# Current density and thermal: Adom Fields on the real copper

The board's copper was solved with [Adom Fields](https://wiki.adom.inc/adom/adom-fields). Fields builds a copper map from the board's filled pours, tracks, pads and vias on a 0.2 mm grid (140 x 100 cells) and computes two things:

- **current**: a DC conduction solve for each loaded net, between its terminals;
- **heat**: a steady-state solve with the layers coupled through the dielectric and every via barrel, losing heat to the air on both faces.

It is a screening model. It covers the rectangular outline, the chip powers from the spec and convection only; there is no radiation, no airflow solve and no junction-temperature qualification. The results guided the layout. None of this has been measured on hardware; the board has not been built yet.

![The Adom Fields app on the final board: current density on F.Cu with the parts shown see-through, SW peak 50 A/mm2, U1 +45 C and L1 +41 C, 0 issues](docs/img/fields-app.png)

## Inputs from the spec

| input | value | source |
|---|---|---|
| copper | 17.5 um (0.5 oz) per layer | the conservative end of the 3rd party fab's range |
| VOUT, SW, GND load | 1.0 A | SW carries the inductor RMS current, 1.01 A |
| VIN load | 0.6 A | input current at the 9 V low end: 5 W / 0.93 / 9 V |
| U1 heat | 0.25 W into its GND pin | conduction 0.15 W at hot Rds(on) + about 0.09 W switching + gate drive ([design](design.md)) |
| L1 heat | 0.12 W into VOUT | 1.01 A RMS squared x 118 mOhm max DCR |
| convection | 10 W/m2K still air, 25 W/m2K moving air, both faces | |
| material | copper 0.40 W/mm K, FR4 0.3 / 0.6 W/m K | |

## Current density: the SW neck, 215 to 50 A/mm2

| board state | SW peak | what changed |
|---|---|---|
| first pours (SW as a trace only) | **215 A/mm2** at U1.2 | the trace necked down at U1's SW pin |
| SW patch added | **60 A/mm2**, at a 0.95 mm throat next to U1.2 | the patch still had a throat right at the pin, at the 60 A/mm2 screening threshold |
| patch polygon opened right after U1.2 (final) | **50 A/mm2** | |

The response to the 215 A/mm2 reading was a small solid SW patch, not an SW pour. The patch adds copper only between U1.2 and L1.1, where the current flows, and the switching node's area (and its radiated field) stays small. [Board](board.md) has the full story.

![Current density in the SW net on F.Cu, from the top: the current spreads from U1 pin 2 through the solid patch into L1, peak 50 A/mm2 at the pin](docs/img/current-density.png)

Final Fields peaks per net:

| net | load | copper carrying it | vias | peak current density | where |
|---|---|---|---|---|---|
| SW | 1.0 A | 20 mm2 patch on F.Cu | 0 | 49.8 A/mm2 | F.Cu at (113.9, 90.5), beside U1.2 |
| GND | 1.0 A | 798 mm2 on both layers | 13 | 14.0 A/mm2 | F.Cu at (112.1, 91.3) |
| VIN | 0.6 A | 111 mm2 | 0 | 13.8 A/mm2 | F.Cu at (110.1, 84.5) |
| VOUT | 1.0 A | 103 mm2 | 6 | 13.2 A/mm2 | B.Cu at (122.9, 88.5) |

The Joule heating in the copper is about 1 mW in total, so the copper itself is not a heat source. The heat comes from U1 and L1.

## Thermal: where the heat goes

![Heat flow in the F.Cu copper with the parts solid: U1 sheds its 0.25 W through its GND pin into the top pour and down the thermal vias, and L1 spreads 0.12 W through the VOUT pour](docs/img/heat-flow.png)

How each part's heat leaves, from the Fields solve (W):

| part | power | along its own copper | down its vias | through the dielectric | to air at the part | other vias |
|---|---|---|---|---|---|---|
| U1 (via GND pin) | 0.25 | 0.145 | 0.069 (3 vias) | 0.026 | 0.011 | 0 |
| L1 (via VOUT) | 0.12 | 0.077 | 0.007 (1 via) | 0.026 | 0.011 | 0 |

U1 is a SOT-23 with no exposed pad, so its heat goes out through the pins, and mostly through GND. The GND vias Fields counts at U1 (two under U1.1, one beside C4.2) carry about a quarter of it straight into the B.Cu ground plane.

![Temperature rise over the board in still air: U1 +45 C, L1 +41 C, board peak +50 C](docs/img/temperature.png)

| part | still air (10 W/m2K) | moving air (25 W/m2K) |
|---|---|---|
| U1, at its GND pin | +44.7 C | +24.6 C |
| L1, at its VOUT pad | +41.4 C | +21.0 C |
| board peak | +50.2 C | +30.0 C |

## The thermal budget: 55 C, still air, junction based

The first target was a 30 C rise, to keep the junction under 85 C at a 55 C ambient. In still air the solve shows U1 +45 C and L1 +41 C, over that target.

Better copper cannot close that difference. The limit is the board's area, not the layout:

- A 28 x 20 mm board losing 0.37 W to still air on both faces rises about **+33 C on average even with perfect copper** (0.37 W / (10 W/m2K x 2 x 560 mm2)).
- Adding 8 more GND vias at the nearest free spots (2.6 to 3.8 mm from U1.1) changed U1 by **0 C** in the Fields solve.

So 30 C in still air cannot be reached on a molecule this size. The board uses a **junction-based budget of 55 C pin/tab rise in still air** instead. The reasoning:

| step | value |
|---|---|
| U1 rise at its GND pin, still air | +45 C |
| pin to junction: psi_JB 33.7 C/W (datasheet) x 0.25 W | about +8 C |
| U1 junction | about ambient + 53 C |
| at a 40 C ambient | about 93 C, 32 C under the 125 C rating |
| L1 rise, still air | +41 C, inside its -40 to +125 C rating including self-heating |
| moving air (25 W/m2K) | both under 30 C (U1 +25 C, L1 +21 C) |

With the 55 C budget, the final solve reports 0 issues. Note that this budget assumes a 40 C ambient, where the first target assumed 55 C. In moving air, both parts are under the original 30 C target.

## Files

| file | what |
|---|---|
| `aiflow/spec.json` | the loads, hot parts, copper thickness and thermal budget the solve used |
