SunSproutHub

Work in progress — never fabricated or tested

No physical board exists yet. Everything on this site is verified in software only: ERC and DRC pass, values and pin assignments are checked against manufacturer datasheets, and the fabrication outputs generate cleanly. Nothing has been powered on or measured.

The design is published at this stage to gather feedback before a fabrication run. Treat it as an untested design rather than a finished product, and review it yourself before building one — errors in the charging or protection circuitry could damage a battery.

A compact power-management and connectivity board built around the ESP32-C5, designed to take power from USB-C, a small DC/solar source, or a Li-ion/Li-Po battery — charge the battery with true maximum-power-point tracking on the solar input, and regulate a clean 3.3V system rail — while exposing two independent I2C buses for onboard and remote sensors.

Every IC's datasheet-required external components are drawn directly into the schematic rather than treated as black-box breakout modules, and the board is hand-routed rather than auto-generated.

Physical

• 6-layer PCB (F.Cu, 4× inner, B.Cu).
• 10 external connectors: USB-C power input, screw-terminal DC/solar input, JST battery input, two STEMMA QT ports, two 8P8C (RJ45-style) differential-I2C jacks, two field-terminable thermistor connectors, and two GPIO expansion headers you fit yourself. See Connectors for the full list.
• Two LEDs: a charge-status LED driven by the charger, and a firmware-controlled status LED.
• Three buttons: reset, boot-mode, and a wake button that brings the charger out of ship mode — or, held for ~10 seconds, hard power-cycles the whole board.
• U.FL external antenna connector (module-side), for installations where the onboard antenna isn't sufficient.

Capabilities

Compute & radio

MCU	ESP32-C5-WROOM-1U
Wireless	Dual-band Wi-Fi 6, Bluetooth 5, 802.15.4, CAN FD
Antenna	U.FL external connector (module variant)

Power management

• Li-ion/Li-Po battery charging with JEITA temperature-qualified fast charge, via an onboard charge-temperature thermistor input.
• Automatic source arbitration between USB-C and the DC/solar input — both can be connected simultaneously; the charger ORs them into a single charge-path rail.
• Fuel gauge for state-of-charge tracking, with its own independent battery-temperature thermistor input (kept separate from the charger's, since each IC applies its own excitation to whatever's wired to its thermistor pin).
• True MPPT on the solar input, performed by the charger itself. It periodically samples the panel's open-circuit voltage and regulates to the corresponding maximum-power point, autonomously — no firmware control loop and no separate monitor IC involved.
• Measurement from the charger's integrated 16-bit ADC — input voltage, input current, battery voltage, battery current and temperature, all readable over I2C. Between that and the fuel gauge, every rail is covered without a dedicated power-monitor part.
• Switched 3.3V "user" rail, independent from the always-on system rail, that firmware can power down to save current when the secondary I2C bus and its downstream devices aren't needed.

Other protections

• Reverse-polarity protection on both the battery and DC/solar inputs, sized to each source's expected voltage range. USB-C doesn't need one — it's a keyed, spec-defined connector.

• ESD/TVS protection on the ports designed to carry a cable off-board: the four differential I2C lines on the two 8P8C jacks, and the USB-C data lines. The USB VBUS rail gets its own TVS as well.

  The remaining connectors have no dedicated ESD parts — both STEMMA QT ports, the two GPIO expansion headers and the debug UART on them, both thermistor inputs, and the USB CC lines. Those are intended for sensors sitting close by inside the same enclosure. If you run a cable from one of them out of the enclosure, add protection at that end.

• Hardware alert/interrupt lines from the charger and the fuel gauge are wired to their own separate GPIOs, so firmware can react to a fault without polling — and without having to interrogate both devices to work out which one raised it.

Interfaces

I2C — two independent buses

• Internal bus — always-on, dedicated to the board's own power-management ICs (charger and fuel gauge) plus a local STEMMA QT connector for directly-attached sensors.
• User bus — switchable, feeding a PCA9615 differential I2C buffer that drives two 8P8C jacks, plus a second, separate STEMMA QT connector.

The differential side is wired to be compatible with the SparkFun QwiicBus — EndPoint: same PCA9615 buffer, same 8P8C pinout, so a standard Ethernet patch cable carries I2C from this board to an EndPoint, and Qwiic/STEMMA QT devices plug into the EndPoint at the far end. That's how you get sensors metres away from the hub instead of centimetres.

This board takes the pass-through role, so it carries no termination resistors on the differential lines — the EndPoint at the far end provides them. It therefore needs to connect to an EndPoint rather than acting as the end of a chain itself.

These buses are deliberately kept apart — see Notes for why that matters when wiring things up.

Connectors

See the dedicated Connectors page for the full list with mating-part guidance.

Power

Input

Source	Connector	Notes
USB-C	USBC1	5V USB-C power/data input. The charger runs BC1.2 and non-standard adapter detection over D+/D- to set its own input current limit. There is no USB-PD controller, so the port is 5V only; HVDCP (9V/12V) is reachable only from a USB-A QuickCharge supply over an A-to-C cable, and only if firmware enables it.
DC / small solar panel	CN5 (screw terminal)	Wide-range DC input suitable for a small solar panel; can be connected at the same time as USB-C.
Battery	J4 (JST PH, 2-pin)	Single cell. Double-check pack polarity against the board's silkscreen before connecting — JST-PH battery polarity isn't universally standardized across cable vendors.

Charging summary

Charger	BQ25798, buck-boost, dual-input
Cell configuration	1S, set in hardware by the PROG resistor at power-on
Chemistry	Li-ion / Li-Po and LiFePO4 (the fuel gauge carries profiles for both)
Input voltage, either source	3.6 V to 24 V for a valid input; 30 V absolute maximum
Practical solar ceiling	24 V, the charger's own recommended maximum — the input protection no longer sets a lower one. The reverse-polarity FET on that input is self-biased, so the full input voltage appears across its ±20 V gate in normal operation as well as under a reversed source; a Zener gate clamp (D2) holds that within rating across the whole range. Size a panel by its cold-weather open-circuit voltage, which rises above the rated Voc. See input protection.
Input current limit	~2.0–2.1 A, set in hardware by a resistor divider. Firmware cannot exceed it. The exact value tracks the charger's REGN rail, which is itself input-dependent — see below.
Charge current	Firmware-set over I2C. Defaults to 1 A; do not exceed 2 A (see below).
Solar tracking	Autonomous open-circuit-voltage MPPT, run by the charger itself
Source arbitration	Both inputs may be connected at once; the charger's internal mux selects between them
Measurement	Integrated 16-bit ADC: input voltage/current, battery voltage/current, temperature
Temperature qualification	JEITA, via a thermistor on J302 (not fitted — see Connectors). With no thermistor fitted the charger reads "too cold" and suspends charging until firmware sets TS_IGNORE; see below.

Output

Rail	Voltage	Description
System rail (always-on)	3.3V	Powers the MCU, the internal power-management I2C bus, and its STEMMA QT connector (J6).
User rail (switched)	3.3V	Powers the I2C differential buffer and its STEMMA QT connector (J203); firmware-controlled, defaults OFF at power-up.

Charging & PMU behavior

• Input current limit: roughly 2.0 A to 2.1 A, hardware-set by the R302/R303 divider on the charger's ILIM_HIZ pin. This is an analog ceiling the charger enforces regardless of what firmware writes to the corresponding register. It caps the combined draw from whichever source (USB-C, DC/solar, or both at once) is feeding the charge path — it isn't the battery's charge rate.

  It is not a single fixed number, because the divider is referenced to REGN, and REGN itself depends on the input voltage. The charger computes the clamp as V(ILIM_HIZ) = 1 V + 800 mΩ × I, and the divider ratio is 130 k / (110 k + 130 k) = 0.542:

  Input	REGN (typical)	Resulting limit
  USB-C at 5 V	4.8 V	2.00 A
  Solar or DC at 15 V	5.0 V	2.14 A
  Across the full REGN spec, 4.6 V to 5.2 V		1.87 A to 2.27 A

  The divider was sized for the 5 V case. On a high-voltage input the ceiling sits about 7% above 2 A, and worst-case silicon reaches 2.27 A — budget the copper for the top of that range, not for 2.00 A. The charger latches this clamp from an ADC reading taken at power-on before the converter starts switching, so whichever source is present at power-up sets it for that session.

• Charging is enabled in hardware, but not currently sufficient on its own. The charger's CE pin is tied to ground, so nothing has to be written to start a charge. The charger's own protections (OVP, OCP, thermal shutdown, UVLO) apply throughout.

  As built, though, a board with no firmware running will not charge, because of the thermistor. J302 ships unpopulated, and the TS divider (R305 5.1 kΩ to REGN, R306 30 kΩ to ground) then sits at 85.5% of REGN. Every JEITA cold threshold is below that — the 0 °C threshold is 73.3% of REGN, and even the −20 °C OTG threshold is 80% — so the charger reads a battery colder than its cold cutoff and suspends charging.

  The network is correct once a thermistor is present: a 103AT-type 10 kΩ NTC at 25 °C puts the divider at 59.5% of REGN, mid-window. Fit one — a Semitec 103AT-11 is the specified part; see Connectors. It is what the charger's datasheet expects, and the only arrangement that gives real JEITA temperature protection.

  To run without one, firmware must set TS_IGNORE (register REG18, bit 0), which tells the charger to treat the temperature as always acceptable. The bit defaults to 0 and TI lists it as reset by the watchdog and by a register reset; this board's firmware disables the watchdog, so in practice only the register-reset control clears it. That route gives up temperature qualification entirely, and the ESPHome driver does not currently expose the bit.

  There is deliberately no fixed resistor standing in for the thermistor. It would make the charger read a permanent 25 °C and report that as a real measurement, which is a worse failure than having no reading at all.

• Battery charge current is firmware-configured over I2C, and resets to that 1A default on power-up, on a watchdog timeout, and on a register reset. Program a charge current appropriate to your specific pack's capacity before relying on fast charging.

• Firmware must not set the charge current above 2000mA. Unlike the input limit, the charge current has no hardware ceiling, and in buck mode it can exceed the input current — so a higher setting can push the charge-side copper past its rating.

• Ship mode and a true hardware power cycle, via an external ship FET (Q3) on the charger's SDRV gate-driver pin. The FET sits in series in the battery path, so the charger can physically disconnect the pack from the system rail. That gives three states beyond normal operation: ship mode (FET off, I2C still alive — for shipping and storage), shutdown (FET off, I2C down, only an adapter brings it back), and a system power reset that drops the FET for ~350ms and actively pulls the SYS rail down before restoring it.

• Holding SW3 (WAKE) for about 10 seconds hard power-cycles the board. This runs entirely inside the charger — no working firmware, no I2C — which makes it the recovery path when the MCU is wedged. A shorter press (~1s) just wakes the charger out of ship mode.

  Two caveats worth knowing. The reset cycles the SYS rail, so it restarts the MCU, both 3.3V rails and everything downstream of them — but not the charger itself (it's the thing performing the reset) and not the fuel gauge, which is deliberately wired to the pack side of the ship FET so it keeps coulomb-counting straight through. And since neither I2C device loses power, the reset can't unstick a hung I2C slave; that still needs the usual nine-clock recovery sequence in firmware.

• Firmware must set the charger's SFET_PRESENT bit (REG14, bit 7) during init. It defaults to 0, and while it's 0 the charger locks out every ship-FET mode above — the board behaves as though no FET were fitted.

Current consumption

Not yet characterized — measurements will be added here once bring-up testing is complete.