Gooser

All 4 motors are controlled by a single STM32G431 in 3PWM mode, and the encoder connectors each have two ADC pins for reading linear hall sensors on the motor.

Soldering heavy copper bars on the back with ceramic capacitors between them worked well before, so I've committed to that design and removed the terminals from the board. The battery wires will attach to the copper bars, most likely using lugs and screws. There are holes for locating pins to ensure the copper bars don't slide around while assembling for reflow, and for mechanical strength. Sized to use electrolytic capacitor leads as the pins, getting them soldered in the same reflow operation (with some care to avoid overheating them).

The drivers are DRV8300, which has the same pinout as FD6288Q but doesn't need diodes. They are powered by the main input voltage. They and the mosfset gates are 20V max, so input voltage must stay below that. I've rated it for 4S lipo, which is 16.8V at full charge, to give some safety buffer.

I've included the buck converter from Valentine's Mosquito SIMPLEFOC_BLDC_MOSQUITO_V20 - EasyEDA open source hardware lab with some edits to output 5V instead of 3.3V, and capacitors recommended in the datasheet. That is regulated down to 3.3V for a cleaner supply to the linear halls.

This is my first 4-layer board design after watching a bunch of Robert Feranec videos. The inner layers are mostly ground fill for EMI reduction. I tried to give every signal via a ground via nearby to reduce field spread when changing ground planes.

The overall board size is 44x48mm, but has a 2mm margin that can be removed from one or more edges for various possibilities of joining multiple boards together. Remove the right-side margin from two boards so the mounting holes get cut in half, and the half-holes match up when joined side-by-side on a single set of 84mm long copper bars. Or trim all the margins from one board and join back-to-back on 44mm long copper bars. The motor wire solder pads on the back of the downward-facing board will then extend 2mm out from the edge for easy soldering, and the encoder connectors can be soldered on the back side to the second set of holes outside the margin. Cleaner and sturdier than having a bunch of cables coming out from under it (although the programming connector on the bottom board will need a cable). For a robot with up to 16 motors, you can even join four Goosers side-by-side and back-to-back, in which case four of the encoder connectors will need cables coming out from under.

Every pin of the MCU is utilized. There were no UART pins available, but USB is, and that should work just as well for communicating with Arduino serial monitor. UART2 could potentially be used via SWCLK and the aux port pin (I plan to use that port to communicate position targets from another STM32 with NRF24 wireless transceiver), but configuring SWCLK for it may result in needing hardware reset when programming. Please ignore the fact that the reset pin's pad is slightly enlarged. That's a contingency since I was going to use a 6-pin connector since I have them on hand, and this way it sticks out from under the connector housing so I can access it as a touch pad.

Compared to Lepton, the mosfets have a larger area to feed the ground pin, with a 3mm round exposed copper area for soldering conductivity boosters (easier to make than rectangular bits, by sawing a slice of copper wire and hitting it with a hammer to flatten).