serial coms - again

[malc-c]

Uhmmm strange things.. I spoke too soon.

If I follow your example code

Term_RX:
HSERIN[nTest]
    SELECT CASE nTest
    CASE "S"
       TempWD = 0
    HSERIN 1000,RX_Bombed,[DEC3 TempWD]
    normtemp[0] = TempWD
    gosub send    
    Goto Term_Rx
RX_Bombed:
    TimeoutCount = TimeOutCount + 1
 lcdOUT $FE,1,"all we got was", dec TempWD,", ", DEC TimeOutCount 
 
    end select 
 
   Goto Term_Rx

FlushBuffer:
While PIR1.5
HSERIN [TempWd]
WEND
TempWd=0
LCDOUT $FE,1,"Buffer flushed"
RETURN

when the program 1st runs and detects somthing in the buffer it gosubs to Term_RX, but with the goto hi-lighted above the program then never returns to the main program... If I change this to go to main I'm back where I started from...

I guess I should ask Charles to design his application so that it opens a second window to do the updates, and when it opens that window it sends a Q or P or something which makes the PIC jump to the Term_RX section, and then sends another character when the updating window is closed (sort of click APPLY )

[HenrikOlsson]

Hi,
That goto was needed in order to test it at my end (I said I modified it), it was never my intention for you to copy/paste that code back into your application - sorry about that.

With this code:

main:
FOR TempWD = 0 TO 1000
        IF RCIF=1 THEN GOsub Term_RX
        PAUSE 1
NEXT TempWD
pause 200
goto main

You are looping around polling the RCIF flag 1000 times in one second (basically). But at 115200 baud you're sending ~11500 bytes or characters per second (one startbit, 8 databots, one stopbot = 10bits per character).

This means you're sending ~11.5 characters every millisecond. The PICs hardware buffer is only 2 levels deep so you're very likely to overrun the buffer before the HSERIN command is able to grab the data.

At 2400baud you're only sending 240 characters per second so your polling loop will have no problem catching the RCIF flag and jump to the Term_Rx routine before the buffer overruns.

Does that make sense?

/Henrik.

[malc-c]

Hi,

You are looping around polling the RCIF flag 1000 times in one second (basically). But at 115200 baud you're sending ~11500 bytes or characters per second (one startbit, 8 databots, one stopbot = 10bits per character).

This means you're sending ~11.5 characters every millisecond. The PICs hardware buffer is only 2 levels deep so you're very likely to overrun the buffer before the HSERIN command is able to grab the data.

At 2400baud you're only sending 240 characters per second so your polling loop will have no problem catching the RCIF flag and jump to the Term_Rx routine before the buffer overruns.

Does that make sense?

/Henrik.

Yes, somewhat - no wonder why I was having problems missing the data... so is it better to reduce the baud rate for the port or reduce the for next loop steps ?

[HenrikOlsson]

Hi Malcolm,
If the PC starts to send S789 right at the beginning of the Pause 1 statment all four bytes will have been sent before the PIC checks the RCIFlag and your data will be messed up. So, as you say, either check the flag more often or reduce the baudrate so that you are sure that a maximum of two characters can arrive while you are not watching - so to speak.

Actually, there can be two bytes in the buffer and a third on its way in without messing it up. But as soon as the last bit of that third byte is shifted in you're in trouble if there's no room in the two byte buffer.

Let me see if I've got the overall scheme correct.
** The PIC does its thing, running the PID and everything. Each time thru its main loop it checks the RCIFlag.
** The PC, when it needs attention sends a single character "Q" to the PIC and waits.
** The PIC sees the RCIFlag and gets the character (there's no problem here since a single character will sit in the buffer untill you get around to read it)
** The PIC compares the received character to "Q" and sends something back to say "OK, hit me" - it then goes to the loop we've been discussing.

That should work..... At 115200 baud two characters takes roughly 175us so you need to check the RCIF flag at least that often. I'd probably go with something like:

WaitHere:
For I = 0 to 10000
  If RCIF = 1 then GOSUB Term_Rx
  PauseUs 100
Next

Keep it up!
/Henrik.

[malc-c]

Hi Malcolm,


Let me see if I've got the overall scheme correct.
** The PIC does its thing, running the PID and everything. Each time thru its main loop it checks the RCIFlag.
** The PC, when it needs attention sends a single character "Q" to the PIC and waits.
** The PIC sees the RCIFlag and gets the character (there's no problem here since a single character will sit in the buffer untill you get around to read it)
** The PIC compares the received character to "Q" and sends something back to say "OK, hit me" - it then goes to the loop we've been discussing.


/Henrik.

Pretty much it. As far as I understand Charles has programmed his code to send Q when the program is first launched, this was so that we could use Q as a trigger to jump to a subroutine and have the PIC send the data string.... then the user over types the value in the target temperature box and then when the Update button is clicked it sends Snnn to the PIC, followed by Q to refresh the PC application.

I've tried migrating the test code over to the main code that you were involved in and have discovered one issue. On launching the application the LCD stops running, but I can set the temperature set point via the application on the PC, which upon pressing the Update button the value on the LCD changes to match. However whilst the application is still open the LCD is still not updated. If I then close the application it still locks up and I have to restart the PIC.

At the moment I've reduced the

FOR TempWD = 0 TO 1000

to

FOR TempWD = 0 TO 10

so at 115200 baud rate that should equate to 115 bytes as second (? yes)

If RCIF=1 it gosubs to the TERM_RX

Term_RX:
HSERIN [nTest]
    SELECT CASE nTest
    CASE "S"
       TempWD = 0
    HSERIN 1000,RX_Bombed,[DEC3 TempWD]
    normtemp[0] = TempWD   
    SetPoints(0)=normtemp[0]
    goto Term_TX
  
return

RX_Bombed:
    TimeoutCount = TimeOutCount + 1 
    end select 
    Goto Term_Rx

Term_TX is simply

Term_TX:

Hserout [DEC3 Temperatures(0)]
Hserout [DEC3 Temperatures(1)]
Hserout [DEC3 Temperatures(2)]
Hserout [DEC3 Temperatures(3)]

HSEROUT [dec3 normtemp[0]]
HSEROUT [dec3 normtemp[1]]
....
...
...... etc etc 

gosub flushBuffer 
goto main

If I close the down the application and even disconnect the cable from the EasyPIC board the PIC is still frozen....

Any idea ?

[HenrikOlsson]

Hi,
Changing the FOR-NEXT loop will only change how many times it loops around checking the RCIFlag and therefor the total time it spends waitng for it - not how often it checks it or how many bytes per second that is sent or received.

For 115200 baud you must check the RCIFlag at least once every ~170us so, like I showed you earlier:

For TempWd = 0 to 10000
  If RCIF then GOSUB Term_Rx
  PauseUs 100     '<----Note PauseUs, as in Pause micro seconds...
Next
Goto Main

This will check the RCIFlag every 100us and it will do it 10000 times, so all in all it will wait one second, if no data is received it'll jump to Main.

If you by, the LCD stops updating, mean that the data written to the LCD by the main program in the PIC it's because once the PIC sees the initial RCIF caused by the the "Q" beeing received the main code stops and jumps to your receive routines. It then stays there untill you return to the main program so the LCD can't update....

RX_Bombed:
    TimeoutCount = TimeOutCount + 1 
    end select 
    Goto Term_Rx

Here, you probably don't want to jump back to Term_Tx but instead jump back to either your main program or RETURN depending on how you got to the routine in the first place. That should prevent the PIC from locking up. And you can remove the TeimOutCount thing, that was for my tests....

You need to implement some form of handshaking when the user presses Update in the application, simmilar to when the app is first started. The PC can't just dump data into the PIC since your application may be off doing something else and miss it. (Since you're not using interrupts).

Take a pen a piece of paper, think it thru.

/Henrik.

[malc-c]

Ok I see the logic behind the pause statement.

I've managed to resolve the stalling...

 CASE "Q"
        Goto Term_TX

Just placed that above the CASE "S" and now it works as it should .... strange how the logical things get overlooked !