add-lab-equipment-device

Add Lab Equipment Device

Add equipment support only after proving the physical interface, host environment, command protocol, and acceptance path. Preserve existing devices and keep model-specific behavior isolated. For recurring failure patterns from the AFG-2125 implementation, read lessons-learned.md before coding.

The manual's remote-control chapter is the first authority

Everything the driver sends, accepts and reports must come from the manual's remote-control or programming chapter, and every such decision must be traceable to a page. Familiarity with other instruments is not evidence.

• Spellings, options, ranges and limits come from the manual. Do not extend an option list with values that are merely conventional for the transport.
• Do not invent descriptive metadata. If the manual describes a setting across several columns, reproduce those columns rather than collapsing them into a plausible-sounding phrase.
• Anything the manual does not state is either omitted or explicitly marked as not documented, in the driver, the tests and the device guide. Never present a guess as a manual fact.
• When the manual contradicts the transport's habits, the manual wins.

Cases that motivated this rule, all from the Fluke 8808A run: an odd parity option the manual never lists (it names only none and E = even); a flow-control option list the manual never mentions at all; and trigger-type descriptions invented from front/rear-panel intuition when the manual's table 4-3 actually has three columns (trigger source, rear-panel trigger, settling delay).

Stage gates

Five checkpoints each require an artefact before the next phase starts. The workflow below says what to do; the gates make each prerequisite provable rather than assumed. Skipping a gate is a failed run even when the code works, and a gate is passed by producing its artefact, never by stating that the work was done.