Newsgroups: comp.sys.next
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!mstar!mstar.morningstar.com!bob
From: bob@MorningStar.Com (Bob Sutterfield)
Subject: Re: Can the Next ports run at 57600 baud RS422?
In-Reply-To: rit@killdeer.Stanford.EDU's message of 13 Jun 91 15:24:26 GMT
Message-ID: <BOB.91Jun14121423@volitans.MorningStar.Com>
Sender: usenet@MorningStar.COM (USENET Administrator)
Reply-To: bob@MorningStar.Com (Bob Sutterfield)
Organization: Morning Star Technologies
References: <1991Jun13.152426.20007@neon.Stanford.EDU> <1991Jun13.182809.23141@zardoz.club.cc.cmu.edu> <SCOTT.91Jun14005609@mcs-server.gac.edu>
Date: Fri, 14 Jun 91 16:14:39 GMT
Lines: 60

In article <1991Jun13.152426.20007@neon.Stanford.EDU> rit@killdeer.Stanford.EDU (Jean-Francois Rit) writes:
   Can the Next ports run at 57600 baud RS422? 

   All I could get from the man pages is that zs can run at any of 16
   speeds and that tty speeds are: [50 thru 38400]

The highest supported async clock speed, with the standard device
driver, is defined in <sys/ttydev.h> as 38400.  I haven't yet tested
to see whether they can achieve that in real throughput.  If one
wished different values than those (higher or lower or in between)
then one would need to supply a custom device driver, complete with
custom ioctls.

The highest sync clock speed we recommend with our X.25 daemon on the
NeXT serial ports is 19200, because that's the highest speed at which
it runs reliably.  We have clocked it higher (e.g. 38400), but it gets
more underruns the higher you push the speed.  X.25 "works" at those
speeds only because of the redundancy of the higher-level protocols,
which cause frames to be resent when needed.

In article <1991Jun13.182809.23141@zardoz.club.cc.cmu.edu> ddj@zardoz.club.cc.cmu.edu (Doug DeJulio) writes:
   I'm pretty sure the hardware can do better.  

The same X.25 daemon code runs reliably at 64K on a SPARCstation-1,
which uses the same Zilog 8530 serial interface chip, so yes the
hardware can do better.  We run X.25 over RS232 (sync) on our VME
boards using the 8530 at speeds over 500K, so yes some parts of the
hardware can do much better.  (You don't want to run RS232 async at
very high speeds, because the waveforms begin to get distorted by the
capacitance.)

Our X.25 daemon's serial interface speed is limited on the NeXT's
native ports because of certain design misfeatures of the NeXT Mach
2.* 8530 interrupt service routines.  This has been reported to NeXT
as a bug.  Also, the SPARC architecture is much better at servicing
interrupts than is the Motorola 68K found in (today's :-) NeXT.

In article <SCOTT.91Jun14005609@mcs-server.gac.edu> scott@mcs-server.gac.edu (Scott Hess) writes:
   The super-serial chips ... are not supposed to run any faster than
   38400, or so I've heard.

38400 is the highest async serial speed found in <sys/ttydev.h> on any
UNIX system I've encountered.

   The problem is mostly that they are simply older technology (albeit
   well-proven).

The 8530 can run fast just fine.  As above, they manage 64K sync just
fine in other UNIX systems, when those systems service their
interrupts correctly.

   I've heard various rumors about various devices that _can_ go
   faster - via the SCSI or DPS ports - but, alas, as shipped Unix
   can't do greater than 38,400.

Our SCSI-attached sync serial interface runs at T1 speeds and better,
but it doesn't use the 8530.  A serial async device driver either for
its ports or for the NeXT-native 8530s could define whatever speeds it
likes, but you probably won't get much over 64K async on the native
NeXT ports.
