METRIC FLIGHT / MF-01 MIRROR COMMISSIONING / DRAFT 08
20 October 2026. NOT REGISTERED. NO EXPERIMENTAL DATA.
This commissioning specification supersedes the Edition 07 1 W primary criterion.
The v0.2 Alpha 1/graphene research report remains a separate model release.

QUESTION
Can an independently calibrated balance recover the signed optical force on a
passive mirror under an external 1064 nm beam, after measured nuisance corrections?
This tests metrology, not a new field or propulsion of an isolated closed system.

PRIMARY ESTIMAND
Full beam-on minus beam-off force D. At 2 W intercepted power, perfect normal
reflection predicts 13.3425638 nN. A collinear transmitting optic instead gives
(1 + R - T)P/c. Actual scatter, angular alignment, clipping and absorption require
a measured vector momentum budget. Source and dump stay off the balance.
1 W (ideal 6.6712819 nN) is a secondary linearity point, not the primary pass/fail.

DESIGN FREEZE / MISSING ITEMS
Before confirmatory acquisition, the named experimental lead and independent
reviewer must approve serial numbers, calibrated effective arm, stiffness,
damping, actuator, optics, sensor locations, power at sample, vacuum/gas,
acquisition bandwidth, dwell and settling rules, alignment, enclosure/interlocks,
calibration certificates and full covariance model. Freeze this completed protocol
and code hash in a timestamped registration. No registration or PI appointment is
claimed. Commissioning data and confirmatory data must remain distinct.

SAMPLING
Propose 70 valid four-window ABBA blocks at the primary point: A=beam on, B=off.
Use equally spaced settled windows. D_i=(A1-B1-B2+A2)/2; average D_i across blocks.
The illustrative 1 nN independent window SD gives u_stat=1/sqrt(70)=0.119523 nN.
This does not establish independence, effective N, coverage or elapsed run time.
ABBA cancels constant/linear drift, not quadratic drift or thermal hysteresis.
Counterbalance ABBA with BAAB in a frozen seeded schedule (invert the estimator
sign for BAAB). Assess autocorrelation and drift on separate commissioning data;
freeze the estimator/covariance method and dwell before collecting decisive data.
Do not keep adding blocks until agreement appears. If 70 valid blocks cannot be
obtained under the frozen protocol, report an incomplete campaign and its reasons.

STANDARD UNCERTAINTY ALLOCATIONS AT 2 W, nN
Optical prediction 0.400277 (3%); independent force calibration 0.667128 (5%);
statistics 0.119523; thermal residual 0.150; electrical residual 0.080;
gas residual 0.080; mechanical residual 0.080. All are k=1 planning allowances.
Combined U(k=2)=1.626367 nN (12.1893%) if independent; illustrative thermal/mechanical
correlation rho=0.6 gives 1.643980 nN (12.3213%). Unknown effects cannot be zeroed.
Use actual calibration certificate coverage factors and avoid counting the same
calibration or nuisance uncertainty twice. Covariance must be propagated for
D_measured - D_optical; coverage factor 2 is not automatically exact 95% coverage.

PASS CRITERION / STOP RULES
All controls below must fit the assessed uncertainty budget. Require signed
closure |D_measured-D_optical| <= U_difference AND U_difference <= 0.15|D_optical|.
For the ideal 2 W point the upper limit is 2.001385 nN. A <=10% interval is a
stretch target, not achieved by these allowances. An interval including zero or
poor signal-to-noise cannot constitute closure by widening its uncertainty.
Repeat qualified closure on at least 3 days and after controlled reassembly.
No pooling that conceals a failed day. Stop for unsafe operation, lost calibration,
instability, unexplained state-correlated force, or violated frozen criteria.
Report mismatch without assigning it to Alpha 1. Graphene testing has its own gate.

FIVE NUISANCE PROBES
1. Matched heater: measure temperature/force transfer, lag, mount and centroid drift.
2. Beam-off modulation and grounded dummy: test electrical/magnetic/cable pickup.
3. Pressure sweep: measure gas-force dependence and plateau; pressure alone fails
   to prove negligible gas force. Illustrative Ap/(4T) at A=1 cm2,T=300K is
   83.333 pN/K at 0.001 Pa, or 83.333 nN/K at 1 Pa. Geometry matters.
4. Pump/isolation/tilt and cable-routing controls: quantify mechanical response.
5. Optical null and reversal plus 0.5/1/1.5/2 W checks: distinguish momentum from
   heating and alignments. Secondary points need their own validated ranges and
   covariance. Freeze linear-fit acceptance criteria after commissioning and
   before confirmatory runs; no post-hoc slope acceptance.

EXCLUSIONS / REPORTING
Only logged faults (saturation, loss of lock, missing timestamps or acquisition
failure) may exclude a window under prespecified thresholds. Exclude the whole
block if one window is invalid. Preserve all raw data, exclusions and an all-data
sensitivity analysis. Do not reject observations based on force sign or agreement.
Publish calibrated inputs, code/version/hash, raw/processed records, covariance,
all tested configurations, residuals and failures. Blind condition labels during
analysis where practicable. Independent replication is a separate accomplishment.

PUBLICATION STATUS
No OSF registration, Zenodo deposit/DOI, CODECHECK certificate, external review,
host-lab agreement, procurement or funding submission has been completed here.
Complete the unresolved design fields before attempting registration.