Serial port problems (including overhead, speed)

[Robert Keith]

Thanks heaps!

Once again. Thanks for the really helpful replies. There is certainly something there to solve this problem and get me entirely bamboozled.

For Cliff: I have been working with various versions of this device for years many different platforms including Palms, Wilke Tiger devices and Bluetooth links. It runs up to 57,600 baud happily but the manufacturer is making it much faster. This timing stuff is becoming more important. I am used to reading large lumps of data in one go, such as a whole logger full, but I am now doing more with smart sensors that have to be prodded just so to get the data out.

Cheers,

Rob

[Dave Biggs]

If the timeout accuracy isn't too important you could use something like this to ensure the receive response is 'snappier'..

#Include "win32API.inc"
 
    IF COMM(#nComm, RXQUE) THEN      'flush the buffer
        COMM RECV #nComm, COMM(#nComm, RXQUE), rStr
    END IF
 
    COMM SEND #nComm, SndString
    t1??? = timeGetTime
    ' The timeGetTime function retrieves the system time, in milliseconds. 
    ' The system time is the time elapsed since Windows started. (NB. Count rolls every 49.71 days)
 
    DO UNTIL i& >= numbytes OR t2??? - t1??? >= 100  'time out after 100 ms (or so)
      k& = k& + 1
      t2??? = timeGetTime
      i& = COMM(#nComm, RXQUE)
    LOOP
 
    COMM RECV #nComm, i&, RcvString

For more info on the limitations of SLEEP and some useful test code, see David Roberts' post here..

http://www.powerbasic.com/support/pbforums/showthread.php?p=316070

[Cliff Nichols]

Hi Rob,
Welcome to the world of the serial port

Some minor things are missing for a better explanation, but having worked with multiple devices the last 10 years with serial ports, I think I can bring some of it into perspective.

1.) MSCOMM in its day was great (For general purposes), but lacking when needing precise timing. (I could say the same about MarshallSoft, but they got the job done better than M$ did)

2.) What is the device you are trying to communicate with? (This could have its own "quirks"

3.) The general format (in your configuration) would be
Baud: 9600 Bits/Sec - Average not dead on timing
Bits per character (ANSI): 8
Characters: 8 bits/byte + 1 start Bit + 1 Stop Bit
Minimum transmission per character: 8.3333333333333333333333333333333e-4 ms (And is really longer due to start/stop bits and whatever else the computer may be doing at the time.)

All the above can account for the "General Rule of Thumb" that timeouts would be 100ms (Anything less than this is bound to NOT produce the correct results unless extreme care is taken for what is happening outside of your code.)

I will see if I can locate some of my code to demo how even 1 character may not be retrieved within an expected 1ms (1Ms is > 0.8 ms so you would think you would have a response, but the real response could be 10ms, 30 ms or even 80ms)

[Paul Dixon]

Rob,
on an average systen, SLEEP 5 will actually sleep for 15ms because 15ms is the default timeslice. 2 of those would make the 30ms delay you mention so maybe an alternative to SLEEP might be better?
You could look at timeBeginPeriod to see if you can set the timeslice to something smaller:

http://msdn.microsoft.com/en-us/library/ms713413(VS.85).aspx

Paul.

[Robert Keith]

Hi All,

I am writing an creating an extensive set of tools, to replace and extend upon those I did using VB6 years ago, and work with a specific group of hardware with RS232 interfaces.

The required port parameters for the hardware are: 8 data bits, 1 stop bit, no parity, no flow control. The RTS and DTR lines (or at least one of them) need to be held high to power the device. My code to open the port is:

FUNCTION StartComms (PortNo AS LONG) AS LONG
LOCAL cPort AS STRING
    EndComms  'flushes the buffer and calls COMM CLOSE #nComm
 
    SLEEP 30
    cPort = "COM" & LTRIM$(STR$(PortNo))
    COMM OPEN cPort AS #nComm
    IF ERRCLEAR THEN EXIT FUNCTION       ' Exit if port cannot be opened
    COMM SET #nComm, BAUD     = 9600     ' 9600 baud
    COMM SET #nComm, BYTE     = 8        ' 8 bits
    COMM SET #nComm, PARITY   = %False   ' No parity
    COMM SET #nComm, STOP     = 0        ' 1 stop bit
    COMM SET #nComm, TXBUFFER = 1024     ' 4k transmit buffer
    COMM SET #nComm, RXBUFFER = 1024     ' 4k receive buffer
    COMM SET #nComm, XINPFLOW = 0        ' Disable Xon Xoff flow control
    COMM SET #nComm, XOUTFLOW = 0        ' Disable Xon Xoff flow control
    COMM SET #nComm, RTSFLOW  = 1        ' Raise RTS line
    COMM SET #nComm, CTSFLOW  = 0        ' Disable CTS flow control
    COMM SET #nComm, DTRFLOW  = 0        ' Disable DTR flow control
    COMM SET #nComm, DSRFLOW  = 0        ' Disable DSR flow control
    COMM SET #nComm, DTRLINE  = %True    ' Raise DTR line
    SLEEP 50 'allow the device to power up
    GetLinkType 'query the device type and version
    IF LCASE$(LEFT$(LinkType,4)) = "link" THEN FUNCTION = %TRUE
END FUNCTION

Most of the communication with the device consists of strings between 1 to 60 characters long. I know what length the returned string will be so here is the code I use to communicate with the device:

 
    IF COMM(#nComm, RXQUE) THEN      'flush the buffer
        COMM RECV #nComm, COMM(#nComm, RXQUE), rStr
    END IF
 
    COMM SEND #nComm, SndString
 
     DO UNTIL i& >= numbytes OR k& = 20  'time out after 100 ms
         k& = k& + 1
         SLEEP 5
         i& = COMM(#nComm, RXQUE)
      LOOP
 
      COMM RECV #nComm, i&, RcvString

The problem with this is that I am getting a 30 millisecond delay or overhead with every send/receive transaction. k& only ever reaches 1 or 2. Some operations require over 10 transactions so the user will be left wondering what is happening. The device itself responds within a couple of ms so I suspect the delay is happening during either the sending or querying the size of the buffer.

My VB code, with either the now useless MSCOMM control or the excellent Marshallsoft serial library, does these transactions almost instantaneously. I would use the Marshallsoft DLL except that I am hoping to
distribute only one DLL for this device.

Can anyone see where I am going wrong?

Thanks and regards,

Rob