v6 T4: Blocks O+P+Q+R - A7670SA LTE island complete in the schematic (33/33 net checks pass)
Block O (modem): U19 A7670SA-LASC (C18548276, LGA-124) - 19 signal nets + 39 GND + 66 NC,
all 124 pads accounted for. UART via 1.8V-side nets (_M suffix); USB joins J9 directly
(VBUS A4/A9/B4/B9 + DP/DM); USB_BOOT -> TP8; VDD_EXT -> 1V8_MODEM (shifter reference only);
NETLIGHT (1.8V) -> R109 4.7k -> Q22 MMBT3904 low-side, reusing the existing Rmled1+D4G1 LED
string re-driven: MODEM_LED label -> 3V3, D4G1 cathode GND label -> MODEM_LED_K.
Block P (3.8V rail): 24V -> D20 SS34 || MAIN5V -> D21 PMEG3050EP (C50603) -> VIN_MODEM ->
U20 AP64350SP-13 (C2071691, 40V/3.5A) -> L20 4.7uH -> VBAT_MODEM. Values from datasheet
DS41976: FB 0.8V (R110 75k / R111 20k = 3.8V), RT R117 200k = 500kHz (pin must not float),
COMP R116 14k + C141 3.3nF + C142 47pF, BST C129 100nF, EP -> GND. EN = MODEM_PWR_EN (PC10)
with R113 100k pull-down: rail DEFAULTS OFF (invariant #1); EN low >= 2s = HARD_CYCLE
(invariant #3). Bulk: 2x 220uF/10V (C2827940) + 22uF/1uF/100nF/33pF/10pF + R112 1k bleed +
D22 SMAJ5.0A TVS (C113952). VERIFY at layout: D22 pad-1-is-cathode against the fp marker.
Block Q (SIM): U22 MFF2_SIM symbol (new in Drogheda-sym; FOOTPRINT TBD "Drogheda:MFF2_5x6_TBD"
- build from the chosen carrier chip datasheet; PCB sync will flag it, expected). SIM_VDD
100nF + PESD5V0S1BA on DATA/CLK/RST (D23-25, C2827694, orientation cloned from D14).
Block R (level shift): 2x 74LVC2T45DC (C730237, direction-fixed). U23 DIR=GND: 3V3 DIN_4G ->
1.8V DIN_4G_M. U24 DIR=1V8: DOUT_4G_M -> DOUT_4G (also feeds U18 RXD2) and MODEM_STATUS_M ->
MODEM_STATUS (PC9). R106 100R taps DOUT_4G -> USART6_RX (PC7). DNP 0R bypasses R107/R108
across the RTCM path. Q20/Q21 2N7002 open-drain on PWRKEY/RESET with 100k pull-downs -
MCU absent leaves the modem free (invariant #1). PWRKEY timing: SIMCom >=50ms on / >=2.5s off.
CORRECTIONS this pass: (1) the schematic has 382 WIRE SEGMENTS - the earlier "zero wires"
finding was a regex artifact (multi-line xy); label-on-pin work was unaffected (all verified
against the PCB), but the LED relabel needed a wire-graph trace, now in the script.
(2) 9 stale no_connect markers removed from MCU pins that now carry labels (PC1-PC3, PA7,
PC8-PC12). (3) A7670SA is a 124-pad LGA, not the 88 claimed in the modem survey.
NEXT: Leo runs ERC + Update PCB from Schematic (U22 will complain - footprint TBD, skip it),
then T5 = LTC4040 input-limit errata fix + U7/L7 disposition, T6 = RF Block S + netclasses.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ4eAB6zjQT49RVwpgfKdt
4f6e1bb347