History shows that plain telephony cabling was a simple tree-and-branch affair. This used a building distribution frame adjacent to the private branch exchange (pbx), with 100 or 200 pair cables leading to patching cabinets and box connections, from which individual outlets were cabled.
The technology was simple. The highest frequency used was 3·5 kHz and virtually all of the flexibility points were insulation displacement connection punchdown modules with jumper wires. Many different proprietary computer networks grew up alongside this, like the IBM ‘pair of co-axials’ for visual display units.
Ethernet eventually emerged, enabling different manufacturers’ equipment to communicate. This was initially implemented in ThickNet, using a 0·5 inch-thick co-axial cable with tap-offs for individual pcs. ThinNet followed, which used a standard thickness co-axial cable in a daisy-chain format, where unplugging one pc took down the whole network!
The major breakthrough was to get 10 Mbps over untwisted pair cabling (utp) – Category 3 cabling was born. It wasn’t long before 100 Mbps Ethernet arrived and a better data-grade utp was required – Category 5.
For a long time the concept of flood wiring ie a double or quad outlet for every potential work area had been applied by telecoms managers. This gave spare capacity and inherent flexibility at three or four points in the extension wiring.
The same concept began to be applied to data cabling – the invention of structured cabling – but there were significant differences. Voice is low frequency and will happily run for half a mile or more. Ethernet, at 10 Mbps or 100 Mbps, was limited to approximately 100 m, so network managers had to deploy comms rooms containing active equipment at convenient points around the network. These were connected back to the main comms room by higher capacity links or backbones, usually optical fibre.
Flexibility was a major requirement in these comms rooms. Unlike telecoms, where jumpering was the norm, IT staff decided to go for an ‘anyone can do it’ approach and chose RJ-45 patching instead of a jumpered solution.
At this time patch panels and Category 5 cabling were expensive so not all data cabling was structured. Also, voice cabling came under the telecoms manager and Ethernet under the IT department, so there was little thought of convergence.
The move to integration
As the number of pcs in enterprises grew to one per desk, the use of structured cabling, with its many Category 5 RJ45 patch panels, grew in large and mid-sized businesses. In larger installations, economics dictated that conventional CW1308 and jumper frames/box connections were installed for voice.
Users and consultants began to think that running two complete sets of flood-wired cabling to every desk was inefficient. They therefore started to deploy what we would now call integrated structured cabling systems.
Integrated cabling systems use components of the highest common denominator, for example Category 5e or 6, as required by the data applications. The lower grade analogue or digital voice is then delivered through the same structured system.
Many systems now integrate voice and data over the same structured cabling system (see figure 1, over). From a technical standpoint, digital and analogue voice will work equally well over CW1308, Category 3, 5, 5e or 6, so the performance level of the cabling system is dictated purely by the data applications required.
If a conventional pbx system is used, it is standard to cable using CW1308 multi-pair from the pbx through a building distribution frame/test jack frame or directly onto Category 5e or 6 patch panels mounted in standard 19 inch cabinets or on advanced patching frames.
In the backbone, Ethernet data requires active equipment less than 100 m from the user. It tends to use fibre optic backbones, which aggregate traffic onto high speed links. For voice it is simpler and cheaper to run multi-pair cables to the local comms closet.
The only area where patch panels differ from the conventional Category 5e/6 is when integrated services digital network (isdn) devices are being deployed, either as direct exchange lines or behind an ISDN pbx. Here special patch panels can be installed, allowing star-connected devices to appear bus-connected to the pbx or ISDN2 terminal adaptor.
So far so good. Voice through the structured cabling system seems easy – except for the fact that most phones have BT phone plugs (type 431/631) and all structured cabling ends in RJ45 connectors.
Outlet converters are needed to adapt from the RJ45 jack to the BT phone socket (except for the ISDN converter). There are four different categories of converter available: PABX master, full master, secondary and ISDN. Some pbxs need special variants and most firms produce outlet converters and flying-leads.
What is VoIP?
In conventional pbx systems the voice is digitised by a phone then transmitted over a low speed (16-64 kbps) dedicated synchronous data channel. With VoIP, voice is digitised, then packetised – just like any other Ethernet data.
If data packets take 2 or 150 ms to go from pc to server it doesn’t matter. If they get to the other end in the wrong order it doesn’t matter either, because Ethernet will wait until its got all of the data then pass it on to the software for action.
Imagine doing the same to voice. It matters greatly if packets are lost and have to be resent, delaying the speed or making it gappy. Humans have a very poor forgiveness quotient if there are delays in sound or speech. In fact, everyday Ethernet isn’t good enough for VoIP because it is a best efforts protocol not a managed one.
VoIP needs to be switched as Ethernet packets through Level 2 Ethernet switches, which most enterprise local area network (lan) switches are. Voice also needs to be given priority over ordinary data using quality of service-managed Level 2 Ethernet switches, which most enterprise switches aren’t.
In VoIP, the voice-as-packet data doesn’t need to go through a pbx at all as long as there is a quality of service-managed Level 2 Ethernet switch in the lan. However, two pbx-like devices are still necessary. One to interface with external exchange lines (the gateway), the other to provide directory lookup and special feature management services (the gatekeeper).
Some pbx manufacturers are currently packaging a gateway and gatekeeper into a VoIP pbx as a halfway house, enabling simple deployment of VoIP in lans where the legacy switching isn’t yet up to the job. Some of these VoIP pbxs also include conventional time diversion multiplexing switching and line cards, which allows a gradual changeover throughout the enterprise.
VoIP convergence – stage 1
The VoIP pbx must interface directly with a quality of service-managed Level 2 Ethernet switch (see figure 2, above). Currently, the managed switch is often used in an overlay lan with separate cables and outlets fed to each workstation for phone and pc. Analogue connections are still needed for faxes, modems, etc.
The effects of VoIP on patch panels and cabling are minimal from a technical perspective. However, both data and VoIP are Ethernet and so require at least Category 5e performance throughout.
Practically, however, up to 5000 VoIP phones theoretically need only a single Category 5e connection to the switch, reducing, to some extent, the number of patch panels needed. However the issue of 5000 phones being dependent on a single patch cord raises significant resilience and reliability issues.
VoIP phones, being Ethernet devices, come with an RJ45 plug and don’t need outlet converters.
VoIP convergence – stage 2
The next interesting idea is that a VoIP phone and a pc could share a single Category 5e cable/outlet (see figure 3, opposite).
There will undoubtedly be computer telephony integration applications for this where phone and pc closely interact, such as in call centers. However, I for one would not want to connect something highly reliable like a phone to something like a pc which can be prone to crashing several times a day.
It would be a very brave network manager who jeopardized the current high levels of telephone service quality for small potential savings on the cabling element of their VoIP installation costs. Especially as this would only save 1-2% of the overall network cost.
From a practical point of view, if they’re connected together there are major demarcation and test issues when fault finding. You would either have to accept additional down time or have a spare adjacent outlet so they could be segregated while fault finding, leading to a similar level of outlet provision.
The cabling conclusion
VoIP is progressing, but it’s currently far from proven or totally stable. High levels of quality of service-managed traffic on busy or heavily loaded lans can be a significant problem. Voice quality is still an issue and, according to Gartner’s Geoff Johnson (PC World, 2 September 2002), most large enterprises will have a higher total cost of ownership for VoIP compared to conventional pbx over the next three years. “This is down to uncertain integration costs,” he said.
VoIP certainly needs to stand on its own based on lifetime costs, and not be justified on small savings on cabling installation to make its business case.
VoIP will have little effect on the cabling market over the next three or four years. Demarcation and flexibility points are still needed – and Category 5e/6 patch panels are the ideal solution.
VoIP technology is still too much in its infancy for phone and pc over a single connection to be a credible option from a reliability perspective – especially given the small savings from throwing away this level of resilience. The VoIP lobby has lessons to learn from the conventional telephony players in terms of reliability and stability.
All structured cabling manufacturers have a vested interest in advising users not to cut down their horizontal cabling and number of outlets. But it also makes good technical and operational sense to mitigate risk and have redundancy/resilience in the horizontal as well as the backbone.
Source
Electrical and Mechanical Contractor
Postscript
Kevin Anderson is product manager (copper) with Krone.
No comments yet