Amiga Serial cant do 28.8?
Michael Shuldman
> hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
> to connect anywhere at anything better than 21.6 (and that was only once). i
> have a 3000 with 3.1 roms and have the serial port settings at the max of
If you by connect mean what you should mean, connecting modem<->modem
(and not Amiga<->modem) then this has nothing to do with your Amiga
but is a bad seting/modem of yours, I would suggest asking in
comp.dcom.modems if you dont figure it out (mind that, GVC is not supposed
to be a big step up from Supra, qualitvice). Could also
be a case of bad lines ofcource, which is difficult to find out unless
you can try an USR or ZyXEL on the line, if those dont work either
you probably have a bad line, complain to teleco.
And there is no problem (:~) setting the serial port to 38.8Kbps on
an A3000 (espically not since you have the 040 acclerator, unless
that somehow makes it worse), download artser from Aminet for
a serial device I would suggest (no, I dont have a location).
Ofcource, dont atempt much multitasking with that speed, the Amiga
serial ports are a sad case.
> Angel Freire
--
_ //
\X/ -- Michael Shuldman <mich...@ifi.uio.no>
Jeremy Rusnak
: hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been
: have a 3000 with 3.1 roms and have the serial port settings at the max of
: 31,250. i'm using terminus and have tried it at every setting from 31,250
: 115,200 and have been unable to even attain 24k. i have even d/l
: and tried running that serial device but have had the same results
: it was more unstable on my system). oh, i have a gvp 040/40 in my system.
: the standard amiga serial port capable of reaching 28.8? i know of a local
: that has a couple of 28.8 lines running on an a3000 but those nodes are
: connected to a gvp serial card. if anyone could shed some light on the
: it'd be a most welcome x-max gift. thanx in advance.
The new 28.8 modems require good line conditions to operate correctly.
Most likely, you have older lines in your home or somewhere along the line
that is causing the problem. I know quite a few users with 28.8's that
don't get 28.8 connects very often just because the telephone system is,
how shall we say, not so good???
Try running new telephone wire inside your home and see if that works, or
better yet try it from another location. If that fails, call the phone
company and complain, and they will usually put noise filters on your line
after you complain enough...
juan perez
to connect anywhere at anything better than 21.6 (and that was only once). i
115,200 and have been unable to even attain 24k. i have even d/l baudbandit
and tried running that serial device but have had the same results (actually
it was more unstable on my system). oh, i have a gvp 040/40 in my system. is
the standard amiga serial port capable of reaching 28.8? i know of a local bbs
that has a couple of 28.8 lines running on an a3000 but those nodes are
it'd be a most welcome x-max gift. thanx in advance.
Angel Freire
Ronald Santos
>Angel Freire
That's because standard RS-232C interface has a maximum bit rate of
20,000 bps. Reaching speeds above this rate is by chance and I doubt it
will happen again.
========================================================================
:
Ron Santos, : Using No Way as Way
Computer Science : Having No Limitation As Limitation
University of Manitoba : - Bruce Lee -
-CANADA- :
:
E-Mail: umsa...@cc.umanitoba.ca :
:
:
========================================================================
Jeffery Carter May
: hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
Yes, the port can handle >28.8. A friend of mine with a USR Courier
regularly does it. The problem you are seeing is probably related to how
well your modem tolerates line conditions.
My friend's Courier had problems with 28.8/21.6 split connects (It would
connect OUT at 28.8, but IN at 21.6) until he put in the new ROM. Now he gets
solid 28.8/28.8 connects every time...
--
AA AA i Jeffery C. May ONLY
A A A A President, Amiga Atlanta, Inc. Users' Group AMIGA
AAAA AAAA i Senior EE, Georgia Institute of Technology MAKES IT
A A A A i gt1...@prism.gatech.edu POSSIBLE
Beth Sawilejskij
: hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
: Angel Freire
I don't think its a problem of your serial speed here. Its the line
quality that is preventing you getting 28800 connects. The serial port speed
only comes into play when you actually start transferring data.
You will not have any trouble getting good speeds when you get a
28800 connect with your setup.
--
B. Sawilejskij Internet: bsaw...@scu.edu.au
Southern Cross University
Lismore NSW Australia
Greg Block
In article <3df86o$m...@newshost.fiu.edu> jper...@solix.fiu.edu (juan perez) writes:
> hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
> to connect anywhere at anything better than 21.6 (and that was only once). i
> have a 3000 with 3.1 roms and have the serial port settings at the max of
> 31,250. i'm using terminus and have tried it at every setting from 31,250 to
First: Don't use 31,250. That's a MIDI speed, and not all modems like it.
Next: An A3000 with 3.1 -SHOULD- be able to handle a speed of 38.4k with
few problems; HOWEVER. If you cause a lot of interrupts, run system hacks
like FBL without setting it to SLOW (although I advise that people not use
FBL, etc, because they're flawed under 3.1, and cause more problems than
they're worth), or load your system heavily, you WILL DROP CHARACTERS.
The internal port causes an interrupt that _MUST BE SERVICED_ for every
character that comes in. It has no buffer on its serial port. If an
interrupt is missed, a character is missed. The higher the speed you need,
the more you need a serial card.
> 115,200 and have been unable to even attain 24k. i have even d/l baudbandit
> and tried running that serial device but have had the same results (actually
> it was more unstable on my system). oh, i have a gvp 040/40 in my system. is
Well, without instantly blaming a GVP product for causing problems with
interrupt servicing, it's possible that something is preventing you from
responding to the interrupts.
The most reliable answer is to use a serial port board, which you should
really do anyways. A 14.4k modem should really be connecting to the
computer at 57.6; a 28.8k, should be connecting as fast as your system can
handle, in order to get the compressed throughput. So, I'd advise getting
something like the GVP card, or the MFC-III (which I use, and I know many
are happy with). I haven't had a single problem with the MFC, and it seems
to support all of the serial modes perfectly.
> the standard amiga serial port capable of reaching 28.8? i know of a local bbs
> that has a couple of 28.8 lines running on an a3000 but those nodes are
28.8, yes. But for a 28.8k modem, a connect at 115k is what you really
want. 76.8k is minimum, I'd say. Anything less, and you're wasting
compression space.
Greg
--
(: (: (: (: (: (: (: (: (: (: (: (: :) :) :) :) :) :) :) :) :) :) :) :)
(: Moral #3: Gregory R. Block :)
(: People who live in glass houses Oubliette Software :)
(: should get dressed with the lights *off*. Win...@Oubliette.com :)
(: -Yakko Warner :)
(: (: (: Have you overdosed on smileys today? Why NOT!?! :) :) :)
Joel Corn
a 28800 connect. My term proggy is set to 57600 and no probs here.
More than likely your GVP card.
Joel
--
**********************************************************************
* Joel A. Corn - Owner/Head Programmer - DarkSoft Computers *
* email:dark...@golden.ncw.net - phone/fax:509-886-0581 *
**********************************************************************
Randy Vice
Amigas. What are the maximum serial rates for the A3000, A1200, and the
A4000? Are they all hobbled to 19.2 max serial speed?
: damo...@nostromo.gate.net : Bruce Morrow,a man before and after his time:
:"The Constitution shall never be construed to prevent the people of the :
:United States who are peaceable citizens from keeping their own arms." :
: - Samual Adams : Morrow Project MARS: Postholocaust Rock and Rollers. :
Todd Johnson
>> to connect anywhere at anything better than 21.6 (and that was only once). i
used to have my Amiga 1200 hooked up to our schools VAX on a direct connect
line at 38400. At that time it was a stock setup, 020 2 megs chip 0 fast.
I never had any problems and usually got 3200-3300 chars/sec when
downloading something.
-----------------------------------------------------------
Todd Johnson Amiga 1200/50mhz 030 6megs
Computer Science The LBM must DIE!
-------------------------------------------------ATO-------
Demetri
: Since we are back on the Serial stuff again, I have a question about the new
: Amigas. What are the maximum serial rates for the A3000, A1200, and the
: A4000? Are they all hobbled to 19.2 max serial speed?
have a 14.4k modem...) I also have a A2000/'040. Maybe you should get a
faster processor!?
: : damo...@nostromo.gate.net : Bruce Morrow,a man before and after his time:
Daniel O'Connor
: hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
*NO* errors, even doing stupid things like jpeg'ing pictures in IFX2
at the same time... I use the v34ser.device instead of the
crappy(IMHO) serial.device....
The reason you do not connect above 21.6k is because of crap telephone
lines(or a crap modem ;), so complain to your telephone company.. :)
Seeya
Darius
~~~~~~
Finger me for my PGP public key!
We secretly replaced the dilithium with Folgers Crystals
--
+------------------------------+-------------------------------------+
|Seeya d...@byron.apana.org.au | AIDS: Natures answer to mankind! :) |
+------------------------------+-------------------------------------+
Jeffery Carter May
: Why would you say that? I have my port at 57.6k and no loss in chars (I only
: have a 14.4k modem...) I also have a A2000/'040. Maybe you should get a
: faster processor!?
In an ongoing discussion through E-mail, it seems that the problem isn't as
much processor speed, but rather you need to have memory that is DMA-able from
the HD without hitting the Z2 bus.
Before I bought my Zeus 040, I was able to produce sustained 57,600 bps
transfers with my unaccel'd 2000 to a GVP HC8+ with 8 megabytes of RAM. NOTE:
I had to use a 4 color screen to do it. With the accelerator, I can get a way
with an 8 color screen at 57,600.
Randy Vice
D> Randy Vice (damo...@nostromo.gate.net) wrote:
D> : Since we are back on the Serial stuff again, I have a question about the
D> new
D> : Amigas. What are the maximum serial rates for the A3000, A1200, and the
D> : A4000? Are they all hobbled to 19.2 max serial speed?
D> Why would you say that? I have my port at 57.6k and no loss in chars (I
D> only
D> have a 14.4k modem...) I also have a A2000/'040. Maybe you should get
D> a
D> faster processor!?
That wouldn't matter since it's the CIA chips that are the limiting factor of
19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
this limit as well. I just wanted to know if I was going to have to buy a
Multiface III card when I upgrade my current pathetic system to a more modern,
and far more powerful Amiga. I would dearly love to see the full capabilities
of my Supra 28.8 :-)
Henric Jungheim
>That wouldn't matter since it's the CIA chips that are the limiting factor of
>19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
Hu? They are quite happy sending stuff at 115k... if you try to recieve
that fast, then the interrupts can't keep up. Depending on what sort of
hard drive controller you have (and other stuff like that), your mileage
may vary.
I run my a4000's serial port at 38.4k. I ran my a500's serial port at
38.4k (sometimes). The CIAs are the same...
>this limit as well. I just wanted to know if I was going to have to buy a
>Multiface III card when I upgrade my current pathetic system to a more modern,
>and far more powerful Amiga. I would dearly love to see the full capabilities
>of my Supra 28.8 :-)
It should be fine at 38.4k. If you set the modem correctly, it will tell
you both the modem<->modem and the computer<->modem rates. Most of the
time, you will probably find your 28.8 connecting at 21.6 or so. That's
totally independent of the computer<->modem rate (try it... set your term
prog to 110 bps and then connect someplace; you may need to use 300 bps
if the modem doesn't go that low. :)
A buffered serial port will definitely improve things (it appears to
reduce system load as well).
--
Henric Jungheim -- uhe...@mcl.mcl.ucsb.edu -- hen...@engineering.ucsb.edu
Maxwell Daymon
: Since we are back on the Serial stuff again, I have a question about the new
: Amigas. What are the maximum serial rates for the A3000, A1200, and the
: A4000? Are they all hobbled to 19.2 max serial speed?
The A3000 serial port can do 230.4 Kbps.
Amiga serial port aren't limited in the way that clone ports are limited.
What you can do WHILE you have the serial port running at xxxKbps is the
question.
On an A2000 with a 33MHz 68040, I was able to do 115.2Kbps null-modem to
an A3000 with no data loss and VERY fast transfers. I couldn't do any
multitasking and ONLY the term program in non-overscan, non-laced
resolution. But I could do it. It all depends on what, when, and how.
There is no "19.2" limitation. I'm running at 57,600Kbps right now on a
768x482 8 color screen with programs running in the background on my A3000.
------------------
Maxwell Daymon One man's "magic" is another man's engineering.
mda...@rmii.com "Supernatural" is a null word.
------------------ - Robert Heinlein (AUISG Page 9)
Maxwell Daymon
: That wouldn't matter since it's the CIA chips that are the limiting factor of
: 19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
That doesn't exist. I have run at 57,600 FULL SPEED with two Amiga 500's.
I have run at 115,200K on an Amiga 2000. This limit is purely in your
imagination, and there are MANY of us running at 57,600 on 500's and
2000's RIGHT NOW - this very _microsecond_ as I write this message.
: this limit as well. I just wanted to know if I was going to have to buy a
: Multiface III card when I upgrade my current pathetic system to a more modern,
: and far more powerful Amiga. I would dearly love to see the full capabilities
: of my Supra 28.8 :-)
Give your "pathetic" system to someone who a) appriciates it and b) knows
more about it and is able to admit when (s|%)he is wrong.
My A500's and A2000 all handle 57,600 just fine, and the A2000 handles
115,200 fine. I'm not using newer or magical CIA's. Furthermore, the SAME
CIA's that are in the A500/2000 are in the A3000!
Bernhard Rieckhoff
In article <3drmsr$r...@potogold.rmii.com>, mda...@rmii.com (Maxwell Daymon) wr
SNIP
> That doesn't exist. I have run at 57,600 FULL SPEED with two Amiga 500's.
> I have run at 115,200K on an Amiga 2000. This limit is purely in your
> imagination, and there are MANY of us running at 57,600 on 500's and
> 2000's RIGHT NOW - this very _microsecond_ as I write this message.
SNIP
> My A500's and A2000 all handle 57,600 just fine, and the A2000 handles
> 115,200 fine. I'm not using newer or magical CIA's. Furthermore, the SAME
> CIA's that are in the A500/2000 are in the A3000!
SNIP
>
Hmmm
on my stock A500 with Xetec+8M ram I can't seem to communicate with my modem
when I set Serial prefs higher than whatever the pref is that is closest to
and greater than or equal to 19.2K. I sadly came to the conclusion that was the
limit but I'd love to learn what I'm doing wrong if it isn't!
--
Bernhard Rieckhoff Sr. Geophysicist, Schlumberger-GeoQuest
Houston,Tx USA Voice (713) 952-2100 x638 FAX (713) 789-1632
*Opinions expressed here are unlikely to reflect those of my employer*
Tony Philipsson
: 19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
: this limit as well.
I just want to put things right,
It is the Paula chip that do the serial transmit, not the CIA.
Paula have the TXD and RXD (5v) and interupt flags.
The CIA-B have the DTR, RTS, CD, CTS and DSR and is just for hand shaking
And it's not the chips that set a max rate, it is the multi-tasking
system and serial-device that set a limit. but the phone line
probably flunk out before that.
Tony Philipsson, Ft.Luaderdale, FL, USA to...@gate.net
Ralph Babel
> There is no "19.2" limitation.
RKM Devices, page 275, io_Baud.
Maxwell Daymon
: In article <3drmsr$r...@potogold.rmii.com>, mda...@rmii.com (Maxwell Daymon) wr
: > My A500's and A2000 all handle 57,600 just fine, and the A2000 handles
: > CIA's that are in the A500/2000 are in the A3000!
: on my stock A500 with Xetec+8M ram I can't seem to communicate with my modem
: when I set Serial prefs higher than whatever the pref is that is closest to
: and greater than or equal to 19.2K. I sadly came to the conclusion that was the
: limit but I'd love to learn what I'm doing wrong if it isn't!
On a stock machine, you'll have problems - but not because of the CIA
chips - because of your processor not being able to handle the amount and
quantity of interrupts being generated. A $299 68030 (Derringer?) with
some 32-bit RAM would probably allow 38.4 EASILY, 57.6 PROBABLY.
One thing I forgot - the highest rate I could get on a stock "old"
(A500/A2000) was 78,600bps for both send AND receive without forbid() or
some other hostile system takeover or freeze.
The serial hardware is actually very flexible, just not horribly powerful
(in terms of what you can do at the same time)
The final thing that you are doing wrong is setting your prefs to a speed
the modem doesn't communicate with (31,250?) which is a MIDI speed. Some
14.4K modems don't even RUN at 14.4 - you have to go at 300, 1200, 2400,
4800, 9600, 19,200, or 38,400. You need to get "Term" and set the port
from there to at least 38,400bps. I've got my port at 38,400 right now,
but if I'm downloading text files I put it to 115,200bps.
Nathan Palmer
Maxwell Daymon
: Maxwell Daymon wrote:
: > There is no "19.2" limitation.
: RKM Devices, page 275, io_Baud.
Amiga Hardware Reference Manual: 255, paragraph 3
"The Paula custom chip contains a Universal Asynchronous
Receiver/Transmitter...any rate from 110 to over 1,000,000 bits per second."
"with a cable of reasonable length, the maximum reliable rate is on the
order of 150,000 - 250,000 bits per second."
So, perhaps there is a 19.2Kbps artificial limitation but it certainly
doesn't affect those of us who prefer 115,200bps. (I admit, my A3000
doesn't like a WHOLE lot of other stuff at 115,200, but it WORKS and I
can download to the hard drive)
Maxwell Daymon
: : That wouldn't matter since it's the CIA chips that are the limiting factor of
: : 19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
: : this limit as well.
: And it's not the chips that set a max rate, it is the multi-tasking
As I stated, the maximum Paula can do is ~1,000,000 bps (1Mbps)
: system and serial-device that set a limit. but the phone line
: probably flunk out before that.
I will acknowledge that the OS/device system may have an artifical
limitation, but that doesn't stop anyone from doing 115,200 if they have
processor speed and a well designed DMA hard drive to back them up.
Ralph Babel
> Ralph Babel wrote:
>
>> Maxwell Daymon wrote:
>>
>>> There is no "19.2" limitation.
>>
>> RKM Devices, page 275, io_Baud.
>
> Amiga Hardware Reference Manual: 255, paragraph 3
Irrelevant. Are you using the serial.device or hitting the
hardware directly?
Tomas Arvidsson
>Ralph Babel (rba...@babylon.pfm-mainz.de) wrote:
>: Maxwell Daymon wrote:
>: > There is no "19.2" limitation.
>: RKM Devices, page 275, io_Baud.
>Amiga Hardware Reference Manual: 255, paragraph 3
>"The Paula custom chip contains a Universal Asynchronous
>Receiver/Transmitter...any rate from 110 to over 1,000,000 bits per second."
>So, perhaps there is a 19.2Kbps artificial limitation but it certainly
>doesn't affect those of us who prefer 115,200bps. (I admit, my A3000
"For the built-in driver, any baud rate in the range of 110 to about 1
megabaud is acceptable. [...] Although baud rates above 19,200 are
supported by the hardware, software overhead will limit your ability
to ``catch'' every single character that should be recieved. Output
data rate, however, is not software-dependent." (note the common baud
vs. bps mistake...)
Obviously, the 19,200 limit is not carved in rock but if you are going
to stay within the limits of the system that is the highest rate you
can expect to work on *all* machines, A500-A4000. However, it is
(again) obvious that it is possible to go beyond that limit, without
messing up the system, if you have a machine that is a little bit more
powerful than a basic A500. For instance, my A3000T is happily talkning
to my modem at a constant rate of 57.6Kbps and still have plenty of
time over for other activites.
>------------------ Maxwell Daymon One man's "magic" is another man's
--
Tomas Arvidson *** md94...@nada.kth.se * d91...@csd.uu.se ***
Maxwell Daymon
Using Term, Terminus, or NComm with serial.device, artser.device,
"USRser.devcice", or baudbandit.device.
Ralph Babel
> Ralph Babel wrote:
>
>> Maxwell Daymon wrote:
>>
>>> Ralph Babel wrote:
>>>
>>>> Maxwell Daymon wrote:
>>>>
>>>>> There is no "19.2" limitation.
>>>>
>>>> RKM Devices, page 275, io_Baud.
>>>
>>> Amiga Hardware Reference Manual: 255, paragraph 3
>>
>> Irrelevant. Are you using the serial.device or hitting
>> the hardware directly?
>
> Using Term, Terminus, or NComm with serial.device,
> artser.device, "USRser.devcice", or baudbandit.device.
Quod erat demonstrandum.
Richard W. Losey
>
>In article <3drmsr$r...@potogold.rmii.com>, mda...@rmii.com (Maxwell Daymon) wr
>
>SNIP
>
>> That doesn't exist. I have run at 57,600 FULL SPEED with two Amiga 500's.
>> I have run at 115,200K on an Amiga 2000. This limit is purely in your
>> imagination, and there are MANY of us running at 57,600 on 500's and
>> 2000's RIGHT NOW - this very _microsecond_ as I write this message.
>
>SNIP
>
>> My A500's and A2000 all handle 57,600 just fine, and the A2000 handles
>> 115,200 fine. I'm not using newer or magical CIA's. Furthermore, the SAME
>> CIA's that are in the A500/2000 are in the A3000!
>
>SNIP
>>
>Hmmm
>
>on my stock A500 with Xetec+8M ram I can't seem to communicate with my modem
>when I set Serial prefs higher than whatever the pref is that is closest to
>and greater than or equal to 19.2K. I sadly came to the conclusion that was the
>limit but I'd love to learn what I'm doing wrong if it isn't!
I also would like to add my "Hmmmm"; I have an accelerated 2000; I am using
Commodore's 'serial.device' with a Supra 14.4k modem (external). I have my
serial port locked at 38.4k and would love to go up to 57.6k; however, I
appear to get errors at this speed. I would appreciate comments on what else
the problem could be -- I use TERMINUS also.
--
Richard W Losey (rlo...@sdf.saomai.org) OR !uunet!saomai.org!rlosey
"Great men are not always wise" -- Job 32:9
Jeffery Carter May
: I also would like to add my "Hmmmm"; I have an accelerated 2000; I am using
: Commodore's 'serial.device' with a Supra 14.4k modem (external). I have my
: serial port locked at 38.4k and would love to go up to 57.6k; however, I
: appear to get errors at this speed. I would appreciate comments on what else
: the problem could be -- I use TERMINUS also.
How many colors are you using? I use 4 or 8 colors, and have no problems at
57,600. With 16 colors, you WILL have problems. Try setting your screenmode
to something less than 16 colors.
Maxwell Daymon
: Commodore's 'serial.device' with a Supra 14.4k modem (external). I have my
: serial port locked at 38.4k and would love to go up to 57.6k; however, I
: appear to get errors at this speed. I would appreciate comments on what else
: the problem could be -- I use TERMINUS also.
The first thing to point out is that you are running at 38,400 (2x faster
than the 19.2 rate).
Next:
"Accelerated" can mean anything from a 14MHz 68000 to a 33MHz 68040.
You'll have to be more specific. Anything above 19.2 _is_ going to be
processor and configuration dependent.
A) Do you have at least a 16MHz 68020? (A 25, 33, or 40MHz 68030 would be
preferable)
B) Do you have 32-bit RAM? (Enough to have leftover after running the
Term program, and enough to have all the buffer memory in 32-bit RAM)
C) Do you have an RTS/CTS hardware handshaking modem cable, and do you
have RTS/CTS enabled?
D) Are you downloading to floppy, RAM, or a hard drive? DMA usage may
vary greatly and some controllers (esp. GVP it seems) will cause many
errors at high serial port speeds.
E) What screen modes and overscan area is selected? Screens that are
pushed to the lower left hand corner of the active area will not
interfere with DMA as much as full overscan screens and screens pushed to
the upper left.
F) What resolution and color mode? NTSC:Hires-Laced 736x482 16-color mode
is going to have a HUGELY detrimental effect on how fast you can go. Set
your term screen to a low res, low color mode.
G) A Picasso-II allows me to have an 800x600 NI 72Hz 16-color terminal
screen with full speed scrolling that keeps up with my modem at 57,600bps
and I've gone as high as 115,200. (This is on a 33MHz 68040 in an A2000)
My point is that the serial port *is* limited, but in the sense that its
limitations are quickly revealed by the environment. The port, in itself,
can achieve very high speeds. It just depends on what ELSE you want to
do. An Amiga DEDICATED to receive data from the serial port at the
expense of video, multi-tasking, etc. can probably get very high rates.
It's a trade-off. You might even consider a seperate "minimal"
startup-sequence for when you want to do high-speed telecommunications.
Maxwell Daymon
(010?) or better. Move your VBR. There are programs out there that do it,
some FastROM options do it automatically.
Ralph Babel
> I use 4 or 8 colors, and have no problems at 57,600.
57600 bps using the Amiga's built-in serial port, eh?
That's 5760 characters (8N1) per second, i.e. one RBF
interrupt every ~174 microseconds, right?
Too bad that Disable() may be in effect for up to 250
microseconds (see the corresponding exec.library autodoc
entry), i.e. your settings will conflict with properly
written applications.
You lose. Better luck next time.
Michael van Elst
>Too bad that Disable() may be in effect for up to 250
>microseconds (see the corresponding exec.library autodoc
>entry), i.e. your settings will conflict with properly
>written applications.
Not even counting that proper written interrupts may
cause even higher delays or that proper generated screenmodes
may stop interrupts for even a larger time.
Fortunately this is pretty much irrelevant in real life.
Regards,
--
Michael van Elst
Internet: mle...@mpifr-bonn.mpg.de mle...@serpens.rhein.de
"A potential Snark may lurk in every tree."
Jeffery Carter May
: Too bad that Disable() may be in effect for up to 250
: microseconds (see the corresponding exec.library autodoc
: entry), i.e. your settings will conflict with properly
: written applications.
: You lose. Better luck next time.
You'd be REALLY mad if I told you that I once did a null-modem transfer of
a 300k file at 115,200 bps using this configuration with absolutely NO
errors.
Glenn M. Saunders
: E) What screen modes and overscan area is selected? Screens that are
: pushed to the lower left hand corner of the active area will not
: interfere with DMA as much as full overscan screens and screens pushed to
: the upper left.
: F) What resolution and color mode? NTSC:Hires-Laced 736x482 16-color mode
: is going to have a HUGELY detrimental effect on how fast you can go. Set
: your term screen to a low res, low color mode.
It's too bad terms don't support a screen blank or reducing feature to
try to cut down on gfx load during downloads.
That would certainly make things easier.
Michael van Elst
>Ralph Babel (rba...@babylon.pfm-mainz.de) wrote:
>You'd be REALLY mad if I told you that I once did a null-modem transfer of
>a 300k file at 115,200 bps using this configuration with absolutely NO
>errors.
Me thinks that Ralph has little interest in real world. It's not easy
to be the guru at the top... sort of.
Michael van Elst
>It's too bad terms don't support a screen blank or reducing feature to
>try to cut down on gfx load during downloads.
No need to blank the screen. With ECS 2 color hires is ok, with AGA
you can do 4 color productivity or 8 color hires with no effect by
bitplane DMA at all.
Ralph Babel
> You'd be REALLY mad if I told you that I once did a
> null-modem transfer of a 300k file at 115,200 bps using
> this configuration with absolutely NO errors.
*yawn* Am I supposed to be impressed? You (and Van Elst)
keep ignoring the simple _fact_ that _latency_ always deals
with _worst-case_ figures. (Besides, I'm using a 16552 with
a 16-byte FIFO anyway.)
Look at it this way: You may be able to run a red light
unharmed. You may get caught by the cops. You may even get
killed. It's _your_ decision, but don't talk others into
trying this as well without explaining the possible
consequences to them first.
Michael van Elst
>keep ignoring the simple _fact_ that _latency_ always deals
>with _worst-case_ figures.
What a poor mind... not that you know everything better, you
also know what we are thinking and ignoring. Of course, you
didn't have to read what we wrote then... omniscience has its
merits.
Virtual Adept
3 BBS's here in York, PA that support that speed, I get a clean 28.8
connect to the BBS I help run, and 26.4k connects to the other two. The
reason I get those speed connects is due to both the line quality and the
relay substations that I get routed through... it is definately not the
modem or the computers serial port that limits me on the two 26.4
connects. I run my serial port at 38.4k with absolutely no errors in
transmission of any kind. :)
I'm happy with that!
Ralph Babel
> [Van Elst's usual diarrhea of the fingers deleted]
Van Elst, do the Usenet community a favor and get a life.
Alternatively, get a rope.
is
mle...@speckled.mpifr-bonn.mpg.de (Michael van Elst) writes:
>In <29...@babylon.pfm-mainz.de> rba...@babylon.pfm-mainz.de (Ralph Babel)
writes:
>>*yawn* Am I supposed to be impressed? You (and Van Elst)
>>keep ignoring the simple _fact_ that _latency_ always deals
>>with _worst-case_ figures.
>What a poor mind... not that you know everything better, you
>also know what we are thinking and ignoring. Of course, you
>didn't have to read what we wrote then... omniscience has its
>merits.
I just might add that
* the serial.device autodocs say 292000 bps, and mention potential
problems above 32 kB ON A BUSY SYSTEM, ... this warning would
account for filesystem and console.device throughput problems and
too small serial input buffers.
* 250 uSec disable() time gives you about 40000 serial async bps
(1 stopbit counted).
Regards,
Ignatios SOuvatzis
Ralph Schmidt
To add something nobody has mentioned yet.
When the scsi.device gets an io-request it has to flush the datacache.
This means for the A3640 card:
1024 longwords(Datacache) * 7 Waitstates(No Burst) * 40ns(25Mhz)
=286720 ns ~ 287 uSec
total buslockup until the processor is able to start the global
interrupt routine.
--
Ralph Schmidt la...@uni-paderborn.de
University of Paderborn (Germany)
Richard W. Losey
>Richard W. Losey (rlo...@sdf.saomai.org) wrote:
>
>: I also would like to add my "Hmmmm"; I have an accelerated 2000; I am using
>: Commodore's 'serial.device' with a Supra 14.4k modem (external). I have my
>: serial port locked at 38.4k and would love to go up to 57.6k; however, I
>: appear to get errors at this speed. I would appreciate comments on what else
>: the problem could be -- I use TERMINUS also.
>
>How many colors are you using? I use 4 or 8 colors, and have no problems at
>57,600. With 16 colors, you WILL have problems. Try setting your screenmode
>to something less than 16 colors.
I use 4 colors when I emulate a VT102; I use 8 colors for IBM-style consoles.
Could the number of lines make a difference? I have some 40-line IBM stuff.
On another note, I haven't tried 57.6k for quite a while; I just remember
getting errors when I first tried it. I don't do scads of uploads anyway, so
perhaps 38400 will be OK for me.
Christopher Diedrich
(Richard W. Losey) writes:
RL> I also would like to add my "Hmmmm"; I have an accelerated 2000; I am using
RL> Commodore's 'serial.device' with a Supra 14.4k modem (external). I have my
RL> serial port locked at 38.4k and would love to go up to 57.6k; however, I
RL> appear to get errors at this speed. I would appreciate comments on what else
RL> the problem could be -- I use TERMINUS also.
Could it be possible that TERMINUS is the cause of all those errors?
I also have an accelerated 2000 (C= A2620) and I am running my
my ZyXEL U1496E+ with NComm at a locked DTE rate of 76.8kbps and have
NO problems at all. I am using the BaudBandit.device.
RL> Richard W Losey (rlo...@sdf.saomai.org) OR !uunet!saomai.org!rlosey
Ciao, Chrickel.
GE/SS d-/---(?) H+ S+:+ g+ p2>!p !au a-- w+ v?/ c++/+++ U- L+ E-
N+++ K- W--- M+ !V -po+ t+ !5 !j R+ G? tv b+ D+ B? e+ u* h/!
f+ !n y?
Mike Ellenberg
: Jeffery Carter May wrote:
To carry you analogy one step further Ralph...
If you get a ticket for running the redlight on your first trip around
the block, stop at the redlight on your next trip. If you get caught for
speeding past your supposed "Serial Port Speed Limit" the simple answer
is slow down next time. I would hope most people would find that to be the
logical solution to "hmm I guess I can't go *THAT* fast".
If you don't believe we can handle 28.8k on the internal serial port with
DTE's up to 115.2k then fine... no one told you (or anyone else for that
matter) that they *HAD* to do so, and you can stay at the 9600bps rate
that the former Commodore said was the maximum "safe" speed.
It also quite true that the "serial port speed limit" is determined by
several things, and I think most people can find a comfortable solution
if they just experiment with a few simple parameters, which I list below.
1) Screen depth. The more chip ram bandwidth you use the lower the
maximum "safe" speed you will be able to handle.
A 4 color non-interlaced high res screen uses less
bandwidth than a 16 color interlaced high res screen.
The 4 color (or 8) is more likely to support higher
bit rates than the 16.
2) CPU. A faster CPU will aid in processing all that incoming data.
3) Number of tasks running. The more you ask of your machine the less
often it can get around to getting that
incoming character.
There are a few other things of course, like other serial device drivers,
terminal programs, and I've even seen some blitter speed up hacks that
totally remove the ability to handle high speed serial I/O on the internal
serial port, they are all things to check and try.
Personally on my 2000 w/3.1, a PP&S Zeus 040, the standard serial.device,
and Terminus 2.0d - running on an 8 color highres interlaced screen, I have
*NO* (read zip, zero, nada, never ever) trouble handling 28.8k V.34 (or V.FC)
connections at 115.2k DTE rates, except in one or two cases which are easily
avoided, one of which is bringing any high chip ram bandwidth screens to the
front while online. I regularly download (directly to my hard drive) at
over 3200 to 3300 cps without a single error.
If I *HAVE* to have a 16 color highres screen, I just drop back to 38.4k and
then all my troubles go AWAY.
BTW- just to back up Mr. May, I have also done null modem transfers via
the stock serial port at 115.2k in Terminus and done them *WITHOUT* error,
to another Amiga similarly configured at 115.2k.
Don't bother flaming me... you won't get a response, and enjoy that
"16552" of yours, a 16550AFN UART does making high speed life easier, but
it's not the *ONLY* way to get reliable high speed serial I/O.
*---------------------------------------------------*--------------------*
| Mike Ellenberg Internet: mi...@crl.com | "Around the World |
| Freelance Photographer/VT Editor/Video Engineer | In 80 Presets" |
| USR Courier V.34 Dual Amiga 2000/Zeus 040 | by: Phil Collins |
*---------------------------------------------------*--------------------*
Ralph Babel
> When the scsi.device gets an io-request it has to flush
> the datacache. This means for the A3640 card:
>
> 1024 longwords(Datacache) * 7 Waitstates(No Burst) * 40ns(25Mhz)
>
> =286720 ns ~ 287 uSec
>
> total buslockup until the processor is able to start the
> global interrupt routine.
Don't confuse the issue with facts.
Ralph Babel
> If you don't believe we can handle 28.8k on the internal
> serial port with DTE's up to 115.2k then fine...
Whether or not I believe you is beside the point.
> just to back up Mr. May, I have also done null modem
> transfers via the stock serial port at 115.2k in Terminus
> and done them *WITHOUT* error, to another Amiga similarly
> configured at 115.2k.
What's your point? I merely noted that - according to
_Commodore's_ documentation - it's not _guaranteed_, and it
may well fail with _properly_ _written_ Amiga applications.
If it nevertheless happens to works on your particular Amiga
configuration: fine, consider yourself lucky; but _if_ a DTE
rate above 19200 bps _does_ cause problems, you've got only
_yourself_ to blame, _not_ the software.
Th.Huber
>Jeffery Carter May wrote:
>
>> You'd be REALLY mad if I told you that I once did a
>> null-modem transfer of a 300k file at 115,200 bps using
>> this configuration with absolutely NO errors.
It`s not possible with common terms, but twinexpress can do it.
I allready transfered files with 300kbaud. That`s pretty fast.
bye
DAVID BALAZIC
> juan perez (jper...@solix.fiu.edu) wrote:
> : hi all. i just upgraded from a supra 14.4 to a gvc 28.8 and have not been able
> : to connect anywhere at anything better than 21.6 (and that was only once). i
> : have a 3000 with 3.1 roms and have the serial port settings at the max of
> : 31,250. i'm using terminus and have tried it at every setting from 31,250 to
> : 115,200 and have been unable to even attain 24k. i have even d/l baudbandit
> : and tried running that serial device but have had the same results (actually
> : it was more unstable on my system). oh, i have a gvp 040/40 in my system. is
> : the standard amiga serial port capable of reaching 28.8? i know of a local bbs
It depends on the processor speed.
With vanilla A500 U can get 190 000 bits/sec !!
I achieved this with A500 connected to A1200 using the TwinExpress program.
It doesn't use the serial.device I think.
The docs say that a A3000 get achieve 290 000 bps !
To get good speed out of serial.device U must use it in HISPEED mode also
known as RADBOOGIE. I'm not sure what software supports/uses this mode.
> : that has a couple of 28.8 lines running on an a3000 but those nodes are
> : connected to a gvp serial card. if anyone could shed some light on the matter
> : it'd be a most welcome x-max gift. thanx in advance.
-----------------------------------------------------------------------
No E-mails unless neccesary , please !
-----------------------------------------------------------------------
David Balazic David....@uni-mb.si
Amiga 1200 "This is a quote !"
Quantum LPS340A Slovenija, moja dezela.
-----------------------------------------------------------------------
"This sentence is wrong !"
-----------------------------------------------------------------------
DAVID BALAZIC
>
> That wouldn't matter since it's the CIA chips that are the limiting factor of
> 19.2! This is the reason I was asking if the later models (3K,4K,1.2K) have
NO! The CIA's have (almost) nothing to do with the serial port. It is Paula !
And it can work at very high rates ! Like 300 000 bps ...
The problem is fetching or writing the data with the processor !
> this limit as well. I just wanted to know if I was going to have to buy a
> Multiface III card when I upgrade my current pathetic system to a more modern,
> and far more powerful Amiga. I would dearly love to see the full capabilities
> of my Supra 28.8 :-)
Maxwell Daymon
: Maxwell Daymon (mda...@rmii.com) wrote:
: : E) What screen modes and overscan area is selected? Screens that are
: : F) What resolution and color mode? NTSC:Hires-Laced 736x482 16-color mode
: It's too bad terms don't support a screen blank or reducing feature to
: try to cut down on gfx load during downloads.
A disable video DMA might be nice, maybe it would only come back if there
were more than 2 errors in a row or something.
: That would certainly make things easier.
I don't think it'd be horribly difficult.
Sylvain Bourcier
\/ Subject: Re: Amiga Serial cant do 28.8? - Dated: 1 Jan 95
rbp> 57600 bps using the Amiga's built-in serial port, eh?
rbp>
rbp> That's 5760 characters (8N1) per second, i.e. one RBF interrupt every
rbp> ~174 microseconds, right?
rbp>
rbp> Too bad that Disable() may be in effect for up to 250 microseconds
rbp> the corresponding exec.library autodoc entry), i.e. your settings will
rbp> conflict with properly written applications.
rbp>
rbp> You lose. Better luck next time.
You are right but they can't see :)
There are PLENTY of people that open their ports at 57k without any
problems. BUT SINCE THE MODEM HARDLY EVER SENDS AT THAT SPEED, there aren't
any problems...
IT WILL CAUSE TROUBLES if the device EVER sends data at that speed but
that's rarely the case. For example, even a relatively fast computer will
have problems fetching data at 57600 bps. Skeptics can try it with a huge
text file and see what happens when they reach >4000 cps and that CTS/RTS
isn't enough to make it through ;)
An '040 equipped Amiga might be able to fetch at 57600bps but I don't have a
modem fast enough to test it...
---
mEMBER:-wORLD-wILDLIFE-fUND-+-gREEN-pEACE-+-pLANETARY-sOCIETY-
--
|GatorNet: Sylvain Bourcier 128:200/96.66
|Internet: sylvain....@deltacom.cam.org
|
Glenn M. Saunders
: A disable video DMA might be nice, maybe it would only come back if there
What I was thinking is closing down to one status line using a custom
copper list. I've seen this done in the Atari8 with a custom display
list. Unarc and Arc uses this, to speed up throughput. It doesn't take
much CPU overhead to just write a few lines and leave the rest of the
screen blank.
Michael van Elst
>What I was thinking is closing down to one status line using a custom
>copper list.
Come on people. There are lots of video modes that do not slow down
the CPU whatever.
Darren MacKenzie
And WHAT would these 'consequences' be? A corrupted file? Shouldn't
some error correction take care of that? Of course...the error correction
would slow down the transfer process... :-)
I'd try this type of transfer with gold-plated D-sub connectors, a 1-foot
length of monster cable (Amigas one on top of the other), and all power
cords in lead shields........then I bet you'd get GREAT transfer rates
(in 2 colors, 320x200, of course...) :-)
-Darren
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
= Darren MacKenzie = Bell-Northern Research =
= dar...@bnr.ca = DMS Global Loadbuild Specialist =
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
= 18 Deerfield Dr. Apt. 110 Nepean, Ont. K2G-4L1 =
= (613)-226-8747 HOME (613)-765-3041 & 763-8135 WORK =
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Mike Hall
>>How many colors are you using? I use 4 or 8 colors, and have no problems
>>at 57,600. With 16 colors, you WILL have problems. Try setting your
>>screenmode to something less than 16 colors.
>
> I use 4 colors when I emulate a VT102; I use 8 colors for IBM-style
> consoles. Could the number of lines make a difference? I have some
> 40-line IBM stuff.
>
> On another note, I haven't tried 57.6k for quite a while; I just remember
> getting errors when I first tried it. I don't do scads of uploads anyway,
> so perhaps 38400 will be OK for me.
This has come up on another network in the past and I still don't
know the answer but both my stock 1200 and 3000 work 100% at 57600
in 16 colour mode.. (with both term and terminus).
--
__
__///
\xX/ ,\\ike Hall mi...@fulink.edex.edu.au fidonet: 3:640/944
Ralph Schmidt
>In article <3e9r0p$6...@taifun.uni-paderborn.de> la...@taifun.uni-paderborn.de (Ralph Schmidt) writes:
>>
>>To add something nobody has mentioned yet.
>>When the scsi.device gets an io-request it has to flush the datacache.
>>This means for the A3640 card:
>>
>>1024 longwords(Datacache) * 7 Waitstates(No Burst) * 40ns(25Mhz)
>>
>>=286720 ns ~ 287 uSec
>>
>>total buslockup until the processor is able to start the global
>>interrupt routine.
>But this is only worst case. In reality only the dirty parts of the
>cache get flushed, and that's most certainly not the whole one.
>Though I don't have any figures about typical percentages.
This is not true. Your reality doesn't exist.
CacheClearE(address,length,caches)
btst #CACRB_ClearI,d1
bne.s Both
CPUSHA DC
move.l a0,a5
rts
Both:
CPUSHA BC
move.l a0,a5
rte
CPUSHA=ALL
so it doesn't care for the address field.
-----------------
If it would care for the address field you need a busy loop that does this
lsr.l #4,d0 ;devide by 16 to get line count
move.l a0,d1
and.w #$fff0,d1
move.l d1,a0
addq.l #1,d0 ;Align ..
0$:
CPUSHL DC,(a0) ;flush line
subq.l #1,d0
bne.s 0$
rte
But this routine isn't there
Dr. Peter Kittel
>
>When the scsi.device gets an io-request it has to flush the datacache.
>This means for the A3640 card:
>
>1024 longwords(Datacache) * 7 Waitstates(No Burst) * 40ns(25Mhz)
>
>=286720 ns ~ 287 uSec
>
>total buslockup until the processor is able to start the global
>interrupt routine.
But this is only worst case. In reality only the dirty parts of the
cache get flushed, and that's most certainly not the whole one.
Though I don't have any figures about typical percentages.
--
Best Regards, Dr. Peter Kittel //
Private Site in Frankfurt, Germany \X/ Email to: pet...@combo.ganesha.com
Currently employed at "Commodore Konkursverwaltung Hembach"
Michael van Elst
>This is not true. Your reality doesn't exist.
>CPUSHA DC
The number of dirty long words in the line to be pushed determines
the size of the push transfer on the bus, minimizing bus bandwidth
required for the push.
Execution time of a CPUSHA is 267 clocks minimum and 11 + 256*line + idle
in the worst case where 'line' is the time needed to transfer a cache line
and 'idle' is the time needed to push all write buffers and finish all
pending bus transfers.
Of course, when doing SCSI transfers the worst case might be the regular
case as bulk transfers thrash the cache quite easily.
Rask Lambertsen
the device driver used. serial.device is not the only choice. You
could use BaudBandit.device if you want fast and safe transfers.
Remember that serial.device needs to support a lot of rarely used
configuration, and that adds overhead time. BaudBandit.device
is really a better choice than serial.device in most cases.
Remember that the software used (term program AND device driver)
has a big effect on the maximum safe speed.
BaudBandit can be found on AmiNet directory comm/misc.
The file to download is `baudbandit1.4b.lha' (for version 1.4b).
-- Rask
--
/ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ\
| Rask Ingemann Lambertsen <gc94...@gbar.dtu.dk> or <e986...@ebar.dtu.dk> |
| |
| Keyboard error: <Ctrl> and <Alt> are stuck - press <Del> to continue |
\___________________________________________________________________________/
Michael van Elst
>What's your point? I merely noted that - according to
>_Commodore's_ documentation - it's not _guaranteed_, and it
>may well fail with _properly_ _written_ Amiga applications.
That's true.
>If it nevertheless happens to works on your particular Amiga
>configuration: fine, consider yourself lucky;
No need for luck. With common setups this happens to work
regularly.
>but _if_ a DTE
>rate above 19200 bps _does_ cause problems, you've got only
>_yourself_ to blame, _not_ the software.
Rubbish. You can very well blame the software (or some hardware).
The C= documentation isn't the law. After all you could have
put chunks of Disable()'d code in your SCSI-Driver and if these
sections wouldn't exceed 250 microseconds this would be very
well within C= specs. But everyone would blame such code and
they were right to do that.
Ryan Kubica
: /\ QUOTING: rba...@babylon.pfm-mainz.de (rba...@babylon.pfm-mainz.de)
: \/ Subject: Re: Amiga Serial cant do 28.8? - Dated: 1 Jan 95
: rbp> 57600 bps using the Amiga's built-in serial port, eh?
: rbp>
: rbp> That's 5760 characters (8N1) per second, i.e. one RBF interrupt every
: rbp> ~174 microseconds, right?
: rbp>
: rbp> Too bad that Disable() may be in effect for up to 250 microseconds
: rbp> the corresponding exec.library autodoc entry), i.e. your settings will
: rbp> conflict with properly written applications.
: rbp>
: rbp> You lose. Better luck next time.
: You are right but they can't see :)
: There are PLENTY of people that open their ports at 57k without any
: problems. BUT SINCE THE MODEM HARDLY EVER SENDS AT THAT SPEED, there aren't
: any problems...
: IT WILL CAUSE TROUBLES if the device EVER sends data at that speed but
: that's rarely the case. For example, even a relatively fast computer will
: have problems fetching data at 57600 bps. Skeptics can try it with a huge
: text file and see what happens when they reach >4000 cps and that CTS/RTS
: isn't enough to make it through ;)
I don't no where this >4000 came from but back when my 3000 only had an
030/25 I was able to connect to a pc via null modem and transfer a file
at 115200. Which in turn was about 8900 characters a sec, no errors. IN
fact when I started doing more things on the Amiga like: playing a mod
and using higher screen rates (otherwise known as; saturating the bus)
the rate went even higher. I think this was due to the PC not being able
to take the full speed of my Amiga and when I slowed my Amiga down it
finally could keep up. Who knows?
Of course when I put an 040 in it, I had problems with even 57.6 so I
bought a GVP io Extender. And know I have absolutely no problems anytime.
Go figure.
: An '040 equipped Amiga might be able to fetch at 57600bps but I don't have a
: modem fast enough to test it...
: |
Ryan Kubica
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
Michael van Elst (mle...@speckled.mpifr-bonn.mpg.de) wrote:
: That's true.
Can't remember what Scotty (in ST-TNG) said verbatum but:
" I wrote the manual, so I know it's possible.. A good engineer is
always careful to set lower maximums when writing the manual. "
-Wighte Wolfe-