ATTiny82 I2C slave, based on wire example code, consistently runs then fails

[mathew314]

Hi

I'm having trouble with an I2C slave on a ATTiny85 using "ATTinyCore" and the "Wire" library that runs for a while then stops. I have a reproducible test that fails, and I can get the same hardware working with a different implementation ("attiny" and "TwoWireS"), so my theory is that it's related to my usage of ATTinyCore + Wire library. Any tips would be appreciated, details follow:

// Based upon ATTinyCore/avr/libraries/Wire/examples/slave_sender/slave_sender.ino
#include <Wire.h>

const uint8_t DIAG_LED = 3;
bool led_state = true;

void setup() {
  pinMode(DIAG_LED, OUTPUT);
  digitalWrite(DIAG_LED, LOW);
  Wire.begin(0xB);                // join i2c bus with address 0xB
  Wire.onRequest(requestEvent); // register event
}

void loop() {
  digitalWrite(DIAG_LED, led_state? HIGH: LOW);
  led_state = !led_state;
  delay(100);
}

// function that executes whenever data is requested by master
// this function is registered as an event, see setup()
void requestEvent() {
  Wire.write("he"); // respond with message of 2 bytes (so it fits in my scope/decoder)
}

(my ref: attiny85_i2c_attinycore_example.ino)

Setup:

1. I2C master (Pico W) scans the I2C bus for clients, and then cycles through the following devices doing 20 reads each and printing diagnostics.
2. I2C Slave at 0xB: ATTiny85
3. I2C Slave at 0x40: INA219 Current Sensor

Expected outcome:

1. Vast majority of reads succeed on both devices

Actual outcome:

1. ATTiny85 I2C slave: Works for all address scans. It works for the first 15 sends and then doesn't respond. Address scans continue to work, and status LED continues.
2. INA219 Current Sensor I2C slave on same bus: works 100% of the time.

Variations tried:

1. ATTinyCore 1.5.2: Works then fails after 15 interactions.
2. ATTinyCore 2.0.0-dev (from Github): Works then fails after 15 interactions.
3. attiny/TinyWires version on same hardware: Works 100%.
4. ATTinyCore 1.5.2: Change size of data. Doesn't change failure point.
5. ATTinyCore 1.5.2: Vary amount of time since reset. Doesn't change failure point.

Current suspicion: It's correlated to the 4 bit counter USIOIF overflow. This goes high at the failure point, but it gets reset during the bus scan and slave interactions don't resume.

Any suggestions on root cause, or what to do next to get the example working?

Regards
M

Versions:

• ATTinyCore v 1.5.2 and 2.0.0-dev.
• Arduino: 2.3.4
• Processor: ATTiny85 (fuses set to 8 MHz)
• Secondary slave device: INA219 DC current sensor
• I2C Master: Raspberry Pico W

[FL550]

I have the same issue with a ATTiny85 which runs as a slave. It only responds to exactly 15 requests.

[mathew314]

I tried changing the TWI_RX_BUFFER_SIZE and TWI_TX_BUFFER_SIZE from the default of 16 (avr/libraries/Wire/src/USI_TWI_Slave/USI_TWI_Slave.h) and observed the failure point was TWI_RX_BUFFER_SIZE - 1. That was a surprise because I wasn't planning on receiving data. After added a receive event response that drains the receive buffer (based on the slave receive example) the slave ticks along merrily sending data.