SAMD51 Thing Plus: I2C Issues

Development Boards
1

I’m connecting a SAMD51 Thing Plus to a Raspberry Pi Compute Module over i2c. When the Pi performs an I2C operation the SAMD51 ceases to respond. I’ve ran the same code with an Uno and had no issues. I’m at a loss for what to do next, any advice?

I2C is wired into directly into SDA(D20), SCL(D21) for the SAMD51, and A4(SDA), A5(SCL) for the Uno. On the Pi I have it wired directly into the hardware I2c bus and configured to show up as dev/I2C-1,

I’ve created a sketch just to demonstrate this issue:

#include <Wire.h>

// I2C address
#define ADDR 0x08

void setup() {
  // Serial.begin(9600); // uno
  Serial.begin(115200);  // thingplus

  Wire.begin(ADDR);
  Wire.onRequest(requestHandler);
  Serial.println("Setup complete");
}

void loop() {
  Serial.println(millis());
  delay(1000);
}

void requestHandler() {
  Serial.println("Req: responding 6 bytes");
  // response message
  Wire.write("ping  ");
}

Compile, upload and serially connect to SAMD51:

arduino-cli compile -v --fqbn SparkFun:samd:samd51_thing_plus ./sketches/i2ctest && \
arduino-cli upload -v -p /dev/ttyACM0 --fqbn SparkFun:samd:samd51_thing_plus ./sketches/i2ctest && \
sleep 2 && \
cat /dev/ttyACM0

After uploading the program to the SAMD51 I verify over serial that the time is being printed then I run i2cdetect on the Pi:

$ i2cdetect -y 1
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- --

As you can see, no devices show up and at this point the SAMD51. As soon as I run i2cdetect the Thing Plus stops printing timestamps. The result is the same for i2cdetect, i2cget or a C program I wrote to handle I2C operations on the Pi.

Here you can see what happens when I run i2cget for the SAMD51:

$ i2cget -y 1 0x08
Error: Read failed

Once again the Thing plus stops logging serially, the message in the request handler, “Req, responding 6 bytes” is never printed.

Whereas when I run the program on UNO, everything is nominal.

After switching to serial at 9600 I compile, upload and listen to serial:

arduino-cli compile -v --fqbn arduino:avr:uno ./sketches/i2ctest && \
arduino-cli upload -v -p /dev/ttyACM0 --fqbn arduino:avr:uno ./sketches/i2ctest && \
sleep 2 && \
cat /dev/ttyACM0

On the Pi we run I2c detect to find the device at address 0x08:

$ i2cdetect -y 1
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- 08 -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- 1a -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- --

After running detect, the Uno still sends timestamps.

Running i2cget also works as expected:

$ i2cget -y 1 0x08
0x50

The Uno successfully logs, “Req, responding 6 bytes” and continues to output timestamps.

2

I’ve tried the following to no avail:

  • Adding 1k, 2.7k, 3.3k, 4.7k pull up resistors to SDA/SCL in case there aren’t any on the board.

  • 5v to 3v level converter with no luck.

  • Rewiring everything, changing breadboards.

I’ve hooked up other I2C devices to the pie and had no issue running I2C detect finding them.

3

Further update:

I’ve ran the same code on another SAMD51 chip, the Adafruit Trellis M4 and am facing the same issue.

4

I tested on an Adafruit ItsyBitys M4 (SAMD51) today and the same issue is present.

I ordered a logic analyzer to see if that will help uncover any issues, should arrive Monday.

5

Turns out there is an issue with the underlying Wire implementation for ArduinoCore-samd. The Adafruit team has an issue opened on their fork of the SAMD core with a workaround posted. The issue can be found here: https://github.com/adafruit/ArduinoCore-samd/issues/45