# adom-fields 0.10.6

Three fixes from Glenn's 772 power module list (issues 27, 28 and 29). Thanks, Glenn.

- **The grid covers the whole board (issue 27).** The column and row counts were rounded to the nearest pair of fine pixels, so a 49.53 mm board got 247 columns of 0.2 mm, 49.4 mm in all, and a pad centred on the right edge had no cell. An explicit terminal on it failed with "a terminal has no copper under it".
  - The grid now rounds up to whole cells. A board that is already a whole number of cells keeps its grid.
  - A pad centred on any edge, or within 1 mm past it, now reaches the copper inside the grid. This also fixes the case rounding up can't: a pad on the far edge of a board that is an exact number of cells, such as 50.00 mm.
  - The last column and row can now reach past the outline by part of a cell. Only the part on the board loses heat to the air, and the viewer shows only that part, with each cell in its true place.
  - `at` on a terminal is unchanged: it still picks one physical pad by its centre.
- **One board thickness (issue 28).** A body under the board coupled to B.Cu only when its top face was at the sum of the dielectrics and the two outer copper layers, while `fields.json` reported every copper layer in `board.thicknessMm`. On a 4-layer board those differ, so a body placed at the reported thickness silently took 0 W.
  - `board.thicknessMm` is now the one number the report, the viewer and the bodies use.
  - A top face couples when it is within 0.05 mm of that thickness, which takes a nominal 1.6 mm on a 1.57 mm stack, or at the older outer-copper sum, so existing specs keep working.
  - A new `solid-uncoupled` issue says when a body is left out. It is `high` when `boardContactM2kW` is set but no face is on the copper, and `medium` when a body has no heat of its own and no path to the board, so it carries 0 W. A solve also logs it, and `solids.still` has `onCopper` per body.
  - A body larger than the board now keeps its top face in the air outside the outline. It used to lose that face altogether.
- **Decoupling counts the chip's own grounds (issue 29).** The review looked for a capacitor from each power pin to one ground, the first in alphabetical order. On a board with AGND and PGND, that gave a false `decoupling-missing` for every rail decoupled to PGND.
  - A capacitor now counts when it returns to any ground the chip has a pin on, or to the review's ground as before. `review.decoupling` records the `return` net of each pin.
  - A pin with no capacitor is still a finding, and so is a capacitor to a ground the chip has no pin on. The suggested capacitor goes to the ground of the chip's nearest ground pin.
  - The choice of the board's ground (issue 25) is unchanged, and so is `--with pdn`, which still looks for a rail's capacitors on that one ground only.

**What moves on your boards.** A board that isn't a whole number of cells gains a column or row, so its numbers shift slightly with the added copper and area. Peak rise moved under 1 % on the real boards below, and a value sitting right on a threshold can cross it: on StickHub, C22 at about 75 °C dropped out of a `capacitor-temperature` finding.

**Checked:**

- Each report reproduced on 0.10.5 and fixed here, on small test boards:
  - Terminals on the far edge at 49.53, 50.00 and 49.46 x 20.03 mm all failed on 0.10.5 and now solve, also with `--with all`. A 49.53 mm Fusion `.brd` does too.
  - A carrier at z = -1.57 took 0 W on 0.10.5. It now takes the same 0.537 W at -1.57, -1.535 and -1.6. At -2.0 it takes 0 W and raises `solid-uncoupled`. Fan levels and a transient ran on it.
  - On a chip with AGND and PGND, VCC2 to PGND is no longer a finding. A rail with no capacitor and a rail whose capacitor goes to a ground the chip isn't on still are.
- On real boards, 0.10.5 against 0.10.6:
  - The 12 V buck (28 x 20 mm): every finding and number unchanged.
  - BQ25792 charger (Fusion `.brd`, AGND and PGND, 50.4 x 26.4 mm): the grid and heat numbers unchanged, and two false `decoupling-missing` gone (LDO_3V3, on chips with only PGND pins, with C3 and C5 to PGND).
  - StickHub (16.5 x 40 mm): peak rise 61.7 to 61.3 °C, one column more. CM5 Minima (59.5 x 62.7 mm): 43.44 to 43.42 °C, one column and one row more.
- The viewer, in a headless browser on a coarse-grid test board: the fields line up with the board and stop at its outline.

**Not checked:** the LM251772 board itself, which wasn't available; `verify` and `report` on the new grid.