I2C Repeated Start

[cdan]

Hi,
the last release of the Wire library has added support for an extra parameter to the Wire.endTransmission and Wire.requestFrom, to specify if a STOP should be sent or not. Nevertheless, it seems it is not working for Due. I have inspected the source code for the two aforementioned methods and they ignore the boolean parameter altogether. I have also inspected the SCL and SDA lines with an oscilloscope   and I could clearly see the STOP condition being sent.

Am I missing something? or is this not supported on Due yet?

Best regards,
Dan.

P.S.
See the attached files.

[chriskner]

Dan,

I'm about to start an I2C project that may require I2C repeated starts.

Have you learned anything about this issue?

-Chris

[cdan]

Hello,
what board are you going to use ? What are the specific I2C transactions that you'll need?

With the Due board, I was able to read/write from/to an internal register of a I2C slave device (MLX90614), using the API in the libsam library. You can find the source in
arduino-1.5.1r2\hardware\arduino\sam\system\libsam\source. Look for twi.c and twi.h
The functions of interest are void TWI_StartRead and void TWI_StartWrite. They both have a parameter called uint32_t iaddress where you can put the internal address/register you want to access. If you specify it, then the controller will automatically  send a REPEATED START condition after writing the address of the slave device and of the internal register on the bus and before reading/writing the data bytes.

Best regards,
Dan.

Dan,

I'm about to start an I2C project that may require I2C repeated starts.

Have you learned anything about this issue?

-Chris

[chriskner]

Hi Dan,

On the Due.

I have some pcb's with both a PCA9539 (from NXP) and LTC2489 (Linear Tech), perhaps others.

Protocols:
LTC2489_I2C.png (12kB)
PCA9539_I2C.png (22kB)

There's a bit of history with these boards (struggling to get other external uC's to issue an I2C repeated start without stop), so your post caught my attention.

I appreciate the pointers, and any other suggestions you may have.

Regards,

Chris

[cdan]

Hello,
I had a quick look over the LTC2489's specs and I think it would be very easy to interface it with the Due board.

• You WRITE to the device's input register to select the input channels for the next selection. Input register is 8bit long.


• You READ from the device's output register the value of the last conversion. The output register is 24bit long


• The device starts a new conversion after the output register is READ and the master has generated a STOP CONDITION


• If you want to select a different channel and read the previously converted value, in a single transaction, then you use a REPEATED START condition

Check out the attachment.
I'll have a look to the PCA9539, later this evening.

[chriskner]

Thanks Dan,

I'm on the road right now.  I'll check it all out in one week.

-Chris

[chriskner]

Hi Dan,
(DUE, 1.5.1r2)

I also have confirmed that a 'Stop' is always sent regardless of including true/false in endTransmission().

I can supply logic traces if anyone's interested.

Now, perhaps this neophyte could have a bit more help? 

How does one invoke 'TWI_Startread' in code?  Specifically, what would I include for the first parameter:
    TWI_StartRead(, i2c_addr, i2c_register, 8 );

Thanks,

Chris

[cdan]

The first param is a pointer to a Twi structure denoting the I2C bus that you want to use.
In the case of arduino due there are 2 I2C buses, therefore you can use WIRE_INTERFACE_ID or WIRE1_INTERFACE_ID. Both of them are defined in hardware\arduino\sam\variants\arduino_due_x\variant.h

[chriskner]

Dan,

Please forgive my ignorance on this matter.

However:

Code: [Select]

TWI_StartRead(WIRE_INTERFACE_ID, i2c_addr, i2c_register, 8 );
results in:
error: invalid conversion from 'int' to 'Twi*'
error: initializing argument 1 of 'void TWI_StartRead(Twi*, uint8_t, uint32_t, uint8_t)'

I have this uncomfortable feeling of not knowing the right questions to ask, and where to look to figure things out...

Regards,

Chris

[cdan]

It's my mistake. Use  WIRE_INTERFACE or  WIRE1_INTERFACE
You can find both defined in variant.h

[chriskner]

Thanks Dan... Onward!

[chriskner]

After some turmoil, I have retreated to the state where only 1 byte at a time may be read from a slave.

Figure 34-20. "TWI Read Operation with Multiple Data Bytes with or without Internal Address" in the SAM3X8E datasheet seems out of reach. (Page 729, of the 28-May-12 revision)

I have found that a NACK is always generated after the 1st read by some unidentified mechanism, making multiple byte reads impossible.
For the desired protocol, see Figure 34-10. "Master Read with Multiple Data Bytes" (Page 721).

It seems that the Arduino community has been driving in circles over this problem for years.  For a sample:
http://code.google.com/p/arduino/issues/detail?id=28

I haven't been able to find a way to specify how the master (the Due) should respond for multiple sequential reads.

I have resorted to supporting only 1 byte reads for the foreseeable future.  This, of course, is a hack in my 50+ device I2C network.

Any ideas/suggestions?

Code: [Select]


    TWI_StartRead(WIRE_INTERFACE, i2c_addr, i2c_register, 1);
    Embiic_WaitByteXferComplete();
    for(i = 0; i < numOfBytes; i++) {
      //Need to set STOP condition WHILE reading the last byte...
      if ((i+1) == numOfBytes) {
        TWI_SendSTOPCondition(WIRE_INTERFACE);
        //DBG_PRINT("Sending last-byte STOP.");
      }
      Embiic_WaitByteReceived();
      //DBG_PRINTLN("Read_I2C_Values: Byte Ready");
      params[i] = TWI_ReadByte(WIRE_INTERFACE);
    }
    if (Embiic_WaitByteXferComplete()) {
      DBG_PRINTLN("Read_I2C_Values: Transaction completed successfully.");
    } else {
      DBG_PRINTLN("Read_I2C_Values: Transaction nottacompleted.");
    }


[cdan]

Can you please post a screenshot of the output from running your algorithm with numOfBytes=2 ?

[chriskner]

The traces above are the same with any numOfBytes value.  That one was generated with nOB=3.

However, I get the same result for nOB= 1.

Of course, as you know, we expect the NACK on only the last read, and ACK's on all 1..n-1 reads.

The way that it is now, all reads beyond the first byte result in a timeout. (code below)

I suspect that there's a state machine somewhere (that I can't find), that I would like to look over.

I did find this:
/arduino-1.5.1r2/hardware/arduino/sam/system/CMSIS/Device/ATMEL/sam3xa/html/TWI0.html
which lists the TW registers in a nice format.  However, there isn't much to play with there.

Code: [Select]


boolean Embiic_WaitByteReceived() {
  unsigned long start, now;
  start = millis();
  while (!TWI_ByteReceived(WIRE_INTERFACE)) {
    //if (WIRE_INTERFACE->TWI_SR & TWI_SR_NACK) { // Failed ACK
    //  DBG_PRINTLN("Embiic_WaitByteReceived: Failed ACK.");
    //  Error_Code = 9010;
    //  return false;
    //}
    now = millis();
    if (now < start) {  // cheap! handle overflow, at at the expense of a longer timeout
      start = now;
      now++;
    } else {
      now = now - start;
    }
    if (now > I2C_READ_TIMEOUTmS){
      DBG_PRINTLN("Embiic_WaitByteReceived: Timeout.");
      Error_Code = 9011;
      return false;
    }
  }
  return true;
}

[cdan]

I have just order a Bus Pirate from a retailer in RO. I'll have it tomorrow and I'll give it a try too.
I have to look closer to this issue.