Product, Configuration & Market Matrix

The golden sample leaves other products unexamined

“The golden sample passed. Can we ship the other five SKUs?”

The condition-monitoring team has a neat report for one node, one antenna and one software build. Beside it sits a release list: a baseline node, an external-antenna node, a different enclosure, a USB gateway, a Wi-Fi gateway and a cellular-option gateway. All use the same node-radio silicon. The report describes a specimen; the release claim describes a family. The missing work is the relationship between them.

A SKU is a commercial stock identity. A configuration is a reproducible combination of hardware, software, accessories and use conditions. A coverage unit is the exact configuration and requirement question in one market under one method scope. Sharing a SKU does not merge those units; having different SKUs does not prove they all need independent repetition.

Keep 09.1’s authority map: the kind of document and its authority still matter. This lesson owns the product boundary and representative rationale. It carries the fictional teaching-node-v1 case, not a real approved radio. Its 2.450 GHz QPSK waveform does not acquire Bluetooth, Wi-Fi or cellular status because a gateway is added.

Define the product and accountable actor

A product boundary tells the reviewer what is being made available together. A radio component, a certified module, the assembled host and a mains adapter are distinct objects. “Gateway kit” may include the antenna, cable, mounting bracket, adapter and installation instructions. If an accessory can change the emission path or exposure condition, its identity belongs in the boundary even when it has its own supplier paperwork.

Identify the commercial manufacturer and the actor for each territory and activity. EU RED Article 10 connects manufacturer duties with identity, documentation and instructions. GB guidance separately describes manufacturer and importer responsibilities. The NI guidance has its own territory scope. These are reasons to record the actor; this exercise has not established a legal entity or market route. [RED] [UK-GB] [UK-NI]

A module record is evidence about its stated configuration, not a substitute for inspecting the host. FCC Part 15 modular conditions and ISED RSP-100 host-integration provisions illustrate that scoped relationship. Actual grant and integration conditions still need review. The FCC KDB index identifies D04 V02, but its attachment was inaccessible during this research; no clause is inferred from the index. [FCC-15] [ISED-RSP] [FCC-KDB]

Intended use changes the evidence question

The same enclosure can be mounted on a wall, held while commissioning, or installed next to a worker. Those uses change distance, orientation, cable routing and access. A label such as “industrial” does not describe who can approach the antenna or whether an installer can defeat a power control. Record intended use first, then foreseeable configurations and the facts still missing.

RED Article 17 requires assessment across intended operating conditions and possible configurations; its reasonably foreseeable conditions provision is specifically tied to Article 3(1)(a). For the Canadian exposure example, RSS-102 connects the operating and installation instructions to supporting configurations and results. An attractive separation assumption cannot replace evidence. [RED] [ISED-102]

Do not promote an unknown installation to a convenient device class. For example, the FCC digital-device class definitions have their own intended-use distinctions; the word “gateway” alone does not select one. U4 retains that classification question for the EMC specialist. [FCC-15]

Inventory actual radio operating states

A list of radios is not a list of operating states. A transceiver can idle, receive, transmit bursts or sustain a test stream. Two installed radios may be mutually exclusive, may share a scheduler, or may transmit together during recovery. The firmware state machine and the physical interfaces determine which combinations can occur; a marketing feature list does not.

The inherited node waveform remains generic QPSK: 2.450 GHz carrier, 20 kbit/s information rate, 10 ksymbol/s, RRC roll-off 0.35, 80 ksample/s at D3, and 256 symbols over 25.6 ms. Those values identify a teaching signal. They neither name a standardized radio mode nor prescribe a legal measurement bandwidth. Use 07.7’s concurrency evidence to distinguish a supported mode from evidence of its performance.

Only G02 Wi-Fi/node and G03 cellular/node concurrency exist in this fixed inventory. That declaration permits an exercise request; it supplies no test result. G03’s cell radio being off in U7 does not prove it can never turn on. Document region selection, service commands, update/recovery paths and the difference between test and production images. A disabled menu is weak evidence if a host command can still enable the state.

Antennas and accessories are causal variables

Changing an antenna changes more than peak gain. Its impedance can alter the transmitter load; its pattern changes field strength by direction; its feed and enclosure can carry unintended currents. A lower-gain antenna at the wanted frequency can have a different harmonic pattern. A comparison needs the relevant frequency, orientation and feed conditions, not a single catalogue scalar.

The supply is another causal input. Switching activity can enter the RF path through rails or ground, while an external cable can provide a radiation path. P0 battery, P1 USB and P2 mains adapter are therefore explicit identities. Their presence does not predict a numerical result; it identifies a mechanism worth testing. Connect this reasoning to 06.3’s antenna integration.

The canonical support table declares N01 and N03 with A0/P0, N02 with A1/P0, G01 with A0/P1, G02 with either antenna and P1 or P2, and G03 with A1/P2. All 36 hardware tuples remain inspectable in the workbench: nine allowed, 27 disallowed, zero unknown initially. Each SKU individually lists F-EU, F-CH, F-GB and F-NA. Existence of an image does not establish its market suitability.

Thus 6 × 2 × 3 × 4 × 5 = 720 unfiltered candidates becomes 9 × 4 × 5 = 180 after the declared hardware constraints. These counts precede modes, installations and requirement families. You have reduced an inventory, not calculated mandatory tests.

Build a reproducible configuration manifest

A manifest is the join between a report and the product it is meant to support. If a report says “firmware 1.0” while two region images share that label, the join is ambiguous. Use stable build identities and preserve the prior manifest when changing a causal fact. The new version explains what changed; the parent link explains which old evidence must be reconsidered.

RED’s technical-documentation provisions include product and software/firmware identity. RSP-100 distinguishes hardware, firmware, host and marketing names; RSS-Gen’s reporting provisions connect equipment identity with operating and test setup details. These are scoped documentary requirements, not endorsement of this lesson’s synthetic schema. [RED] [ISED-RSP] [ISED-GEN]

The parent is SYN-P09-M01-MANIFEST-v1. Preserve teaching-node-v1, NEED-08, C-MARKET, M08-01-RAW-A, M08-01-DERIVED-A, P09-M01-C09-v1, P09-M01-C10-v1 as provenance. The earlier measurement records identify a bench witness; they do not become a final-product approval. The fixture freeze is 2026-09-08; later portfolio observations are not imported into that date. A checksum could strengthen a real build record, but this lesson uses clearly synthetic version IDs rather than inventing a binary or checksum.

For each regional build, establish how image selection is controlled, how recovery behaves and whether test commands can reach production-forbidden states. Unknown control integrity belongs in the manifest. 08.1 owns measurement traceability; 08.6 owns the debugging and pre-compliance gate.

Split obligations and evidence states

One configuration can face several questions. Spectrum concerns radio operation and emissions under a stated scope; EMC considers disturbance and susceptibility questions; health/safety and exposure need their own use and physical models; firmware/security applicability may depend on features and intended activity. A scheduling bundle may contain these separate reviews, but one field reading cannot answer all of them.

The subset is intentionally small. Health/safety has no cell here; that omission is an explicit full-scope gap, not a reviewed “not applicable” result. Other firmware/market combinations, receiver states, installations and G03 concurrent operation remain outside this eight-cell demonstration. A real plan must expand or justify that scope before claiming sufficiency.

FCC §2.947 and RSS-Gen make scope and adequate supporting data consequential to equipment evaluation. Their provisions do not authorize the cross-family sets in this lesson. All SYN-methods and ER-records below are fictional planning inputs. [FCC-2] [ISED-GEN]

Worst case depends on the requirement

Comparison contract: R1-TX is the node’s conducted output reference plane into the same nominal 50 Ω setup; power is an on-time mean over the same declared transmit interval. W3’s 18 dBm is the node-radio value, not the combined power of both radios. S0 is a fixed 3 m observation of a synthetic 1 GHz spur: peak detector, 100 kHz resolution bandwidth, 300 kHz video bandwidth, same corrected receiving antenna at 1.5 m, horizontal polarization and the same azimuth. These declared choices define the exercise comparison; they are not a legal test method or a search for the maximum over all directions.

The normalized internal exposure screen is an invented dimensionless input. It is neither SAR, power density nor an official normalized sum. The values are supplied, not derived from the three dBm values. No uncertainty budget or pass/fail rule is supplied, so compare only the nominal rankings. Do not add dBµV/m readings or per-radio dBm scalars. Immunity, receiver and thermal worst states remain unknown.

A complete directional rationale

A → U2 proposes transferring a spectrum evaluation from N01/A0/P0/F-EU to N02/A1/P0/F-EU under SYN-SPECTRUM-v1 in EU-DE. The target differs in antenna/feed and hardware identity. SYN-BASIS-A-U2/v1 asks the fictional RF method reviewer to bound pattern, matching, cable loss, harmonics and radio commands. Review is proposed; source review is current through 2026-12-07. No accepted equivalence or report is supplied for U2.

The dependency list includes hardware, firmware/region, radio mode and duty control, antenna/version/cable, enclosure, supply/version, installation/user/environment and test-image relation. Any relevant change reopens this edge. Its direction is N01 → N02 for this scope only. There is no automatic N02 → N01 edge or path through a third product. A → U4 is a separate host-load EMC proposal; D → U6 is a separate simultaneous-exposure proposal. Their causal rationales are fully inspectable in the workbench.

Reduce a plan without hiding its gaps

Once the units and support basis are visible, choose evaluations efficiently. Let U be the requested universe. Each named bundle has a set of member evaluations. Its union tells you which questions it proposes to address. It says nothing about whether a report exists, whether the equivalence argument is accepted, or whether U includes every real obligation.

The transparent greedy rule selects the greatest new coverage, breaking ties by ascending ASCII bundle ID. A first adds U1–U4 (gain 4); D adds U5/U6/U7 (gain 3); E adds U8 (gain 1). Its U7 membership is repeated, so the union still has eight units. Exhaustive checking of all 64 subsets finds minimum cardinality 3, with {A,D,E}, {A,D,F}, {B,C,E} as all equal minima. Lexicographic order selects A/D/E as canonical. Direct-only recomputation instead selects B → C → E.

For selected A/D/E, direct planned coverage is {U1,U3,U5,U7,U8}. Conditional-only coverage is {U2,U4,U6}; optimistic uncovered is ∅. The schedule has three bundles and 4 + 3 + 2 = nine member evaluations. The optimization counts bundles, not elapsed time, sample count, money or member evaluations. Removing a selected bundle does not erase an already supplied record.

A greedy plan can use too many bundles

Greedy X → Y → Z uses three bundles. Exact {Y,Z} covers all six with two. No single bundle contains all six, so two is minimal. This counterexample explains why the suggestion is not an optimizer guarantee. For an empty universe the result is “no requested units”; when a requested unit has no candidate the result is “no complete proposal,” with the original denominator retained.

Audit a change against the whole matrix

A release change should ask “which dependency changed?” before “which reports can we reuse?” U9 adds a new A1 revision and concurrent-state firmware to G02. It has a new manifest ID and a parent link to U6. The old U6 request remains; no candidate silently grows to include U9. This creates a visible new gap rather than overwriting history.

Dates answer different questions. Publication identifies an edition; entry into force and application/transition fields concern a legal instrument’s stated effect; actual access and review dates describe the review process. Here publication/access/review metadata are synthetic inputs and legal dates are explicitly unassigned. The fixture is current through its review due date, then requires recheck. A later application date can be recorded without pretending a future review already happened.

For your own portfolio, reduce a variant set and defend each retained or removed evaluation. Begin with 05.1’s verifiable requirements. Name the full-scope exclusions, explain each representative’s causal basis, and assign U2 to the RF lab, U4 to the EMC lab and U6 to the exposure specialist plus firmware lead. A perfect set cover remains inadequate if its universe, assumptions or methods are wrong.