23 Jul
2006
23 Jul
'06
7:39 p.m.
Hi,
Does anyone know what the available bandwidth of "standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
Thx,
--don
24 Jul
24 Jul
1:41 a.m.
Don Y wrote:
Hi,
Does anyone know what the available bandwidth of "standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
I think fairly slow. The printer port strobe has to be done
in software.
3:15 a.m.
From "Parallel Port Complete" (1996) by Jan
Axelson:
If it's an ISA-bus adapter, the bus speed is essentially the limiting
factor, no matter what the type of port is. At an 8.33 MHz ISA bus speed,
that translates to about 1.3MB/sec. SPP, because of the requirement of 4
writes to do a handshake, can go at best 300KB/sec. ECP and EPP can move
at 1.3MB/sec because the handshake's done in hardware. If ISA shortened
cycle timings (0WS) are used, one can hit about 2.7MB/sec. If DMA is used,
this slows to about 1 MB/sec.
Somewhere, I've got a copy of the IEEE spec, but I don't know if it calls
out an upper lmit on transfer rates.
Cheers,
Chuck
12:59 p.m.
Chuck Guzis wrote:
Yes, this agrees with my analysis: *if* I/O's in ALL pc's are
throttled to that 8MHz bus rate (i.e. even on 3GHz machines!).
And, it also doesn't address how parallel ports work on *other*
machines without the ISA legacy "problem".
From
"Parallel Port Complete" (1996) by Jan Axelson:
If it's an ISA-bus adapter, the bus speed is essentially the limiting
factor, no matter what the type of port is. At an 8.33 MHz ISA bus speed,
that translates to about 1.3MB/sec. SPP, because of the requirement of 4
writes to do a handshake, can go at best 300KB/sec. ECP and EPP can move
at 1.3MB/sec because the handshake's done in hardware. If ISA shortened
cycle timings (0WS) are used, one can hit about 2.7MB/sec. If DMA is used,
this slows to about 1 MB/sec.
1:42 p.m.
On 7/24/2006 at 9:59 AM Don Y wrote:
Yes, this agrees with my analysis: *if* I/O's in
ALL pc's are
throttled to that 8MHz bus rate (i.e. even on 3GHz machines!).
And, it also doesn't address how parallel ports work on *other*
machines without the ISA legacy "problem".
I did a little research on PCI parallel ports. The highest data rate
claimed on the ones I've seen is 5.5-6 MBps in EPP mode.
There's a slide show on 1284 that includes speed numbers on the Warp Nine
web site:
http://www.fapo.com/files/IC99.pdf
Cheers,
Chuck
25 Jul
25 Jul
12:24 a.m.
Chuck Guzis wrote:
On 7/24/2006 at 9:59 AM Don Y wrote:
You are correct; in fact, Disk2FDI uses PCI LPT cards and a special
cable to dump diskettes completely (GAP bytes, everything) by using the
host machine+PCI LPT to emulate a controller.
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Help our electronic games project: http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
A child borne of the home computer wars: http://trixter.wordpress.com/
Yes, this agrees with my analysis: *if*
I/O's in ALL pc's are
throttled to that 8MHz bus rate (i.e. even on 3GHz machines!).
And, it also doesn't address how parallel ports work on *other*
machines without the ISA legacy "problem".
I did a little research on PCI parallel ports. The highest data rate
claimed on the ones I've seen is 5.5-6 MBps in EPP mode. 
24 Jul
24 Jul
2:27 a.m.
From: Don Y <dgy at dakotacom.net>
Reply-To: "General Discussion: On-Topic and Off-Topic
Posts"<cctalk at classiccmp.org>
To: Classic Computers <cctalk at classiccmp.org>
Subject: Parallel port bandwidth
Date: Sun, 23 Jul 2006 16:39:33 -0700
Hi,
Does anyone know what the available bandwidth of "standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
Thx,
--don
Hi Don,
I seem to remember the following.
in SPP (Standard Parallel Port) mode the max transfer rate is 150 kbytes/sec
mostly because
the handshake is done in software. It is a "forward" only protocol (data
from computer to a
peripheral).
in ECP/EPP mode the transfer rate can go up to 2Mbyte/sec, and that depends
among others
on the cable quality you use (IEEE-1284 IIRC), and both are bi-directional.
Tha handshake is
done in *hardware*.
The last mode is called nibble mode, and is SPP, but with a data channel
from the peripheral
back to the computer. In the
beginning the parallel port had uni-directional
databus drivers,
so a return channel was not possible over the data lines. But SPP has a few
(5) wires that
are *input* on the computer side (like ERROR*, PAPER_EMPTY*, BUSY). 4 of
these wires
were used to transfer 4-bit (nibble) data back to the computer. In nibble
mode, the forward
transfer rate is identical to SPP, the reverse comm channel can do 50
kbytes/sec.
- Henk, PA8PDP.
_________________________________________________________________
Iets leuks meegemaakt? Zet het op je Space! www.spacemomentje.nl/start.aspx
3 a.m.
Henk Gooijen wrote:
Yes, IIRC, you write data, write to assert strobe, and write to release
strobe (in my case, the interface is always ready so no need to poll
it's status).
(ignore affect of cabling as I sit *on* the output connector)
But, I figure that should be about 500K/s (in a tight loop) *if*
the bus is deliberately slowed to emulate 8MHz legacy machines.
(if allowed to run "flat out" at FSB speeds, then you can
build a little FM Xmitter! :> )
I am concerned as to how this works on other machines (e.g.,
SPARCs) with parallel interfaces. (I don't expect them to
suffer from the PC's braindamage).
[other modes snipped -- I'm just designing for "dumb" ports]
Thanks!
--don
From: Don Y
<dgy at dakotacom.net>
Does anyone know what the available bandwidth of "standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
I seem to remember the following.
in SPP (Standard Parallel Port) mode the max transfer rate is 150
kbytes/sec mostly because
the handshake is done in software. It is a "forward" only protocol (data
from computer to a
peripheral). 
5:58 a.m.
Does anyone know what the available bandwidth of
"standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
At one time I wrote a "PC logic analyzer", which monitors 13 input
lines from the parallel port (two reads per sample) in a tight assembly
loop to watch for a trigger condition, and then reads and buffers a
whole lot of samples at a user defined speed in an equally tight loop.
On most machines of 486 or better, the loop topped out at somewhere
around a 500khz sample rate. This is entirely bus speed limited - some
P4s do worse than some 486s.
Thats a raw two-read tight loop with no handshaking - once you put
some strobes and acknowleges in there, the actual throughput you
will realize will rapidly drop.
Dave
--
dave06a (at) Dave Dunfield
dunfield (dot) Firmware development services & tools: www.dunfield.com
com Collector of vintage computing equipment:
http://www.classiccmp.org/dunfield/index.html
1:01 p.m.
Dave Dunfield wrote:
The P4 comment implies that *all* PC's have this upper limit
(enforced in hardware? or, emulated in software??)
Does anyone
know what the available bandwidth of "standard"
parallel ports might be? E.g., on PC's I believe bus speed
is emulated at 8MHz (?) for I/O instructions (legacy). But,
what about other machines with parallel (printer) ports?
At one time I wrote a "PC logic analyzer", which monitors 13 input
lines from the parallel port (two reads per sample) in a tight assembly
loop to watch for a trigger condition, and then reads and buffers a
whole lot of samples at a user defined speed in an equally tight loop.
On most machines of 486 or better, the loop topped out at somewhere
around a 500khz sample rate. This is entirely bus speed limited - some
P4s do worse than some 486s. Thats a raw two-read tight loop with no handshaking -
once you put
some strobes and acknowleges in there, the actual throughput you
will realize will rapidly drop.
2:19 p.m.
On most
machines of 486 or better, the loop topped out at somewhere
around a 500khz sample rate. This is entirely bus speed limited - some
P4s do worse than some 486s.
The P4 comment implies that *all* PC's have this upper limit
(enforced in hardware? or, emulated in software??)
6577
days inactive
6579
days old
10 comments
6 participants
participants (6)
-
bfranchuk＠jetnet.ab.ca -
cclist＠sydex.com -
dave06a＠dunfield.com -
dgy＠DakotaCom.Net -
henk.gooijen＠hotmail.com -
trixter＠oldskool.org