A BERK-TEK WHITE PAPER |

Maximizing Cable Performance for Voice, Data and Power in the Real World

PERFORM BEYOND EXPECTATIONS

In the past decade, the Information Communications Technology (ICT) industry has undergone a continuous metamorphosis of emerging applications, better known as “IP Convergence.” Caught in this wake has been the structured cabling system that continues to be tasked as the vehicle to transport and connect these many applications onto the network – from voice to data to power.

Voice and data networks employing structured cabling emerged in the late 1990’s and early 2000’s, as copper cable standards were being defined and ratified. At that time, data and voice communications employed similar twisted pair cabling and associated connectivity, but ran as separate parallel cabling systems. Simple voice traffic was transmitted over plain old telephone systems (POTS) that used Category 3 cabling. Data applications ran over Category 5, 5e then 6 cable. Today data and voice use the same Category 5e or Category 6 cabling system, and the network has evolved to support more complex applications.

Transmitting “data” has transitioned from transferring simple text files to now include large, graphically intense documents and files, as well as transmitting data packets in the form of Voice over IP (VoIP) and streaming video such as for IP cameras. In addition, more and more once disparate applications are becoming IP enabled and the data streaming over the network encompasses complex surveillance systems, access control and building automation applications. Many of these devices also employ Power over Ethernet (PoE) to provide electrical power to the device using the same structured cabling that enables data connections, eliminating the need for separate AC outlets.

This article explores the real world challenges now taxing copper-based systems as applications have evolved from simple voice and data to converged IP applications. Backed by lab testing, but emulating the real rigors of the installed cable plant, the Nexans Data Communications Competence Center (DCCC) has developed a new test that closely replicates the challenges faced by current and future converging networks. The test results provide insight into how cables will perform when faced with the challenges of simultaneous transmission of voice, data and power in stressed environments with stressors including fluctuating adjacent power and variable operating temperatures.

Page 2

PERFORM BEYOND EXPECTATIONS

Figure 1: Defining Bandwidth Size

CHALLENGE #1: BANDWIDTH GROWTH

According to Cisco, bandwidth requirements are expected to increase three-fold from 2012 to 2017. For business networks, there will be a need to transmit and store 22 exabytes worth of data on a monthly basis by 2017. To better understand the true scale of these predictions, 50,000 years worth of high-quality video would fit within a one exabyte hard drive. So, that the predicted 22 exabytes worth of information is equivalent to filling five and a half billion DVDs worth of information, monthly.

In the physical layer, data is actually transported in bits, where 8 bits make up 1 byte. When data reaches the data layer link, these bits are assembled into frames containing between512 and 12,144 bits (or 1,518 bytes). These frames, also referred to as “packets” (see fig. 2), as defined by IEEE 802.3. Once these frames are constructed, source and destination IP addresses are attached to enable transport in the network layer.

Ethernet frames are then transmitted using one of a variety of networking protocols, the most common of which are TCP/IP (Transmission Control Protocol/Internet Protocol) and UDP (User Datagram Protocol). Each protocol has different strengths and weaknesses and the choice of which to employ is dependent upon the requirements of the service being supported.

TCP/IP includes error-correction mechanisms that allow for retransmission of data should frames become lost in transport or arrive at their destination address damaged or corrupted in some way. Damaged or dropped frames can be resent until they are received and understood by the receiver. During this process however, the network becomes filled with “re-send” transmissions along with new transmissions and network throughput rate drops significantly. This protocol provides a high-reliability solution that all packets will ultimately arrive in the manner in which they were sent.

In contrast to the high-reliability and time-insensitive nature of TCP/IP, UDP prioritizes timeliness of delivery, potentially sacrificing some degree of fidelity. UDP does not include error-correction, meaning there is no re-transmit of data, even if there are errors in the initial transmission. UDP is frequently used to support voice and video services because the lag caused by time-insensitive TCP/IP transmission would be more detrimental to the provided service than the issues potentially cause by dropped or corrupted frames. However, data corruption in UDP transmissions does have an impact on the quality of the provided service, like causing the choppiness of broken words or sentences in a VoIP application for example.

Name Numerical Description Kilobyte 1000 Thousand Bytes Megabyte 1000² Million Bytes Gigabyte 1000³ Billion Bytes Terabyte 1000⁴ Trillion Bytes Petabyte 1000⁵ Quadrillion Bytes Exabyte 1000⁶ Quintillion Bytes Zettabyte 1000⁷ Sextillion Bytes Yottabyte 1000⁸ Septillion Bytes

Figure 2: Standard Ethernet Frame

Page 3

PERFORM BEYOND EXPECTATIONS

To better manage the increasing volume of data, advances are being made in data compression technology. But these advances in compression can magnify the effect of lost or retransmitted frames because so much more is packed into a frame, as illustrated in Figure 3. Most Ethernet traffic today (mainly video and data) is compressed in some way by using various compression codecs. The tradeoff is that much more information is lost if there are errors and a frame is dropped. Just looking at what is lost if a single frame is dropped or corrupted also misses the potential overall impact since some individual frames that are transmitted correctly will not operate correctly without their preceding or following frames depending on the compression technology used. As a result, frame errors can have devastating consequences on current and future networks, especially as compression technology continues to advance.

CHALLENGE #2: POWER OVER ETHERNET (PoE)

There are two main external inhibitors that influence a cable’s performance – heat and noise. Heat is generated within a cable when power is transmitted via the conductors of a category cable. Heat generation is a function of the amount of power consumed by the end device and the cable resistance. As more current is supplied to the end device, temperatures rise, thereby increasing attenuation and reducing the network signal to noise ratio. This makes voice and data traffic more susceptible to frame errors.

The DC power supplies used in PoE can and do generate noise, especially as they age. Sudden changes in current, such as when a device first powers-up, servo movement, or a fan cycling will generate electromagnetic fields which can affect voice and data traffic if they are not isolated and protected from these fields. Additionally, environmental noise sources from non-network devices can vary and can be transmitted by nearby power cables, office equipment, elevators or even fluorescent light ballasts, and disrupt network performance.

Future PoE standards will allow more devices to operate using PoE technology by increasing the amount of supplied power. PoE is a simple, convenient and financially rewarding way of powering devices. Deploying PoE to power IP devices costs only a fraction of what it would cost to run separate electrical outlets to power the same devices. A new PoE standard is in development now with the IEEE (as of 2014). The goal is to significantly increase the available power over structured cabling systems. See

Figure 3: Increased Data Density Due to Compression

Page 4

PERFORM BEYOND EXPECTATIONS

Figure 4: IEEE 802.3; Four Pair Power over Ethernet; Call for Interest March 2013

figure 4 for IP enabled devices requiring more power than is currently available. A new PoE technology of at least 50W would allow many more devices to be powered using this technology.

Berk-Tek representatives are contributing to the IEEE development of this Next Generation IEEE PoE standard. Current projections indicate that this standard will most-likely be at the 60W level to then be followed by a 100W version.

The potential addressable market for a Next Generation PoE is shown in figure 5 below. Research indicates a current demand for 20 million ports, should this technology be available today. That potential is projected to more than double to 45 million ports by 2018. Currently available PoE and PoE+ supported devices are not included in these projections.

The development of this Next Generation PoE provides a good deal of opportunities for increased network convergence and simplification of management. However, it is important to not let these future benefits distract from the potential implications of increasing power as a disruptor of network performance. It is also vital to remember that as the current supplied using PoE increases, the amount of heat generated by the cable itself will also increase. In turn, this increased heat has the potential to negatively influence network performance, so care must be taken to ensure that cabling infrastructure installed today has the ability to manage this potential future challenge.

System Market Power Nurse Call Systems Healthcare 30-50W Point of Sale Retail 30-60W IP Turrets (Trading Floor Phones)

Financial/Retail 45W

Lighting, HVAC, Access Control

Building Management

40-50W

IP Cameras (PTZ, Heaters) Security 30-60W Video Conferencing Hospitality 45-60W Motor Control Unit, Drives

Industrial >30W

Figure 5: Forecasted Growth of Proposed PoE++ (IEEE 802.3 Call for Interest)

Page 5

PERFORM BEYOND EXPECTATIONS

TESTING DESIGN: DEFINING THE PROPER MEASURE OF THE NETWORK

The ANSI/TIA-568-C.2 Standard defines both the performance measures that must be met by cables and the testing mechanism and configuration to be employed when testing for those parameters. Category 5e and 6 cables are built to meet a 100-meter, 4-connector channel test configuration as shown in figure 7. The tests measure specific parameters (crosstalk, return loss, insertion loss, DC impedance, etc.) defined by the electrical characteristics that affect data transmission. The test configuration does not address simultaneous operation of voice, data, power or any other IP application.

Additionally, testing is generally performed in the controlled environment of a third-party lab. This controlled environment helps create a level playing field and allows for consistent testing, both between sample sets and between various manufacturers. This ensures that customers have access to “apples-to-apples” data for evaluating their product and system options. Tests are usually “one-offs,” neither repeated tests with a single cable nor multiple tests with a variety of similar but unique samples. This single test approach provides a very clear indication of how a specific cable or system works at a given moment in time, but eliminates the possibility of establishing an average performance value that can only be realized after multiple tests. Finally, the current TIA test configuration does not have a mechanism for addressing the impact of real world inhibitors such as additional internal and external noise and heat.

The existing test structure does have a place and provides a benchmark against which cable and system designs can be measured to establish whether or not a product meets the requirements to be recognized as category compliant. However, with the ever-increasing burdens and stressors that are being added to the structured cabling infrastructure, this single test configuration no longer provides sufficient evidence to conclude that a particular cable or system will be able to provide optimal capabilities for data transmission once installed and fully burdened.

To address this disconnect between the established testing parameters and what is truly necessary to support networks for the next several decades, the Nexans Data Communications Competence Center has developed a new test configuration and has broadened the evaluation of cabling systems beyond simple electrical transmission characteristics. This new paradigm provides a more robust and diverse set of evaluative mechanisms, painting a more vivid and detailed picture of system capabilities that is more suited to the varied and critical nature of the contemporary IT infrastructure. Equally as important, the new testing approach can be consistently applied, ensuring that the results produced are both accurate and valid.

The new test configuration can be seen below in figure 8. While superficially similar to the TIA test, the new approach employs simultaneous transmission of voice and data in conjunction with power. To emulate the applications, an Ixia data source generates both VoIP and data traffic and transmits the

Figure 7: Industry standard ANSI/TIA-568-C.2 Configuration Test Set-up for a 100-meter, 4-connector channel

Page 6

PERFORM BEYOND EXPECTATIONS

network traffic through VoIP phones on both ends of the channel for one hour. An endspan switch also injects PoE into the channel. This new test also layers the two most significant real-world inhibitors into the process, heat and noise. To address the impact of heat, it subjects 75 meters of the horizontal cable to elevated heat in a test chamber to simulate a plenum area. External noise is induced into the system to simulate EFT interference from outside noise sources such as copy machines, and fluorescent light ballasts, by running 15 meters of the horizontal cable adjacent to electrical cables being injected with 250V spikes.

Using this test set up, several application based tests were developed to determine the impact that these stressors have on the performance of a converging network in a real world environment. The tests measure:

• Frame Error Rates (FER) - a measure of Data quality • Mean Opinion Score (MOS) – a measure of VoIP quality • Media Loss Rate (MLR) - a measure of IPTV and video quality

Frame Error Rate

There are multiple ways in which the quality of networking products can be assessed. Two clearly defined techniques are the measurement of bit error rate (BER) and frame error rate (FER). Many networking products are specified with the bit error rate that they achieve or support. BER is simply the ratio of bits that have been incorrectly received to the total number of bits sent. The test is performed by sending strings of pseudo-randomly generated bits through the device.

Instead of providing the ratio of incorrect bits to total bits transmitted, Frame Error Rate testing provides the ratio of the number of incorrect frames to the total number of frames sent. Since frames are the basic data unit in real-world networks, they are a better indicator of network performance. If a switch or a NIC receives a frame with one or many incorrectly transmitted bits, the complete frame will be discarded. A frame could contain in excess of 12,000 bits. The degradation of network performance due to dropped frames is the same for one or several thousand bit errors within a frame. BER does not account for this fact.

FER testing takes longer to complete than BER testing. A typical FER test can be one hour in duration, whereas a BER test requires about two (2) minutes. This increased test time is significant because it

Figure 8: DCCC 100-meter, 4-connector test configuration for simultaneous applications with added noise & heat

Page 7

PERFORM BEYOND EXPECTATIONS

allows the hardware and cabling to reach true operating temperatures. The performance of networking products decreases as the temperature increases. Robust performance is therefore better verified during longer tests that allow for transceivers, switches cabling, etc., to reach their true operating temperature.

The impact of dropped frames varies depending upon Ethernet protocol under consideration. If using UDP protocol, for example in a VoIP call, dropped packets cause choppiness, like bad cell phone reception. If the TCP/IP protocol is being used, as in most data applications, a lost frame means the receiver asks the transmitter to “please re-transmit.“ If errors continue, the network can quickly fill with retransmit requests causing the network to slow to a crawl, and eventually time-out if the delays become excessive.

The Nexans Data Communications Competence Center conducted FER testing for multiple, one hour tests using the new test configuration and methodology (simultaneous applications with environmental stressors) for a variety of cable and connectivity configurations. Figure 9 presents the results, as follows:

• CAT 5e cables averaged over 5600 errors, • Minimally compliant CAT 6 cables averaged about 3000 errors, • Competitor 1 CAT 6 enhanced cable had almost 21,000 frame errors • Competitor 2 CAT 6 enhanced cable had over 11,000 frame errors • Competitor 3 CAT 6E (Enhanced or premium cable) had approximately 750 errors • LANmarkTM-1000 and LANmarkTM-2000 -- 39 and 14 errors respectively

Figure 9: Results of Frame Error Testing Over One Hour

Page 8

PERFORM BEYOND EXPECTATIONS

VoIP – Mean Opinion Score (MOS)

Voice is not considered high bandwidth, but it is very sensitive to electronic noise and heat, because both can cause errors and interfere with timing. When timing is disrupted, there are delays and the call can sound choppy with broken words and sentences. Voice over IP (VoIP) moves voice applications from conventional commercial telecommunications line configurations to data network lines. While some may consider VoIP an old technology, shipments of VoIP phones are expected to grow 35%, to 35 million units per year by 2018. New VoIP phones are incorporating larger LED screens for video conferencing, and requiring more power as a result. Additionally, they have also transitioned from 10/100 Mb/s to 1 Gb/s, thereby eliminating the bandwidth limitation they once imposed.

To test for voice quality in a VoIP system, the Data Communications Competence Center employed an established test known as Mean Opinion Score (MOS). This is a commonly used Quality of Service (QOS) measurement for voice with a rating system of 1 through 5, with 5 being the best. The original scoring system was based on user-evaluation of call quality, but today’s testing relies upon standardized software, ensuring consistent results. A score of 3.6 is generally considered a minimally acceptable level of call quality.

Commercially-available testers and software packages account for packet loss (missing words or phrases), latency (propagation delay from traveling through network), and jitter (delay from packet to packet). The Data Communications Competence Center tests for MOS using an Ixia IP Performance Tester under the conditions previously described.

Test results for a variety of cable and connectivity combinations can be seen in figure 11. The red line indicates a score of 3.6, which is considered the minimum call quality level that should be accepted, although in some organizations, a MOS of 4.0 is the minimum acceptable score. LANmark-1000 scored approximately 4.15 and LANmark-2000 scored about 4.4.

MOS Expected User Satisfaction 4.34 - 5.00 Very Satisfied 4.03 - 4.34 Satisfied 3.60 - 4.03 Some Users Dissatisfied 3.10 - 3.60 Many Users Dissatisfied 2.58 - 3.10 Nearly All Users Dissatisfied < 2.58 All Users Dissatisfied

Figure 10: MOS Measurement System

Figure 11: MOS Test Results for VoIP

Page 9

PERFORM BEYOND EXPECTATIONS

Media Loss Rate Testing For Video

The third and most complex application test is for IPTV, a video application. IPTV is an emerging technology used in many market segments and is defined as multimedia services, such as television/video/audio/text, etc., delivered over IP-based networks. IPTV allows content to be centrally controlled and managed over the network.

Media Loss Rate (MLR) is a way to measure the quality of IPTV. MLR is defined as the number of lost or out-of-order packets over a certain time frame. Because any frame error will affect video quality, the MLR should be zero. However, service providers can usually negotiate a maximum average MLR of around 0.004 per hour, which equates to about two packet errors per hour.

In Figure 12 are examples of MLR greater than (>) than 0 and an MLR equal to (=) 0 . Even if the photo on the left lasts for less than a second, if it was to happen repeatedly every hour, this would not be acceptable to the viewer.

Figure 13 shows the MLR test results using the new Competence Center test, incorporating IPTV transmission. Again, the testing was conducted for over one hour, and the results shown are the average of five tests.

An MLR greater than zero is not recommended as errors degrade the video stream, and like UDP, there is no re-transmit of data packets. The real world limit with standard definition is usually a maximum average MLR of about 2.0 errors per hour. The results show that many solutions had an MLR maximum of anywhere from 6.4 to 25.1 errors per hour. Berk-Tek LANmark-1000 and LANmark-2000 both had zero errors for every test.

MLR>0 MLR=0

Figure 13: Test Results for MLR for Video IPTV over One Hour

Figure 12: Media Loss Rate Effects at =0 and >0 - Actual screen captures from Nexans Data Communications Competence Center testing.

Page 10

PERFORM BEYOND EXPECTATIONS

When reviewing all these tests in combination, it becomes clear that LANmark-1000 and LANmark-2000 have consistently outperformed other products in all the application tests, including frame error tests, MOS scores and IPTV, while being tested under the rigors of real world environments. See Table x below for a review of these details. So what is it about our Category 6 products that allows them to perform so well in these stressed situations?

Comparing Category Cable Performance

Product MOS min*

Frame Errors

MLR Errors Tek-Twist Proprietary

Materials 116 M Channel**

Min 5e 3.46 5,683 7.7 NO NO NO LANmark-350TM 3.20 2,471 7.0 NO NO NO Min 6 3.48 2,919 5.7 NO NO NO Comp Enhanced 6 3.23 9,752 7.2 NO NO NO Comp Premium 6 3.68 748 25.1 NO NO NO LANmark TM -1000 4.14 39 0.0 YES YES NO LANmark TM -2000 4.37 14 0.0 YES YES YES

BUILDING A BETTER CABLE FOR TODAY AND TOMORROW

To address the issue of heat, the Berk-Tek’s Materials Engineering team developed proprietary insulating compounds, specially designed to protect and insulate the copper conductors from heat and maintain signal strength. These materials are referred to internally as “less lossy” compounds, because there is less signal loss (attenuation) when they are used compared to standard, off-the-shelf compounds used throughout the industry.

To address noise, Berk-Tek again developed something new. Through extensive R&D work by our Process and Design Engineering team and our IT group, Berk-Tek developed “Tek Twist Technology,” a new approach to suppress noise through proprietary algorithms that optimizes twist. These algorithms are a series of instructions that continually adjusts the twist rate of each pair to reach a desired level of optimal performance for voice, data and power, versus employing a static twist rate for each pair designed to maximize data only performance. Small adjustments like this can make a huge difference.

When only data was transmitted, the noise generated was mostly in the form of crosstalk (or internal noise interference), which was relatively small. Crosstalk was controlled by twisting each pair at different rates. With the increasing loads of more data, greater compression, more voice and especially more power, twisting to static twist rates is no longer enough. Tek-Twist Technology was developed to ensure our customers that their converging networks will continue to operate trouble-free, even as more power is transmitted through the cable.

Additionally, the Berk-Tek engineering teams developed a process control system that works in conjunction with Tek-Twist Technology to monitor the helical concentricity during the inline insulation process. Helical concentricity means that both insulated conductors are exactly the same as each other with the same diameters and with the conductors perfectly centered within the insulation. If a problem does occurs during the manufacturing process, it is immediately addressed and resolved, versus waiting until final product testing, or worse, when the product is installed in the customer’s network.

Figure 14: Cable Attribute Comparison

Page 11

PERFORM BEYOND EXPECTATIONS

Figure 15 below illustrates helical concentricity and why it is important. Diagram 1 illustrates a twisted pair with helical concentricity and Diagram 2 shows a twisted pair without helical concentricity. Both diagrams show a differential signal transmission with +/- voltages as used in Ethernet networks. In diagram 1 on the left, the strong helical concentricity allows the voice and data signals on each conductor to travel at exactly at the same speeds and if noise is introduced, it affects both “sides” (+/- voltage) of the transmitted signal in exactly the same way. When the signal reaches the receiver, the receiver can easily identify the noise, filter it out and understand the transmission. In Diagram 2, without helical concentricity (note the insulated conductors are not identical), the noise is affecting the negative side of the signal differently than the positive side. In this case, the receiver cannot identify and filter the noise, and therefore cannot understand the transmission. Although the receiver will request the transmitter to resend, in an effort to receive a transmission it can understand, this process can slow the network and eventually lead to time-out errors.

DIAGRAM 1 DIAGRAM 2

All of these materials, design and process elements are focused on addressing the challenges unique to today’s cabling infrastructure. By reviewing and optimizing the interactions of all these elements holistically, Berk-Tek is able to provide cable to the market that is measurably better and that can easily address whatever challenges arise.

GETTING THE MOST OUT OF YOUR IP INFRASTRUCTURE INVESTMENT

There are literally hundreds of new IP enabled devices that require structured cabling. Cable which was once defined for data, now faces many unforeseen network applications introduced over the last decade. The cabling landscape has changed dramatically as the demands of the network have increased. The development of new standards for higher bandwidth applications, including mobile devices, cloud computing and high definition video is sitting on the horizon. But, who can predict the many applications that will be added to the network in the next decade?

Today’s IT managers must be able to realize maximum return-on-investment from a converged network for today and in the future. Upgrading to a higher grade cable today will cost a little more now, but the benefits are countless – clearer voice and video, faster data throughput with little to no frame error rates to encumber the network

Figure 15: Effects of Helical Concentricity

Page 12

CONTACT INFORMATION Corporate Headquarters 132 White Oak Road

New Holland, PA 17557 USA TEL: 717-354-6200 TEL: 800-237-5835 FAX: 717-354-7944 www.berktek.com In Canada, please contact: Nexans Canada Inc. 140 Allstate Parkway Markham, Ontario L3R 0Z7 Canada TEL: 905-944-4300 TEL: 800-237-5835 FAX: 905-944-4390 www.berktek.com

MaxVDP 23G14