==-SERIAL KILLER-(revisited)============================================

 Often times, COM3 and COM4 will not open, and you will get a Device Unnavail-
 able error, even if you have a serial port at the address!  This is not
 PowerBASIC's fault, however.  PowerBASIC looks at the "BIOS Data Area" for
 the actual addresses of the ports.  These addresses are 4 consecutive 2-byte
 integers beginning at segment &H40, offset 0 (absolute address &H400).
 Because the BIOS is responsible for putting these addresses here, and often
 times COM3 and COM4, the last two integers, are left at 0, PowerBASIC becomes
 under the impression that these ports do not exist.  You simply need to Poke
 them into being:

         DEF SEG = &H40 : POKE$ 4, CHR$(&HE8, &H03, &HE8, &H02)

 These are the "standard" addresses for COM3 and COM4.  Of course, you may
 stuff any address you like.  If you read bytes 4 and 5, and stick them into
 the offsets 0 and 1, then suddenly COM3 becomes COM1 (This won't work with
 QB, since QB's port addresses seem to be hard coded).  Typically, COM3 and
 COM4 use the same IRQ's as COM1 and COM2, that is IRQ 4 and IRQ 3 respectively,
 but most modems these days, like the little $50 internal 2400 from Boca
 Research, can be configured to use IRQ 2,3,4,5 or 7.  Since PowerBASIC allows
 you to _select_ the IRQ in the OPEN COM statement (like no other language),
 and since PowerBASIC does not limit you to just one com port open at a time,
 you can easily use all four serial ports simultaneously, or, like me, have
 a PowerBASIC BBS running on COM3 and COM4, a mouse on COM1, and Little Big
 Lan running on COM2.  Cool.


 Incidentally, here is a terrific little modem terminal routine:

 ' normal COM2 at 2400 baud
 OPEN "COM2:2400,N,8,1,RS,CS,CD,DS,ME,FE" FOR RANDOM AS #1
 ' alternate COM4 on IRQ 5 at high speed
 'OPEN "COM4:38400,N,8,1,ME,FE,IR5" FOR RANDOM AS #1

 OPEN "CONS:" FOR OUTPUT AS #2
 DO
         A$=INKEY$
         IF A$=CHR$(27) THEN EXIT LOOP  ' abort program on escape
         IF LEN(A$) THEN
                 PRINT #1,A$; ' send keystroke to modem
         ELSE
                 IF NOT EOF(1) THEN IF LOC(1) THEN A$=INPUT$(1,#1)
                 IF A$=CHR$(10) THEN A$=""                  ' filter linefeeds
                 IF A$=CHR$(13) THEN A$=CHR$(13)+CHR$(10)
                 IF A$=CHR$(8) THEN A$=CHR$(8)+" "+CHR$(8)  ' process backspace
                 PRINT #2, A$; ' output to ANSI driver
         END IF
 LOOP
 CLOSE

 I have noticed that adding the "RS,CS,CD,DS" helps to avoid timeout problems
 with some serial ports in direct-connect situations, or with some older
 external modems, but it appears to prevent high-speed error-correcting modems
 from working.  Any commercial package should have both OPEN COM formats
 available as options.  RS and CS without arguments (disable RTS/CTS) affects
 modems which perform buffering, as in the case of modems which adjust to
 the caller's baud rate but maintain a fixed baud rate connection to the
 PC.  ME (thank you Bob!) masks out framing and overrun errors, a very common
 occurance especially where modems are involved.  These errors usually occur
 at the worst possible time and tend to blow apart even the most logically
 designed program.  ME causes PowerBASIC to simply ignore these nit-picky
 little UART snits.  FE is an option which flushes the byte currently in the
 UART holding register when an error occurs (ever see û û suddenly appear
 when line #2 rings, when a jet flies overhead, or when the A/C cuts on? FE
 fixes that!).

 Do not confuse framing and overrun errors with error 69, "buffer overflow".
 ME and FE deal with problems caused by a single byte entering the UART
 chip.  Error 69 deals with where the data goes from there.  Since your
 program cannot always be monitoring the com port and reading in characters
 as fast as they arrive, PowerBASIC buffers incoming characters up to 32K,
 so you can go off and do something, then come back and read in the data
 which arrived while your program was busy with other chores.  When this
 32K buffer is exceeded, there is nothing you can do to avoid getting
 this error.  You must drain the buffer immediately by reading it (or
 closing and opening the port) and even in the split second it takes your
 program to do this, many characters will be lost, like people trying to
 pile into overfilled lifeboats on the titanic.

 Even though PowerBASIC supports multiple port I/O and baud rates up to
 115200, there may be hardware limitations to your equipment that will
 prevent serial communications from operating properly.  The early IBM
 PC, XT and probably most of the XT clones produced throughout the 80's
 used 8250 UART chips for their serial ports.  I believe that this chip
 is also widely used in extremely cheap internal modems.  According to
 the specs on this chip, it cannot run any faster than 57,600 baud, half
 of PowerBASIC's maximum speed.  PowerBASIC will not give you any sort
 of error message (as far as I know) if you attempt to open an 8250
 serial port at 115,200 baud.  This type of port in a fast computer can
 also cause problems, as it would be possible to write to it's control
 registers faster than the chip itself can handle the instructions,
 which would seem likely to blow it apart (figuratively).  Also, even
 with an advanced UART, PC and XT machines may run too slowly to
 process high baud rates in some cases, and characters may be lost.

 Later computers, modems and serial boards also use a sixteen bit version
 of what is essentially the same as the 8250 and 8250A, called the 16450.
 The maximum baud rate for this kind of serial port is also 57,600 and
 attempting to use it at 115,200 will likely cause serious errors as
 well.

 Most serial ports designed today use the 16550, but be careful.  You
 should actually have a 16550A, AF or AFN.  The regular 16550 was actually
 recalled by National Semiconductor.  The 16550 series was designed to
 support some extended features, including a built in 16-byte buffer as
 well as a maximum baud rate of 256 kilobaud!  The original 16550 (the
 one that was recalled) is said to work well as long as the 16 byte
 buffering is not invoked.  Most current PC software does not support this
 extension, and the current version of PowerBASIC does not either, however
 PowerBASIC most definately will in the future, and many popular
 communications add-ons sport this feature.

 I have read that although the maximum baud rate for the 16550x is
 specified at 256kb, the maximum baud rate is fixed to 115kb in PC
 machines because of the hardware design, and higher speeds are not
 physically selectable.  I have entreated Lloyd to attempt to rig
 a set of 256 kilobaud serial ports with a magnifying glass and a
 soldering iron, and perhaps we'll get some literature from him on
 that (pro or con).

 It is also said that 16550s manufactured by Western Digital do not
 operate correctly at or below 2400 baud.  I have not been able to
 discover any other reference as to the nature of this malfuction.

 PowerBASIC will open up to four serial ports at a time.  If all of
 the ports occupy the same IRQ, they must be on the same physical
 board (except on a PS/2).  Interrupt configurable modems and serial
 cards can usually be found, and are generally less expensive than
 a single 8-port board.  Multitasking environments like Desqview or
 PC-MIX (available on our BBS) will allow you to open as many ports
 as you can install with PowerBASIC, provided you understand a few
 minor limitations.  Although PowerBASIC can interact with up to
 four ports on the same IRQ, you may not open two ports on the same
 IRQ inside different windows (or tasks).  If many ports occupy the
 same IRQ, they can only be utilized by a single program running
 alone or in a Desqview window (I sure wish someone had told me
 that years ago!).  In order to use many ports on the same IRQ, you
 must use a "Fossil Driver".  Such an animal exists on our BBS as
 X00124.ZIP, and a discussion of it is better left out of this text
 or we'll never get to the end.

 An alternative to normal serial ports is "intelligent coprocessing",
 which is simpler than it sounds.  An intelligent coprocessing serial
 board is one which contains its own memory and CPU.  The 80186 seems
 to have found a home here.  To interact with this board, you need no
 communications software, no interrupt handler, and IRQs don't even
 enter into it.  All you have to do is (via CALL INTERRUPT or with
 INP and OUT) ask the card "Is there any data on channel 1?".  If
 the card says "yes", then you say "Give it to me", otherwise, you
 go on to channel 2, then 3, and keep going like that.  Parts of
 Cakes.

 For serial co-processing, an 8-channel intelligent serial board is
 available from GTEK, inc in Bay St. Louis, MS (601-467-8048).  A complete
 set of PowerBASIC drivers (written by Yours Truly) is available from
 GTEK or from the PowerBASIC tech support BBS.  Boca Research, in Boca
 Raton, FL, offers an 8 port 16550 board for around $129.  I have not been
 able to verify this.  Decision Computer of Taipei, (011 886 2 766-5753
 voice, or 011 886 2 766-5702 fax) as a 4 port board for (reportedly)
 around $36 and an 8 port board for roughly $100.  At the time of this
 writing I am awaiting further information from this company.

 Although there are a great many other similar products out there, there
 is one that stands out from all the rest:  Galacticomm.  The standard
 Galacticomm 8-port 16550 UART, interrupt and address configurable
 board is functionally identical to all other non-coprocessing boards
 and like the others, works pretty well with PowerBASIC on all 8 ports,
 up to 4 open at a time in a single program (or all 8 running under
 something like desqview with a fossil driver) but what makes the
 Galacticomm board unique is the company's incredible marketing abilities
 which allow them to get away with selling the board for around $750
 (it probably costs about $15 to make) and absolutely refusing any sort
 of tech support unless you buy their own overrated BBS software.  After
 much prodding, I finally got them to fax me a bunch of pages which
 they tore out of a National Semiconductor spec book, and since I was
 new to working so closely with serial ports back then, it took me a
 little while to realize that there was nothing special (or even very
 useful) about this incredibly expensive common serial port circuit board,
 except that it cost me the equivalent of a whole 'nother PC.  I ended up
 trading it to Lloyd Smith's son Michael for a Joe Weider weight machine.
 Michael sold it to a friend up North who actually had some Galacticomm
 software.

 For PowerBASIC examples of serial communications, particularly modem based,
 see James Davis' DWDOORxx series of door libraries, Dave Navarro's
 soon-to-be-released PC-Board toolkit (PB-KIT?), the files MODEM.BAS
 (author unknown, looks Navarro-esque), TERMINAL.BAS and XMODEM.BAS on
 our BBS demonstrate various ways of writing terminals and file
 transfer protocols, and HBASIC.ZIP is a dBASE compatible BBS program
 with source code, all of which can be found on our BBS, and other
 BBS's which support PowerBASIC.

 Erik Olson
 6/11/93 (Jurassic Park opens TODAY!!!)