Microtracking/plan_article_2026_11_04.md

# Article Revision Plan: 4 November 2026 Field Evaluation

## Status And Objective

The revision specified below was implemented on 4 November 2026. The execution
record at the end identifies delivered files, verification and remaining
experimental limitations. The plan is retained as the revision specification.
All manuscript text, captions, axes, annotations, tables and analysis reports
are in English. Authors' names and official institutional names are preserved.

Make the 4 November 2026 outdoor acquisition the principal experimental case
for the WiT-CPV controller. Build the argument around sustained focal tracking,
aperture-referenced electrical performance, observed operating conditions and
the practical limits revealed by the complete record. A strong central result
must retain initialization, interventions, missing intervals and performance
losses rather than selecting only the most favorable period.

Final title, selected after auditing firmware provenance:

**Outdoor Characterization of Photocurrent-Based Focal Tracking in a
Tracking-Integrated Micro-Concentrator Photovoltaic System**

## Source Of Truth

- Main manuscript: `articulo/PRINCIPAL.tex`.
- Bibliography: `articulo/cas-refs.bib`.
- Principal experiment: `logs_nuevos/log_2026-11-04.txt`.
- Solar-position reference: `logs_nuevos/suncalc_log_2026-11-04.pdf`.
- Existing analysis patterns: `analyze_2026_11_03.py` and its numerical tests.
- Supporting, not headline, experiment: the quality-controlled 3 November record.
- Historical September/October experiments: retain only independently supported
  information necessary to explain development or a distinct control behavior.

The current main article has not yet incorporated November results. Its abstract
and highlights (approximately lines 116-134) lead with September 2025 and a
selected October window; most results (approximately lines 400-738) develop
those earlier records. The generated November 3 section exists separately and
is not included by the main TeX source. This calls for replacement and editorial
restructuring, not another appended daily case study. Line ranges refer to the
source inspected during planning and will change during revision.

Do not overwrite raw logs, silently rescale historical experiments, delete old
analysis artifacts, change firmware or deploy the hub as part of article work.
Preserve the existing main PDF until a successfully compiled revision exists.

## Preliminary Numerical Audit

These are read-only planning calculations, not final publication results. The
integrated energy, irradiation, duration, aperture ratio, basic ADC quality and
extreme responsivity observation were independently checked with Python 3.12
during planning. The implementation must still produce
the audited artifacts and validate sensor assumptions before manuscript insertion.

| Item | Preliminary finding | Consequence for the revision |
| --- | --- | --- |
| Acquired record | 5,049 electrical/coordinate samples; 09:52:50-16:05:13 | Describe the acquired daytime span, not sunrise-to-sunset coverage. |
| Recorded firmware | 1.1.3-ws | Do not credit subsequent software changes. |
| ADC quality | No clipped, near-current-rail or negative main-channel observations in this audit | This is cleaner than the clipped November 3 record, but absence of clipping does not validate every sensor observation. |
| Integration support | Approximately 6.121 h using adjacent intervals no longer than 15 s | Report support and missing time explicitly. |
| Full-record modeled energy | Approximately 3.746 Wh at FF = 0.82 | Available-MPP estimate, not measured delivered energy. |
| Full-record aperture estimate | Approximately 23.41%, energy-weighted | Uses confirmed coplanar irradiance; conditional on calibration and FF, not intrinsic cell efficiency. |
| Control conditions | 61 logged remote commands, including tuning and manual repositioning | Do not label the full record an uninterrupted autonomous, fixed-parameter trial. |
| Late behavior | Substantial output degradation toward the end | Retain it in the central result; examine operating limits without asserting an unmeasured cause. |
| Responsivity reference | The maximum I/G is dominated by an irradiance anomaly near 13:01 | A maximum-derived "tracking efficiency" is not a defensible headline without anomaly/reference analysis. |

Five sample gaps exceed 15 s in the independently checked parse:
09:53:10-09:53:51 (41 s), 09:54:11-09:57:17 (186 s),
10:34:38-10:34:59 (21 s), 16:03:43-16:04:23 (40 s), and
16:04:43-16:05:01 (18 s). No decreasing main-sample uptime was found in that
check; reset/reconnection events and TEL counters still require separate auditing.
These event boundaries must be regenerated by code rather than manually
encoded as an undocumented exclusion list.

At 13:01:07 the preliminary audit finds G near 313 W/m2 with current near
337 mA, producing a maximum responsivity near 0.001078 A m2/W, versus a
99th percentile near 0.000420 A m2/W. This large reference sensitivity is a
diagnostic, not grounds to delete an inconvenient observation. Inspect nearby
raw channels, timing and events; possible sensor shading, weather variability
or asynchronous measurement remain hypotheses until supported. Do not claim
that this log demonstrates every control regime or independently establishes
tracking accuracy.

## Established Calibration And Interpretation

Use signed 24-bit ADC values and positive full-scale code D = 8,388,607:

```text
Isc [A] = C_signed / D * 2.048 / (20 * 0.130)
Voc [V] = V_signed / D * 2.048 * 2
G [W/m2] = P_signed / D * 2048 / 0.2
P_hat [W] = Isc * Voc * FF, with nominal FF = 0.82
eta_ap_hat = P_hat / (0.002497 * G)
```

- The physical 130-milliohm shunt applies to the 4 November record. A hub
  conversion update that morning is not a physical shunt intervention.
- The pyranometer divisor is 0.2, not the obsolete value 2.0.
- The lens aperture is 0.002497 m2. Aperture efficiency includes the optical
  train and receiver; it is not intrinsic multi-junction cell efficiency.
- The X axis is azimuth; the Y axis is elevation. Audit nomenclature,
  electronics labels, equations, position plots and prose consistently.
- Alternating Isc/Voc sensing does not measure MPP delivered to a load.
  Use "estimated available MPP power" and "integrated FF-based energy
  estimate", not measured harvested energy or demonstrated net production.
- A constant FF is an assumption, not a field I-V measurement. Show linear
  sensitivity to FF without presenting a chosen range as a confidence interval.
- I/G normalization does not remove cell-temperature drift or spectral effects.
- Correcting irradiance offline does not retrospectively alter firmware
  thresholds, standby decisions or the control law used during acquisition.
- Attribute results to the firmware actually recorded, not a later compiled
  version or the subsequently developed configurable X:Y allocation.
- TEL counts fine-tuning steps only. Coordinate differences capture sampled
  net displacement and can miss reversals between observations. Neither alone
  establishes total motor travel or whole-system energy consumption.

## Work Package 1: Freeze And Audit The Principal Experiment

1. Record source-file hashes, byte sizes, date, parsing rules, library versions
   and units in a machine-readable analysis summary. Preserve source line numbers.
2. Parse electrical/coordinate observations, uptime, TEL messages, firmware
   versions, controller states, parameter updates and manual coordinate commands.
3. Report acquisition span, typical sampling interval, duplicates, missing
   intervals, reboots, clipping, near-rail values and invalid polarity. Flag
   questionable observations without reconstructing clipped current.
4. Build a chronological event table. Distinguish initialization, commanded
   repositioning, parameter changes, tracking intervals and the final low-output
   period. Do not call the whole record autonomous or fixed-parameter unless
   supported by that audit.
5. Define primary integration support before reporting favorable subwindows.
   Require adjacent accepted endpoints, positive elapsed time, no uptime reset
   and a maximum interval of 15 s. Never join rows across rejected observations.
6. Report the complete acquired record and any event-defined operational phases
   together. Give both elapsed duration and accepted integration duration, with
   coverage. A working-range subwindow must not replace the full-record result.
7. Test sensitivity to gap limits and illumination thresholds. Do not describe
   the acquisition as sunrise-to-sunset, all-day energy or a replicated campaign.

Deliverables: `analyze_2026_11_04.py`, focused numerical/source-integrity tests,
an audit CSV under `figures/2026-11-04/`, JSON summary and event table.
The exact output names may follow existing analysis conventions.

## Work Package 2: Quantify Performance And Solar Context

### Electrical And Relative-Tracking Metrics

- Compute coincident Isc, Voc, G and estimated MPP power; report timestamps at
  peaks rather than combining noncoincident maxima into an artificial operating
  point.
- Integrate estimated power and measured irradiance on identical accepted
  support. Report energy-weighted aperture efficiency as the ratio of integrals,
  separately from median or maximum instantaneous efficiency.
- Define an explicitly empirical same-record responsivity reference and an
  integrated relative-yield indicator. Compare maximum and 99th-percentile
  references. Do not call this independently validated ideal tracking efficiency.
- Make the observed reference anomaly a release gate: document it and test
  defensible reference definitions before choosing a headline tracking metric.
  Do not automatically prefer P99 merely because it yields a higher score.
- Keep sample occupancy, time occupancy and integrated relative yield separate.
  State thresholds, denominators, reference windows and interpolation rules.
- Report median, spread and time evolution of I/G. Interpret changes jointly
  with irradiance, voltage, position and logged interventions, without assigning
  an unmeasured thermal or optical cause as established fact.
- Evaluate electrical range and end-of-session loss of response. Distinguish
  observed operation from rated travel limits or a validated angular acceptance
  envelope.

### Solar Position And Weather

- Transcribe the supplied SunCalc table into a traceable CSV with its exact
  printed date and coordinates: 21.87304 degrees N, 102.28282 degrees W.
- Preserve azimuth convention and label elevation separately from zenith.
  Verify the relation between the PDF's local clock and acquisition timestamps;
  the log does not itself encode a time-zone offset.
- Use the provided 15-minute positions for context. If interpolating between
  entries, label interpolation and do not imply sub-minute astronomical accuracy.
- Plot the acquired interval against azimuth and elevation, including the
  observed useful-response range and any subsequent decline.
- SunCalc supplies solar direction, not receiver pointing error, optical
  incidence angle or proof of maintaining perfect focus. Module tilt and
  orientation are needed to convert solar position into incidence geometry.
- Record that the experimenter described the day as clear, distinguishing that
  field observation from an independently instrumented weather classification.
  Assess the irradiance trace and its brief anomalies before adopting an
  unqualified clear-sky claim. A smooth pyranometer signal alone does not certify
  cloud-free conditions or provide DNI. "Experimenter-reported clear conditions"
  is acceptable provenance; stronger radiometric claims require further evidence.
- A clear-condition record cannot by itself validate cloud-transient shielding,
  storm operation or recovery under rapidly changing irradiance. If one earlier
  experiment provides genuinely independent, quality-controlled evidence for a
  retained algorithm claim, keep that compact supporting test; otherwise narrow
  the claim instead of filling the paper with historical traces.

### Core Figures And Tables

1. **Field-day overview:** aligned panels for irradiance, current/voltage,
   estimated power/aperture efficiency, and X/Y coordinates. Mark gaps and
   intervention events without smoothing away losses.
2. **Solar path and operating context:** SunCalc azimuth/elevation over the
   acquisition interval, linked to measured response. Separate absolute solar
   position from any unmeasured collector incidence or tracking error.
3. **Tracking and operating envelope:** irradiance-normalized response and
   event-defined tracking phases, with a short representative convergence
   detail only if it adds evidence not visible in the overview.
4. **Fixed-silicon benchmark:** a clearly labeled modeled comparison on common
   temporal and irradiance support, showing assumptions and sensitivity.
5. **Primary results table:** full acquired support plus disclosed operational
   phases; samples, integrated hours, irradiation, estimated energy,
   energy-weighted aperture efficiency and relative-yield indicator.
6. **Event/quality table:** a compact main-text version or supplementary table
   giving interventions, exclusions and missing intervals.

Prefer these focused figures over a gallery of historical daily traces.

## Work Package 3: Build A Defensible Fixed-Silicon Comparison

The WiT-CPV system has fixed external optics and internally moving receivers.
Do not equate it with an externally dual-axis-tracked module receiving DNI at
normal incidence. Both the collection geometry and the comparison basis must
be stated explicitly.

### Comparison Available Without A Reference Panel

Use an **equal-aperture, common-irradiance engineering scenario**, not a claim of
measured superiority over silicon:

```text
P_Si_scenario(t) = A_reference * eta_Si_assumed * G_common(t)
E_Si_scenario = integral[P_Si_scenario(t) dt] on CPV accepted support
R_aperture = E_CPV_hat / E_Si_scenario
```

Set A_reference = 0.002497 m2 for aperture-normalized comparison. Explain that
this is an area-equivalent scenario, not an actual fabricated silicon panel.
Use a transparent, source-backed silicon-efficiency assumption or sensitivity
range. Keep the assumed module efficiency distinct from cell record efficiency.

The verified PVWatts module-type documentation provides representative Standard
and Premium assumptions of approximately 19% and 21% nominal efficiency, with
power-temperature coefficients of -0.37%/degree C and -0.35%/degree C,
respectively. These are model categories, not a measured reference panel or
universal bounds on crystalline silicon. Cite the version/access date and replace
them with a real module datasheet when one is available. Do not apply the thermal
coefficients without measured or explicitly modeled silicon-cell temperature.

The simple scenario holds efficiency constant; it is not a weather- and
temperature-corrected yield prediction. If thermal data become available, use
a documented silicon module coefficient and measured/model-qualified cell
temperature. Never assign unmeasured temperature as fact.

The experimenter confirmed on 4 November 2026 that the pyranometer is coplanar
with the concentration lens. The measured G therefore represents global
irradiance in the fixed aperture plane; G times A_L is the incident-power
estimate and must not receive a second cosine projection. This does not make
G direct-normal irradiance, nor imply ideal sensor angular/spectral response.

The silicon scenario assumes a hypothetical fixed surface in that same plane.
Numerical CPV-to-silicon ratios remain conditional, not an observed prototype
advantage, because silicon output is modeled and CPV MPP output is estimated.
A more complete PVWatts-style model additionally needs module optical corrections,
temperature and a documented model version, including its low-irradiance
treatment when applicable.

Equal rated power is a different question. Add a rated-power-specific comparison
only if a validated CPV rating exists; do not invent Wp from a selected peak.
Do not compare the tiny CPV output directly to a commercial panel's absolute
wattage, annual yield, cost or lifecycle performance.

### Minimum Dataset For An Experimental Comparison

Request or obtain the following without delaying the modeled scenario:

- Timestamped fixed-silicon DC power at MPP, or simultaneous voltage/current
  under a specified MPPT/load; acquisition interval, units and local time offset.
- Panel model, active/aperture area definition, rated Pmax, nominal efficiency
  and power-temperature coefficient from its datasheet.
- Absolute tilt and azimuth of the shared lens/pyranometer plane and the silicon
  panel, plus sensor calibration. Relative lens/sensor coplanarity is confirmed.
- Simultaneous POA irradiance in the relevant plane, ideally with DNI/DHI if
  separating concentration and diffuse-light collection is a goal.
- Panel/cell or backsheet temperature, ambient temperature if available, and
  shading/soiling information.
- CPV board and motor power if a net-energy comparison is intended. A genuine
  comparison of delivered energy also requires CPV operation into a measured
  electrical load, not only alternating Isc/Voc samples.

Nonconcurrent silicon records can support a weather-normalized contextual
comparison, not a simultaneous head-to-head experiment. Report the remaining
geometry, thermal and electrical-measurement asymmetries explicitly.

### Verified Reference Sources

- PVWatts module-type assumptions:
  https://pvwatts.nlr.gov/docs/help_module_type.php
- Sandia PV Performance Modeling Collaborative, plane-of-array irradiance:
  https://pvpmc.sandia.gov/modeling-guide/1-weather-design-inputs/plane-of-array-poa-irradiance/
- Sandia PV Performance Modeling Collaborative, PVWatts DC model:
  https://pvpmc.sandia.gov/modeling-guide/2-dc-module-iv/point-value-models/pvwatts/
- PVWatts V8 model inputs and outputs:
  https://developer.nlr.gov/docs/solar/pvwatts/v8/
- Apogee SP-212/SP-215 pyranometer manual, including angular and spectral
  response limitations:
  https://www.apogeeinstruments.com/content/SP-212-215-manual.pdf

Convert retained sources to verified bibliography entries during implementation.
SunCalc provenance must cite the supplied dated PDF and website, not an invented
irradiance or module-orientation dataset.

## Work Package 4: Refocus The Manuscript

### Proposed Main-Text Structure

1. Introduction: the internal focal-tracking problem, prior prototype and the
   specific control/evaluation contribution; compress generic solar-tracker review.
2. System and control: optical/mechanical context, correct physical axis mapping,
   evaluated hardware/firmware, implemented control and acquisition sequence.
3. Experimental methods: site/date, SunCalc reference, configuration chronology,
   calibration, data-quality rules and honest definitions of estimated metrics.
4. Results: complete 4 November record, event-defined operating phases,
   aperture performance, solar-position context and observed limits.
5. Fixed-silicon comparison and discussion: common-area scenario first;
   measured comparison only if the required reference dataset is supplied.
6. Conclusions: quantitative demonstrated capabilities, observed limitations
   and specific next validation steps, all consistent with the abstract.

### Targeted Editorial Changes

- Replace the abstract and highlights currently led by September 2025 and a
  selected October two-hour window. Use the new primary results only after
  numerical verification, including their FF-based and coverage qualifiers.
- Replace the October-focused outdoor section and extensive September results
  with the principal November case. Do not merely append another daily report.
- The main replacement targets in the currently inspected source are the
  October section (about lines 400-479), September cross-day comparison
  (481-546), detailed September session (548-738), and conclusions (740-751).
  Resolve the empty dynamics/mechanical-model headings near lines 242-245;
  they must not remain as unfulfilled sections.
- Remove repetitive cross-day tables, redundant old power/position plots and
  disturbance anecdotes that do not answer a distinct research question.
- Retain concise laboratory validation only if its methods and evidence are
  traceable; clearly separate previously reported component characterization
  from new outdoor results. No unsupported solar-simulator performance claims.
- Keep November 3 at most as a brief calibration/ADC-range supporting case or
  supplement. Preserve its piecewise 260-to-130-milliohm analysis and clipping
  exclusions; it is not a clean paired causal test of hardware improvement.
- Remove unused cloud-recovery definitions, clear-sky fitted models and
  nomenclature once their corresponding experiments leave the main text.
- Correct stale 75/260-milliohm descriptions for the evaluated November hardware,
  without rewriting historical experiments to use the current shunt.
- Reconcile algorithm description against the evaluated implementation.
  Remove claims that later software features were demonstrated by this log.
- Replace assertions of eliminated hunting, reduced wear, causal cloud recovery,
  perfect lock or net energy benefit with measurements or appropriately bounded
  interpretations. Remove them when neither is available.
- Rewrite conclusions after results and discussion, then update abstract,
  highlights, title and nomenclature last to prevent contradictory headline claims.

### Conclusion Contract

The revised conclusion should state four things with measured or explicitly
modeled quantities: the duration/support of the demonstration, estimated
aperture-level performance, the observed solar-position/operating range and
the outcome of the assumption-labeled silicon scenario. It should also state
what remains unproven: delivered MPP energy, net system energy, independent
angular accuracy, multi-day repeatability and simultaneous silicon superiority.
Strength comes from a coherent supported result, not stronger adjectives.

## Work Package 5: Reproducibility And Release Checks

1. Add tests for signed conversion, current scaling, pyranometer scaling,
   clipping, rejected rows, gaps, uptime resets, event-window selection,
   trapezoidal integration and identical comparison support.
2. Test fixed-silicon scenario scaling with area and assumed efficiency, and
   FF scaling of CPV estimates. Preserve raw-code/source-line integrity.
3. Regenerate all numerical tables and figures from the audited raw source;
   include source hashes and assumptions in the machine-readable summary.
4. Independently check a sample conversion, a coincident peak, one integration
   and the energy-weighted efficiency before inserting headline numbers.
5. Check every in-text number against generated output and every caption
   against its plotted support. Retain failure periods and disclose exclusions.
6. Audit visible English throughout, including embedded figure text. Proper
   names and official institution names are not translation errors.
7. Update bibliography with verified primary references for the silicon model,
   sensor/model assumptions and SunCalc provenance; remove unused references.
8. Compile from `articulo` with the available MiKTeX/Biber toolchain:

```text
pdflatex -interaction=batchmode -halt-on-error -file-line-error PRINCIPAL.tex
biber PRINCIPAL
pdflatex -interaction=batchmode -halt-on-error -file-line-error PRINCIPAL.tex
pdflatex -interaction=batchmode -halt-on-error -file-line-error PRINCIPAL.tex
```

9. Inspect warnings, undefined references/citations, page breaks, floating
   figures, table width and readability in the final PDF. Successful compilation
   alone is not scientific or visual validation.

The working scientific interpreter is `py -3.12`. The default `python` currently
resolves to Python 3.11 without Matplotlib, so use the existing 3.12 environment
instead of assuming the default command has the required analysis packages.
Both `pdflatex` and `biber` are available; the planning pass has not recompiled
the unchanged manuscript.

## Completion Criteria

- The principal experimental narrative is November 4, not a sequence of
  unrelated historical sessions.
- No numerical headline is unsupported by a reproducible output and clearly
  defined observation/integration window.
- The manuscript never confuses modeled MPP output with harvested energy,
  relative response with angular accuracy, or SunCalc context with measured lock.
- The silicon comparison has explicit geometry, area, electrical-measurement
  basis and uncertainty; modeled results cannot be mistaken for observations.
- The revised manuscript, figures, captions, tables and analysis reports are
  in English, with coherent axes, units and calibration.
- Relevant numerical tests pass and the final main PDF compiles without
  unresolved references, followed by a visual and independent scientific review.

## Parallel Work And Integration

The planning audit uses three independent internal subagents: raw-log analysis,
manuscript evidence review and fixed-silicon comparison research. No visible
Agent Manager sessions or additional worktrees are needed.

During implementation, one worker may own the November analysis and generated
assets while another researches/verifies comparison references. Keep one owner
for `PRINCIPAL.tex` to avoid conflicting editorial changes. The integrating
agent verifies numbers, edits the narrative and compiles the final article.

## Execution Record: 4 November 2026

### Delivered Revision

- Rebuilt `articulo/PRINCIPAL.tex` around the November 4 acquisition, retaining
  authorship and affiliation metadata. September/October development narratives,
  unused metric definitions and inconsistent historical diagrams no longer
  compete with the principal case. Original logs and historical analysis assets
  were not modified.
- Generated `articulo/results_2026_11_04.tex` and
  `articulo/metrics_2026_11_04.tex` directly from the new analysis. The abstract,
  highlights and conclusions share generated numerical macros with the results.
- Produced the main `articulo/PRINCIPAL.pdf`: 14 pages, five figures including
  the revised functional architecture, two results tables and 14 cited sources.
  A separately built dated validation PDF is also available.
- Updated bibliography metadata against primary records, including the published
  tracking-integrated CPV review instead of its preprint, conference author order,
  verified DOIs, and the new SunCalc, pyranometer and silicon-model references.
- Added `analyze_2026_11_04.py` and `test_analyze_2026_11_04.py`.
  `figures/2026-11-04/` contains the raw-code audit, accepted-edge support,
  event chronology, SunCalc transcription, window metrics, sensitivity checks,
  source/software provenance in `analysis_summary.json`, and PDF/PNG figures.

### Verified Results And Scope

- 5,049 main observations and 5,026 separate telemetry messages; five sample gaps
  longer than 15 s; zero clipped, near-current-rail or negative main observations;
  no decreasing main or TEL uptimes.
- Full-record accepted support: 6.121 h, 98.63% of the acquired elapsed span.
  Reference irradiation: 6.408 kWh/m2. Available-MPP estimate: 3.746 Wh at
  FF = 0.82. Energy-weighted aperture estimate: 23.41%.
- The event-defined 12:32:02-13:53:00 phase yields 28.47% and 1.126 Wh over
  1.349 h, without remote commands inside that interval. Earlier tuning,
  including the 12:27:20 delay command, is explicitly disclosed.
- The full 15:00-16:00 clock-hour result and all subsequent observations remain
  included. No causal explanation for the late deterioration is asserted.
- The irradiance anomaly remains in the source and analysis. Relative-yield
  sensitivities are 28.90% for the maximum reference and 74.07% for P99;
  neither is promoted as validated angular tracking efficiency.
- The 19% and 21% equal-area silicon assumptions give 3.040 and 3.360 Wh on the
  shared pyranometer reference. Ratios 1.232 and 1.115 are explicitly algebraic
  scenario results, not measured comparative yields.

### Verification Completed

```text
py -3.12 analyze_2026_11_04.py
py -3.12 -m unittest -v test_analyze_2026_11_03 test_analyze_2026_11_04
```

All 27 tests pass: 13 existing November 3 tests and 14 November 4 tests.
Checks cover source quality and event provenance, signed calibration, invalid
observations, gaps, resets, trapezoidal support, FF invariance/scaling,
equal-area scenario scaling and SunCalc interpolation without extrapolation.
An independent artifact check also verified all 5,049 raw observations against
the exported audit CSV, the 5,043 accepted integration edges, and both the raw-log
and analysis-source fingerprints.

The main PDF was compiled with the documented pdflatex/Biber sequence. The final
LaTeX log contains no warnings, overfull/underfull boxes or unresolved references;
the Biber log contains no warnings/errors. Visual review covered all-page layout,
architecture labels, tables, the full acquisition plot, solar-response colors,
reference sensitivity and scenario chart. The final PDF contains the new
23.41%, 28.47%, 1.232 and 1.115 values, not the retired 67.3%/88.5% headlines.

### Remaining Experimental Information

The manuscript does not claim simultaneous silicon superiority, delivered MPP
energy, net system energy or independent angular accuracy. A stronger physical
comparison still requires silicon MPP monitoring and temperature, a documented
matched panel plane, plus actual CPV load and supply-power
measurements when delivered or net energy is claimed. The recorded 1.1.3-ws
deployment is not presented as a validation of later firmware corrections.

### Coplanar Sensor Confirmation

The experimenter subsequently confirmed that the pyranometer and concentration
lens share the same plane. This resolves the earlier uncertainty about whether
the irradiance reference represents the aperture plane. Methods, discussion and
reproducibility metadata now identify G as global plane-of-aperture irradiance.
No numerical recalibration or additional cosine factor is introduced. Absolute
tilt/azimuth are still needed for solar-incidence-angle reconstruction, not for
the aperture-energy denominator obtained from the coplanar sensor.
The result text now explicitly distinguishes the 33.21% instantaneous estimate
at the maximum-power observation from the 28.47% phase estimate and 23.41%
full-record estimate. The analysis metadata records the experimenter's geometry
confirmation and the absence of an additional cosine factor. All 27 tests pass;
the main and dated validation PDFs were rebuilt with unchanged numerical results.