Xref: utzoo comp.dcom.modems:4645 sci.electronics:8191
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!apple!sun-barr!decwrl!jumbo!murray
From: murray@jumbo.dec.com (Hal Murray)
Newsgroups: comp.dcom.modems,sci.electronics
Subject: Re: SW56/ISDN for Macintoshes??
Summary: software limit is ~50K bits/sec
	 hardware limit is under 1 megabit/sec
Keywords: 56KB, ISDN
Message-ID: <14120@jumbo.dec.com>
Date: 12 Oct 89 03:49:35 GMT
References: <1989Sep29.114152.4344@ux1.cso.uiuc.edu> <2439@ucsfcca.ucsf.edu>
Organization: DEC Systems Research Center, Palo Alto
Lines: 25

There are two limits that I know about for the Mac serial ports: software
and hardware.

If you use the standard driver, the interrupt overhead limits you to ~50K
bits/second total throughput on all active lines. I think that's in the
hardware book. I have a hack test program that could mostly keep up in
send-only at 56K.  When sending and also receiving what it was sending
(via a paper clip in back) it only got 30K even though the hardware was
still running at 56K.

The trick for going faster than that is to poll rather than take an
interrupt on each character. That's how AppleTalk does it. I can get 400K
bits/sec using code written in LSP. (Disable interrupts to avoid
missing characters.)

I think you could poll over 1 megabit/second, but the SCC won't go that
fast, even with external clocking. New data sheets contain an abscure
footnote that limits the data rate to 1/4 of pClk. PClk is nominally
3.672MHz so you might get 900KHz.

3.672MHz is slightly less than 1/2 of the Mac processor clock (7.8336
MHz), so something fishy is going on.  I've seen slight jitter on a scope
when watching a single bit produced from the internal clock. I assume
somebody is stretching cycles for memory refresh or display or... I don't
know what the actual limit is, but I'll bet it's below 3.672MHz/4.