I2C >> Bit Error when compiling for 160 MHz

[detamend]

Basic Infos

• This issue complies with the issue POLICY doc.
• I have read the documentation at readthedocs and the issue is not addressed there.
• I have tested that the issue is present in current master branch (aka latest git).
• I have searched the issue tracker for a similar issue.
• I have filled out all fields below.

Platform

• Hardware: Wemos D1 mini pro (ESP8266)
• Core Version: 2.5.2
• Development Env: Arduino IDE
• Operating System: Windows 10 x64

Settings in IDE

• Module: Wemos D1 mini pro
• Flash Mode: not sure (using IDE)
• Flash Size: 4MB
• lwip Variant: v2 Higher Bandwidth
• Reset Method: not sure
• Flash Frequency: not sure (IDE)
• CPU Frequency: 80Mhz AND 160MHz
• Upload Using: SERIAL
• Upload Speed: 115200

Problem Description

Trying to let two ESPs talk to each other via I2C - easy, right? Obviosly not, when boths CPUs running at 160 MHZ.

SENDER SKETCH

#include <Wire.h>

#define SDA_PIN 4
#define SCL_PIN 5

#define sender 0x42
#define reciver 0x08

void setup(){
  Serial.begin(115200);
  Wire.begin(SDA_PIN,SCL_PIN,sender);
}

void loop(){
  Wire.beginTransmission(reciver);
  Wire.write("aaAA");
  Wire.endTransmission();
  delay(1000);
}

RECIVER SKETCH

#include <Wire.h>

#define SDA_PIN 4
#define SCL_PIN 5

#define sender 0x42
#define reciver 0x08

void setup(){
  Serial.begin(115200);
  Wire.begin(SDA_PIN,SCL_PIN,reciver);
  Wire.onReceive(receiveEvent);
}

void receiveEvent(size_t bytesRecived){
  (void) bytesRecived;
  String command;
  byte c;

  while(Wire.available()){
    c=Wire.read();
    Serial.print(c);
    Serial.print("\t");
    Serial.println(c,BIN);
    command+=(char)c;
  }
  Serial.println(command);
  Serial.println("---------");
}

void loop(){
  delay(1);
}

Compiled with both CPUs are running at 80 MHz, the reciver gets "aaAA" from the sender.

If I compile both scripts at 160 MHz CPU speed, the reciver gets

?aAA

instead.

To get an idea, whats been send/recived, I dumped every recived by in the reciver script:

225	11100001
97	1100001
65	1000001
65	1000001
?aAA

So the first bit of the data should have been 0 - but gets recived as 1. This sets the MSB of the first byte to 1, changing the expected 97 to a 225.

I use two 2k2 resistors as pullups on the SCL/SDA lines. Sadly no logic analyzer around, so I can't pinpoint the sender or reciver as the culprit.

The boards are connected via a triple base board, only two CPU Modules are plugged in.

Any help would be greatly appreciated! I know: simply go back to 80 MHz... but I really like to get this setup working.

[detamend]

Got my trusty Oscilloscope going and compared the signals for CPU@80MHz and CPU@160MHz. I'm sending to Adress 0x08 and transmitting a single Character: "S". S in binary reads 01010011.

On 80MHz I recive 01010011 (as expected).
On 160MHz I recive 11010011 (fist bit should be 0).

Besides the error, the timing is quite different. Running 80MHz, one cycle is roughly 23us. @160MHz a typical cycle is 17us.

On the screenshots of the oscilloscope it' s pretty obvious, that the first bit of 01010011 is not low when the CPU is running at 160MHz.

[Jason2866]

Maybe the same issue as #5762

[detamend]

Maybe the same issue as #5762

I read that issue - definitly related.

[Jason2866]

Can you try this?
arendst/Tasmota#6360 (comment)

[detamend]

Removing those two interrupts breaks the reciver completely even at 80 MHz (the script outputs nothing).

Sadly, that's not it, but thanks a lot, @Jason2866 !

[TD-er]

Maybe this one is related?
#6592

The PR description has a lot of links referring to the original issue and a detailed readme page describing what the original author tries to fix.

[detamend]

@TD-er - definitly related. Just replaced the core_esp8266_si2c.cpp with the one from the PR - now works like a charm (even at 160MHz).

Great Job, guys!

[devyte]

#6592 is merged and released.
Closing.

[laercionit]

I am using CORE 2.7.2 I have the same problem, anyone else?