Layer 1 — Wikipedia: Physical Layer

The Physical networking layer contains specifications for both the actual physical media, and the means to transmit data on that media.

Physical Media

That said, the choices of the media itself is another layer of almost mind-boggling complexity yet again, however there is no official dogma defining this. Basically, if you choose a physical media there is a Layer 1 protocol to use it.

I don’t mean to scare you, but some of the choices are:

Fiber:

glass or plastic
single mode or multi-mode
core diameter 50 or 62.5 microns
shape of the connectors

Copper:

solid or stranded core
riser cable
Cat 3, Cat 5, Cat 5e or Cat 6
shape of the connectors

Wireless:

antenna shape and size
frequency
laser or radio

And those are just the choices of your physical medium. Luckily, once you have selected a physical medium you really don’t have any choice over the Layer 1 signaling protocol you will use; unless you’re doing carrier grade stuff in which case you shouldn’t be reading this blog — go back to work!

Some network consultants, particularly those who focus on wireless or fiber deployments (both of which depend heavily on the physics of wave theory) will spend a lot of time in this layer. This layer is mathematically heavy, and could just as easily be (and probably is) populated by PhDs, instead of regular humans.

The rest of us (like me) will be happy knowing that we can just change a fiber patch cable, or swap out a transceiver. For these elements, you need to know the limitations of the hardware you’re working with.

For example, here is a handy document on fiber transceivers that shows the frequencies used, the fiber required and the distances that can be transmitted for each grade of transceiver.

http://www.cisco.com/en/US/docs/routers/7200/install_and_upgrade/gbic_sfp_modules_install/5067g.html

Logical Protocols

What you should know is that these protocols define very important elements of Ethernet behavior, at a basic level the signaling frequency, and extending so far as defining the collision detection from when we used to use hubs instead of switches.  These days, you’re mostly going to see auto-negotiation and full/half duplex protocols.

The auto negotiation protocols are used to determine the speed, and duplex of an ethernet connection. Gigabit ethernet interfaces always use auto negotiation, while 10/100 interfaces can be configured manually, or use auto negotiation. Unfortunately, sometimes auto negotiation can occasionally fail presenting partial or complete network failures. Because of this, auto negotiation can present a risk to critical network infrastructure links.

In general it is a good idea to follow these practices.  Never use auto-negotiation for infrastructure links like servers, switch-to-switch, or switch-to-router links; infrastructure physical connections almost never change and the risk of failure is high due to the critical nature of the devices.  Use auto negotiation for user ports because user ports change frequently with laptops and desktops, and the risk of failure is low as only a single user port is affected.

The 802.3ab (1000BaseT) standard does require auto negotiation for gigabit connections, but some hardware (including Cisco) will allow you to set these parameters manually.

Duplex — Half and Full

Half duplex in any medium, implies that only one device may send at any given moment; effectively the network must take turns. This mode of operation comes from the early evolution of ethernet, when networks operated on hubs with a wide collision domain. These days we are using switches which limit collision domains to individual segments, and transmission protocols that send and receive at the same time (using a copper pair to send and a pair to receive).

Today, network consultants rarely find a system intentionally configured for half-duplex — but some WAN (wide area network) services are provided at half-duplex as a way to limit bandwidth. More often, a half-duplex connection is found as a result of failed auto negotiation — the problems happen when this causes a duplex mis-match.

Duplex is a negotiated protocol, and the default behaviour of a Cisco device if negotiation fails, is to set half-duplex; the Cisco device is assuming the the device on the other side is a hub, the lowest common denominator in networking.  Negotiation can fail if one side is configured manually, and the other side is configured for auto-detection, or sometimes different vendors are unable to negotiate properly due to interoperability bugs.

A half-duplex connection expects to never receive while it is sending data, and interprets this as a collision. When a half-duplex interface detects a collision, it stops sending and waits for a silent moment in which to send again. The problem with a mis-match, is that the other side is set at full-duplex, which means it is able to send and receive at the same time, and will continue to send traffic which overwhelms the half-duplex side of the segment. Ultimately, a duplex mis-match will degrade network performance — if this happens on a WAN or some other infrastructure segment it could cause serious network performance problems.

Speed

Ethernet connections often may be configured to run at 10, 100, or 1000 mbps — and auto negotiation may set an interface at any one of these speeds. Speed is either negotiated, or it can be detected by the pulses on the line.  A speed mismatch only happens when two devices are manually configured for different speeds.

One Response to “The Physical Layer — Layer 1”

1. The Physical Layer – Layer 1 « WOZNEY enterprises ltd says:

   April 21, 2010 at 10:44 am

   [...] And here is a link to the original article that you were looking for! http://wozney.ca/2008/05/16/physical-layer/ [...]