On Thu, 29 Aug 1996, Michael L. Hitch wrote:

> > >    Have you got option DEFER_LEVEL6 (sp?) enabled?
> > > 
> > > LEV6_DEFER
> 
>   You do not want this option when using the Multiface III card.  The
> serial ports on the Multiface cards use the level 6 interrupt and if you
> defer the interrupt, you will almost certainly lose lots of characters.
> I think I even saw lots of errors when running at 1200 bps.
> 
>   One thing that could cause fifo overflows at any baud rate would be a
> bad or miswired cable, or a misconfigured modem.  The modem must have
> hardware flow control (CTS & RTS) enabled, and the cable must have RTS
> and CTS wired correctly.
> 

Yes, any solution is predicated on a fully wired cable (8 wires at least)
and the modem must be operating with RTS/CTS handshaking enabled.  The
modem commands to do this vary with each modem brand.

Additionally, modem cable length can be a factor at high baud rates.  The
RS-232 spec calls for a length of 50' at 9600 baud, 25' at 19200, and so
on.  This means that to be completely within spec, a cable for a 115200
baud port speed needs to be 4' 1" long :-)  (Sorry, I don't speak metric)

However, in real life, you can usually get away with about twice this
length if you have a good quality cable.

It could be worth trying the idea of invoking Xdaniver with -depth 1 to
save on chip memory bandwidth, just to see if that influences the serial
port performance for the folks who are just using the built-in serial
port.  I would not be surprised if doing that would also help the
MultiFace III too (just a hunch).

------------------
Bruce Drake
Synectics, Inc.
(919) 872-1275 USA