I2C scanner ?

Comments

• kwinn Posts: 8,697

  2017-02-10 20:27

  Shouldn't be hard to scan through all the addresses to see which ones return an ACK, but I'm not sure how difficult it would be to identify the chip at that address.
• Erlend Posts: 612

  2017-02-10 21:37

  I made one in Spin, but it should be easy to transcribe:

  PUB WhoOnBus(ptrOnBusArr) | onbus, addr                             'Fills an array with max 119 elements with addresses that get a response
                                                                      'and writes how many is onbus to the 0th element
     onbus:= 1
     
     REPEAT addr FROM %0000_1000 TO %0111_0111                        'Scan the entire address space, exept for reserved spaces
       IF CallChip(addr << 1)== ACK                                   'If a chip acknowledges,
         LONG[ptrOnBusArr][onbus]:= addr                              'put that address in the callers array
         LONG[ptrOnBusArr][0]:= onbus                                 'and update the total count of chips on the bus
         onbus++                                                      
         IF onbus> 119                                                'until loop expires or maximum number of elements in the array is reached
           Stop
           QUIT
       Stop                                                           'After each call send a stop signal to avoid confusion
  

  
  Erlend
• john_s Posts: 369

  2017-02-10 22:50

  Erlend, don't you need to put some provision for time gap expired while waiting for ACK?
• Erlend Posts: 612

  2017-02-11 10:21

  True, but I2C is so fast, and Spin is so comparatively slow, that the few executions in the loop is sufficient. Same with any intrepeted lanuage I would think.
  
  Erlend
• Tracy Allen Posts: 6,667

  2017-02-11 18:32 edited 2017-02-11 18:33

  While it is not documented, the I2CIN and I2COUT commands will themselves time out if they do not receive the expected data. I forget the exact duration. Unlike SERIN there is no provision for a timeout vector, and there is no error code associated with a timeout. It is easy enough to test for i2c inputs that you expect to be present by reading a register for valid response. However, just to say "hello" to an arbitrary address, I think maybe the only thing to go on would be the timing of the response, and the Stamp does not have a built-in timer to support that. Good question.