AnasSarkiz/usb-c-pd-brushed-motor

docs/RAIL-DISCHARGE.md

# A11 motor-rail discharge and power-change review

R68 is the unchanged official C2074262 / ROHM ESR18EZPF1001, 1 kΩ ±1%, 0.5 W/1206, connected VM to GND. It supplies a predictable passive discharge path without relying on a connected motor, driver/comparator operating below their supply range, or an MCU-controlled discharge transistor. This adds one essential protection component: **129 components /55 supplier codes**. It is a power/voltage-change mechanism, not a direction/reversal delay.

Before this change there was no dedicated bleed resistor. At low VM, the certain feedback and ADC paths are approximately 125.1 kΩ and 110 kΩ in parallel. With a 650 µF conservative bank, their 2.5→1 V tail alone screens near 35 s when other loads are not credited. Actual powered monitor/LED/driver loads can shorten it but are not a guaranteed below-range discharge specification. The old 1 s software timeout could latch a normal voltage change or replug off. R24 is the buck-enable pull-down, not a VM bleeder.

The bank envelope uses C18's 330 µF initial +20% and temperature +20%, three nominal 47 µF ceramics at +20%, and bypass reserve: ≤650 µF is a design bound, not measured capacitance. R68's 940–1060 Ω screen reserves 1% initial, ±100 ppm/°C over 80°C and 3% endurance variation within 6% overall. The maximum passive RC time from 15→1 V is **1.866 s**, so the sequence timeout is now **3 s**. The state machine still requires a fresh actual VM≤1 V measurement before changing feedback. It never enables from elapsed time alone. External back-drive/regeneration can prevent decay; that correctly times out to latched inhibition. Firmware must inhibit nSLEEP before eFuse power-off and retain the old feedback while VM is charged.

At 12.6 V (5% above 12 V), worst initial bleeder dissipation screens at **169 mW**. At 14.432 V (nominal OV trip plus a provisional 5% reserve) it is **222 mW**. ROHM derates 0.5 W linearly above 70°C to zero at 155°C; at **105°C local ambient** the allowed power is approximately **294 mW**. These declared cases pass that screen. This does not approve sustained abnormal overvoltage, repeated dump heating, an arbitrary ambient, or an actual resistor temperature. Resistor thermal isolation/temperature, measured bank tolerance and unplug/replug/voltage-change transients remain prototype requirements.

## Independent supplier inspection

Official exact-footprint import, warning-free network-enabled build, all five native checks and strict full-document/schema/electrical-pin/pad audit pass. Exactly two resistor pins map to two top SMD lands; two positive rectangular paste apertures remain inside them, with no holes. Imported pads 1.208×1.701 mm centered ±1.479 mm give 1.750 mm gap and 4.165 mm span. The manufacturer's 3.20±0.15×1.60±0.15 mm body/terminal drawing fits nominally; terminal-bearing end lengths are 0.50±0.25 mm. The manufacturer's example lands (0.9×1.5 mm/gap2.2/span4.0) are examples, not identical to the library. The imported pattern has at least 0.25 mm worst-terminal longitudinal overlap at zero placement error, additional toe space and greater land width than the example. Finished paste/solder joints and body/courtyard fit require assembly/placement review; no imported geometry is edited.

Direct live supplier page at the retained timestamp: **101,150** exact C2074262 units. The earlier C859155 Yageo candidate imported and passed electrical/native checks but its live exact-SKU stock is zero; the cached 600-unit result was stale. Recommendation stock numbers are not assigned to the main SKU. Raw pages, supplier hashes and exact datasheet are retained. Inventory is not reserved; JLC assembly eligibility still needs confirmation.

The ROHM manufacturer dimension and derating pages and generated PCB probe were visually reviewed. Reports: evidence/discharge-import-audit-A11.json, supplier-C2074262-checks-A11.json, discharge-review-A11.json and ROHM-ESR-*. No product placement or copper was authored.

## PD budget change

Both host calculations and the C policy now include **+5% VM** within the permitted rail window, plus the bleeder's worst load at 940 Ω outside the existing 1 W auxiliary allowance. At 2.423 A peak they screen **1.347 A for 5 V/15 V, 2.226 A for 9 V/15 V, and 2.145 A for 12 V/20 V**, versus the 2.243 A minimum eFuse limit. The 9 V/15 V margin is only about **17 mA** at this combined corner; transient/thermal/current-limit accuracy must pass before declaring that contract reliable. Adequate 20 V remains the lower-mode alternative. A 15 V-only source remains inhibited for a 12 V motor. Fault-current contract checks are unchanged because the bleeder is downstream of the eFuse.

References: [ROHM exact manufacturer datasheet](https://datasheet.lcsc.com/datasheet/pdf/47b0f000b8781f6578038fbca26cc49a.pdf?productCode=C2074262), [exact supplier part](https://www.lcsc.com/product-detail/C2074262.html), [Panasonic SVPK](https://industrial.panasonic.com/cdbs/www-data/pdf/AAB8000/AAB8000C226.pdf). Physical tests are unperformed; routing stays disabled.