Issue 153 April 2003
E-Chips

Start • MAC Attack • Kitchen Sink • The 80 Way • Tini the Tiger • Sources and PDF

Those of us, ahem, of a certain age may recall a childhood visit to Disneyland. Back when Walt was in charge, Disneyland would sell you a book of ride tickets comprised of various combinations of denominations from A through E. The scheme had the effect of turning zillions of visiting kids into budding entrepreneurs, middlemen, and market makers. At a tender age, we learned the basics of free enterprise, everything from options and futures to black markets and monopoly power. 

A-Tickets were only good for the Mickey Mouse amusements your kid brother might have liked. All the better if you could’ve conned him into trading a couple of deflated As for a crisp new pass to the most thrilling rides, the coveted E-Ticket.

Of course, you could’ve just tried lying, cheating, or stealing to get your hands on your little bro’s tickets. But you didn’t, knowing full well you’d get busted by the ever-vigilant FTC—father in total control—who’d likely lay a bit of cruel and unusual punishment on your behind. There’s probably a lesson there.

Back in my February column, I pondered the choices designers face when deciding how to “network-enable” their embedded applications (“Working the ’Net,” Circuit Cellar 151). I hope the message you took away was that it’s important to understand the complexities and subtleties hidden behind the simple-sounding catchphrase. In that column, one thing all the solutions had in common was Ethernet, setting the stage for the emergence of a new class of Ethernet chips, or E-Chips—the E-Tickets you need to board that thrill ride known as the I-way.

PARC PLACE

Most of you are familiar with the Ethernet story, it being another of the pivotal advances in computing spawned during the glory years at Xerox’s heralded Palo Alto Research Center, a.k.a. PARC. [1] That was a time when much of what we experience today as computing—networking, object-oriented languages, laser printers, bitmap displays, and even the ubiquitous mouse—floated to the top of the primordial brew at PARC.

In fact, the roots of Ethernet go even further back, in particular to the Aloha network developed at the University of Hawaii. [2] Aloha actually evolved as a packet radio protocol, but the fundamental issue of multiple users contending for access over a shared medium translated nicely to Xerox’s goal of hanging a bunch of Alto workstations and laser printers on a single wire.

Ethernet passed many milestones during the long journey from Bob Metcalfe’s mid-70s brainstorm and 2.94-Mbps prototype. The Ethernet ball really got going in the late 1970s thanks to the so-called DIX consortium (DEC, Intel, and Xerox) that successfully standardized and commercialized the concept. 

The next level came as a blessing from the IEEE with the first of the now myriad 802.x LAN standards. Actually, the standard codified a number of changes to the DIX version that led to a period of confusion as the number of implementations and upgrade proposals proliferated.

However, within the last few years, the situation has been resolved—first in favor of 10BaseT and subsequently 100BaseTX. One or both of these is behind the “Fast Ethernet” RJ-45 jacks on typical PC and network gear of recent vintage. The compelling advantage for these two versions of the standard is that they are upward/downward-compatible (i.e., ports are increasingly dual-speed 10/100 capable) thanks to the fact that they use the same wiring (shielded twisted pair) and pin assignment.

It’s still called Ethernet, but along the way even the fundamentals have been changed. For instance, taking advantage of the extra wires in the cable, modern Ethernet configurations are increasingly point-to-point, full-duplex links versus the shared-wire, half-duplex scheme of the original. With full-duplex links, you get twice the speed and don’t have to worry about carrier sensing, multiple access, and collision detection (i.e., CSMA/CD).

Of course, the headlines these days are about the new 1-Gbps (and even 10-Gbps) versions of Ethernet on the horizon. For now, I’ll leave those to the performance-at-any-price and big-iron infrastructure crowd. On the embedded front, the challenge is to find a simple and inexpensive way to drop a standard 10/100 Ethernet interface into practically anything. And, thanks to the Silicon wizards, it’s easier than ever to do just that.