[michaldengusiak]

Documentation-only. Record: documentation/Licensed-TasOpenStudio-Validation-Q4.md; evidence: documentation/evidence/licensed-e2e-20261019/. Validated content: Q4 4ee45fef (SAM_UI), 68eded0 (SAM_Validation), 618db819 (SAM), a34a298d (SAM_Tas), d4216f3 (SAM_OpenStudio). No product, test, tolerance, physics or baseline change.

Results (fresh licensed runs, real SAM Analytical UI, Run Simulation > Both)

Test	Tas	OpenStudio	Comparison	Result
box-aperture, embedded weather	exit 0, Success	exit 0 (E+ 25.1.0)	6/0/10/2, 4 contributing + 6 informational	PASS (workflow)
box-aperture, explicit Budapest EPW	exit 0	exit 0	6/0/10/2	PASS (workflow)
HungaryHouse (14 spaces), explicit EPW	exit 0, crash not reproduced	exit 0	48/3/41, 14/14 spaces	PASS (workflow)

Reports are byte-identical to benchmark-compare.exe; values and differences are independently rechecked against TSD/TBD/ESO output. There is no time offset between the engines.

Main findings (details and proposed fixes in the record)

• F1: SAM's EPW writer (SAM.Weather.Query.DataString) writes global horizontal and horizontal IR as 0, and horizontal direct into the DNI column. With embedded weather, Tas simulates without solar gain.
• F2: per-person outdoor air reaches EnergyPlus Ideal Loads 24 h but not Tas, which has no Ventilation profile and no AHU. This is about 99% of E+ box heating. Removing only that input brings heating to within 11%, with the same peak hour.
• F3: OpenStudio cooling is total (includes latent), while Tas is sensible.
• F4/F5 (SAM_Tas): the space peak hour is the raw 1-based Tas index (the 871/872 pair), and space peaks are design-day governed rather than annual. Not patched: SAM_Tas is read-only while the HVAC conversion is in progress.
• F6/F7: a zero-peak hour is reported as 0, and the canonical hash includes UI-only ParameterSet GUIDs.
• Local build: the local SAM_UI\build has a stale deps.json from the pre-Simulate > OpenStudio Simulation: minimal OpenStudio/EnergyPlus run from SAM_UI (#229) #237 baseline build, so Both/OpenStudio cannot load there. A Rebuild is needed. This is not a product defect.

Unresolved: the box cooling residual (~20%) after aligning ventilation, and the remaining HungaryHouse heating gap (whole-house temperature and internal gains; see F8/C6).

• F8 / C6 (new, 19 Oct, head 0a569236): HungaryHouse GF_08 heating is mostly Tas window opening (zdwno Part O aperture function, not converted by SAM_OpenStudio). Tas-only experiment: baseline re-run reproduces the original TSD bit-for-bit; closing only GF_08's window gives GF_08 1425.6 → 387.6 kWh and whole-building 4507.4 → 3388.9 kWh (−1118.5 kWh, 40.6% of the 2756 kWh gap). The Air-panel hypothesis is refuted for the bathrooms.

Checks: git diff --check clean; docs/evidence only (4 analysis .cs tools with SPDX headers, not part of any project). Full runs are machine-local under C:\TasOut\licensed-e2e-20261019.

Not for merge without owner review.

• C7 (new, 19 Oct): all-window Tas sensitivity: heating 4507.4 → 3190.5 kWh (−1317.0, 47.8% of the 2756 kWh gap; GF_08 alone 40.6%, the six other windowed zones +7.2%); 1439.3 kWh remains vs E+. Part O audit: one zdwno,0,22,24,99 string on all 9 windows; GF_08's ramp starts below its 24 °C setpoint; string has one field fewer than EDSL documents; Unrestricted night opening is a loader default. Recommended milestone: ZoneVentilation:WindandStackOpenArea + EMS (design only, nothing implemented).

🤖 Generated with Claude Code