RX Buffer...futile search so far

[arga]

The problem I am having is this:

I have a circular buffer in the form of an array of 55 elements (type int). I use int_rda, which simply moves incoming data to the buffer. Not much code in the int_rda routine. I made it as simple as possible. Manipulation of received data in handled in the main().

The maximum data the PC can send at any one time is 50 bytes.

It seems that my code can only successfuly process 20 bytes.

I think it could be a timing issue.

Has anyone encountered a problem before? How did you overcome such a problem?

[Paladdin]

Not much code in int_rda but... how much? It looks like you are having an overrun problem. A little example of your source code would be very useful to help finding the issue.

[neil]

Here is my ISR, this is what I would call small code and this does work. How big is yours?

#int_RDA
RDA_isr() {
rx_buffer[buff_in++]=RXREG; // direct register access, don't like getc()!!
buff_in &= BUFF_LIMIT; // Wraps around the buffer pointer to make it circular.
rx_free--; // a variable which tells me how much free space is left in the buffer.
}

Neil.

[Mark]

You should be checking for errors. If you get an error and do not clear it, then the isr will stop responding. What is the value of BUFF_LIMIT? If your array is 55 bytes, then you can't do it the way you are trying. You would need a "nice" binary number like 1,2,4,8,16,32,64,128,ect...

[neil]

Mark, that was my code, not arga's! I am using a binary number instead of using the % operator. Keeps things neater. I think I'm using 16 bytes, but haven't looked; This is code I did months ago.

I hadn't thought of checking for errors, I do have the ERRORS directive in the #use though. I must have been lucky, I'm using 19.2Kbaud over a 1M cable between racks.

[Ttelmah]

[neil]

Hmmm, see above!

[Mark]

Didn't check the poster!!! My mistake and RJ's

[arga]

Thanks for all your replies.

By me saying not much code in the rda routine could still be too much.
So, in view of this here's the code :

#int_rda
void serial_isr()
{
if(!flag.m_FifoFull)
{
*pFifoHead = getc();

if(oerr)
{
nTemp = RCREG;
nTemp = RCREG;
nTemp = RCREG;

cren = 0;
cren = 1;

putc(nBoardNo);
printf("Receive buffer overrun");
}

flag.m_Fifo = 1;
m = (m + 1)%MAX_RCV;
pFifoHead = aRcv + m;

if(pFifoTail == pFifoHead)
{
flag.m_FifoFull = 1;
}
}
else
{
nTemp = getc();

if(oerr)
{
nTemp = RCREG;
nTemp = RCREG;
nTemp = RCREG;
cren = 0;
cren = 1;

putc(nBoardNo);
printf("Receive buffer overrun");
}
}
}


MAX_RCV is a constant with a value of 55
pFifoTail and pFifoHead are pointers. The former points to the last memory space filled with data and the latter points to the next available memory that can be written with data.

flag.m_FifoFull is a flag that, when set, indicates that the buffer is full, which happens when pFifoTail catches up with pFifoHead.

When an overrun do occur the rx buffer is cleared. the nTemp is just a temporary buffer used for dumping whatever data is in the rx buffer when an overrun occurs. It has no other purpose.

Perhaps, the printf statement shouldn't be there. Somehow, I must send a message to the PC if an overrun occurs, but it's not sending any message at all. Transmit works for everything else. So, the tx setup is correct.

I have the ERRORS directive in the #use rs232

[arga]

To simplify reading the code I attached previously here are the relevant declarations made outside main()

#byte RCREG = 0x1A // offset of RCREG register

#bit oerr = 0x18.1 // OERR bit of RCSTA register
#bit cren = 0x18.4 // CREN bit of RCSTA register

The flag.m_Fifo, when set, indicates that there is data in the circular buffer. This is reset in main() once all data in the circular buffer have been processed.