BLACKFRAME P03 is not a marketing feature list. It is the current research architecture used to separate ideas from evidence.
The central rule is straightforward: a system diagram proves that a system has been designed. It does not prove that the system performs in the real world. A bench prototype proves that components can operate together under controlled conditions. It does not automatically prove a field claim. MURMUR documents those states separately.
FOUR EVIDENCE STATES
01 / ARCHITECTURE
The intended function, boundaries, interfaces and failure assumptions are defined. At this stage we can describe what the design is supposed to do and how the parts are intended to relate. We cannot legitimately convert that intention into measured performance.
02 / BENCH
Reference hardware or construction has been brought up under controlled conditions. Electrical rails, sensors, interfaces, isolation boundaries, firmware roles or garment construction details can be checked individually. Bench success is important because it proves the architecture can leave the page, but it remains controlled evidence.
03 / LAB
A specific versioned configuration is subjected to repeatable evaluation with recorded conditions and expected results. Where RF test sources are involved, active adversary validation remains constrained to the shielded laboratory environment with fail-off controls and leakage monitoring. The objective is reproducibility, not spectacle.
04 / FIELD
The same defined configuration is evaluated under representative operating conditions. Field observations do not retroactively validate a different hardware revision, different garment integration or different firmware build. Configuration identity matters.
WHY BLACK / FIELD / RED / LAB ARE SEPARATE
BLACKFRAME uses explicit boundaries because trust is part of the engineering problem.
BLACK is the trusted control domain. Its job is to remain bounded and recoverable even if another subsystem misbehaves. The current reference direction uses a Raspberry Pi Pico 2 non-W class controller for trusted local control rather than placing that responsibility on a radio-facing processor.
FIELD contains passive sensing and measurement functions. These provide context without requiring every observation to be interpreted by a network radio.
RED contains radio-facing processors and receive functions. RED is treated as less trusted: it may encounter malformed traffic, excessive load or software faults. That means RED should not share BLACK secrets, storage assumptions or a casual common power/ground path.
LAB is where deliberately adversarial validation belongs. The laboratory boundary exists so active test conditions do not quietly become field functionality.
THE ONE-WAY BOUNDARY
The BLACK/RED relationship is intentionally narrow. RED can report bounded telemetry toward BLACK through galvanic isolation; BLACK does not depend on RED being well-behaved in order to preserve its own core state.
This principle is more important than any particular microcontroller. Hardware will change. The trust model should survive those changes.
THE WI-FI OBSERVER TRIAD
The current Rev C.1 direction uses three ESP32-S3-class observer nodes so the 2.4 GHz Wi-Fi reference channels 1, 6 and 11 can be watched simultaneously rather than forcing a single radio to hop between them. This is an observation architecture, not a claim of complete spectrum coverage. It does not add 5 GHz or 6 GHz Wi-Fi visibility.
That distinction matters because diagrams often make systems appear broader than they are. BLACKFRAME documentation deliberately states the boundary of each measurement rather than implying universal coverage.
RAW ENERGY AND PROTOCOL OBSERVATION ARE DIFFERENT
Protocol-aware radio observers can identify activity they understand. A separate RF measurement path can answer a more basic question: how much energy is present in a selected band? Those are different observations and should not be conflated.
The current reference design therefore treats wideband log-detector measurements as metrology, not as a substitute for protocol analysis. Expensive evaluation hardware can be staged during bring-up rather than purchased simply because a complete block diagram contains four channels.
PUBLIC CLAIMS
MURMUR product and research pages should say what evidence state a statement belongs to. Phrases such as “designed to,” “prototype,” “bench verified,” “lab measured” and “field evaluated” are not interchangeable.
This discipline is intentionally conservative. It protects the project from becoming a collection of impressive renders and unsupported claims. More importantly, it gives future revisions a clean record of what was actually known at the time.
THE PURPOSE OF THE PROCESS
Evidence gates are not bureaucracy for its own sake. They make iteration faster because failures become local and diagnosable. A garment fit problem should not be mistaken for an electronics problem. A radio observation problem should not contaminate the trusted controller. A new module should not inherit claims earned by an older configuration without retesting.
BLACKFRAME P03 exists to make those distinctions explicit. Build. Measure. Verify. Then say exactly what the evidence supports.
