I2C communication problem with two slaves (atlas scientific) and one Nano (official)

Question

I'm working on a project using an arduino Nano (official) and two peripherals from Atlas Scientific (pH and DO) communicating via i2c. Everthing works fine when i use a usb alimentation connected to the PC or to a 5V adaptater (typically a phone charger).

I want to use a 12V battery and a 12V DC 1.5A adaptater to power up my system, the problem is that it works fine for a moment (30 to 40 seconds) and then the pH start having problem with the communication and the led becomes red and the TX/RX leds from the arduino (which are connected to nothing start going crazy. I use 4.7k pull up resistors and an L7805ACV voltage regulator to go from 12 to 5V.

Attach, a part of the system schematics.

If you have any clue, starting to be new idealess

EDIT : Here is the part of the program concerning the i2c communication :

#include "Wire.h"

/*********************************************************************************************************
 *                                               Address
 *********************************************************************************************************
 */
int pH_address = 99;
int DO_address = 97;

/*********************************************************************************************************
 *                                         i2C CONFIGURATION
 ********************************************************************************************************* 
 */

byte received_from_computer = 0;                                                                           
char sensor_data[20];                                                                                      
byte in_char_sensor = 0;                                                                                   
byte j = 0;                                                                                                
byte code = 0;                                                                                             
int time_ = 850;     

void setup()
{
  Wire.begin();
}
void loop()
{
  pH = Get_i2C(pH_address);    
  delay(200);
  DO = Get_i2C(DO_address);
  delay(200);      
}

/**********************************************************************************************************
 *                                                  i2C
 * ********************************************************************************************************
 */

float Get_i2C(int address)
{
  char computerdata = 'R';  
  Wire.beginTransmission(address);
  Wire.write(computerdata);
  Wire.endTransmission();

  delay(time_);

  Wire.requestFrom(address, 16, 1);
  code = Wire.read();
  while (Wire.available())
  {
    in_char_sensor = Wire.read();
    sensor_data[j] = in_char_sensor;
    j += 1;
    if (in_char_sensor == 0)
    {
      j = 0;
      break; 
    }
  }
  return atof(sensor_data);
}                                                                    

Answer 1

Two ideas.

1: Count your grounds. If you count more than 1 you have a problem. Your system should have a single ground, so ANY connection on every component labeled GND should be connected to ground.

The reason multiple grounds are a problem is that your signals (other than differential signals) are measured relative to ground, so if two components have different ideas of what "ground" is they start to disagree about what constitutes a 1 and a 0.

2: Scope your power supply. Check that your 5V supply is still providing 5V. It's possible (although unlikely) that the L7805ACV can't supply enough current and your system is browning out. If possible have a look with an oscilloscope to check for noise on the 5V supply.

Your boards should already have decoupling capacators, but adding another won't hurt, so also try adding a 0.1uf cap between 5V and GND to eliminate any noise.