Is anyone interested in Apples' Appletalk network? I am very interested
and have just discovered how to implement a system for a fraction of the cost
of the standard Appletalk hardware components. Check out this comparison:
Old Appletalk New Appletalk
$50 node connection $15 node connection
$1.50 foot wire $.11 foot wire
$3.60 connector $.16 connector
1000' max length 5000' max length
How is this done? Well, it turns out that Apple went for some overkill
when designing their net. The shielded twisted wire cable and the beautiful DIN connectors are not only unecessary, but can introduce a noisy ground
connection. What are the best components to use? The old standby, regular
Ma Bell 4 conductor telephone wire and the modular phone plugs.
If all you want to do is connect two Macs together or a Mac and the LaserWriter, all you need is a radio shack Joy Stick cable wired as a null modem. You do NOT need either the ATALK or BMUG hardware to connect 2 machines. If your network will have more than two machines connected together than check this out.
If you want a permanent network, you first layout the normal 4 conductor phone wire thru the walls or conduit like you would for phone connections. Actually you can use a regular phone setup that is already installed. Just make sure that it is disconnected from the phone system. The Junction Boxes where you will plug Macs together are the regular Modular junction boxes. You will use only 3 connections. The Yellow phone wire is the same as the ATALK blue wire, the Red phone wire is the same as the ATALK White wire, and the Black phone wire is the same as the ATALK shield. The Green phone wire is not connected. After the junction boxes are set up you run a phone extention wire from the junction box to the BMUG connection box connected to the Printer port on the Mac. This box looks like a long phone junction box with a modular plug on one end and a DB9 connector on the other.
The guts of the translator box are essentially the same as in the ATALK box except the BMUG box uses a better quality isolation transformer. The next paragraph details the results of a VAR who installs networks for a living.
DEEPER INSIDE APPLE TALK
Infotek Inc. 56 Cammile, E. Patchougue, NY 11772
Appletalk is a good, low cost computer data networking system, but like most computer system elements, you should know more than just how to plug it in. The design and literature implying any idiot can instantly be running 32 Macs, Laserwiters, file servers and mainframes on the networks probably accounts for many of the problems and rumored problems with Appletalk. We will try in the following notes to give some background and recommentations based on years of communications, design, development, experience and a few month use of Appletalk.
The transmission of digital information is usually by hard wire or over voice networks via modems. We are specifically covering hard wired and twisted pair differential communications (RS422) as used in Appletalk. Most people have the idea that coaxial or shielded, single ended transmission is the best, but years of experience shows that balanced differential twisted pair can have mush lower error rate per magabit mile than coaxial or shielded pair if it's done right. Millions of telephones have been running voice frequencies or 256,000 bits/second voice data for years with unshielded twisted pair and I wonder why people disregard these facts when designing. Apple states that unshielded cable can be an unreliable transport medium but Datmouth college, Corvus, DNA, Orchid and the telephone industry seems to say different. Corvus often runs at 1 megabit per second rather than the Appletalk under .25 megabit/second rate. Flexible Coaxial cable depends on a braided shield that is usually called 85% shielded. Coaxial cable is excellent for 10 magabits/second broadband or RF carrier uses but working baseband with a signal relative to ground brings up the question of what is ground. A good broadband ground is perhaps the most ellusive thing in field engineering. The lab worker avoids the question by testing on a thick copper bench with a central ground point. A lab design is then optimized with a near perfect ground and the engineer then wonders why field results are so bad. Is the green wire on the home or office wall receptacle really the good, low noise ground that designers rely on? I've had much better results over the years assuming a high impedance, noisy ground was all thats available in "real life".
Looking quickly at VSWR (voltage standing wave ratio) I find people apply 100 megahertz or one gigahertz thinking to 230 KHZ and end up compromising in the wrong direction. Transmission line impedance or characteristic impedance is another factor to be considered. Attenuation or loss can be excessive if 500 megahertz power is sent on a mismatched transmission line but at 250 KHZ it's not that serious and it pays to experiment. Unbalanced or twisted pair transmission can be critioal if multiple repeaters are used but since no repeaters are used, this is not a problem if the right isolation transformer is used. Lets look at what Appletalk specs say and experience has shown;
Appletalk: a bus network (Theory)
(1) Retail per connection $50.00
(2) Service - Laser printer, file server, bridge to other Appletalk networks, ethernet gateway, disk servers, IBM PC interface, electronic mail etc.
(3) 32 devices per network
(4) Installation - easy
(5) Baseband with no tuning
(6) Network software - built into applications, architecture - Layered open
(7) Max. length 1000 ft.
(8) Transmission speed - 230 Khz
(9) Access method CSMA/CA
(10) Cable type. Twisted pair, precut cables in 2 meter, 10 meter and 100 meter lengths.
Experience: Apple has a connection to the Laserwriter and third parties are just coming out with a few of the remaining items needed. Maximum cable length seems to be under 400 ft for any reliability and it's difficult to get 32 devices into 400 ft.
To find the reasons for the 400 ft. and to find how to reliably extend this to our presently available over 5000 ft. we measured the internal Mac circuit and then came up with an equivalent circuit. We then built up several long distance (to 6000 ft.) and multi-Mac networks and tested them while injecting typical electrical noise as found in an office. We also installed some networks in a business invironment.
The real world and theory agreed within less than 10% for the ungrounded telephone wire systems but varied by 300% amd more for the Apple connectors. It appears that beauty won out over function and the nice chrome plated ground connections of the Apple version develops noise far better than any of our labs noise generators, and leaving out the ground through the 3rd pin makes it worse. The crazy glue used to hold these Apple connectors together make this poor shell connection far worse on some boxes, thus some networks work and others don't. Our reliability experience over the years with telephone modular connectors has been excellent, so we usually design with these field proven connectors to avoid being the pioneers with the arrows in our back.
To run a network without grounds or at least without shielding, we use isolation transformers with less than 10 picofarads of primary to secondary capcitance rather than to use transformers like Apple that have about 36 times this value and depend on a good ground to solve the problem.
The surprising result of using regular telephone components is an easier to install, much lower in cost, 5000 ft.+ reliable network that allows the full potential of a really good low cost networking idea (Appletalk) to be used. GET ON LINE!
The above was taken from the Fall 1985 issue of Berkeley Mac User Group Newsletter. This thing is 220 pages of really useful information and a good public domain utility disk. This thing comes out once a semester. Well worth the cost. Subscriptions can be had by writing
Berkeley Macintosh Users Group, 1442A Walnut St., Suite #153, Berkeley, CA 94709. Phone # is (415) 849-9114
