| |
PA2AGA > HDDIG 25.09.99 03:22l 182 Lines 7855 Bytes #-9766 (0) @ EU
BID : HD_99_239F
Read: GUEST
Subj: HamDigitalDigest 99/239F
Path: DB0AAB<DB0PV<DB0MAK<DB0ERF<DB0HSK<PI8DRS<PI8DAZ<PI8GCB<PI8HGL<PE1NMB<
EA7URC<PE0MAR<PI8VNW
Sent: 990924/2108Z @:PI8VNW.#ZH2.NLD.EU #:2296 [HvHolland] FBB7.00g $:HD_99_239
From: PA2AGA@PI8VNW.#ZH2.NLD.EU
To : HDDIG@EU
Received: from pa2aga by pi1hvh with SMTP
id AA19766 ; Fri, 24 Sep 99 19:57:41 UTC
Received: from pa2aga by pa2aga (NET/Mac 2.3.67/7.5.3) with SMTP
id AA00015850 ; Fri, 24 Sep 99 21:07:28 MET
Date: Fri, 24 Sep 99 21:01:22 MET
Message-Id: <hd_99_239F>
From: pa2aga
To: hd_broadcast@pa2aga
Subject: HamDigitalDigest 99/239F
X-BBS-Msg-Type: B
PCs so common and cheap, but it would be feasible. The TAPR NNC
didn't have any *technical* reasons why it wouldn't work. It just got
bypassed by the economics of clone PCs. (I have one of the six
extant fully populated and functional TAPR NNCs.)
>> Basically, you're looking at $600 to set up for 56 kb. That's more than
>> it costs to dedicate a TNC and radio to 1.2 kb, but not much different
>> from what it costs to set up to do 9.6 kb correctly (that's because you
>> generally have to use an IF modulated multimode to get an acceptable
>> BER with G3RUH style 9.6 kb).
>
>Things are a bit cheaper for 9600. Any old serial card will work.
>Radios are available cheap (TEKK e.g).
When I started fooling with 9.6 kb, I thought the same. But the TEKK
radios aren't a panacea. The receivers are pretty deaf, yet they have
wide open front ends. They are IMD generators on a high RF site
without cavities in front of them, and if you add cavities, that just makes
them deafer. So they are really only useful at electrically quiet sites, and
paired with a station capable of delivering a strong enough signal to
overcome their lack of sensitivity. Thus they can serve in some cases
as user radios, but they won't do in general as network radios.
You need a more robustly designed radio to service the network. Now you
*can* convert commercial surplus such as a Micor, Mitrek, or GE Mastr II,
but that's not *off the shelf*, and I've seen a resistance by you and Charles
to the idea that you might actually have to do some radio work. So that
limits you to appliances, and the only appliances sold for amateur radio
which work adequately out of the box are the multimodes.
You still have the problem of dealing with HDLC encoded signals too.
So ordinary serial interfaces won't do either. You either need a HDLC
hardware card in the computer, or some sort of software bit banger.
You wouldn't want to use a TNC (how quaint) to handle HDLC to serial
RS232 conversion because you double latency when you do that. You
can't afford that when you start out with such a low speed.
Trying to make a G3RUH encoded 9.6 kb system work well is neither
simple nor cheap when you begin to look at the problem from a network
perspective (and limit yourself to appliance solutions).
Gary
Gary Coffman KE4ZV | You make it |mail to ke4zv@bellsouth.net
534 Shannon Way | We break it |
Lawrenceville, GA | Guaranteed |
>.
------------------------------
Date: Wed, 22 Sep 1999 18:30:07 -0700
From: "Hank Oredson" <horedson@att.net>
Subject: Let's look at real numbers for TNC software sales
Gary Coffman <ke4zv@bellsouth.net> wrote in message
news:AAbpN1T7Qxjj1htLwGtHBWlt64mY@4ax.com...
> On Mon, 20 Sep 1999 20:02:33 -0700, "Hank Oredson" <horedson@att.net>
wrote:
> >Gary Coffman <ke4zv@bellsouth.net> wrote in message
> >> Perhaps in the densely populated little nations of Europe national
> >> networks can be done, but not across our vast rural spaces. That's
> >> sad, but it is real. We need to face up to reality and do what we are
> >> best suited to do, and that is the smaller scope metropolitan and
> >> relatively small regional networks that we can practically fund,
> >> build, and maintain.
> >>
> >> The GRAPES network covers an area of roughly 45,000 square
> >> miles, and that's about as big as is practical for a fast amateur
> >> network in our part of the US. We don't have the people, money,
> >> or sites to grow it much bigger. California might be different. They
> >> have 10 times our population, and a bunch of natural high sites. I
> >> haven't seen an organization with the energy and dedication of
> >> GRAPES building a fast network out there, however.
> >
> >California is mostly open space with no people, just like the rest
> >of the West. It's a BIG place. The big cities are, north to south,
> >with approximate distances:
> >
> >Vancouver, BC - Seattle, WA - 50 miles.
> >Seattle, WA - Olympia/Tacoma, WA - 50 miles.
> >Olympia/Tacoma, WA - Portland, OR - 100 miles.
> >Portland, OR - San Francisco, CA - 600 miles.
> >etc.
>
> My, my, how California has grown. My atlas says that of those
> cities, only the latter is in California. :-)
Referent was "the West", transitive through "It's".
I'll let others comment on your strange ideas about California.
45,000 square miles is a very small area. The Western network
covers only 1,050,000 square miles now. It used to cover nearly
1,500,000 square miles, until we lost the links east from Montana
and the links across Texas. This does not count the Canadian
or Mexican portions of the network ...
Don't even bother with the "realtime" comments. It ain't, it
was not intended to be, and the resources to build such a
thing do not exist. It's store and forward.
--
... Hank
http://horedson.home.att.net
>.
------------------------------
Date: Wed, 22 Sep 1999 18:21:42 -0700
From: "Cathryn Mataga" <cathryn@junglevision.com>
Subject: Let's look at real numbers for TNC software sales
Gary Coffman wrote in message ...
>>My horizon to the north, however, is about
>>one mile. But I still have a 9600 baud link in that direction, of
>>about 25 miles. 25W, 10 el yagis on both ends. Works well,
>>even though we are way out of the first Fresnel zone.
>>
>>> You need a BER of under 1 in 10E6 or your network
>>> performance becomes dismal out past 3 or 4 hops.
>>> That means you have to engineer things *right* on
>>> every link.
>>
>>Not possible. We have these hills and mountains.
>>The 100 mile link to the south works very very well.
>>However that link of 25 miles to the north would take
>>three relay points to make it line of sight. Oops.
>
>Yeah, and if the two hops were linked as part of a network path,
>all the traffic traversing that part of the net would be limited by the
>BER of the worst link in the path. That's why you can't have bad
>paths in your network. They kill more than just the one hop's
>performance. They drag down a large segment of the network's
>traffic along with them.
Okay, I'm still trying to figure out why his 25mile 'bad path' link
would necessarily have a BER, as you say, above 1e6. Is there
any reasonable way for him to really measure this? Maybe a
problem here is that we 'think' we might have links that are good,
but they have bit error rates, good enough so the the link 'seems'
okay, but really the errors are bad enough to cause network problems.
And that what's really needed are some good tests.
1e6 bits, that's 125,000 bytes or about 490 256 byte packets. Now,
let's say, he digipeated 500 256 byte packets of random data from his
system, to the other end and back, we should find that he should miss
more than a couple, if his BER is worse than 1e6 -- right? Since that
would have been 2*500*256*8 bits going through the link. That is twice
for the trip up and back. I'd be curious to know the results of this kind
of measurement to know how good or bad the link really is. Could
this be a way to decide if new links are going to hurt or help the network?
That is simply do measurements of bit errors, and if they're low enough
it's okay, however the thing works.
Also, I'm also trying to figure out what exactly is the mechanism that
brings in the errors -- assuming that if he did the measurement, we'd
find BER's much worse. For the speed of light of 186000mps, the
To be continued in digest: hd_99_239G
Read previous mail | Read next mail
| |
