Linux-Hams archive - July 1998: Re: Baycom problem - comment and question

Re: Baycom problem - comment and question

Al Woodhull (gticrjb@relay.tunkki.fi)
Wed, 1 Jul 1998 11:16:45 -0400 (EDT)


On Wed, 1 Jul 1998, Duvall, Michael J. wrote:

> Someone told me that the UART was used and as a result the Baycom
> required little CPU time, just the normal UART interrupt overhead. I
> guess my original fears were correct.
>
> I think they call this a bit banger UART. Cheap and DIRTY! Unless CTS
Note-> ^
The point is, for sending packets you can't use a U*A*RT. The 'A' stands
for asynchronous, that is, a form of data transmission in which there are
synchronizing start and stop bits framing each byte of data. But a
packet is a relatively large number of bytes sent as a *synchronous*
stream, i.e., with no start and stop bits between the bytes of
data.

To generate packets in hardware you need a USRT or USART chip (the 'S'
meaning it is capable of synchronous operation). This is not standard
equipment on PCs. A TNC has hardware or firmware that converts between
asynchronous bytes from a serial port to synchronous packets that are then
sent to the modem which is also part of the TNC. The classic Baycom
interface is not a TNC, it is only a modem.

Yes, it is very CPU intensive. But when I used it with an XT-class laptop
under MS-DOS with the original Baycom around 1988 or so it worked OK.
Even an 8088 can keep up with a 1200 baud data stream. I haven't used
Baycom recently, but apparently a Baycom driver under Linux on any CPU
competent to run Linux (386 or better) doesn't feel the strain very
much. There is a hardware timer chip to do the hard part, it isn't as
if the CPU has execute timing loops like the cassette tape interfaces
on some of the old 8-bit single board development kits.

73, Al N1AW
+----------------------------------+
| Albert S. Woodhull |
| Hampshire College, Amherst, MA |
| terhi.victor@logonet.com |
| http://minix1.hampshire.edu/asw/ |
+----------------------------------+