# Validation report — gates 1 and 2

Date: 2026-09-08  
Engine: `0.1.0-alpha`  
Reference driver: Dayton Audio UMII12-22, dual 2-ohm coils in series

## Input normalization

The manufacturer's independent primary values Fs, Vas, Qms, Qes, Sd, Re, Le and Xmax
are preserved with manufacturer provenance. Because the datasheet's redundant Mms, Cms,
Bl and Qts values are rounded and mutually differ by a few percent, the circuit fixture
derives one internally exact redundant set from the primary values. Both the reported and
normalized values remain in each manifest.

## Gate 1: OpenISD black-box reference

The driver was manually entered in the public OpenISD UI. No OpenISD source code or driver
database was read into the implementation. All comparisons used 2.83 V RMS, zero source
resistance, full-gyrator Le handling, half-space radiation, ρ=1.2041 kg/m³ and c=343.2 m/s.

| Topology | Configuration | Compared outputs | Result |
|---|---|---|---|
| sealed | 28.3168 L, Ql 10, Qa 100 | SPL, cone X, |Z|, group delay ≥40 Hz | pass |
| bass reflex | 70.792 L, Fb 25.2 Hz, Ql 10, Qa/Qp 100 | SPL, cone X, |Z|, port velocity, group delay ≥30 Hz | pass |
| passive radiator | 70.792 L, one synthetic 528 cm² PR | SPL, cone X, PR X, |Z|, group delay ≥30 Hz | pass |
| infinite baffle | no rear load | analytic impedance/wiring invariants | pass |

OpenISD reports port velocity as sinusoidal peak; Core reports RMS. The comparison explicitly
multiplies Core's value by √2. Physical vent length was normalized to an effective acoustic
length of 0.521242 m from OpenISD's displayed Fb, avoiding an end-correction mismatch.

### Known OpenISD difference

For the sealed fixture, OpenISD's group delay below 40 Hz is about 0.7–1.4 ms higher while
its pressure, excursion and impedance match. The low-frequency group-delay points remain in
the dataset but are excluded from the pass gate pending a documented convention comparison.
The Core result is calculated directly as `-dφ/dω` from its full complex pressure transfer.

## Gate 2: VituixCAD 3 black-box reference

VituixCAD 3.0.1.7 was installed from the official site. Its Enclosure tool supplied native
text exports for total half-space SPL and complex input impedance. The fixture uses one
bundled Dayton Audio Ultimax UM12-22 DVC parameter row, transcribed manually with explicit
`independent-reference-software` provenance. No source code or complete driver database was
copied. Each curve has a sidecar manifest and SHA-256 integrity check.

| Topology | Configuration | Compared outputs | Observed maximum difference | Gate |
|---|---|---|---|---|
| sealed | 30.0 L, Ql 10, Qa 100 | SPL, impedance magnitude and phase | 0.724 dB, 0.307 Ω, 2.032° | pass |
| bass reflex | 70.792 L, Fb 25.2 Hz, Ql 10, Qa/Qp 100 | SPL, impedance magnitude and phase | 0.799 dB SPL at ≥15 Hz; 0.445 Ω; 2.186° | pass |

The vented fixture differs by 1.609 dB at 10 Hz, where output is over 40 dB below the
passband; its explicit 10–15 Hz allowance is 1.7 dB. Core uses an effective acoustic vent
length of 0.521242 m derived from the shared Vb, Fb and vent area. VituixCAD displayed a
43.1 cm physical vent with 0.85 end correction.

VituixCAD acoustic phase is retained but not pass-gated because its source/polarity reference
does not match Core's documented monopole convention. Impedance phase is directly gated.

After export, VituixCAD was closed and its original driver database was restored; the restored
file's SHA-256 exactly matches the pre-test backup.

## Outcome and next gate

All 16 executable tests pass, including seven immutable black-box curve files from two
independent applications with SHA-256 integrity checks. The next reference gate is Hornresp
for an exactly equivalent radiation/topology setup, followed later by measured impedance and
near-field fixtures. Neither blocks the v0.1 architecture or Unity integration work.