There are the specific standards by which physical layer and physical medium(cabling infrastructure) devices must be qualified to achieve error-free data transmission. The IEEE is the organization that provides standards to address specific Ethernet applications and devices, while the TIA/EIA is the standards body that provides standards to address physical layer (layer 1 of 7 OSI model) performance and requirements. To ensure sufficient signal-to-noise ratio and bandwidth necessary for operation of Layers 2-7 devices and applications, IEEE and TIA cooperated in the development of IEEE 802.3ab for 1000 BASE-T operation and TIA/EIA-568-B for Category 5e and Category 6 electrical performance.  In other words, compliance with TIA/EIA-568-B specifications should ensure a BER of less than 10^-10. Since Gigabit Ethernet was a new application, consultants, contractors, and end-users wanted a method to validate the level of cabling required to support Error-Free Transmission. As a result, several manufacturers began BER testing to verify Gigabit Ethernet operation over physical layers. End users have adapted these BER test results as a method of validating application support.

Manufacturer BER Testing Results
Hubbell Premise Wiring (HPW), a recognized leader in connectivity and cabling performance, took a leadership role in offering error-free (per IEEE 802.3ab) assurance to end users. HPW performed the first Gigabit Ethernet testing on cabling plants in the fourth quarter of 1999. Performed at an independent test lab, The Tolly Group, the testing demonstrated Gigabit Ethernet performance over qualified cabling systems. The results were reported in the terminology specified by the IEEE as a maximum BER of 10^-10.Since the HPW testing, additional manufacturers have followed suit and conducted their own BER Gigabit Ethernet testing over cabling plants. These reports claimed increasingly higher performance values, such as a BER of 10^-12 and higher, using minimum IEEE 802.3 defined packet size with 64-byte data frames. HPW decided to verify test methodologies and offer clarification on what is involved in the testing process and how to interpret results.

Clarification of Testing Methodologies
The majority of the tests conducted used the minimum allowable frame size of 64 bytes. This is mainly due to time constraints. At its highest throughput, Gigabit Ethernet operation can transmit 10¹⁰ data packets with 64-bytes size data frames in approximately 2 hours. At this rate, running a test with a significant number of cabling configurations is feasible. Furthermore, since error occurrence is a random process, the probability of obtaining errors during only 2 hours of transmission is minimized. In contrast, transmitting 10¹⁰ data packets with the maximum allowable frame size (1518 bytes), requires approximately 34 hours of data collection. In addition to increasing test time, running the test with worst-case data packets is more challenging and increases the probability for errors. Meeting IEEE requirements of a maximum of 10^-10 errors with 1518-byte size frames means that less than or equal to 1 error is allowed in almost 2 full days of operation. This is the equivalent of 4 workdays of continuous data communication versus the 2 hours for minimum-size packets.

Performance Comparison of Maximum and Minimum-size Packets
HPW developed a test program to compare the performance of maximum and minimum-size data packets. The tests measured Gigabit Ethernet frame errors and loss as worst-case calculations of BER over a sample of at least 10 billion successive packets of both 64-byte and 1518-byte frames. The testing was performed on 90 meters of Category 5e and Category 6 cabling systems in a two-connector channel configuration at Hubbell Premise Wiring’s Interoperability Laboratory. Prior to the IEEE specified application testing, the cabling systems were tested for electrical performance as specified by TIA/EIA-568-B for Category 5e and TIA/EIA proposed Category 6 standard (Draft 10) for permanent link limits. All systems passed with significant margin.

The testing was conducted using SmartBits™, an advanced multiport performance analysis system from Spirent Communications, equipped with GX 1420B 1000BASE-TX interface cards as shown in Figure 1.

The Gigabit Ethernet interface cards offer full duplex testing capabilities over copper wire as specified by the IEEE 802.3ab standard. Tested links included HPW Category 5e and 6 connecting hardware and various Category 5e and 6 cables. After being tested for electrical transmission performance, each permanent link was connected to the SmartBits system using corresponding category-qualified patch cords to form a 2-connector channel. Figure 1: Test Configuration.

Bit Error Rate
The bit error rate was determined using the following formulae:

*BER = Transmitted Frames - Received Frames > 
                Transmitted Frames (10 billion)

or

BER = Total # of Errors (Packet loss, CRC etc)  
                Transmitted Frames (10 billion)

BER (In Bytes) = Total # of errors x 8

                Transmitted Frames x 8

HPW Performance Comparison Test Results
BER testing was performed using GX1420B Gigabit Ethernet UTP cards, transmitting and receiving packets with 64-byte and 1518-byte size data frames in continuous full duplex mode with an Interpacket Gap of 0.096 msec. The test utilized a period of 1 hour and 52 minutes to collect 10 billion 64-byte size data frames and approximately 34 hours to collect 10 billion 1518-byte size data frames. Rx Bytes represents the total number frames received multiplied by the frame size. Multiplying Rx Bytes by 8 yields the total number of bits received. In Table 1, the rows labeled BER indicate the sum of Cyclic Redundancy Check (CRC) errors, oversize-packets, undersize-packets, and fragmented errors recorded and divided by the number of transmitted frames.

In addition to the zero BER results, several other observations are noted regarding the experiments. First, HPW was able to confirm that the worst-case test condition appears with the maximum packet size. When sub-standard cabling was used, problems appeared while running the 1518-byte size data frames, but not with 64-byte size data frames. The first two data columns in Table 1 were obtained by running the test over off-the-shelf cabling and connectivity that was marginal Category 5.

The second observation is with respect to cabling configuration. Channel length did not appear to affect network performance, although it will be explored in a future study. The lack of difference in this experiment can be attributed to good Far End Crosstalk (FEXT) and Return Loss (RL) performance, which assured that minimal secondary crosstalk noise and signal loss were created in shorter channels.

High-power transient noise in the environment was found to cause CRC errors. By operating devices, such as hair-driers, or by cycling the on/off state of a heating fan in close proximity to the transmitting or receiving end, a small number of CRC errors were created. This is being investigated further, as it may be the result of secondary effects such as the current draw on the test setup.

Conclusions
1. TIA/EIA-568-B defines the cabling plant specifications that should ensure error-free Gigabit Ethernet transmission as defined by IEEE 802.3ab. BER testing validates this assumption, providing end-users with additional assurance.
2. The majority of BER testing has been based on packets of 64-byte size data frames. True, worst-case tests use 1518-byte size data frames.
3. Marginal Category 5 cabling systems that are not TSB-95 compliant may have difficulty supporting Gigabit Ethernet error-free transmission. Category 5e should be the minimum performance cabling plant deployed for new installations.

While most manufacturers run tests with minimum 64-byte frame size, HPW runs worst-case maximum 1518-byte size frames over their cabling systems as a standard test procedure. The test setups achieve the required BER with no missing or erred frames, meeting and exceeding the IEEE 802.3ab specification of BER (a maximum of 10-10). Test data shows that the HPW Category 5e and 6 cabling systems performed within an acceptable BER limit per IEEE 802.3ab for both minimum and maximum size packets.