How CDDI and 100BaseT got started by a bored engineer
The engineer was bored.
His latest product was sailing through manufacturing, no surprising hitches or glitches to resolve, and he came to work each day wondering what to do to keep busy. Management and Marketing had not yet decided on the new projects, and so he had a couple months of free time on his hands. With nothing better to do, he spent a few days dreaming up a serial backplane interface based on passive directional couplers to provide slot-to-slot port isolation.
As far as he knew this had never been tried before, and he was curious as to whether or not it would actually work. With this technique a multiple-slot, high-speed serially-encoded data bus would not be susceptible to loading variations caused by filled or empty card slots, and each individual card could include true end-of-line matched termination instead of forming an open stub.
The bored engineer rummaged through his junkbox and pulled out several directional couplers left over from a previous defunct project.
He soldered their innards to a copper-clad breadboard to simulate a 16-slot tree and threw together a quickie random data generator with transmit pre-emphasis equalization to accommodate the frequency rolloff of the cascaded directional couplers. As expected, the received eye patterns were beautiful and did not vary with slot loading.
This took a couple of weeks. Then after the experiments worked successfully he was bored again, and no one else was much interested in the idea. With no immediate new projects on the drawing board, it was considered “a solution in search of a problem”. Some did not fully comprehend what a directional coupler was, nor how it could be used in a digital domain.
Then the bored engineer began to wonder what would happen if the transmission path was extended. Instead of a small tree topology of directional couplers on a backplane, could 125 Mb/s data be transmitted usable distances point-to-point on copper cable? How much distance? This was at the time when FDDI (Fiber Distributed Data Interface) was in development and all the hype was for optical fiber. But it seemed worthwhile to see what could be done with copper and equalization at the FDDI bit rate and double that. After all, there wasn’t much else for him to do.
So the no-longer-bored engineer designed and constructed a bit error rate test system that generated the FDDI DDJ (Data Dependent Jitter) test pattern at both 125 and 250 Megabits/second.
The bit error detection at 250 Mb/s used the wired-XNOR technique described here.
Since this was being done without official corporate management approval, or even with their knowledge, the budget was somewhat limited. The lab slush fund allowed purchase of some more ECL and RF amplifier chips and line transformers. TTL chips were easily available from existing stock. The junkbox had some leftover 20 MHz crystal oscillators. Raw double-sided copper-clad scraps were available from the pcb house.
The test system was built into tin cans for RF shielding to avoid the need to purchase shielded metallic enclosures; at this data rate logic signals are RF (radio frequency) and the various functional blocks required shielding from each other, especially the low-signal-level receiver.
With the basic test equipment up and running, the happy engineer needed a lossy copper transmission medium. He scrounged a pair of 1000 foot reels of RG-62 coaxial cable from another team leftover from a previous project (real engineers NEVER throw anything away). Ugly stuff, wrong impedance (93 ohms), poor single braid shield, but good enough to provide the attenuation for initial tests. At this point in time Category 5 twisted pair copper did not exist.
The lab setup is shown in Figures 1,2,and 3. Since this was an unofficial project it was not possible to create printed circuit boards, so the clock generators, data pattern generator, transmit driver, receiver and equalization, clock recovery, and sync/bit error pattern comparison circuits had to be breadboarded by hand on copper-clad. This was a blessing since circuit modifications could easily be made on the fly to get everything to work together. The engineer was finally having a lot of fun and was no longer bored. The hardware cost was relatively cheap, and the engineer’s time would otherwise have been wasted anyway.
Figure 1. The “Can Lan,” a home-brewed 125 and 250 Mb/s bit error rate test system for coaxial cable
Figure 2: The data pattern generator inside the can.
Figure 3: The primary clock generator inside the can. See circuit detail.
The results were quite surprising and encouraging. Even at 2000 feet of RG-62 quarter-inch coaxial cable the received equalized eye pattern at 125 Mb/s was very nice. At 1000 feet of RG-62 the 250 Mb/s eye pattern was the same.
To get some more information as to the response of other types of coaxial cable the happy engineer called a local CATV company and asked if they had reels of long length trunk cable stockpiled in their backyard. They did, and were very willing to install connectors on these reels to take part in a data transmission test. Some of the eye patterns from those measurements are shown in Figure 4. (The eye patterns are the AMI/NRZI technique described in EDN's Design Ideas.)
The CATV people were treated to lunch and beers for their generous assistance. Low budget, but the tests got done.
Figure 4. Eye patterns at 125 and 250 Mb/s on various coaxial cables. Polaroid Type 667 film.
Now about this time it started to appear that this might just be worthwhile investigating in the upcoming batch of new projects. The engineer’s immediate bosses agreed that the ability to transmit data at this rate on copper cable could be a serious competitor and price reduction over optical fiber for FDDI use, especially short runs. For want of a better name, the engineer called the idea “Coaxial Distributed Data Interface”, or CDDI.
But Marketing and Management were aghast! Why was Engineering fooling around with obsolete copper cable? Everybody knew that optical fiber was the LAN medium of the future! Nobody else was doing CDDI, so obviously not profitable. The engineer’s new data sampling technique was deemed non-patentable because it was not part of the company’s core business.
So now there was a conundrum within the company. Technical management felt that this concept should be pursued. MBA management felt it would lead to nothing. Some of their reasons were interesting and fairly logical, they felt that the industry perception of coaxial cable was that of “special” and “obsolete”. Existing unshielded twisted pair was not capable of this transmission rate due to crosstalk, and they were reluctant to develop a technology that required a “special” cable. At one meeting with Marketing the engineer was asked how long it would take to develop a product. He picked a number out of the air and said “Six months”. “Too long” said Marketing.
With no clear answer, the engineer was allowed to prepare a presentation detailing the test results to promote coaxial cable at the next FDDI X3T9.5 Conference to gauge industry response. Not much happened immediately after that, and the engineer was then assigned to another “me-too” 10 Mb/s Ethernet design. At least that was better than boredom.
Six months after this presentation a small company that continued the research came out with the first CDDI on twisted pair device using some additional and well-thought-out innovations. They were later bought out by a much larger networking company. The concept of CDDI (now evolved to “Copper Data Distributed Interface”) avalanched and was developed into a new PMD standard. Category 5 UTP cable was developed to handle the increased bandwidth. The successful CDDI technology was re-used for 100Base-T Ethernet.
And the company that the bored engineer once worked for? It went bankrupt.
Author Glen Chenier is a design consultant based in Allen TX