---
{
  "path": "/en/research/data",
  "title": "Data and model exports",
  "category": "Research resources",
  "summary": "Provenance and limits of existing model exports, geometric results, synthetic inputs and empty templates; real-data gates remain explicit.",
  "related": [
    "/en/theory/lve",
    "/en/tests/case-studies",
    "/en/research/code"
  ],
  "id": "ASD-DOC-RESEARCH-DATA-EN",
  "template": "document",
  "version": "1.1",
  "updated": "2026-10-03",
  "canonical": false,
  "source_tier": "website-guide",
  "depends_on": [],
  "open_problems": [],
  "equations": [],
  "sources": [
    {
      "label": "ASD 建站蓝图与页面模板（用户提供）",
      "kind": "supplied-blueprint"
    }
  ],
  "theory_status": "OPEN",
  "evidence_status": "DOCUMENT",
  "title_en": "Data and model exports",
  "locale": "en",
  "translation_path": "/research/data",
  "translation_of": "/research/data",
  "translation_version": "1.1",
  "canonical_url": "https://asd-physics.org/en/research/data",
  "section_anchors": [
    "s0",
    "comparison-data",
    "s1",
    "s2"
  ]
}
---

## Available data

Available downloads include formula-generated LVE tables, directional-clock geometry results, experimental comparison baselines and empty input templates. Each answers a specific calculation question; they cannot be combined as one observational dataset. Scripts, run commands and environments are listed in the [computation catalog](/en/research/code).

The [LVE demonstration script](/downloads/ASD_LVE_demo.py) generates CSV values from $\gamma=(1-\beta^2)^{-1/2}$ and the force ratio for the specified equal-speed parallel configuration. Running `python ASD_LVE_demo.py --beta 0.6 --csv lve-model.csv` always writes 96 model rows from $\beta=0.00$ to $0.95$; `--beta` controls the single terminal point. The [LVE](/en/theory/lve) interactive export likewise uses published relations, without original measurements or experimental-error columns. The force quantity is defined on the [magnetic-force page](/en/theory/electromagnetism/magnetic-force).

The [clock pilot](/en/research/clock-response-pilot) distributes its script, apparatus record and frozen results under `tasks/` in [ASD_AI_Research_Pack.zip](/downloads/ASD_AI_Research_Pack.zip). Its output records ranks and symmetry checks for a 96-sample design matrix; `asd_frequency_amplitude` is `null` and `observations_used` is `false`. This is an intermediate calculation for an assumed geometric template, not a frequency time series or experimental validation.

## Comparison-model inputs and outputs

After extracting [ASD_Experimental_Reproducibility_Kit.zip](/downloads/ASD_Experimental_Reproducibility_Kit.zip), run `python run.py --output results` for the following three sets of outputs, and `python verify.py` for file hashes, frozen values and limits. The bundled `equations.md` records equations, reference constants and missing ASD mappings; `environment.json` describes the environment. All example inputs are bundled, with no network data retrieval.

Clock geometry evaluates the projection of a user-selected background axis using $Q=(\mathbf n\cdot\mathbf b)^2-1/3$. Defaults are latitude $40^\circ$, horizontal-axis azimuth $0^\circ$, declination $30^\circ$, and right ascension and starting phase $0^\circ$. `clock_geometry.csv` has 97 half-hour samples from 0 to 48 hours. Phase uses a nominal sidereal day, not precise UTC/UT1 conversion; this is a different window from the 96-sample two-clock rank analysis above. Combining four illustrative independent $10^{-18}$ inputs gives $2\times10^{-18}$ in `summary.json`; it is only an uncertainty-combination example, not a harmonic-fit error or required instrument precision.

Muon calculations use the standard reference relations $P=\exp[-L/(\beta\gamma c\tau_\mu)]$, $L=h/\cos\theta$ and $N=N_0\epsilon P$, assuming a flat atmosphere and constant energy without energy loss or scattering; zenith is restricted to $0^\circ$–$75^\circ$. Defaults are kinetic energy 3 GeV, height 15 km, zenith $0^\circ$, efficiency 0.8 and initial cohort 1000, giving $P=0.4605851746$ and $N=368.4681397$. The three bins in `input/muon_bins_synthetic.csv` are synthetic pre-decay cohorts; the example sum in `muon_bins.csv` is 216.2416691. It includes no measured production spectrum, ground flux or energy migration. ASD path relations have not independently established a decay hazard; see the [muon requirements](/en/tests/predictions#ASD-PRED-002).

High-γ charge calculations use the standard normalized transverse field of a uniformly moving point charge, $E(t)/E(0)=[1+(\gamma\beta ct/b)^2]^{-3/2}$. Defaults are electron kinetic energy 35 MeV and impact parameter $b=2$ mm, with 601 samples over −6 to 6 ps in 0.02 ps steps. `charge_response.csv` also contains an assumed readout obtained by convolution with a unit-area Gaussian of $\sigma=0.2$ ps: the kernel is truncated at $\pm5\sigma$ and renormalized, the field is zero outside the window, and the filtered peak is not renormalized to 1. The ideal peak is 1 and the assumed filtered peak is 0.3338975146. Precision use still requires window, step and kernel-truncation convergence checks. No bunch, conductor boundary, crystal dispersion, phase matching, laser or gain calibration is included, so this does not reproduce Ota data. The actual source/probe problem is recorded in the [high-γ case study](/en/tests/case-studies/high-gamma-charge-field).

`asd_output`, `predicted_difference` and `fit` in `results/summary.json` are all `null`. JSON, CSV and diagnostic figures in `expected/` are reproducibility benchmarks; their model/synthetic identity must be preserved rather than relabelled ASD observations.

## Experimental data availability

These downloads currently contain no real observational-comparison project. Original S301 records, moving-charge/electro-optic measurements, precision-clock time series, apparatus covariance and calibration files are not included; the website reports no new fits made from them. A public literature reference does not mean raw data have been imported here.

`schemas/s301.csv` contains headers only. Object identity, epochs, time scales, coordinate conventions and covariance still need verification. There is no orbital solver or fit output, and the [registered data gate](/en/tests/predictions#ASD-PRED-004) remains. Ota raw data and code are also absent from these packages; a standard point-charge curve and assumed Gaussian response cannot replace its actual measurements. Source and reuse conditions, apparatus models and analysis protocols must be checked after materials are obtained, before a real comparison can begin.

## A reproducible data package

`schemas/clock.csv`, `muon.csv`, `high_gamma.csv` and `s301.csv` in the kit are four header-only templates, with units and field definitions in `schemas/README.md`. They contain no data rows. Missing physical inputs must remain missing rather than be filled with zero. Each real dataset needs a separate `provenance.json` recording URL/DOI, author or institution, acquisition date, license, original-file SHA-256, processing-code version, selection rules, units and missing-value conventions. `data_kind` distinguishes `OBSERVED`, `PROCESSED`, `MODEL` and `SYNTHETIC`.

Also document reference frame, time scale, zero points, calibration/test split, statistical/systematic errors and covariance indexing. Clock ratios retain correlations of reference-clock, link and servo corrections. Muon data specify kinetic/total energy, whether acceptance already includes efficiency, and migration matrices, avoiding duplicated efficiency factors. High-γ data separate raw/reconstructed signals, transfer functions, gain and bunch inputs. S301 data state whether RA offsets include $\cos\delta$, the origin and radial-velocity applicability. Processed data require runnable transformation steps; review and contribution use the existing [reproducibility standard](/en/tests/reproducibility). Filling a table does not itself complete theory or data verification.