[comp.sys.att] Need help with 3b1 modem control registers

ostroff@oswego.UUCP (Boyd Ostroff) (01/29/88)

I am running a specialized BBS for the Performing Arts on a 3B1 system (running
version 3.51 Unix).  I have recently ported my software to the 3B1 from 
a Macintosh, and am very pleased with the system's performance, but using the 
OBM (on-board-modem) for dial-in has created a problem.  

Specifically, some of my users at the University of Rochester are unable to
establish a good 1200 baud connection - typically they get "bracket" 
characters and after a few minutes lose the carrier completely.  Evidently
the source of the problem lies in poor-quality long distance lines, since
I frequently call the system at 1200 baud without any problem.  Part of the
problem may also be related to their phone system which consists of an
analog phone line from the user's terminal connected to a Concord series 2
modem (part of a modem-pool).  The modem is routed through the University's
switch via a 12 khz digital line.  Their telcom people tell me they have
had problems before with outgoing modem calls, but I'd still like to do all
I can to help them out on my end.  300 baud calls get through with no problem.
The odd thing is that they didn't have nearly as much trouble before I 
switched to the 3B1.  My previous system used a (cheap) Avatex 1200 external 
modem.  It seems that the OBM has no tolerance for poor quality connections.

I suppose one solution is to go back to using the external modem, but I'm
wondering if changing the configuration of the OBM might improve things any.
Has anyone played around with any of the constants defined in <sys/modem.h>?
[BTW - what happens if you set "Baud2400_SCM_Wr1" and "Set2400Baud_SCM_Wr8"?
Can the OBM really work at 2400 baud?]  Does anyone have an example of using
the modem control registers - is it done with ioctl() calls?   What registers
would be worth playing with to improve tolerance for poor quality lines? 

I assume that the file /etc/.modeminit normally sets up the OBM. Since there
isn't any source for this, could a call to a short program which resets some 
of the OBM registers to added to /etc/rc?

Any experience, insights, or whatever would be greatly appreciated!

----------------------------------------------------------------------------
Boyd Ostroff * Technical Director * Department of Theatre * SUNY Oswego
Voice: (315) 341-2138  *  BBS "The CallBoard": (315) 947-6414 300/1200 baud
Internet: ostroff%oswego.oswego.edu@devvax.tn.cornell.edu
----------------------------------------------------------------------------

jhc@mtunx.ATT.COM (Jonathan Clark) (02/02/88)

In article <631@oswego.UUCP> ostroff@oswego.UUCP (Boyd Ostroff) writes:
>Specifically, some of my users at the University of Rochester are unable to
>establish a good 1200 baud connection - typically they get "bracket" 
>characters and after a few minutes lose the carrier completely.  Evidently
>the source of the problem lies in poor-quality long distance lines, since
>I frequently call the system at 1200 baud without any problem.  Part of the
>problem may also be related to their phone system which consists of an
>analog phone line from the user's terminal connected to a Concord series 2
>modem (part of a modem-pool).

This sounds very similar to a problem which we found the hard way.
The SCM chip used in the unix pc has a very low tolerance for shaved
stop bits - it insists on having at least one full stop bit. The symptoms we
saw were that a uucp transfer would start up fine, then time out completely
and die. Typically no recovery was possible, and if a transfer did succeed
then the throughput would be atrocious (6 bps was typical). For complex
reasons terminal emulation worked perfectly, it was just uucp transfers
which died (and ?MODEM transfers too).

We eventually fixed the problem by putting another modem pool which was
specially tweaked to provide a full stop bit.

However, the business with the "bracket" characters (typically '{' and '}')
sounds like extreme line noise, possibly coupled with the notorious T-1
sync slip probem (if the users are seeing '}i' pairs). There is nothing I
can think of which will correct this problem.

>Can the OBM really work at 2400 baud?

No. Or rather, not without help. Some other AT&T products use the SCM chip
and work at 2400 Baud, but not the unix pc. Sorry.
-- 
Jonathan Clark		jonathan.clark@mtune.att.com, attmail!jonathan
Any affiliation is given for identification purposes only.

The Englishman never enjoys himself except for some noble purpose.