Files
essim/doc/user/analysis.md
François 300e871aed Help screen, explore→set-connector-type Enter, settype UI polish.
- New `screen_help.cpp` (`screen_idx = 6`). Left column: menu of 13
  topics (Overview, Dashboard, Console, Palette, Explore,
  Connect/plug, set-connector-type, Signal types, NC pins, Analyze,
  Scripting, Save/restore, Quitting). Centre column: paragraphs of
  the focused topic, word-wrapped via `paragraph()` and scrollable.
  Right column: standard help panel.
- `help` bare → opens the screen; `help <name>` keeps the existing
  textual command-help behaviour for scripts.
- Dashboard `[h]` shortcut opens the screen, and the dashboard help
  panel (both the loaded and the no-system branch) lists it.
- Console: title gets the standard breadcrumb (`essim → console —
  type commands, read textual output`). Module/connection counters
  moved off (they live on the dashboard now).
- Explore Enter on a part jumps to `set-connector-type` with the
  exact-match index pre-computed in the filtered list (avoids the
  substring-match collision where `J20` would land on the wrong
  row when J200/J21 also matched).
- set-connector-type screen: bind `focused_entry` to `selected` on
  both menus so the cursor `>` tracks the selected row when state
  is pre-seeded from outside. Right column drops its strict
  `size(WIDTH, EQUAL, 40)` in favour of `flex`, and the `new type`
  input uses `xflex` so it actually stretches across the column.
- Esc on `set-connector-type` honours `screen_back_idx` — when
  entered via Enter on a part in `explore`, Esc returns to explore;
  otherwise it returns to the dashboard like every other screen.
  Standalone command entries explicitly reset the back-link.
- Net-member rows in the explore detail pane carry a
  `module\tsignal` payload so Enter opens the popup scoped to the
  peer module rather than mis-firing on the locally selected one.
  Same scheme for local-pin rows.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-15 10:31:09 +02:00

145 lines
6.3 KiB
Markdown

# essim — how the analysis classifies things
essim looks at signal names and the way pins are wired to decide
whether a net is a **power rail**, a **ground**, a **diff pair**, a
**bus**, etc. This page summarises those rules in plain language so you
know what to expect when you run `analyze` (the `[a]` shortcut on the
dashboard) or when you read the numbers on the home screen.
Nothing here mutates anything you cannot fix manually: every
inference can be overridden with `set-signal-type`, and the rules are
re-run on every `load` so the picture stays consistent with the
netlists currently in memory.
## Signal type — Power / Gnd / Other
Every signal is classified into one of three buckets.
**Gnd** if the name matches one of:
`GND`, `GROUND`, `EARTH`, `SHIELD`, `CHASSIS` (or starts with any of
those followed by `_`). The name alone is enough — false positives
here are essentially nil.
**Power** is a two-stage decision:
1. The name has to suggest power — it contains `PWR`, `POWER`,
`VCC`, `VDD`, `VEE`, `VSS`, `VBAT`, or starts with `VS_`, `VS3_`,
`+5V`-style or `-12V`-style prefixes.
2. The wiring has to corroborate it. essim requires at least one of:
- the signal lands on **4 or more pins** (a real rail goes to
decouplers + ICs + connectors, so it almost always has many
pads), or
- the name contains a **voltage value**`3V3`, `5V`, `12V`,
`0V9`, `5V0`, etc. (any `V` next to a digit).
*Hard floor*: a signal touching **fewer than 3 pins** is
**never** Power, even if both 1 and the voltage motif are
present. Physically you cannot have a rail on 1 or 2 pads.
**Other** in every other case.
This rule deliberately rejects things that look like power but
aren't: `PWR_OK` (status), `VSEL_0` (voltage select), `VDD_SENSE`
(sense feedback) — they all match step 1 but fail steps 2/3. The
analyze screen lists them under **Suspect Power** with the reason
attached (`fan-out 1, no voltage` etc.). Inspect, then either accept
the suspect status or force it back with `set-signal-type`.
## NC (no-connect) pins
A pin is shown as `(NC)` in the explore detail when it has no signal
attached. essim distinguishes three reasons:
- **Imported NC** — the netlist explicitly says the pin is
unconnected (Mentor format: signal name `unconnected` or
`unconnected (by TERM)`; Altium format: the pin is simply omitted
from every signal block).
- **Dropped singleton** — after import, essim removes every signal
that touches exactly one pin. A net with a single endpoint cannot
carry signal anywhere, so the pin is detached and tagged. This
catches both intentional sentinels and the per-IC `NC_*` labels
that customers often put on dead pads.
- **Filled at connect** — when you `connect` two parts that don't
agree on which pins exist (a Mentor part may have all pads, an
Altium part only the wired ones), essim materialises the missing
pads on the smaller side. They are unconnected *locally* on that
module but are bridged to a real signal on the other module via
the connection — so they do not count as orphans.
The dashboard's "NC" row summarises orphan counts (imported and
dropped only; filled-at-connect pins are excluded). The analyze
screen's "Types" tab adds a trailing line with the totals.
## Signal groups
essim groups signals that share an obvious structural pattern. They
are detected per module — a multi-card bus on the system is the BFS
union of the per-module groups it touches.
**Diff pair** — two signals named `STEM_P` and `STEM_N`
(case-insensitive, `_` required before the polarity letter). Both
halves must be present. Lone `_P` halves are flagged as orphans;
lone `_N` halves are *not* flagged (the `_N` suffix is overloaded
with active-low semantics — `RESET_N`, `BOOTMODE_N` — and flagging
them would flood the report).
**Diff bus** — at least two diff pairs whose stems share a common
prefix and only differ by a trailing index: `MDI0_P`/`MDI0_N`,
`MDI1_P`/`MDI1_N`, … → `MDI[0..3]_P/N`. Both `STEMN` and `STEM_N`
forms work (`MDI0`, `PCIE_TX_0`).
**Bus** — at least two signals with a common stem and a trailing
integer index. Two notations: `DATA[0]`, `DATA[1]`, … (bracketed)
or `ADDR_0`, `ADDR_1`, … (underscore — *strict*: an underscore is
required between the stem and the digits, so a name like
`GETH_01_VDD12` is *not* a bus).
**Anomalies** are emitted alongside groups:
- *Diff pair orphan*: a `_P` with no matching `_N`.
- *Diff bus gap*: e.g. `MDI[0..3]` has `MDI0`, `MDI1`, `MDI3` (`MDI2`
missing).
- *Bus gap*: same idea on plain buses.
Internal Mentor net names that start with `$` (like `$N12345`) are
skipped from every group/bus detection.
## Issues reported by `analyze`
The Issues tab of the analyze screen aggregates everything that
deserves attention:
| Tag | What it means |
|---|---|
| `[pin-role]` | A connector pin is typed (via `set-connector-type`) as Power or Gnd but the actual signal landing on it disagrees. |
| `[net-mix]` | A net bridged across modules carries both Power and Gnd signals — almost always a topology mistake. |
| `[diff-pair-orphan]` | `STEM_P` with no `STEM_N` in the same module. |
| `[bus-gap]` | A bus is missing one or more index values inside its range. |
| `[diff-bus-gap]` | A diff bus is missing one or more lane indices. |
Zero issues = the module passes every structural check essim knows
how to run today.
## Overrides
Every classification is advisory. To force a different type:
- **Signal type**: from the `explore` screen, press Enter on a
signal entry → a popup lets you pick `power` / `gnd` / `other`.
The same popup also opens when you press Enter on a pin row in
the parts detail (changes the pin's signal type) or on a net
member row in the signal detail (works across modules). Or type
`set-signal-type <module> <signal> <type>` in the console (or
from the palette).
- **Connector type**: from `explore` with `type = parts`, press
Enter on a part to jump to the dedicated `set-connector-type`
screen with that part pre-selected and the cursor on the type
input. The `set-connector-type` command also keeps working from
the console / palette.
- **Connector type**: `set-connector-type <module> <part> <connector-kind>`
(also via the dashboard `[t]` shortcut). This drives the pin role
expectations, which feed the `pin-role` check.
Overrides survive `save`/`restore` but are recomputed at every
`load` (i.e. the inference re-runs).