home *** CD-ROM | disk | FTP | other *** search
- Path: inforamp.net!ts42-02
- From: crs0794@inforamp.net (Geoffrey Welsh)
- Newsgroups: comp.dcom.modems
- Subject: Re: 8250/16450 uart blues
- Date: 17 Mar 1996 06:19:51 GMT
- Organization: InfoRamp Inc., Toronto, Ontario (416) 363-9100
- Message-ID: <4igau7$gjs@sam.inforamp.net>
- References: <Pine.PTX.3.91.960315100006.17058F-100000@odin.cc.pdx.edu>
- NNTP-Posting-Host: ts42-02.tor.inforamp.net
- X-Newsreader: News Xpress Version 1.0 Beta #4
-
- In article <Pine.PTX.3.91.960315100006.17058F-100000@odin.cc.pdx.edu>,
- "M. Crane" <psu01055@odin.cc.pdx.edu> wrote:
- >I've heard rumors that MSD reports the 16450 as an 8250 UART.
-
- I believe that this is correct; as far as MSD is concerned, it's either an
- 8250 (UART with single-byte buffer) or 16550AF (UART with a multibyte buffer).
- Although the part number reported may be incorrect, MSD has the right idea
- from a functional point of view; the difference between the 8250 and 16450
- UARTs are negligible for 99% of systems out there.
-
- >I've also heard (via the NAVAS faq) that it's possible for my 8250 to
- >exceed 19200 bps using "well written software"
-
- He's correct; the overrun problem stems not from bugs in the UART, but from
- poor or inconsistent reaction time within your system. It's like blaming your
- mailbox for being too small and overflowing if you only remove your mail once
- a month. Buffered UARTs such as 16550As are the functional equivalent of
- replacing your mailbox with a barrel so that monthly mail pickups don't result
- in lost letters due to mailbox overflow.
-
- >Currently I use a 386 laptop with either an 8250 or a 16450. I believe
- >it's an sx/33. In dos, the fastest throughput I've achieved is about
- >1600 Cps, sending and receiving .zip compressed files. I have a linelink
- >14.4 external modem. I also use an old version of Telix for dos as my
- >commware.
- >
- >Given the above configuration, is there anyway I can approach the Holy
- >Grail of 57,700 bps using my existing configuration? Is there a
- >freeware/shareware package out there that will improve my throughput?
-
- You're trying to get something for nothing, and you should not be surprised
- that you won't get it.
-
- The modem transfers data at 14400 bits per second. If there were no error
- correcting protocol, your modem would send the data exactly as received: start
- bit, eight data bits, stop bit. This would result in a maximum throughput of
- 1440 CPS, with real transfer rates slightly lower than that. Thanks to the
- error correcting protocol, start and stop bits are not necessary but there is
- some overhead for error correction frame info, etc., resulting in
- approximately 19% throughput improvement over no error correction, or a
- maximum value of about 1715 CPS, or slightly lower thanks to ZMODEM (or
- whatever) overhead. Higher throughputs can be reached with data compression,
- bit .ZIP files rarely compress well, so you can basically discount data
- compression for this particular test.
-
- You're currently getting about 1600 CPS. That's close to the 1715 CPS
- theoretical maximum (over 93% efficiency, about par for basic free ZMODEM
- implementations), and this says to me that your UART is _not_ your limiting
- factor. In fact, I would suspect that nothing short of buying a faster modem
- would result in faster transfers of .ZIP files on your system... but, with the
- modem's speed limit raised, you may then find that your UART has become your
- limiting factor.
-
- >Thanks in advance for your recommendations (even if they consist of "shut
- >up!")
-
- No need for anything that drastic. <grin>
-
- --
- Geoffrey Welsh, Developer, InSystems Technologies Inc.
- Temporary: crs0794@inforamp.net; At work: insystem@pathcom.com
- At home: geoff@zswamp.uucp or [xenitec.on.ca|m2xenix.psg.com]!zswamp!geoff
- Capitalism is a cold-hearted system which guards the interests of whoever's
- at the top, yet hypocritically claims that it offers everyone a fair shot.
- So is every other system ever put in place by man.
-