The networking industry is currently in the midst of the next major LAN upgrade cycle. Enterprise networks are upgrading from 10/100 Fast Ethernet networks to Gigabit Ethernet networks. One of the key features of Ethernet technology that has led to mass deployment is the "plug-and-play" aspect of installing and configuring the network. The legacy of Ethernet is that when end-users or IT personnel purchase and install Ethernet equipment, it simply works properly and transparently.
By ensuring robust operation over a wide variety of cable types and installations, the physical layer (PHY) portion of the system has played an integral part in delivering a "plug-and-play" Ethernet network. However, in the event that faults exist in the cable plant, such as cable discontinuities (opens) or shorts, a network connection cannot occur. In this scenario, it is critical that the source of the cable fault is quickly found and corrected as network downtime translates to lost productivity and/or revenue for corporations.
In this article, we'll explore a new technology for detecting faults on cable lines that is built directly into the Ethernet PHY. Through this technology, dubbed Virtual Cable Tester (VCT), designers can build NIC cards and switches that pinpoint problems by remotely analyzing the quality and attributes of the attached cable plant. Let's take a deeper look at how this technology works.
TDR Provides the Key
VCT technology uses time domain reflectometry (TDR) to diagnose the attached cable plant. Similar to the principle of radar, TDR is the analysis of a conductor by sending a pulsed signal into the conductor, and then examining the reflection of that pulse. When the transmitted pulse reaches the end of the cable, or a fault along the cable, part or all of the pulse energy is reflected back to the source. The VCT algorithm measures the time it takes for the signal to travel down the cable, see the problem and reflect back. This measured time is converted to distance and made available through internal registers in the Ethernet PHY.
Figure 1 shows some example waveforms for various conditions of an open cable. Referring to the waveforms in Test 1 (100 meter open cable), the first waveform shown is called the "source waveform." The second waveform shown is called the "reflected waveform," and, as its name implies, it is the reflection of the source waveform after it has traversed the cable, reflected at the end of the open cable and returned. The "smoothing" of the reflected waveform is due to the low-pass filter characteristics of CAT5 cable.
Example waveforms for an open cable (disconnected cable).
As shown in Tests 2 and 3, the reflected waveform is closer to the source waveform as the round-trip propagation distance is shorter. Tests 3 and 4 show that for very short cable opens, the energy of the reflected waveform is "added" to the source waveform. Essentially the round-trip propagation is shorter than the width of the source waveform. VCT technology is very precise and can measure these slight propagation delays, and hence the distance to the fault (or open in this case).
Dealing with Impedance Problems
Any time two metallic conductors are placed close together, they form cable impedance. A correctly terminated line is defined as a line or cable with impedance that is equal to the source's impedance as well as the impedance of the load.
For a perfectly terminated line, the reflected waveform is zero. In this case, the load absorbs all of the energy of the source waveform. When the cable is disconnected (or open) at the far end, the load impedance is infinite and the reflected waveform is equal to the source waveform. The following equation defines this dynamic further; it is a calculation of the reflection coefficient, π:
where ZL is the load impedance, Z0 is the cable impedance, and the impedance of CAT5 cable is 100 ohms (see Figure 2).
Figure 2: Source and reflected waveforms.
Table 1 shows the Reflection Coefficient for several load conditions.
Table 1: Calculated Reflection Coefficient for Various Load Impedances
|Load Impedance, ZL (Ohms)ink
||Reflection Coefficient (pL)
Several observations may be made from the above data:
- When the load impedance is greater than the cable impedance, a positive reflection results, and conversely, when the load impedance is less than the cable impedance, a negative reflection occurs (i.e., the reflected pulse is of a magnitude below zero). VCT technology uses this information to help determine the load impedance.
- When the load impedance is 300 ohms, the reflection coefficient is 0.5, which implies that the reflected waveform is one-half the magnitude of the source waveform. VCT technology uses the polarity and magnitude of the reflected waveform to precisely calculate and report the load impedance. Figure 3 shows the examples of a short-circuit (zero ohms load impedance) and a 50 ohm load impedance.
- When the load impedance is 100 ohms, the reflection coefficient is zero, which implies that the load absorbs 100% of the energy of the source waveform, and there is no reflected waveform. In the absence of a reflected waveform, the VCT algorithm knows that no cable faults exist.
Figure 3: Example waveforms showing negative reflections.
The velocity of propagation (VOP) is a specification of the cable indicating the speed at which a signal travels down the cable. Different cables have different VOPs. The VOP is defined relative to the speed of light in a vacuum, which is 186,400 miles per second. The speed of light has a VOP of one (1) all other signals are slower. For example, a cable with a VOP of 0.71 would transmit a signal at a velocity of 71% the speed of light. CAT5 twisted-pair cabling has a VOP of 0.71, which translates to a propagation delay of 4.7 ns/meter.
Using TDR technology to measure the propagation delay of the reflected waveform, and by knowing the calculated VOP of CAT5 cable, it becomes a relatively simple exercise to calculate the length of the cable, or the distance to the cable fault.
Applying it to GigE
The IEEE 802.3ab Gigabit Ethernet standard represents truly extraordinary technology in that the standard defines 1000 Mbit/s data transmission over the same copper media defined for Fast Ethernet at 100 Mbit/s. Thus, Gigabit Ethernet streams will travel over a cat5 unshielded twisted pair (UTP) cable that contains four twisted-pair lines or eight conductors in total.
An important distinction, however, is the fact that Gigabit Ethernet transmission requires four twisted-pairs where Fast Ethernet transmission requires only two twisted-pairs. Since Fast Ethernet systems only require two twisted-pairs, the third and fourth pairs of the CAT5 cable have, more than likely, been dormant for years. Thus, there is a probability that these pairs will require some attention from IT personnel. Gigabit systems integrating VCT technology will quickly notify the user or IT manager of such faults for rapid correction.
As VCT technology is integrated into the PHY silicon of the network equipment, it enables the IT manager or end-user to non-intrusively diagnose the attached cable plant, quickly identifying the failing mechanism and isolating the source of the problem. Essentially, VCT technology integrates, on a single chip, the functionality of a several thousand-dollar cable meter system typically required by IT professionals to handle corporate network service and support issues. Moreover, with integrated VCT technology each device port can independently detect and report cabling issues without the need to unplug cables, connect cable testers and install loop-back modules at the far end all of which are necessary when using traditional cable meters.
About the Author
Jason Knickerbocker is a product manager at Marvell Semiconductor. Jason holds a BSEE from the University of California, San Diego and can be reached at firstname.lastname@example.org.