Software System Specification (SSS)

Software System Specification (SSS)#

Project: Sentinel-2 MSI Synthetic Raw Data Generator (s2_msi_raw_generator) · DRD: ECSS-E-ST-40C Rev.1 Annex B (SSS). System-level requirements (SYS-NNN) are refined into the software requirements of the SRS; interface requirements are detailed in the IRD and controlled in the ICD.

1. Introduction#

1.1 Purpose#

Specify, at system level, the reverse End-to-End performance Simulator (E2ES) of the Sentinel-2 MSI: a ground software system that runs a Sentinel-2B L1B product backwards through the exact inverse of the operational L0→L1B radiometric chain — inverting offset, relative-response/PRNU, dark, un-bin, SWIR re-stage, defective, crosstalk and on-board-equalization, with MTF-deconvolution OFF so PSF and noise are not re-applied — to reconstruct its L1A → L0plus → Synthetic L0 RAW. The reconstructed L0 is validated against the reference ESA L0 img (10/20 m bands ≤ ~4 DN), so that the forward msi-processor chain can be exercised and validated end-to-end when true Sentinel-2 L0 data is unavailable.

1.2 Scope#

One Computer Software Configuration Item (CSCI): the Python package s2_msi_raw_generator with its driver scripts. Radiometric-only reverse chain (ATBD §5): the invert steps of the operational L0→L1B chain (invert offset, relative-response/PRNU, dark, un-bin, SWIR re-stage, defective, crosstalk, on-board-equalization); geometry inversion is cancelled (L1B entry, not applicable). This SSS is deliberately compact for a single-CSC system — the requirement mass lives in the SRS; tailoring is recorded in the SDP.

1.3 Applicable & reference documents#

AD 1

ECSS-E-ST-40C Rev.1 — Space engineering · Software

AD 2

ECSS-Q-ST-80C — Software product assurance

RD 1

SRS — software requirements baseline (51 requirements)

RD 2

ATBD — algorithm theoretical basis (issued v1.0)

RD 3

ICD / IRD — interfaces

RD 4

Sentinel-2 L1 ATBD (S2-PDGS-MPC-ATBD-L1) — forward radiometric model

2. General description#

2.1 System context#

The system runs the operational Sentinel-2 MSI radiometric chain backwards: where the msi-processor (the EOPF Sentinel-2 MSI processing prototype) inverts instrument effects L0→L1B, this system applies the exact inverse of that same chain L1B→Synthetic L0, reconstructing the L1A → L0plus → Synthetic L0 by inverting each operational step (PSF and noise are not re-applied):

        flowchart LR
  L1B0["S2 L1B<br/>(EOPF Zarr, public bucket)"] --> GEN["s2_msi_raw_generator<br/>reverse chain: invert offset / PRNU /<br/>dark / un-bin / SWIR / defective /<br/>crosstalk / on-board-eq · PSF+noise OFF"]
  GIPP["Operational GIPP / SRF<br/>dark · PRNU · offset · defective · crosstalk"] --> GEN
  GEN --> L1Ar["Synthetic L1A"]
  L1Ar --> Synthetic L0P["L0plus<br/>(CCSDS-122 ISPs)"]
  L0P --> Synthetic L0C["Canonical L0<br/>(CCSDS-122 compressed ISPs)"]
  GEN --> Synthetic L0O["Open-container L0"]
  L0O --> PROC["msi-processor<br/>l0_decode → radiometric → toa"]
  L0C -. "compare Synthetic L0 vs reference ESA L0 img (≤ ~4 DN)" .-> ESA["Reference ESA L0 img"]

External actors: the public Sentinel-2 sample-data S3 bucket (inputs), the EOPF Sample Data Environment (SDE, execution host for the full-frame runs), the msi-processor CSCI (consumer of the open-container L0 and the calibration-database ADFs), and the GitLab platform (CI, Pages, package registry).

2.2 General capabilities#

  1. Synthetic L0 RAW generation — reconstruct a focal-plane L1A→L0plus→Synthetic L0 (12 detectors × 13 bands, 12-bit DN) from a S2 L1B entry by inverting relative-response/PRNU, dark, offset, on-board equalization, crosstalk, defective, un-bin and SWIR re-stage from S2B-sourced ADFs, then quantizing to 12-bit DN. PSF (MTF-deconvolution) and sensor noise are not re-applied.

  2. Downlink-representative packaging — carry the image data as CCSDS-122-lossless-compressed payloads in CCSDS space packets (canonical Synthetic L0), with a bit-exact ground decode; emit the open-container L0 the msi-processor ingests.

  3. Synthetic L0 vs reference ESA L0 validation + codec transparency — validate the Synthetic L0 against the reference ESA L0 img (10/20 m bands ≤ ~4 DN), and demonstrate packaging/codec transparency: decode(L0plus) == L1A bit-exact and L1A′ ≡ L1A through the full package → decode → l0_decode chain.

  4. Calibration sub-set — derive dark/relative-response/absolute coefficients from synthetic CSM sun-diffuser + dark acquisitions (inverse-crime cure).

2.3 General constraints#

  • Runtime dependencies limited to numpy (+ zarr for I/O); no EOPF CPM required by the core (REQ-QUAL-001). The E2E drivers use the CPM only on the processor side.

  • Originality: no external-processor source code; source-repository names do not appear in the deliverable (REQ-QUAL-003).

  • All instrument data is S2B-sourced (official PSF, SRF, product noise model, operational GIPP); nothing fitted in the referenceized path.

2.4 Operational environment#

Linux, Python ≥ 3.11. Development/CI on a public python:3.12-slim image (no credentials); authoritative full-frame runs on the EOPF SDE. Products and reports are published through GitLab CI (Pages, generic package registry).

2.5 Assumptions and dependencies#

  • The public bucket L1B is DN-scaled (not physically-calibrated radiance); radiometric correctness is checked by comparing the Synthetic L0 against the reference ESA L0 img (10/20 m bands ≤ ~4 DN).

  • The ESA PSD L0 SAFE image-ISP .bin objects are GET-403 under the bucket policy; reference image-packet accounting is therefore informative-only (see S2 L1B E2E validation, Known limits).

3. System requirements#

System-level requirements; each is refined by the listed SRS requirements, where it is verified.

  • SYS-01 — Reverse E2ES function. The system shall reconstruct L1A→L0plus→Synthetic L0 from S2B L1B by inverting the operational L0→L1B radiometric chain (PSF/noise not re-applied). → REQ-FUNC-010/012/013/015/016/017/018/019/020/022 (invert chain), REQ-FUNC-030…034 (product). V: T.

  • SYS-02 — Real-ADF fidelity. All radiometric instrument characteristics inverted by the system shall originate from S2B-published data (dark, PRNU, offset, defective, crosstalk from the operational GIPP, plus SRF). → REQ-FUNC-015/046, REQ-FUNC-044 (fallback flagged as such). V: T/I.

  • SYS-03 — Processor interoperability. The system’s L0 outputs shall be consumable by the msi-processor without modification of that processor. → REQ-FUNC-042, REQ-IF-002. V: T/I.

  • SYS-04 — Downlink representativeness. The canonical Synthetic L0 shall represent the downlink shape: band data as losslessly compressed payloads in CCSDS space packets, ground-decodable bit-exactly. → REQ-FUNC-092. V: T.

  • SYS-05 — End-to-end verifiability. The system shall demonstrate, on S2 data, that its packaging and the processor’s decode are exactly transparent to the science data (L1A′ ≡ L1A on kept lines, via the L0plus codec round-trip) and that the Synthetic L0 matches the reference ESA L0 img (10/20 m bands ≤ ~4 DN). → REQ-FUNC-093. V: T/I.

  • SYS-06 — Product identification. Every emitted product shall carry a unique, parseable identification per the EOPF PSFD §3 coding system (ECSS-M-ST-40C identification requirement). → REQ-FUNC-091. V: T.

  • SYS-07 — Autonomy of the core. The core system shall run with minimal dependencies and no credentials, so it is executable in any public CI environment. → REQ-QUAL-001, REQ-QUAL-004. V: I/T.

4. Verification#

System-level verification is subsumed by the software V&V programme: the V&V plan (methods T/A/I/R), the SUITP (unit/integration test plan), and the results in the V&V report, SUITR and S2 L1B E2E validation. The qualification status is assessed in the QR report.