µs; 3.300→2.900 V
Firmware	9800 / -200	10000	±40 µs; Command 7 unchanged

Think about itRF aligns at 10.000 ms and supply at 10.050 ms, each with ±.040 ms bounds. Is RF known to precede supply?

Answer

No. RF lies within 9.960–10.040 ms and supply within 10.010–10.090 ms. The bounds overlap, so the 50 µs nominal separation does not establish order. The same conclusion holds before any causal explanation is attempted.

Cross-domain evidence reconciliation
Evidence	Same-state question / eligible inference
Supply and R1-TX power	Aligned fixed-command signatures support P or a shared influence; CONF-P-D supplies the controlled causal contrast.
R2 impedance and S0 directional level	Separate calibrated feed and spatial contrasts support the mounting mechanism; neither equals TRP, TIS or radiation efficiency.
EVM/PER and firmware logs	Normalize processing and packet population; verify actual RF state independently of command logs.
Spectrum and clock state	A moved spur under a clock-only perturbation can support C; simultaneous LO/supply changes invalidate that contrast.

Engineering decision → record update

Retain raw log clocks, offsets, residual bounds, trigger markers and path latency. Mark unaligned records ineligible for timing claims. Preserve correlation as a finding and seek the discriminating intervention.

08 / 10

Capture intermittent failures with a declared observation plan

A long capture contains no event. Was the product stable during the parts you could not observe?

Freeze the observation plan before looking at data. Define the trigger, monitored quantity, threshold, hysteresis, pre/post windows, live time, dead intervals, event-duration requirement, reproduction count and exact state sequence. Observing nothing constrains only the eligible live population. Waiting for a browser animation would not create experimental time.

Intermittent-capture plan · integer-µs half-open intervals
Plan item	Fixed illustrative entry
Horizon / reproduction	[0,30000) µs for each of 3 fixed-seed runs; no early stopping. State script D/1.0, PRBS reset 0x1FF.
Trigger	Supply falling below 3.05 V with 50 mV hysteresis, synthetic review at 20 MHz supply bandwidth; 2 ms pre /4 ms post window. Threshold is not a safety rating.
Timestamp alignment	TIME-D2 offsets and ±40 µs bounds; preserve path latency and RF/control marker IDs.
Live / dead	[0,4000) · [7000,14000) · [17000,24000) · [27000,30000) µs live; dead [4000,7000),[14000,17000),[24000,27000). Total 21000 µs live, 9000 µs dead.
Known counterexample	Events [5000,6000),[15000,16000),[25000,26000) µs lie completely in dead time; all three missed despite 21 ms live.
Alternate event-triggered capture	Separate verified trigger acquisition around 10 ms yields [8000,14000) µs including pre/post evidence. This is not the dead-time schedule above.

The interval logic is the same as 08.3 capture timing. Intervals include their start and exclude their end; overlapping windows are unioned before counting live time. The dead-time counterexample does not estimate an event probability.

\begin{aligned}P(0\text{ failures}\mid p)&=(1-p)^n\\p_U&=1-\alpha^{1/n}=.002991249545095\end{aligned}

Separate PER-COUNT-D: 0 failures in n=1000 fixed IID Bernoulli packet trials. α=.05 is the one-sided tail probability, not a model confidence score. A packet-error bound is not an intermittent-event arrival model.

The upper bound exceeds the .001 target even though the observed error proportion is zero. The calculation needs the fixed count, trial definition and stationarity/IID assumptions; correlated packets or opportunistic stopping do not inherit it. [PER-EXACT]

Common misconceptionNo captured failure means the failure rate is zero.

Dead time, missed triggers and an unrepresentative state script can hide failures. Report the live population and detection limits. An improved capture needs a new acquisition ID and a predeclared horizon.

Engineering decision → record update

Publish the planned horizon, live/dead windows and event eligibility beside the absence claim. Preserve all three reproduction runs, including misses. Assign the measurement owner to verify the trigger and timing bounds.

09 / 10

Build a bounded pre-compliance evidence package

The pre-scan is 1 dB below the screen. Is that enough evidence to send a “pass” to the formal lab?

Pre-compliance is an engineering preparation activity. A useful package explains what was tested, what the apparatus can resolve, and what remains missing. Local probing helps identify sources when probe orientation, height, loading and DUT state are controlled. Its voltage is not directly comparable with a far-field emission threshold. [EMI-PRE] [EMI-TIME]

A separate field pre-scan arithmetic fixture

FIELD-SCREEN-D is a continuous-wave 300 MHz local variant, separate from the 2.450 GHz debug waveform. It assumes a calibrated receiving antenna and a compatible free-space far-field/site model: 5 m separation, maximum source/receive dimensions .20/.50 m, fixed horizontal polarization, both phase centers at 1.5 m, specified direction, validated low-reflection range. A one-point RMS-equivalent indication is measured in 120 kHz RBW, VBW≥RBW, 1 s observation. These supplied conditions support arithmetic only; they are not an execution of a CISPR or ETSI procedure.

M0 is a real matched 50 Ω receiver input. Antenna factor AF=+20 dB/m maps antenna-terminal voltage to field for this specific calibrated setup. Cable loss L=+2 dB and preamp gain G=+10 dB are external path corrections, each applied once. Internal receiver correction is already in its −60 dBm indication. Re-derive the voltage constant from watts and resistance. [LEVEL]

\begin{aligned}V&=P+20\log_{10}\left(10^6\sqrt{.001\times50}\right)\\V&=46.98970004336\,\mathrm{dB\mu V}\\E&=V+\mathrm{AF}+L-G=58.98970004336\,\mathrm{dB\mu V/m}\end{aligned}

P is receiver indication in dBm; V is RMS-equivalent voltage in dBµV referenced to 1 µV. AF has dB/m units; L and G are power/voltage-ratio dB at compatible matched planes. E is in dBµV/m. No antenna mismatch or internal correction is applied a second time.

FIELD-SCREEN-D · local interval containment, not a legal threshold
Quantity	Derived / supplied value
Hypothetical upper engineering screen	60 dBµV/m
Nominal margin	1.01029995664 dB
Expanded uncertainty	U=4 dB supplied, k=2 assumption; component budget and coverage justification not supplied.
Upper interval endpoint	62.98970004336 dBµV/m
Decision	Inconclusive under the chosen containment rule; formal applicability and method evidence remain open.

Think about itDoes the attractive nominal margin clear a rule requiring the entire uncertainty interval below 60 dBµV/m?

Answer

No. 58.989700043360+4=62.989700043360 dBµV/m exceeds the screen. This does not establish that the true field exceeds 60; it means the selected containment rule is inconclusive. The policy must be declared before examining the result. [GUM-RULE]

For a scoped official-method reading exercise, EN 300 328 V2.2.2 §5.4.9.2.2 points radiated measurements to its site and procedure annexes B/C. §5.1.2 requires recording actual environmental conditions. [METHOD-READ] The repository has no completed Path 09 applicability decision for this generic waveform. Treat that method as a candidate question for its owner, with current citation/version, radio classification and full procedure still unresolved. No standard limit is imported into FIELD-SCREEN-D.

Independent readiness axes · 3 scoped reviews · no aggregate pass
Question	Requirement / applicability	Configuration coverage	DUT stability	Instrument / site	Calibration / uncertainty	Margin / rule	Repeatability / raw IDs	Deviations	Next evidence / owner	Lab handoff
Local conducted debug requirement	REQ-DEBUG-D minimum 19.5 dBm at R1-TX; local method only.	D-QPSK, one node, fixed channel/antenna; four supply/duty cells.	Frozen command and supply verified in CONF/VERIFY/REG-P-D.	Synthetic sensor/CAL-D2 evidence; real ratings and capability unknown.	Separate supplied U=.36 dB; real component budget/calibration chain open.	Worst regression 19.90−.36−19.5 = .04 dB lower-endpoint margin.	Three confirmation repeats plus four declared cells, IDs retained.	Other specimens/temperature/channels not covered.	Measurement engineer: actual budget and corner coverage.	Bounded engineering record; no formal conformity conclusion.
S0 field pre-scan	FIELD-SCREEN-D 60 dBµV/m hypothetical internal screen; exact formal applicability unresolved.	One frequency, polarization, height and orientation; alternate accessory/antenna/channel missing.	Continuous-wave local variant, stable source before/after; not packet-mode coverage.	Supplied calibrated far-field site/antenna arithmetic; formal detector/site evidence missing.	Supplied expanded U=4 dB, k=2 assumption; components and coverage justification absent.	Nominal +1.010299956640 dB; upper endpoint 62.989700043360 exceeds screen. Inconclusive.	One supplied level; no independent repeatability population.	Cannot transform a near-field loop voltage with the far-field antenna factor.	Pre-compliance lead + lab: method, detector, site, uncertainty and repeat matrix.	Send raw indication/correction ledger and gaps; request a scoped lab plan.
Formal method review / Path 09	Candidate EN 300 328 V2.2.2 §5.4.9.2.2 reading exercise; not selected for this generic mode.	Product × intended mode × market × antenna/accessory matrix pending.	Frozen debug fixture does not establish representative product stability.	Review §5.4.9 procedures and annex B/C site/method with the lab.	Method-specific calibration/uncertainty and decision policy pending.	Unknown; no official threshold imported.	No formal-lab acquisition supplied.	Jurisdiction, radio classification and current applicability review unresolved.	Path 09 compliance owner: approve exact applicable source chain; Path 10 design owner: validated corrections/regression.	Planned destinations; no certificate, test booking or market authorization.

Local conducted debug requirement · same readiness axes, print layout
Axis	Recorded status
Requirement / applicability	REQ-DEBUG-D minimum 19.5 dBm at R1-TX; local method only.
Configuration coverage	D-QPSK, one node, fixed channel/antenna; four supply/duty cells.
DUT stability	Frozen command and supply verified in CONF/VERIFY/REG-P-D.
Instrument / site	Synthetic sensor/CAL-D2 evidence; real ratings and capability unknown.
Calibration / uncertainty	Separate supplied U=.36 dB; real component budget/calibration chain open.
Margin / rule	Worst regression 19.90−.36−19.5 = .04 dB lower-endpoint margin.
Repeatability / raw IDs	Three confirmation repeats plus four declared cells, IDs retained.
Deviations	Other specimens/temperature/channels not covered.
Next evidence / owner	Measurement engineer: actual budget and corner coverage.
Lab handoff	Bounded engineering record; no formal conformity conclusion.

S0 field pre-scan · same readiness axes, print layout
Axis	Recorded status
Requirement / applicability	FIELD-SCREEN-D 60 dBµV/m hypothetical internal screen; exact formal applicability unresolved.
Configuration coverage	One frequency, polarization, height and orientation; alternate accessory/antenna/channel missing.
DUT stability	Continuous-wave local variant, stable source before/after; not packet-mode coverage.
Instrument / site	Supplied calibrated far-field site/antenna arithmetic; formal detector/site evidence missing.
Calibration / uncertainty	Supplied expanded U=4 dB, k=2 assumption; components and coverage justification absent.
Margin / rule	Nominal +1.010299956640 dB; upper endpoint 62.989700043360 exceeds screen. Inconclusive.
Repeatability / raw IDs	One supplied level; no independent repeatability population.
Deviations	Cannot transform a near-field loop voltage with the far-field antenna factor.
Next evidence / owner	Pre-compliance lead + lab: method, detector, site, uncertainty and repeat matrix.
Lab handoff	Send raw indication/correction ledger and gaps; request a scoped lab plan.

Formal method review / Path 09 · same readiness axes, print layout
Axis	Recorded status
Requirement / applicability	Candidate EN 300 328 V2.2.2 §5.4.9.2.2 reading exercise; not selected for this generic mode.
Configuration coverage	Product × intended mode × market × antenna/accessory matrix pending.
DUT stability	Frozen debug fixture does not establish representative product stability.
Instrument / site	Review §5.4.9 procedures and annex B/C site/method with the lab.
Calibration / uncertainty	Method-specific calibration/uncertainty and decision policy pending.
Margin / rule	Unknown; no official threshold imported.
Repeatability / raw IDs	No formal-lab acquisition supplied.
Deviations	Jurisdiction, radio classification and current applicability review unresolved.
Next evidence / owner	Path 09 compliance owner: approve exact applicable source chain; Path 10 design owner: validated corrections/regression.
Lab handoff	Planned destinations; no certificate, test booking or market authorization.

Engineering decision → record update

Send the indication, complete correction ledger, supplied uncertainty, conditions and independent gaps. Ask the lab to confirm method, detector/site capability and representative configuration coverage; do not label an internal screen as formal approval.

10 / 10

Confirm causes, regress, and hand off

What must be true before a mechanism can be closed—and what remains open afterward?

Separate six states: compatible, required within the catalogue, independently confirmed, correction verified, relevant regression checked, and residual gaps owned. A single surviving mask supplies only the second. Symptom disappearance supplies none of the missing configuration or comparison controls automatically.

Corrective verification and regression · distinct evidence for each cause
Record / cause	Before → after	Prediction / actual outcome	Controls / scope / remaining gaps
VERIFY-P-D / P	TX-DEBUG-D → TX-SUPPLY-VERIFY-D4	Stable supply reproduces the independently confirmed power restoration. Second acquisition 19.99,20.01,20.00 dBm; U=.36 dB; state captured independently.	Command 7, identical waveform, cable/calibration, temperature and mounted antenna. Scope: DEBUG-QPSK-D. Path 10 owns the implemented product supply correction.
REG-P-D / P	TX-SUPPLY-VERIFY-D4 → REG-SUPPLY-D5	No conducted deficit in the declared affected matrix. Four supplied cell minima 19.94,19.92,19.91,19.90 dBm; U=.36 dB; minimum lower endpoint 19.54.	D-QPSK command 7; normal/10% duty × 3.1/3.3 V supply input, 298.15 K, same specimen/load; recalibrated each setup. Scope: Normal duty / 3.1 V; Normal duty / 3.3 V; 10% duty / 3.1 V; 10% duty / 3.3 V. Other channels, temperature corners, firmware and production specimens untested.
VERIFY-A-D / A	VNA-REPEAT-D → ANT-MOUNT-D4	Chosen spacer repeats the feed shift removal and local directional improvement. Three independent remounts Δ\|Γ\| .010,.012,.011 ±.02; local ≤.05 containment rule clears; directional improvement ≥3.3 dB after bound.	Supply D3 held; same feed/CW calibration and separate S0 fixture; command, channel and reference alignment unchanged. Scope: Mounted direction D / 2.450 GHz. Path 10 must validate the physical mounting correction.
REG-A-D / A	ANT-MOUNT-D4 → REG-MOUNT-D5	Both defined mounting torque corners retain the local corrected feed/directional behavior. Two supplied remount cells: Δ\|Γ\| .015,.018 ±.02, both upper ≤.05; S0 retained improvement 3.8,3.9 dB ±.6.	Same spacer, feed/calibration, product orientation D, 298.15 K, command and supply; torque corners MIN/MAX are fixture-defined labels, not physical torque specifications. Scope: Mount MIN / direction D; Mount MAX / direction D. Second antenna, accessory, channel, orientation population and environmental extremes remain gaps.

Acquisition manifest for all confirmation, correction and regression records

Before/after IDs and actual changed variables
Record	Before / after IDs	Intended / actual change	Evidence verification
CONF-P-D	TX-DEBUG-D / TX-SUPPLY-D3	Supply impedance / Supply impedance	Supplied synthetic reviewed record; confirmation; Supply minimum 3.28 V; three R1-TX on-time means 20.00,20.02,19.98 dBm, U=.36 dB. Each lower endpoint ≥19.5. Mounted Δ\|Γ\| still .25.
VERIFY-P-D	TX-DEBUG-D / TX-SUPPLY-VERIFY-D4	Supply impedance / Supply impedance	Supplied synthetic reviewed record; correction; Second acquisition 19.99,20.01,20.00 dBm; U=.36 dB; state captured independently.
REG-P-D	TX-SUPPLY-VERIFY-D4 / REG-SUPPLY-D5	Declared regression corner / Declared regression corner	Supplied synthetic reviewed record; regression; Four supplied cell minima 19.94,19.92,19.91,19.90 dBm; U=.36 dB; minimum lower endpoint 19.54.
CONF-A-D	VNA-REPEAT-D / ANT-MOUNT-D3	Mount spacer / Mount spacer	Supplied synthetic reviewed record; confirmation; Δ\|Γ\| .250→.010→.249, bound ±.02. S0 same-direction received levels −58.0→−54.0→−58.1 dBm, supplied difference bound ±.6 dB.
VERIFY-A-D	VNA-REPEAT-D / ANT-MOUNT-D4	Mount spacer / Mount spacer	Supplied synthetic reviewed record; correction; Three independent remounts Δ\|Γ\| .010,.012,.011 ±.02; local ≤.05 containment rule clears; directional improvement ≥3.3 dB after bound.
REG-A-D	ANT-MOUNT-D4 / REG-MOUNT-D5	Declared regression corner / Declared regression corner	Supplied synthetic reviewed record; regression; Two supplied remount cells: Δ\|Γ\| .015,.018 ±.02, both upper ≤.05; S0 retained improvement 3.8,3.9 dB ±.6.
CONF-PF-D	TX-DEBUG-D / TX-SUPPLY-FW-D3	Supply impedance / Supply impedance; Firmware RF command	Supplied synthetic reviewed record; confirmation; Symptom disappears after simultaneous supply and firmware command 7→8 change. Single-mechanism contrast lost.

After IDs are immutable supplied acquisition summaries embedded in the corresponding record above. CONF/VERIFY-P deliberately use separate repeated values at the same restored configuration; they are distinct acquisitions, not reprocessed views. The multi-variable preset uses distinct CONF-PF-D / TX-SUPPLY-FW-D3 evidence for the supply-plus-firmware intervention and leaves P unconfirmed.

Measurement and debug readiness review

Audit the bundle, define measurands and planes, verify setup/analyzer/VNA/vector methods, preserve uncertainty and count limits, keep competing hypotheses, choose a short discriminating sequence, then require confirmation and regression for every claimed mechanism. Finish with a reproducible lab request instead of a certificate.

Ungraded self-review rubric · no score is calculated or stored
Emphasis	What good evidence looks like	Reasoned model review
Definition / planes · 20%	State quantity, population, interval, reference plane and compatible mode.	Keep PWR-CAL-A, DEBUG-QPSK-D, VNA-CW, algebraic EVM and FIELD-SCREEN-D separate.
Settings / calibration · 25%	Check actual state, calibration/unused verification, fixtures and residual limitations.	Exclude VNA-BAD-D and unqualified leakage; enter only distinct clean D2 records. Calibration does not remove all error.
Uncertainty / statistics · 20%	Use the right covariance, interval and fixed-count assumptions.	17.36<19.5 supports a local deficit; 0/1000 cannot support .001; field containment remains inconclusive.
Causal reasoning · 25%	Retain simultaneous faults, predict outcomes, apply hard gates before ranking.	E3/E4 retain [3,18]; E7=false isolates [3]; E8 corroborates. Confirm P/A separately with unchanged controls.
Records / safety · 10%	Preserve raw IDs and define safe contrasts, owners and recheck triggers.	No unsafe reversal, magic closure or pre-compliance approval. Real-source/hardware and lab gaps remain visible.

The model review supports a bounded engineering closure for the supplied P and A mechanisms in their stated cells. The partial-closure preset honestly closes only P. That is useful progress: it directs the next experiment without pretending that mounting, alternate channels, environmental corners, a new accessory or formal-method coverage has been demonstrated.

Completed measurement record · M08-06-HANDOFF-D

p08-measurement-record-v1 · complete static snapshot
Field	Completed illustrative entry
id	M08-06-HANDOFF-D
parentIds	M08-05-TX-C · M08-05-RX-C · TX-DEBUG-D · VNA-REPEAT-D · ISO-D2 · STATE-TIME-D
owner	08.6 measurement/debug lead; fictional role
question	Which next eligible test distinguishes the range-loss mechanisms, and which gaps block formal-lab readiness?
hypotheses	P/A/B/C/F/L/M; all 29 zero/single/pair masks. Guided evidence requires {P,A} only within the bounded catalogue.
requirement	REQ-DEBUG-D R1-TX ≥19.5 dBm; local feed-shift and FIELD-SCREEN-D rules separately declared.
decisionRule	p08-m06-evidence-rules-v1; eligibility before Boolean consistency and separated-pairs/cost ranking. p08-m06-readiness-v1: independent readiness axes, no aggregate pass.
specimen	p08-m06-fault-cases-v1; SYN-NODE-D, G-D; illustrative engineering case.
configuration	CFG-DEBUG-CLEAN-D2; HW-D1 / FW-D1; command7; DEBUG-QPSK-D/1.0. Earlier method variants are not this operating point.
state	3.300 V nominal, transient 2.900 V in preserved baseline; 298.15 K; mounted assembly D. Interventions change only declared supply impedance or spacer.
timestamp	Frozen synthetic 2026-09-09T10:00:00Z; event timestamps integer µs; actual source access separately 2026-09-09.
environment	Indoor fixed ambient AMB-D; 25 °C; humidity unknown. S0 comparisons require their separately declared site and reference.
stimulus	2.450e9 Hz generic QPSK, 10e3 symbols/s, 20e3 uncoded bits/s, RRC .35/span8, PRBS-9 seed0x1FF. CW VNA/leakage/field and four-symbol EVM are explicit separate variants.
statistic	R1-TX logarithm of gated linear mean watts; R2 dimensionless complex Γ; R3 CRC/timeout count; S0 RMS-equivalent field level. No peak/period-average or PSD substitution.
population	16 fixed-command bursts for power; three feed pairs/remounts; separate 1000 IID packet trials and declared four/two regression cells; no production yield inference.
acquisition	Power 2.0–23.6 ms gate, 1 s burst cadence; VNA 101 points 2–3 GHz; event horizon30 ms with 21 ms live and9 ms dead. Preserve raw/derived/repeated IDs.
planes	R1-TX/R1-RX component ports; R2 feed; S0 spatial; R3 packet decisions. RF connectors real positive 50 Ω.
chain	R1-TX → CABLE-D2 1.000 dB → M-PWR-D matched sensor; R2=M-CAL through characterized F-D2; R1-RX wanted70 dB and independently bounded ISO-D2 bypass. Protection suitability is fixture-supplied; real ratings unknown.
instrument	Fictional SYN-SENSOR-D SIM-2 gated option, SYN-VNA-D SIM-2, independent VERIFY-D2, source/receiver ISO-D2; no invented commercial branding or safe rating.
settings	Power passband ±100 kHz; VNA −10 dBm CW/IFBW1 kHz; isolation10 Hz detection bandwidth; supply20 MHz; separate field pre-scan conditions in section09.
calibration	CAL-PWR-D2 / CABLE-D2, CAL-VNA-D2 unused standard residual≤.01, ISO-D2, TIME-D2. Valid only at named mode, plane, level, connections, temperature and acquisition conditions.
rawEvidence	TX-DEBUG-RAW-D; VNA-BAD-D retained; new ACQ-VNA-D2; LEAK-CONTROL-D retained; new ISO-D2; raw phase symbols and packet counts unchanged; STATE-TIME-D aligned views preserve logs.
processing	rf-debug-evidence-graph/2.0; each external correction once, internal corrections already in indication. Derived views share acquisition identity; invalid evidence remains visible.
uncertainty	PWR-CAL-A independent/correlated budgets preserved; debug U=.36 dB supplied separately; feed bound±.02; timing±40 µs each; fixed-n pU=.002991249545095; field U4 dB supplied, actual components unknown.
result	Guided 15→9→2→1 masks; [3,18]→[3]. E7 is highest initial score 25; prescribed E3/E4 teaching order is a different contrast. P/A separate confirmations and bounded regression supplied.
inference	P and A independently confirmed only under their stated synthetic controls; selected corrective contrasts and regression cells verified. No universal diagnosis, complete antenna metric or conformity verdict.
unresolved	Formal applicability; real instrument/site/calibration capability; uncertainty components; alternative antenna/accessory/channel/mode/thermal/population coverage; unlisted/third causes.
next	Measurement owner: real budgets and missing comparisons. Lab/pre-compliance owner: capability/method matrix. Path 09 owner: applicability. Path 10 owner: validated correction implementation and regression scope.
review	Reopen after specimen, firmware/mode, cable, calibration, supply, mounting, channel, source or method changes; review sources before actual/formal use.

Formal-lab questions, next evidence and owners
Open question	Next evidence	Owner / review trigger
Which requirements and exact methods apply?	Product/mode/antenna/accessory/market matrix and current official source chain.	Path 09 compliance owner; before a conformity claim or scope change.
Can the bench/site produce that evidence?	Real manuals, calibration/uncertainty components, detector/bandwidth/site validation and repeatability.	Measurement/pre-compliance lead and formal lab; before acquisition.
Which corrections should enter the product?	Controlled implementation verification plus affected-mode/corner/specimen regression.	Path 10 design owner; every supply, mounting, firmware or component change.
What was preserved?	Raw and derived IDs, source/state manifest, ineligible findings, confirmation and omissions.	Debug lead; at every handoff and after any configuration repair.

Go deeperFrozen algorithm, provenance and precision

rf-debug-evidence-graph/2.0, p08-m06-fault-cases-v1, p08-m06-evidence-rules-v1, p08-m06-readiness-v1; Class 1 lesson-only exercise. At most 29 candidate sets, nine implemented tests within a 16-test bound, and 24 evidence records. No priors, Bayesian diagnosis, general simulator, live instruments or storage. Independent set enumeration and all Boolean signatures are exact; rational score tolerance 10⁻¹² and field-level tolerance 10⁻⁹ dB verify computation, not measurement uncertainty. Full precision drives decisions. Events use integer µs. [GUM-INDEX] and the actual [GUM-AMD] keep inherited metrology guidance versioned; nonlinear uncertainty cannot be replaced with a generic RSS shortcut.

Engineering decision → record update

Deliver the completed record, hypothesis matrix, next-test rationale, separate confirmations/regressions and lab questions. Preserve all unresolved evidence. This capstone does not establish acceptance of every earlier module or a real product.

Ungraded review

Check your understanding

Answer each question in your own words, then reveal the model answer.

01Is “the node lost range” a causal claim?
Model answer
It is an observed service symptom only when the route, traffic, receiver criterion, environment and configuration are recorded. A requirement failure additionally needs an eligible measurement and rule. “The PA supply droops and reduces R1-TX power” is a mechanism needing direct evidence and a controlled confirmation; range loss alone does not identify it.
02Should a high-separation test proceed with unknown safety or missing E1/E2 controls?
Model answer
No. Test value is computed only after hard prerequisites. Even E7 at cost .1 cannot enter an eligible ranking while safety is unknown or calibration/isolation prerequisites are missing. Obtain the missing independent verification/control evidence first. Real safety review is separate from the fictional flags.
03What is known at masks [3,18], and what if all remaining signatures are equivalent?
Model answer
The sets are {P,A} and {A,F}. A is required within this catalogue; P and F remain possible alternatives. E7 and E8 split the pair, with scores .5 and 1/6. If every eligible unperformed test has overlapping outcome sets for two mechanisms, retain both and request new evidence or an expanded model. Neither a score nor a surviving count is a probability.
04Are raw and phase-corrected EVM views independent repeat evidence?
Model answer
No. They share the same acquisition. PHASE-RAW-D and its corrected child preserve different processing contracts; ideal known-phase removal changes the reported metric without proving a hardware change. Independent repetition needs a new acquisition and defined question. Contradictory content under one immutable ID is invalid input, not last-write-wins.
05Does simultaneous supply and firmware repair confirm a supply cause?
Model answer
No. Both mechanisms predict improvement under that combined change. Hold actual RF state, load, calibration and waveform fixed, use a reviewed safe supply substitution, and independently verify power restoration. Confirm mounting separately. Avoid unsafe fault recreation; each cause needs correction verification and a declared regression matrix.
06Does a nominal field margin of 1.0103 dB establish lab readiness?
Model answer
No. With supplied U=4 dB the upper endpoint is 62.9897 dBµV/m, above the hypothetical 60 dBµV/m screen, so the chosen containment rule is inconclusive. Applicability, representative modes, accessories, detector, site capability and calibrated uncertainty remain independent gaps. A near-field loop voltage is not a far-field strength result.

References and further study

Primary-source access: 2026-09-09. Fixture freeze: 2026-09-09. Review on a method, instrument, fixture, source revision or product-state change, and before any physical or formal use. Chosen synthetic numbers come from the disclosed lesson fixtures and independent derivations, not manufacturer ratings. Only consulted source sections support claims.

GUM-INDEX · BIPM / JCGM. Guides in Metrology. Living official index. Status inspected 2026-09-09; lists 100:2008/Amd.1:2026 and GUM-5:2026. Consulted: GUM, amendment and conformity guide listings. Informative catalogue. Document identity/status only. Catalogue is not full technical text; GUM-5 body not used.
GUM-AMD · JCGM. 100:2008/Amd.1:2026 — Nonlinearity in measurement models. First edition 2026. 2026 publication; body read 2026-09-09. Consulted: Amendments to §4.1.4 and H.1.7, p.1. Informative method. Nonlinearity may affect both estimate and uncertainty; inherited 08.1 covariance conditions remain explicit. No new general uncertainty solver or physical calibration chain.
GUM-RULE · JCGM. 106:2012 — The role of measurement uncertainty in conformity assessment. First edition 2012. Published 2012; body read 2026-09-09. Consulted: §§3.2.4–3.2.9, 8.3.2 guarded acceptance. Definition / Informative. Separate indication/result and acceptance policy; local interval containment is explicitly chosen. No risk probability, legal applicability, approval or universal k-to-confidence conversion.
VNA-ERROR · Keysight. PXIe VNA Help — Measurement Errors. Living undated help. Read 2026-09-09. Consulted: Drift, random and systematic error sections; monitoring error terms. Informative manufacturer guidance. Residual errors and changed measurement setup limit corrected evidence. No specific safe rating, real standard model, complete commercial calibration or universal recalibration interval.
EMI-TIME · Rohde & Schwarz. EMI Debugging with Oscilloscopes — 1TD05. 1.2 / 1TD05-12e. Official listing 2021-08-03; body read 2026-09-09. Consulted: §3.3 / 3.3.1 pp.16–17; gated FFT pp.26–27. Informative manufacturer method. Probe orientation/distance and time-correlated spectral windows matter. A heuristic or shared timing does not establish causation; no far-field conversion from near-field loop voltage.
EMI-PRE · Rohde & Schwarz. Precompliance EMI Debug. Version 1.00. August 2020; body read 2026-09-09. Consulted: §3.3 and §§4.2.1–4.2.4, pp.7–9. Informative troubleshooting guidance. Understand DUT states, ambient interference and local probe comparisons before preparing formal work. Historical marketing/legal generalizations, failure rates, prices and simplistic ambient subtraction are not adopted.
PER-EXACT · NIST / SEMATECH. e-Handbook §7.2.4.1 — Confidence intervals. Living web edition. Access 2026-09-09; revision not asserted. Consulted: Exact binomial tail method. Informative statistical method / Derived local example. Fixed-n independent Bernoulli trials; zero-error bound independently derived. Not correlated packets, opportunistic stopping, a confidence score for causes, or a conformity verdict.
LEVEL · Analog Devices. CATV dBm, dBmV, and dBµV Conversions. Technical article. Displayed publication 2002-07-17; read 2026-09-09. Consulted: Power/voltage conversion and explicit 50 Ω versus 75 Ω distinction. Definition / Derived. Re-derive the matched real 50 Ω RMS-equivalent voltage conversion, retaining full precision. No copied conversion table or assumption that all RF interfaces have the same impedance.
METHOD-READ · ETSI. EN 300 328 — Wideband transmission systems, 2.4 GHz. V2.2.2 (2019-07). Official PDF and compared-version listing checked 2026-09-09. Consulted: §5.1.2 and §5.4.9.2.2; annex B/C dependencies identified, not implemented. Normative document used for an informative method-reading example. Radiated method calls for its specified site and measurement procedures; record actual test environment. No Path 09 applicability selection exists yet. Current legal citation, amendments, product classification, full procedure and limits unresolved; generic QPSK and FIELD-SCREEN-D are not ETSI test results.