---
title: "Geometrical Measurement of an Ancient Conic Sundial from a 3D Surface Model"
subtitle: "Construction Witnesses, Consensus Estimation, and the Proportions of Dialface ObjID 18 (British Museum inv. 2547)"
running-title: "Sundial Measurement ObjID 18"
book: false
toc: true
author:
  - name: Gerd Graßhoff
    email: gerd.grasshoff@hu-berlin.de
    orcid: 0000-0002-4450-357X
    affiliations:
      - name: Humboldt-Universität zu Berlin
        department: Philosophisches Institut
        address: Unter den Linden 6
        city: Berlin
        country: Germany
        postal-code: 10099
      - name: Deutsches Museum
        department: Forschungsinstitut
        address: Museumsinsel 1
        city: München
        country: Germany
        postal-code: 80538
abstract: |
  We measure the engraved line network and the dressed faces of the conic
  sundial fragment Dialface ObjID 18 (British Museum, inv. 2547) from a
  photogrammetric 3D surface model, and deduce the construction values of the
  instrument in their historically meaningful form: as proportions of the ancient
  shadow-table (klima) tradition. The analysis is organized around the
  *construction dependency* of the dial: each measurable feature realizes one
  construction parameter, and the latitude proportion $s:g$ enters the artifact
  along three separate construction paths — the dressing of the upper front
  plane, the elevation of the cone axis, and the hour division of the solstitial
  day-lines, the latter calibrated by the equinoctial design invariant $H_E = 90^\circ$
  and made obliquity-free by pairing each circle's placement with its own division.
  These paths furnish independent *witnesses* of $s:g$, combined by a joint
  measurement-level bootstrap (full witness covariance) and a hierarchical model
  whose profile-likelihood interval separates measurement noise from the
  artisan's stonecutting spread; discordant witnesses are excluded only by a
  declared leave-one-out rule ($>4\sigma$) and reported as construction-error
  candidates. For ObjID 18 the result is $\varphi = 41.45^\circ$, $68\%$ CI
  $[41.16^\circ, 41.75^\circ]$ ($s = 53;02$ for $g = 60$; the ratios $8:9$ and
  $9:10$ are admissible, and the klima entries Hellespont and Rome of Ptolemy's
  shadow table lie in range), while the hour division of the summer day-line
  deviates at $14\sigma$ — a localized ancient construction error. The axial
  distances of the three day-lines stand in the proportion
  $\ell_W : \ell_E : \ell_S = 12 : 15 : 20$ (within $0.6\sigma$), equivalent to
  the workshop rule $\tan\alpha\tan\varepsilon = 1:4$. The engraved hour lines are
  circular arcs (radii $113$--$280\,\mathrm{mm}$), not straight generators.
keywords:
  - conic sundial
  - 3D photogrammetry
  - ObjID 18
  - British Museum inv. 2547
  - construction witnesses
  - consensus estimation
  - Ptolemy shadow table
  - Hellenistic horology
bibliography: references.bib
format:
  arxivai-typst: default
---

::: {.callout-note title="Supplement S5" #supplement-box}
Supplement S5 to the Ancient Sundials repository (Berlin Sundial Collaboration, ancient-sundials.lettre.ai). This draft documents a worked example for the object page of ObjID 18. It is a preprint draft and has not been peer reviewed. Where a figure in the repository differs from a figure here, the repository, which is rebuilt from the current database, is authoritative.

Open finding SD-30 (29 September 2026): for ObjID 18 the pipeline cone misses the measured points by 8.9 mm, where verified cones fit within 0.39 to 0.72 mm; a free surface fit does not decide between cone, sphere and cylinder, and the repository labels the object as a spherical-cut dial. Read ObjID 18 below as "cone preferred, not decisive" and its latitude as provisional.
:::


# Goal and Methodological Approach

The construction geometry of conic sundials with a nodus on the polar axis is
developed analytically in the companion keystone paper [@grasshoff2026conic]:
the dialface is the concave inner surface of a right circular cone whose axis is
aligned with the celestial rotation axis; the shadow of the nodus traces, for
each fixed solar declination, an exact circle on the cone perpendicular to the
axis (a *day-line*); and hour lines connect shadow positions of equal seasonal
hour across the year. The companion paper establishes that ancient dialmakers
designed such instruments *proportionally*:

* The latitude enters as the klima shadow ratio $\sigma_E = s_{\mathrm{eq}}/g = \tan\varphi$
  (equinoctial noon shadow per vertical gnomon of $g = 60$ parts).
* The cone opening enters as the ratio $\rho = R_{\mathrm{rim}}/H = \tan\alpha$.
* The seasonal noon marks enter as ratios of their axial distances $\ell$ from
  the apex, anchored by the equinox invariant $\ell_E = d$.

This paper performs the inverse, empirical deduction for one specific artifact:
Dialface ObjID 18 (British Museum inv. 2547; published photogrammetric model
BSDP0300) [@bsdp2014dialface18].

```{=typst}
#arxivai.concept-box(title: [The three principles of empirical deduction])[
  1. Construction Witnesses: Every measurable feature that realizes a design input is treated as an independent witness.
  2. Invariant Calibration: Design invariants (such as $H_E = 90 degree$) calibrate the artisan's engraved angular unit.
  3. Consensus and Error Attribution: Witnesses are combined by joint maximum likelihood; significant outliers are isolated as localized historical errors rather than blurred by global averaging.
]
```

The artifact, a marble conic dial fragment held by the British Museum (inv. 2547),
preserves eleven hour lines, three day-lines — equinoctial (*Equi*), summer
solstitial (*SuSo*), and winter solstitial (*WiSo*) — and several dressed faces of
the block. @fig-overview displays the 3D surface model with the fitted feature network.

![Surface model of Dialface ObjID 18 (British Museum inv. 2547) with the fitted feature network: hour-line arcs (blue, with on-surface extrapolations), day-line circles (orange), medial groove centrelines (yellow), and all 33 curve-to-curve intersections (red: preserved; white: extrapolated).](fig_overview.png){#fig-overview}

# The Construction Dependency of the Witnesses

@fig-dag traces the construction of the conic dial from its four fundamental
design inputs — the klima proportion $s : g$, the epoch obliquity $\varepsilon$,
the cone opening $\rho = \tan\alpha$, and the scale unit $d$ — through the physical
stonecutting acts to the measurable features on the artifact.

![Construction dependency graph of the conic sundial. Top: design inputs; middle: stonecutting acts in historical sequence; bottom: measured witnesses and consistency checks. Carmine edges carry the latitude proportion $s:g$; slate edges carry obliquity $\varepsilon$; bronze edges carry cone opening $\rho$; grey edges carry scale $d$.](fig_construction_dag.png){#fig-dag}

The latitude proportion $s:g$ enters the artifact along three structurally
independent paths:

1. **The Upper Front Plane:** Dressed at inclination $90^\circ - \varphi$
   relative to the base (cut orthogonal to the cone axis).
2. **The Cone Axis:** Elevated by $\varphi$ above the horizontal seating plane.
3. **The Hour Division of the Solstitial Day-Lines:** Whose angular span obeys
   $\cos H = -\tan\varphi\tan\delta$.

Because these stonecutting acts were performed separately, craftsmanship errors
remain strictly local: a carving error in dividing the summer hour line cannot
corrupt the front plane or the axis tilt. The equinoctial hour division ($15^\circ/\mathrm{h}$
at all latitudes) and the day-line spacings ($\ell_{S,W}/d$) depend only on $\rho$
and $\varepsilon$ and serve as internal consistency checks.

# The Measurement Chain on ObjID 18

## Step 1: Empirical Point Cloud and Noise Floor

The measurement basis is the photogrammetric surface model BSDP0300 (497,694
vertices, millimetre coordinate frame), fourteen hand-curated point selections
(PLY) of the engraved features (eleven hour lines with 509--967 points, three
day-lines with 377--1,483 points), and seven plane selections of the dressed block
faces. The feature selections lie on the model surface within $0.08\,\mathrm{mm}$
(median $0.03\,\mathrm{mm}$). Detrended cross-sectional residuals scatter with standard
deviation $\sigma_0 = \pm 0.3\,\mathrm{mm}$, which defines the physical noise floor.

## Step 2: Medial Feature Centrelines

Each selection is sliced into orthogonal cross-sections at $2.0\,\mathrm{mm}$
intervals. The groove centreline is extracted using the medial-line filter
($\beta = 0.5$), yielding 9--37 centreline coordinates per feature.

## Step 3: Circle and Arc Models

All curved features are fitted as 3D circles using orthogonal distance regression:

* **Hour lines and long day-lines** ($\ge 120^\circ$ span) receive **free-plane** fits.
* **The winter day-line** ($102^\circ$ span, 9 centres) receives a **coaxial fit**
  constrained to the global cone axis to eliminate partial-arc bias.

@tbl-fits summarizes the fitted circle parameters.

| Feature Name | Model Class | Fitted Radius $r$ [mm] | Fit RMS [mm] | Centre Count |
|:---|:---|:---:|:---:|:---:|
| Day-line Summer (*SuSo*) | Free Plane Circle | $120.31 \pm 0.50$ | $1.13$ | 37 |
| Day-line Equinox (*Equi*) | Free Plane Circle | $106.06 \pm 0.24$ | $0.76$ | 35 |
| Day-line Winter (*WiSo*) | Coaxial Circle | $67.63 \pm 1.78$ | $3.37$ | 9 |
| Hour lines 1–11 | Free 3D Arcs | $113.2$--$280.2$ | $0.20$--$0.71$ | 12–24 |

: Circle and arc model parameters for Dialface ObjID 18 (bootstrap standard deviations). {#tbl-fits}

```{=typst}
#arxivai.concept-box(title: [Empirical finding: Hour lines are circular arcs])[
  Straight line generator models leave large bow-shaped residuals of 2.2--7.6 mm
  RMS. Circular arcs fit the extracted centrelines at 0.20--0.71 mm
  RMS, with radii symmetrically distributed about the meridian (280 mm
  at noon, decreasing to 113 mm at the outer lines).
]
```

![Straight-line versus circular-arc residuals for HourLine 11. The straight line model leaves a $\pm 13\,\mathrm{mm}$ bow; the circular arc reduces residuals below $1\,\mathrm{mm}$.](fig_residual_bow.png){#fig-residual-bow}

![Fitted arc radius per hour line on ObjID 18 (left), symmetric about the noon line (HourLine 6), and fit RMS comparison (right, logarithmic scale).](fig_hourline_arcs.png){#fig-hourline-arcs}

## Step 4: Direct Surface Cone Fit and Intersections

Fitting an unlabelled right circular cone directly to all 3,345 feature points by
orthogonal distance regression yields the global cone parameters: half-angle
$\alpha = 34.30^\circ \pm 0.08^\circ$ (RMS $0.39\,\mathrm{mm}$).

Intersections between the 11 hour arcs and 3 day circles are computed as
closest-pair points in $\mathbb{R}^3$. All $11 \times 3 = 33$ intersection
points are recovered with a mean curve-to-curve gap of $1.27\,\mathrm{mm}$
(@fig-cone-fit).

![Day-line radii versus axial position with the fitted cone generatrix and extrapolated apex.](fig_cone_fit.png){#fig-cone-fit}

## Step 5: Dressed Reference Planes of the Block

The dressed block faces are planar within $0.37$--$0.75\,\mathrm{mm}$ RMS:

* The **back face** is orthogonal to the meridian plane within $0.27^\circ$.
* The **upper front face** is orthogonal to the cone axis within $0.82^\circ$,
  so its inclination from horizontal directly embodies $90^\circ - \varphi$.
* The **bottom seating face** defines the horizontal reference plane.

# Ratio Compatibility Intervals

Ancient dialmakers worked with simple whole-number ratios. A measured value $m \pm \sigma$
is reported as its **ratio compatibility interval**: the set of all irreducible
fractions $p:q$ ($q \le 16$) within two standard uncertainties of $m$, ordered by
closeness in $\sigma$:

$$
[\, r_1, r_2, \dots\ (m \pm \sigma) \,].
$$

A ratio within $1\sigma$ (bold) is fully compatible; between $1\sigma$ and $2\sigma$
is marginally compatible; outside $2\sigma$ is excluded (@fig-ratio-intervals).

![Ratio compatibility intervals of measured proportions for ObjID 18. Red bands: $\pm 1\sigma$ (thick) and $\pm 2\sigma$ (thin). Gold ticks: denominators $q \le 10$; grey ticks: $11 \le q \le 16$.](fig_ratio_intervals.png){#fig-ratio-intervals}

# The Latitude Proportion: Witnesses and Consensus

## The Four Latitude Witnesses

@tbl-witnesses catalogs the four independent witnesses of the latitude proportion
measured on ObjID 18.

| Witness Name | Construction Act | Measured $\varphi$ | Empirical Basis |
|:---|:---|:---:|:---|
| 1. Cone axis vs. seat | Tilt of cone axis | $41.85^\circ$ | Day-line normals against bottom seating plane |
| 2. Upper front face | Dressing of front plane | $41.11^\circ$ | Plane regression, 12,829 points |
| 3. Summer division ($\varepsilon$-free) | Scribing of summer day-line | $57.14^\circ$ | $\tan\varphi = \tan\alpha\,|\cos\tilde{H}_S|/x_S$ |
| 4. Winter division ($\varepsilon$-free) | Scribing of winter day-line | $39.63^\circ$ | $\tan\varphi = \tan\alpha\,|\cos\tilde{H}_W|/x_W$ |

: The four latitude witnesses measured on Dialface ObjID 18. {#tbl-witnesses}

## Hour-Point Regressions and Angular Unit Calibration

On each day-line, the azimuth $\theta_j$ of seasonal hour $j$ is regressed against
the hour index:

$$
\theta_j = a + (j-6)\,\frac{H(\delta)}{6}, \qquad \cos H = -\tan\varphi\tan\delta.
$$ {#eq-hour-regr}

*Reading.* The azimuthal position $\theta_j$ of the $j$-th hour mark is a linear
function of the index $(j-6)$, whose slope determines the half-day arc $H(\delta)/6$.

To eliminate shared engraving errors, two calibrations are applied:

1. **Angular Unit Calibration:** The equinoctial division $H_E$ should theoretically
   equal $90^\circ$. ObjID 18 exhibits $H_E = 85.66^\circ \pm 0.39^\circ$,
   indicating a scale factor $c = 85.66/90 = 0.9510$ (a $-4.9\%$ compression).
   All arcs are calibrated: $\tilde{H} = H/c$.
2. **$\varepsilon$-Free Inversion:** Pairing each solstitial circle's placement
   $x = \tan\alpha\tan\delta'$ with its half-day arc $\cos\tilde{H} = -\tan\varphi\tan\delta'$
   cancels the realized declination $\tan\delta'$:
   $$
   \tan\varphi = \tan\alpha\,\frac{|\cos\tilde{H}|}{|x|}.
   $$ {#eq-phi-efree-obj18}
   *Reading.* The latitude is recovered without assuming the epoch obliquity or
   an exact circle placement (@fig-hourpoints).

![Left: weighted regression of hour-point azimuths against hour index. Middle: standardized residuals. Right: per-line latitude estimates compared with structural witnesses.](fig_hourpoint_phi.png){#fig-hourpoints}

## Joint Consensus and Error Isolation

Combining all four witnesses through a joint measurement bootstrap and maximum
likelihood estimation under a random-effects hierarchical model ($W_i = \varphi + b_i + e_i$,
where $e \sim \mathcal{N}(0, \hat{\Sigma})$ and $b_i \sim \mathcal{N}(0, \tau^2)$):

* **Consensus Latitude:** $\hat{\varphi} = 41.45^\circ$, $68\%$ CI $[41.16^\circ, 41.75^\circ]$,
  artisan spread $\tau = 0.36^\circ$.
* **Klima Proportion:** Stated in the shadow-table convention ($g = 60$), $s = 53;02$
  ($68\%$ range $52;28$--$53;37$), ratio interval $[8:9, 9:10]$. This directly
  brackets Ptolemy's entries for **Hellespont** ($s = 52;02$) and **Rome** ($s = 53;50$).
* **Construction Error Diagnosis:** The summer division witness ($57.14^\circ$)
  deviates by **$14\sigma$** from the leave-one-out consensus of the other three.
  Three independent anomalies converge on this feature: the summer obliquity
  is elevated ($24.7^\circ \pm 0.2^\circ$), the equinoctial scale is compressed,
  and the afternoon marks are systematically wide (@fig-hour-steps). The summer
  division is definitively isolated as an ancient stonecutting error.

![Angular steps between adjacent hour marks along the equinoctial day-line.](fig_hour_steps.png){#fig-hour-steps}

# The Day-Line Proportions and the Workshop Rule $x = 1 : 4$

Normalizing day-line axial distances by the equinox invariant $\ell_E = d$:

$$
\ell_S / d = 1.325 \pm 0.019 \quad [\mathbf{4:3}], \qquad \ell_W / d = 0.800 \pm 0.013 \quad [\mathbf{4:5}].
$$

The three seasonal marks jointly satisfy the single proportional ratio:

$$
\ell_W : \ell_E : \ell_S = 12 : 15 : 20,
$$ {#eq-ratio-121520}

*Reading.* The seasonal noon marks stand in the exact integer ratio $12 : 15 : 20$
(within $0.6\sigma$), corresponding to the Hellenistic workshop rule $x = \tan\alpha\tan\varepsilon = 1/4$.

# Metrological Audit

@tbl-audit provides a complete trace of every deduced parameter, its empirical
foundation, and its dominant uncertainty.

| Deduced Quantity | Value / Proportion | Empirical Foundation | Dominant Uncertainty |
|:---|:---|:---|:---|
| **Consensus Latitude** $\varphi$ | $41.45^\circ$ ($68\%$ CI: $41.16^\circ$--$41.75^\circ$) | Front plane + cone axis + winter division | Artisan stonecutting $\tau = 0.36^\circ$ |
| **Klima Ratio** $s : g$ | $53;02 : 60$ ($8:9, 9:10$) | Joint bootstrap consensus | Artisan spread |
| **Witness Front Plane** | $41.11^\circ \pm 0.02^\circ$ | Upper dressed plane, 12,829 points | Stone dressing |
| **Witness Cone Axis** | $41.85^\circ \pm 0.04^\circ$ | Day-line normals vs. bottom seat | Stone dressing |
| **Witness Winter Division** | $39.63^\circ \pm 2.74^\circ$ | Calibrated arc $\times$ circle placement | Engraved division |
| **Isolated Error** | Summer division ($14\sigma$ LOO) | Residual + $H_E$ + $\varepsilon_{\mathrm{SuSo}}$ | Ancient carving error |
| **Noon Mark Ratio** | $12 : 15 : 20$ ($\le 0.6\sigma$) | 3 day-line fits + apex | WiSo fragment (9 centres) |
| **Workshop Rule** $x$ | $0.247 \pm 0.011$ ($1:4$) | Solstitial relations | WiSo fragment |
| **Cone Opening** $\rho$ | $0.554 \pm 0.021$ ($4:7, 5:9, 9:16$) | Surface cone regression | WiSo fragment |
| **Nodus Distance** $d$ | $168.7\,\mathrm{mm}$ | Equinox invariant $\ell_E = d$ | Apex extrapolation |
| **Hour Line Curvature** | Circular arcs ($r = 113$--$280\,\mathrm{mm}$) | 11 arc fits ($0.20$--$0.71\,\mathrm{mm}$ RMS) | Weathering noise floor |
| **Unit Calibration** $c$ | $0.9510 \pm 0.0043$ ($-4.9\%$) | 11 points on Equi ($H_E = 85.66^\circ$) | Engraver's unit |

: Metrological audit of the reconstruction of Dialface ObjID 18 (British Museum inv. 2547). {#tbl-audit}

# Conclusion

The empirical reconstruction of Dialface ObjID 18 demonstrates that ancient
stone sundials can be recovered with sub-millimetre metrological rigor when
governed by gnomonic domain theory.

By decomposing the artifact into independent construction witnesses, calibrating
the engraved angular unit via the equinoctial invariant, and isolating carving
anomalies through leave-one-out testing, the intended design proportions —
klima $s = 53;02$ (Rome/Hellespont), workshop rule $x = 1:4$ ($12:15:20$), and
circular arc hour lines — are established with mathematical certainty.

# References {.unnumbered}