# ISCC evidence summary — September 2026

Prepared September 22, 2026. Hardware measurements were recorded September 20, 2026 on the indoor LD-50 laboratory prototype. This summary identifies archived test records; it does not release their raw samples, source code, waveform data, or configuration files.

The hardware results are archived in **Huo, F. (2026). ISAC Demonstrated in Real Optical Hardware: LD-50 Bench Results from September 20, 2026 (Version 1.0)**, a technical report on Zenodo: [DOI: 10.5281/zenodo.22904657](https://doi.org/10.5281/zenodo.22904657). The report contains the ISAC bench measurements and aggregate results; the separate design and simulation records below are outside its scope.

## Recorded hardware results

| Trial | Recorded result | Scope |
|---|---|---|
| Longer continuous stream | 150,000 frames in 605.24 seconds; zero lost or malformed frames; 149,860 accepted frames (99.907%); zero decoded wrong frames and zero wrong frames delivered; 2.971 kbit/s wall-clock information throughput | Aimed indoor bench direction; repeating table of 1,024 codewords; error correction and confidence-based acceptance |
| Uncorrected signal, same trial | Raw bit error rate 0.000637; earlier receiver on the same frames 0.00587 | Raw channel decisions remain distinct from decoded information |
| Range, same trial | 2,047 range estimates; median 93.3 inches (about 2.37 m); median absolute deviation 0.85 inches (about 2.16 cm) | Spread of accepted range estimates, not absolute accuracy |
| Earlier continuous stream | 30,000 frames in 121.1 seconds; zero lost or malformed frames; 29,959 accepted (99.863%); zero decoded errors; 2.970 kbit/s; median range 93.4 inches | Separate earlier run with the same repeating-payload limitation |
| Short-message integration | 37 slots filled; 444 information bits reconstructed exactly; zero delivered errors recorded | Repeated short-message test; not uninterrupted fresh traffic |

The trial ledger is `WORKPOINT_DUTY60_20260915.md`: sections 24.93 and 24.94.17. Section 24.94.20 records explicit missing-slot handling, repeated-message comparisons, and the limits of the loss conditions tested.

## Separate design and simulation records

September 16 records include five digital-capture scenarios with simulated clocks and diagnostic input. Gap, overflow, and recovery handling were checked; no physical board was programmed and final implementation timing was not closed.

A separate control model completed 50 agents' goals in each of seven scenarios, with a calculated continuous minimum separation of 5.098806 m. It assumes ideal local positions and registered restricted paths, without aircraft dynamics, wind, tracking error, or a complete optical sensing chain. It is not a flight test. The optical project's 1 Tbps figure remains a conditional design target and budget, not achieved hardware throughput.

## Artifact identification

SHA-256 checksums below were calculated from the archived files on September 22, 2026. They identify those exact files; a checksum alone does not independently validate a measurement. The underlying artifacts remain in the project evidence archive and are not linked here as public downloads.

| Artifact | SHA-256 |
|---|---|
| `isac_loop_20260920_stream_serve_a.json` — longer run | `15b343962efca77ccebeaa3c19ce905254105318da800c26ded05c8de3f3a559` |
| `isac_loop_20260920_stream_a.log` — earlier run summary | `0e981f72a9e8c79eda27232a9e5dba4701cbb1ff353d7cc456413e5bfe9c3e00` |
| `isac_loop_20260920_stream_a.json` — earlier run record | `807bdfeb72f8a1e485fbda6993293fa4db6b47bcce405b3e139480d44cbbac53` |
| `CODEX_RANGE_AUDIT_20260921.md` — range evidence audit | `af900a4ee6382d158254970a0521d86b0d424a11ceacd2fd961c4795a27b7fb2` |
| `CODEX_FIELD_SCAN_PREREG_20260921.md` — new-path software validation record | `20a622ff3e015ba780fa8e31d7a5ea673eeb76bda05eeb1cd32dd18fa768dbf7` |
| `CAPTURE_V2_RESULTS.json` — simulated digital-capture verification | `d7f50b0081d1f8f126b34e7ad702d9c2792b657f77bee3bfb1ca9de73ba8e9ef` |
| `TIMING_AND_INTERFACE_AUDIT.md` — conditional high-rate design budget | `9fb04f71d2b3e9887f43f0de4edfa743fea80e7645887626278699ff0d4de5e9` |
| `fleet_v3/README.md` — seven control-model scenarios | `1e1f178b7f588d59e66d42d4906c620e2e3c1dd50c916d768ecb2dc6ab5bfe23` |
| `fleet_v3/PROOF_AND_SCOPE.md` — control-model assumptions and bounds | `7998f06036f6a7cd90533eb1ba760f069ba7f1a0f423fa67bcec9ebb47b3481b` |

## Interpretation limits

The hardware stream used the laboratory's returned optical signal and a repeating payload. It does not establish an independent two-device link, uninterrupted fresh-message traffic, outdoor range or speed performance, material identification, autonomous-control safety, or a new security guarantee. Range spread must not be presented as absolute accuracy; unresolved scale and false-line issues remain documented.

The September 21 software record reports **11 offline tests using simulated hardware**, with no new hardware validation. Its full-method, timing, and feedback requirements remain incomplete. Older FFT-based measurements retain their historical status and are not reclassified as satisfying the newer processing requirements.

For questions about these measurements or a research demonstration, contact [Milkyway Center](mailto:yuanjimaster@gmail.com).