Hardware V2.0
If I were to design a second version of this board, there are several things I would change based on what I learned during bring-up, integration, and the final demonstration.
Replace Micro USB with USB-C
The Micro USB connector worked fine for programming but it feels outdated and the connector is more fragile than USB-C. A USB-C receptacle with the correct CC resistors would be a straightforward swap that makes the board more durable during repeated demo handling and more compatible with modern laptops that no longer ship with Micro USB cables. The ESP32-S3 native USB differential pair would still be used the same way, only the physical connector would change.
Add a Dedicated Camera Power Rail
The OV5640 camera currently shares the main 3.3V rail with the ESP32, the debug LEDs, and everything else on the board. During camera operation the current draw spikes significantly. In a V2 design I would add a small secondary LDO, something like the TLV70033, specifically for the camera module. This would isolate camera noise from the ESP32 power rail and reduce the risk of brownouts during frame capture. It would also make it easier to power-gate the camera when it is not in use, which would help with overall power consumption.
Fix the Camera Connector and Verify Pin Mapping
The J6 camera connector on V1 used a 2x9 pin header with a pin mapping derived from the schematic. The mismatch between the schematic pin assignments and the MicroPython esp32-camera library's expected pin ordering was one of the main reasons the camera was not demonstrated. In V2 I would verify the exact pin mapping required by the firmware library first, then design the connector and schematic around that, instead of the other way around. I would also add a small silkscreen label on the PCB for each camera pin to make bring-up easier.
Add Decoupling Capacitors Closer to the ESP32 Module
The current design has decoupling capacitors on the 3.3V rail but they are not placed as close to the ESP32 module pads as they should be. During WiFi transmission the ESP32 draws current in short bursts and the decoupling needs to be right at the module to be effective. In V2 I would place 100nF and 10uF capacitors directly adjacent to the 3V3 and GND pins of the module following the Espressif layout recommendations more strictly.
Switch to a Higher Current Regulator
The AP63203WU-7 at 2A worked within the power budget, but the margin between the worst-case load and the regulator limit was comfortable but not generous. If V2 adds a dedicated camera rail and any additional peripherals, upgrading to a 3A regulator like the AP63205WU-7 in the same SOT-23-6 package would provide more headroom without any layout changes.
Add an LED for Camera Status
The current board has LEDs for MQTT activity, WiFi status, UART RX, and UART TX, but nothing specifically for camera state. In V2 I would add a fifth indicator LED on a dedicated GPIO to show whether the camera is actively streaming. During demos this would make it immediately obvious whether the camera stream is running without needing to check the laptop viewer.
Improve the Antenna Keepout Zone Compliance
The ESP32-S3-WROOM-1 module requires a keepout area under the antenna section of the module where no copper, traces, or components should be placed. The V1 layout followed this generally but could be improved. In V2 I would be more careful about enforcing the keepout on all layers including inner layers and the ground plane, following the Espressif PCB design guidelines exactly.
Add JTAG Debug Header
The current board relies entirely on USB serial for debugging. In V2 I would add a small JTAG header broken out from the ESP32-S3 JTAG pins. This would allow hardware-level debugging with OpenOCD if the firmware ever gets into a state where USB serial is not responsive, which happened a few times during bring-up when the firmware crashed before the USB stack initialized.