Logic Analyzer

USB Logic Analyzer (Raspberry Pi Pico)

A 6-channel USB logic analyzer built on the Raspberry Pi Pico H, using PIO + DMA for capture and a custom binary protocol over USB serial. Includes protocol decoders for UART, SPI, and I2C, each verified against real hardware.

A custom 2-layer PCB was designed in KiCad for this project (fab-ready gerbers included), but has not yet been manufactured. All functional testing and protocol verification below was done on breadboard, using the same firmware that targets the PCB -- fabbing the board is the natural next step once further validated on breadboard hardware.

Status

How it works

1. Capture (firmware, firmware/) -- A PIO state machine samples 6 GPIO pins every cycle and pushes them into the RX FIFO. DMA moves samples from the FIFO into a RAM buffer with no CPU involvement in the sampling loop itself, so capture timing is not affected by anything else the Pico is doing.
2. Transfer -- Once a capture completes, the Pico sends a binary frame over USB serial: a SAMP magic marker, a 4-byte little-endian sample count, then the raw packed samples (1 byte per sample, low 6 bits = the 6 channels). No text parsing, no ambiguity about where the data ends.
3. Host tools (host-tools/) -- Python scripts read that binary frame, save it as .bin and .csv, and decode it according to the target protocol.

The I2C debugging story

The I2C decoder initially looked broken -- captures came back completely flat, zero transitions on SDA/SCL, no transactions found. Chasing this down turned into a real debugging exercise worth documenting:

1. Confirmed pull-up resistor wiring (4.7k to 3.3V on SDA/SCL) was correct.
2. Rebuilt the breadboard wiring from scratch -- same result.
3. Removed the Pico from the circuit entirely, tested ESP32 + resistors only -- still failing (Wire.endTransmission() returning error code 4).
4. Ran a bare-board I2C scanner sketch with zero wiring connected -- worked cleanly. This proved the ESP32 pins and I2C library were fine in isolation.
5. Reconnected full wiring, re-ran the scanner -- still worked. This proved the wiring wasn't the problem either.
6. Went back to the custom sketch and bisected it line by line: address-only probe (worked), added setClock() (worked), added a single write() byte (failed, error 4).

This isolated the fault to a known regression in recent versions of the arduino-esp32 core's rewritten I2C driver, not the hardware or the wiring at all -- multi-byte Wire.write() calls were triggering spurious bus timeouts on this core version. Once identified, testing continued on a known-stable core version, and the bus began working correctly.

Separately, the very first "working" captures still came back flat because the test sketch used delay(500) between transactions -- with such a low duty cycle, a fixed-length trigger-based capture window had low odds of landing inside the brief active period. Removing the delay (keeping the bus continuously busy) fixed this immediately.

Both of these turned out to be more interesting than the decoder logic itself, which worked correctly the whole time.

Repository structure

firmware/         Pico C firmware (PIO capture + DMA + USB binary protocol)
host-tools/        Python capture + decode scripts
test-sketches/      ESP32 Arduino sketches used to generate known-good UART/SPI/I2C traffic
debug-tools/        spi_diag.py -- per-channel activity check, used to separate
                    wiring/mapping issues from decode logic bugs (same technique
                    used in the I2C bisection above)
hardware/           KiCad project (schematic, PCB layout, fab-ready gerbers)

Building the firmware

Requires the Pico SDK set up per the standard Raspberry Pi Pico C/C++ SDK instructions.

cd firmware
mkdir build && cd build
cmake ..
make

Flash the resulting .uf2 to the Pico in BOOTSEL mode.

Running a capture

cd host-tools
python capture_read_binary.py      # triggers a capture, saves capture.bin / capture.csv
python decode_uart.py              # or decode_spi.py / decode_i2c.py

Each decoder has its channel mapping documented at the top of the file -- update SDA_BIT/SCL_BIT (I2C), channel constants (SPI), or CHANNEL (UART) to match your wiring.

Hardware

• Custom 2-layer PCB, 75x80mm, designed in KiCad
• 5V-tolerant voltage dividers on all 6 input channels
• Onboard I2C pull-ups, status LED
• Not yet fabricated -- design is complete and DRC-clean, gerbers are included in hardware/gerbers/. All testing to date was done on breadboard using the same firmware.

Future work

• Fabricate and assemble the PCB
• Add malformed-frame / truncated-capture handling to decoders
• Support I2C clock stretching (not exercised by the MPU6050 test used here)
• Automated regression tests against saved capture files