Three objects, three research problems

Cuntz: connect an audio configuration to a historical instrument#

The Cuntz positive organ, attributed to Nicolaus Cuntz, Nuremberg, 1610, is held in the Leipzig musical-instrument collection as inventory 243. The thesis combines recording, parameterized geometry and modeled interaction around this object. The signal analysis describes 225 single-tone recordings arranged as five register labels across 45 keys, recorded with a Neumann KM 184 pair and an RME Babyface Pro at 192 kHz and 24 bit.

The physical description in Appendix E.7.1 instead lists four registers and 180 pipes. Section 8.2.2.1 explicitly notes the missing source-and-recording mapping needed to explain the fifth label, Regal 8′. The digital configuration therefore cannot be read as a register-faithful measurement inventory of the historical instrument. The differing counts alone demonstrate neither a rebuild nor an additional original register.

The interactive case adds browser and Quest 3 access, including a life-size mixed-reality presentation and hand interaction. The reported display rate and latency statements do not substitute for a complete end-to-end timing protocol. This notebook consequently treats the experience as an implemented research encounter without assigning it a verified latency benchmark or a completed usability-study result.

Ariston: read a control surface and model its mechanism#

The Ariston case relates a rotating punched disc, a crank-driven wind system and 24 free reeds. The thesis describes scans of 22 original discs within a corpus of 40 disc models. Polar correction and track recognition turn the physical layout of holes into a timed control sequence. Rotational speed is part of that interpretation: the image alone does not determine a unique musical tempo.

For 5,000 manually adjudicated holes, the thesis reports 4,840 true positives, 78 false positives and 160 false negatives. These yield precision about 0.984, recall 0.968 and F1 about 0.976. They evaluate hole recognition, not the fidelity of the complete virtual musical performance. The thesis does not provide the item-level annotation and run manifest accompanying those totals, which limits independent reproduction and uncertainty analysis.

The 24 reeds were recorded at 192 kHz and 24 bit using the original crank and wind system. Separate recordings of the idle mechanism support a useful synthesis distinction: mechanical noise should not be added anew for every sounding reed. The wind-dependent model remains a parameterized account unless backed by an independent pressure-and-response calibration. A readable disc, a plausible control sequence and a validated acoustic mechanism are separate achievements.

Propstglocke: make simulation assumptions visible#

The Landshut St Martin Propstglocke case combines photographic evidence, historical context, reconstructed geometry and mechanical modeling. The thesis discusses cross-polarized photography and geometry estimation with Neuralangelo and MASt3R before geometry preparation for finite-element analysis. This historical case pipeline is distinct from the current Processor Modelgenerator, which uses a documented DA3-based photographic point-cloud workflow.

Material properties, boundary conditions and geometric completeness influence the bell’s modeled modes and transient sound. A plausible rendering therefore cannot by itself establish the model’s quantitative accuracy. Atmospheric absorption under ISO 9613-1 represents one propagation effect; it does not model an entire urban environment with building reflections and acoustic shadows. The scope of each physical model must stay attached to its output.

What transfers between the cases#

Across all three objects, the transferable result is a method for keeping identities, observations, transformations and executable representations connected. Each case also leaves different questions open. The organ needs a resolved historical-to-digital register mapping; the Ariston needs the detailed recognition evidence for independent replication; the bell needs material, boundary and measurement evidence for quantitative model comparison. Those are concrete next research tasks, not reasons to collapse all three cases into one fidelity score.