I2C slave writing troubles....

[tedlarson]

I have been trying to write a simple I2C master/slave communications program, and I am having all kinds of difficulties.

Both are PIC18F252's. I am on CCS PCH 3.155.
I am using the hardware I2C, with 1k pullups.

I have some 24LC65's on the same bus. I can talk to them fine. I am pretty sure it is my slave code that is the problem, but I cannot isolate it.

The behavior I see is that I never receive the 'R' from the master. I always miss it. It's like some kind of timing problem.

Any help would be greatly appreciated.

Thanks,

- Ted


Here is my slave:
--------------------

#use I2C(slave,sda=I2C_SDA,scl=I2C_SCL, address=SLAVE_ADDR,FORCE_HW)

#INT_SSP
void ssp_interupt ()
{
static int lsb=0;
static int msb=0;
int cmd;
int readbyte=0;


if(!i2c_poll())
{
if(icstate==1)
{
i2c_write(msb);
i2c_write(lsb);
msb=lsb=0;
}
icstate=0;
}
else
{
readbyte = i2c_read();
switch(icstate)
{
case 0:
icstate=1;
return;
case 1:
cmd = readbyte;
if(cmd=='R')
{
fprintf(DEBUG, "YES!!!\r\n");
msb = 3;
lsb = 3;
}
else
msb=lsb=99;

icstate=0;
return;
}
}
}

Here is my master:
-----------------------

#use I2C(master, sda=I2C_SDA, scl=I2C_SCL, FORCE_HW)

i2c_start();
i2c_write(MOTORBOARD);
fprintf(DEBUG, "Sending command\r\n");
i2c_write('R');
i2c_start();
i2c_write(MOTORBOARD+1);
msb = i2c_read(1);
lsb = i2c_read(0);
i2c_stop();
fprintf(DEBUG, "Motorboard said (%D)(%D)\r\n", msb, lsb);

[tedlarson]

I am really starting to get frustrated....

After fiddling with this problem for 2 days straight now....I am getting the feeling that the #use i2c stuff in the compiler doesn't work properly for building a slave that is timed properly. I went back and read MANY posts on this message board where everyone solved the problem by just taking control of the registers and DIYing a I2C slave.

All I want to do is build an I2C slave that can read 4 sequential bytes from a master, and send 4 sequential bytes in response. This sounds so simple....but it apparently isn't . The ex_slave.c example makes it look like it should be easy....unfortunately it isn't.

Does someone have a slave example USING the #use I2c that actually works for more than 1 byte up and 1 byte back that they could post?

Thanks,

- Ted

[Mark]

Well I posted something on the old board:

http://www.ccsinfo.com/forum/viewtopic.php?t=6821&highlight=exslave+c

The first example is doing it yourself using the registers (the way I would do it). The second example I modified the ex_slave.c file.

For your example, did you try an remove the fprintf statement from the slave code. That will cause a delay and probably a failure.

[Mark]

One other thing, does SLAVE_ADDR == MOTORBOARD? If not, that will also be a problem.

[tedlarson]

[Mark]

[Neutone]

Because the master triggers everything through the clock all delays between bytes should be in the master. When I do syncronous transfers using SPI the master over runs the slave if I dont place delays between bytes.

In the slave

[tedlarson]

CCS tech support just sent me the following code.
It works very well. My problem is now solved. You can send 2 bytes up, and get 2 bytes back, with a 1 byte address...although it would be easy to extend to 2. I have this running between two PIC18F252's running at 40mhz and it seems solid.

Enjoy!

- Ted

<PRE>
typedef enum {NOTHING, CONTROL_READ,
ADDRESS_READ, READ_COMMAND_READ, DATA_READ_LOW, ADDRESS_READ_LOW} I2C_STATE;

I2C_STATE fState;
byte address;
long buffer[0x10]={0x103, 0x205, 0x307, 0x409, 0x50A, 0x60C, 0x710, 0x812, 0x914, 0xA16};

#INT_SSP
void ssp_interupt ()
{
byte incoming;
static byte dl, dh;

if (i2c_poll() == FALSE) {
if (fState == ADDRESS_READ) { //i2c_poll() returns false on the
i2c_write(make8(buffer[address], 1));//interupt receiving the second
fState = ADDRESS_READ_LOW; //command byte for random read operation
}
else if (fState == ADDRESS_READ_LOW) { //i2c_poll() returns false on the
i2c_write(make8(buffer[address], 0));
fState = NOTHING; //command byte for random read operation
}
}
else {
incoming = i2c_read();

if (fState == NOTHING){
fState = CONTROL_READ;
}
else if (fState == CONTROL_READ) {
address = incoming;
fState = ADDRESS_READ;
}
else if (fState == ADDRESS_READ) {
dh = incoming;
fState = DATA_READ_LOW;
}
else if (fState == DATA_READ_LOW) {
dl = incoming;
buffer[address] = make16(dh,dl);
fState = NOTHING;
}
}
}
</PRE>

[Mark]

Do a test where you send only one byte and then stop (simulating some sort of error). Then try and do a correct sequence. I believe that you will find their state machine gets trashed! That is why it is better to look at the flags when reading the data.

[tedlarson]