Ethernet Design – Rules of Engagement
Ethernet Design – another trip back to the Eighties
I was about to post a blog about the Internet of Things during my research I stumbled upon an article called ‘Remembering… 10BASE2, Coax and BNC’ by David Hayward reminiscing about his time at Wakefield College. As this follows on from my last blog where I mentioned some of the different cabling technologies we worked with over the past 30 years I thought I would discuss a few of the points raised here. I really enjoyed the article in MicroMart (issue 1340) as it was a trip down memory lane for me. The Astro cabling team installed hundreds of 10BASE2 and 10BASE5 networks around the UK for the likes of Central Electricity Generating Board (and later National Grid and PowerGen) and Bourne Leisure. I have also developed and ran many training courses teaching technicians and engineers how to design, install and troubleshoot thick and thin Ethernet installations. Although this all relates to a bygone age the principles behind the content – in particular – following design rules, are still relevant today.
The designation of 10BASE5 and 10BASE2 came from 10Mbps network speed using Baseband signalling over a distance of 500 metres and 200 metres of coaxial respectively. However, the 10BASE2 was an estimation as the actual maximum cable length was 185 metres rather than 200.
10BASE2 used a thick and unwieldy RG8 type coaxial cable specifically designed for 10BASE5 and it was commonly referred to as ‘thick or thick-wire Ethernet’ when it’s younger sibling 10BASE2 was introduced. 10BASE2 used a specifically designed RG58 type cable with foil and braid screen and PTFE dielectric and was referred to as ‘thin or thin-wire Ethernet’. Both supported IEEE 802.3 and Ethernet variants and in any case most people referred to both protocols interchangeably.
”Ethernet devices were attached by vampire connectors that clamped onto and drilled into the cable…”
Coaxial Ethernet was a bus system where devices were attached along the length of the cable. For thick-wire Ethernet devices were attached by vampire connectors that clamped onto and drilled into the cable at 2.5m or multiples of 2.5m spacing. In effect the cable was a single length of continuous cable terminated in 50 ohms at each end with a number of attached devices along the length of the cable. A thick-wire segment or length of cable up to a maximum length of 500m could have 100 devices attached. Thin-wire Ethernet worked on the same principle but devices had to have a minimum of 0.5m spacing with a maximum of 30 devices per segment or length of cable up to a maximum of 185m.
All IEEE 802.3 standards have strict rules to ensure the protocol functions as intended. In Mr Hayward’s article he says 10BASE2 was ‘simply amazing’ implying that you could almost throw it in and even coil it around power lines and it would still give good service. However, later in the article he says that he never managed to achieve maximum length networks. My experience could not have been further from this, when Ethernet design rules were adhered to there were no problems running the cables to their limits.
”…running copper data cables in parallel alongside power cables has never been acceptable practice.”
Some of the installations I worked on needed every last metre we could get out of the cable. I always stuck to the rules rigidly and had many successful and reliable installations at maximum length. In telecommunications some design rules can be broken but you really have to know what you are doing to ensure the network will operate as intended. However, some rules simply must not be broken. For example, running copper data cables in parallel alongside power cables has never been acceptable practice. Even shielded cable is not immune to interference when the shield conductor resides inside the electromagnetic field surrounding the length of a power cable. An appropriate distance should always be maintained.
During the early years of coaxial Ethernet systems I was engaged to troubleshoot many failing networks. My preferred diagnosis tool back then was a Time Domain Reflectometer with an oscilloscope type display that enabled me to examine the length of cable from either end. From one end of the cable I could see along the length of the cable to see where all of the transceivers were connected whether the cable was terminated with the correct value terminator and calculate reasonably accurately the distance along the cable to any damage.
“…I still find myself thinking of the coaxial Ethernet of days of old affectionately. But why?”
Over 80% of the troubleshooting engagements were due to Ethernet design rules not being met. The problems we faced were varied but the most common problems being over-length cables, more segments than the design rules allowed and non-IEEE specified RG8 or RG58 cabling. On this last point, IEEE cable was a very high quality specification with high velocity of propagation and 100% foil and braid screens and PTFE dielectric. Standard grade RG8 and RG58 cabling with lower velocity of propagation and partial screening were certainly one of the reasons we experienced where networks would not operate at full length or with the maximum number of devices connected. Mr Hayward mentions cheap cable and connectors that were easier to terminate that unshielded twisted pair cabling used today which makes me think the college were using the standard cables and connectors rather than the specific IEEE specified components.
My memories of terminating thin Ethernet are not fond ones and my preference is to terminate UTP. If my memory serves me well field termination was a fiddly job. We also came across many poor quality field terminations causing short circuits and intermittent disconnections even with the correct cable and connectors. Having said this I still find myself thinking of the coaxial Ethernet of days of old affectionately. But why? The simple answer is, I have absolutely no idea. Perhaps because it was the pioneering days of Local Area Networking and credit where credit is due it was a great enabler back then
”Over 80% of the troubleshooting engagements were due to Ethernet design rules not being met…”
The missing terminator?
As Mr Hayward mentioned in his article, the cable terminators at the end of the segments used to go missing with alarming regularity. I remember one occasion where we sent an engineer down to West Sussex to investigate a network failure only to find the office manager had thrown the terminator in the bin because he “didn’t like the look of it” on the back of his PC. I also remember flying an engineer up to Scotland on New Year’s Eve to investigate a network failure and we lost him for three days as he needed time to recover from the Hogmanay party. On that occasion some electricians had managed to connect the mains live to earth and all of the radiators and PC chassis were live. Ouch!
So what has all this got to do with today’s networks?
Ever since the International Organisation for Standards (ISO) introduced the Open Systems Internet Connection (OSI) model I have used this as a tool to teach troubleshooting techniques. The whole communications stack rests on the physical layer where the cabling resides. The principles still apply today in that our physical cabling must adhere to design rules. Given that our data cabling is now passing significantly higher data rates than their coaxial ancestors it is as important as ever to ensure the rules are adhered to. Not only do the design rules need to be adhered to but the quality of the installation must also be to a professional standard.
There has always been a tendency to disregard the cabling as being a simple but costly part of the network infrastructure. However, if there are even slight failings in the cabling the physical layer will not support the layers above and performance and network integrity will suffer. After more than 30 years of Ethernet installations we still occasionally see sub-standard cabling causing serious business and operational problems.
”…it is as important as ever to ensure the rules are adhered to.”
So whether your requirement is for Cat5e, Cat6, Cat6A and any optical fibre cabling it is essential to ensure the design rules for the cabling system are adhered to. This includes any testing procedures and documentation. There is also some not so tenuous links here to the Internet of Things as many of the networks we were involved with were telemetry and control monitoring systems. Early machine to machine communications that some, including myself, regard as ancestors of the Internet of Things. But, more on that in the next blog.