ITERATION 4: NO-PROBE FULL-BUDGET BASELINE AND CONTROLLED FACTOR EFFECTS
October 6, 2026. Numerical reference-beam study; no physical fracture validation.

PURPOSE

The previous pre-probe comparison charged every profile for a probe, including
the single source that did not use its estimate. This iteration adds a genuine
no-probe single-source baseline that puts its entire 10 mJ mechanical budget
into the main collision. It also separates local single-factor effects and
isolates the nominal flight/gap timing perturbation in the previous F/H cases.

The earlier contact model, estimator, case matrix and report artifacts are
imported or read unchanged. This version writes only new iteration4 evidence.
The script does not control hardware, access accounts, publish, or send data.

INTERNALLY SPECIFIED EXPERIMENT

The matrix was selected before executing this version. This is internal
prespecification, not formal/external preregistration. It has 15 conditions:

  I4-00      Reference: frequency scale 1, damping ratio .005,
             contact stiffness 30,000 N/m, contact loss 100,000 N*s/m^2,
             no noise/clock/gap error, probe acquisition beginning at time 0.
  I4-01      Only frequency scale changes to .915.
  I4-02      Only frequency scale changes to 1.085.
  I4-03      Only modal damping changes to .007.
  I4-04      Only contact stiffness changes to 33,000 N/m.
  I4-05      Only contact loss coefficient changes to 120,000 N*s/m^2.
  I4-06      Only sensor noise fraction changes to .15.
  I4-07      Only sensor clock-rate error changes to -.002.
  I4-08      Only probe acquisition start changes to .015 s.
  I4-09      Only nominal second-source gap/flight perturbation becomes -.002 s.
  I4-10      Only nominal second-source gap/flight perturbation becomes +.002 s.
  I4-F-error Previous combined F plant/record, including its -.001 s perturbation.
  I4-F-zero  Identical F inputs except its nominal gap/flight perturbation is zero.
  I4-H-error Previous combined H plant/record, including its -.002 s perturbation.
  I4-H-zero  Identical H inputs except its nominal gap/flight perturbation is zero.

Every one-factor row uses the reference random seed. The paired F/H rows retain
their respective original seed and generate identical synthetic probe arrays
within each pair. Thus removing the flight perturbation does not secretly change
the observation noise, estimated frequency, contact properties, or damping.

The code variable flight_time_error_s changes the initial gap by speed times
the stated time. It is a NOMINAL flight/gap timing perturbation. Actual contact
time can change further because the beam moves; each source's first_contact_s
records the actual modeled event. The fixed and estimated pairs receive the
same perturbation. Ideal simultaneous/matched controls have no relative error.

SIX PROFILES AND FAIR MECHANICAL BUDGETS

Every profile starts with the same 20 g total moving striker mass and the same
10 mJ IDEAL TOTAL MECHANICAL energy allocation:

  same_probe_single        0.1 mJ prescribed probe + 9.9 mJ main, one 20 g mass.
  fixed_period_pair        Same budget, two 10 g masses, frozen nominal period.
  estimated_period_pair    Same budget/masses, estimated period from prior probe.
  ideal_simultaneous_pair  Same budget/masses, zero relative timing error.
  same_probe_matched_single Same budget, 20 g single, doubled contact coefficients.
  full_budget_no_probe_single No probe: the entire 10 mJ enters one 20 g mass.

The no-probe single is independently integrated at 1.0 m/s. The probed sources
start at sqrt(.99) m/s. Its result is NOT obtained by scaling a 9.9 mJ result:
the contact damping law is nonlinear, so that shortcut would be inappropriate.
Peak target-mode energy is divided by the same 10 mJ denominator in every row.

The frozen fixed-period pair is also a SAME-PROBE control: it pays for a probe
whose estimate it does not use. This study does not include a full-budget,
unprobed fixed-period pair. It therefore cannot establish that sensing itself
has earned back its overhead compared with that direct timing comparator.

All SIX profiles are run with 20 and 10 microsecond steps. Energy accounting,
contact/displacement domain checks and sensitivity to timestep are retained.
Numerically unresolved small differences are separated from increases or
decreases at the tested timesteps. The sensitivity screen is not a rigorous
error bound or a physical/statistical confidence interval.

WHAT THE BUDGET DOES NOT INCLUDE

Probe modal energy is assigned ideally; actual electrical/actuator-to-specimen
transfer efficiency is unknown. Electrical power, sensor acquisition, launcher,
positioning, resets and losses of a complete device are not modeled. The main
beam is ideal and uncalibrated. A numerical ratio is not wall-plug efficiency,
energy-per-mass fragmentation, or a product performance rating.

Probed profiles retain the 35 s probe-settling interval and its explicit bounded
residual energy allowance. The no-probe profile starts from an assumed resting
beam without that probe/settling protocol. Its modeled interval is .17 s versus
35.17 s for a probed profile. These are INITIALIZATION/OBSERVATION intervals,
not demonstrated device cycle times; handling, launch/reset, initial-rest
requirements and actual throughput remain unknown.

CAUSALITY AND ESTIMATOR QUALITY

The estimator sees only previously acquired noisy timestamps/sensor samples
and a common broad frequency band. It does not receive the true case frequency
or flight perturbation. The repaired baseline-residual quality gate, numerical
amplitude checks, fallback, and measured local computation time are reused.
Actual transfer latency remains assumed rather than physically measured.
Retiming sets the initial gap BEFORE the main event. It is pre-probe feed-forward,
not online feedback, crack detection, or an actuator launch-control model.

HOW EFFECTS ARE REPORTED

Each case reports absolute target energy in joules, its fraction of the common
10 mJ budget, adaptive/baseline ratios, and refined comparisons. One-factor
effects separately report the change in EVERY profile's absolute target energy
and budget fraction, plus the change in the adaptive/fixed ratio. An increasing
ratio caused only by a worse denominator is therefore visible rather than
being mistaken for more energy delivered by the retimed source.

Paired F/H interventions use exactly the same frozen plant and synthetic record
and remove only the nominal gap/flight perturbation. These identify a local
within-model effect, not universal physical causation. One-factor results are
local to the reference; they are not a full interaction/global sensitivity map.
No new parameter choice is tuned after viewing outcomes. Negative results stay
in the output. Conditions are deterministic and are not a statistical sample.

REPRODUCE AND VERIFY

From custom-poc, with Python and NumPy installed:

  python efficiency/iteration4.py
  python -m unittest discover -s efficiency -p test_iteration4.py -v

Run both steps. The post-run tests compare current source hashes with those
captured for the run and reject mismatches before accepting the evidence. They
also verify report/input hashes, the one-factor design, identical paired inputs,
all six budgets/refinements, and honest separation of numerator/denominator effects.

New outputs:

  efficiency/results/ITERATION4_summary.json
  efficiency/results/iteration4/case_*.json
  efficiency/results/iteration4/SYNTHETIC_probe_*.csv
  efficiency/results/iteration4/TEST-report.json
  efficiency/results/iteration4/TEST-log.txt
  efficiency/results/iteration4/AUDIT.json

The summary embeds exact inputs, source hashes and environment versions. Every
case and synthetic CSV has its own SHA-256 digest. CPU/runtime metadata vary
between runs; compare numerical metrics and tolerances, not entire JSON bytes.

This module performs no publication itself. A reviewed release of these numerical
materials can be prepared separately under the owner's current authorization.
Credentials, administrative records and live hardware controls are not part of
this numerical model or its required reproduction inputs.

OBSERVED NUMERICAL OUTCOMES, OCTOBER 6, 2026

All 15 conditions and six profiles completed in 298.566 seconds. All profiles
passed the declared numerical feasibility/energy/refinement checks. Ten new
experiment-control and saved-evidence tests passed with zero skips. Current
model, estimator and iteration4 source hashes match those captured at run start.

Against the full-budget unprobed SINGLE, the estimated-period pair ranges from
2.933% less to 5.636% more peak first-mode energy. Three conditions decline
(I4-09, F-error, H-error). This deterministic matrix is not a probability of
success. A difference above the timestep sensitivity screen does not establish
practical importance, an experimental result, or a fracture capability.

At reference, estimated pair energy is 3.370081 mJ and the unprobed single is
3.201696 mJ. That is +5.259% relative or +1.684 percentage points of the common
10 mJ budget. Its difference from the SAME-PROBE FIXED PAIR is numerically
unresolved. This reference result does not establish a benefit from sensing.

Removing only the nominal gap perturbation in F increases estimated-pair energy
by 0.173513 mJ and fixed-pair energy by 0.004558 mJ. The estimated/fixed ratio
rises from .939984 to .990097 and still favors the fixed pair. Removing only
the corresponding H perturbation raises estimated energy by 0.204248 mJ while
lowering fixed energy by 0.383557 mJ. H's ratio rises from .921812 to 1.112556;
that large ratio change partly reflects a poorer denominator, not just a gain
in the estimated source. Both pairs reuse identical synthetic probe records.

Frequency shifts of -.085 or +.085 leave estimated-pair energy near its
reference level (changes -0.006529 and +0.006444 mJ). They lower the fixed pair
by .239065 and .349119 mJ respectively. This narrow model supports frequency
retiming as a way to limit mistuning losses. It does not show unlimited growth,
energy creation, a universally optimal delay, or improved fragmentation.

Noise and clock-rate perturbations each reduce estimated-pair energy in the
specified reference experiment. Contact dissipation and timing perturbations
also materially affect the outcome. A direct full-budget unprobed fixed-pair
comparison is the next required overhead check; it must be specified separately
instead of being silently added to this completed matrix.
