AN ETHERNET CABLE DISCHARGE EVENT (CDE) TEST AND MEASUREMENT SYSTEM
Abstract — A Cable Discharge Event (CDE) is an electrostatic discharge between a cable and a connector. CDEs occur on unshielded Ethernet based communication interfaces and inject currents into the pins directly [1-3]. The charging processes are in general understood; however, the discharge processes are complicated due to the number of pins involved and their connections to a system. Based on an understanding of the factors which determine the severity of a CDE, this article describes how to setup a variety of repeatable CDE tests and how to analyze the measurement results.
Keywords — Cable Discharge Event (CDE) Test; Cable ESD;
关键词：线缆放电事件（CDE）测试、线缆静电放电
I. INTRODUCTION
Both occurrence rate and severity of a CDE needs to be considered when determining the importance of dedicated CDE tests to ensure a reliable system. The occurrence rate of a CDE depends on the type of connector used and the environment the system is used. While a USB connector on a laptop will receive many more ESD events than a LAN connector, the LAN connector still poses a larger risk to the system as it is usually not shielded and often used in applications that require high reliability such as back bone internet routers. In contrast to a USB connector, a LAN connector can have a spark from the connector to a pin during a CDE. In the case of USB connectors the connector shells will mate first. Provided that the shells are connected to the shield and to the system ground, most of the discharge current will flow on the connector shells.
For the LAN CDE case a good understanding of the dominating processes provides a mean for developing and validating models. These models will allow computer simulation, and laboratory test setup formulation for reproducing real CDEs. This is crucial for IC and system level engineers to study and optimize the immunity of Ethernet based communications interface designs.
It is well known that triboelectric charging is the culprit that generates the charge on a cable [1],[2]. This fact is especially important in Ethernet communication systems due to the long lengths of UPT cable that must be pulled through conduit, which may result into large charges. A model for describing the interaction of one twisted pair in a UTP (Unshielded Twisted Pair) cable is presented in [2] using a three body model analyzed from an electrostatic point of view. In [3] a system for discharging a cable using a relay on a test bed is presented, and experimental results are shown. In regards to the measurement analysis side of Ethernet CDE, there are few publications discussing the phenomena. Some ESD test setup have been developed to repeat ESD current transfer through Ethernet magnetic effects in [4] and [5].
The focus of this article is to present a CDE testing system that will allow for real world test conditions to be reproduced in the laboratory, thereby providing design engineers a quick and reliable method for testing new hardware designs. The test system will allow for different types of copper based Ethernet cables, and twisted pair termination strategies. Further, full control of the charge and discharge sequence of each cable line allows for all possible cases to be explored.
II. IMPORTANT PARAMETERS IN ETHERNET CDE
Prior to discussing the CDE test system, a few important parameters of copper based Ethernet cable CDE, especially UTP must be reviewed.
A. Parameters of the charged Ethernet cable affecting CDE
Several important parameters must be considered in UTP cables. The first parameter is the charging processes in UTP cables. In particular, the wires can be charged due to charges on the outside of the jacket, due to migration of charges through the jacket and insulation to the wires, or due to direct contact of the wires with a charged object. The voltage magnitude can be as high as 2 kV. In most cases all wires will be charged to the same voltage. If the spatial arrangement of the cables is changed, the capacitance between the charged wires and ground will also change, which can further increase the charge voltage.
Another important parameter to consider is that there are shielded and unshielded cables for Ethernet. These cable types are illustrated in Fig. 1. They all maintain 100 Ohm differential impedance for each twisted pair, but the unshielded cable will have relatively large common mode impedance (100-300 Ohm) versus the shielded cables because their current path is physically far from the ground as compared to the shielded wires. This common mode impedance plays a critical role in the magnitude and to a lesser extend in the shape of the CDE discharge current waveform.
B. The termination of the Ethernet device
Load terminations also play an important role in the characterization of a CDE. Many different types of termination schemes exist, however, the most commonly used is the Bob Smith Termination [5]. This termination uses a 75 W resistor for a common mode impedance match at each signal pair, and they all connect via a high voltage capacitor to chassis ground as shown in Fig 2. The isolation between the Ethernet connector/chassis and internal PHY circuit is established through a transformer and some designs incorporate common mode chokes to further reduce common mode current motivated EMI concerns. Further variations of the Bob Smith circuit can be found in power over Ethernet applications, and strongly cost reduced designs.
One pin will make the first contact and any other pin will contact next. The initial contact leads to a charge redistribution, a partial discharge of the total charge on the cable, and it can charge the capacitor used in the Bob Smith circuit. Depending on the contact sequence, various common mode and differential mode termination plus the PHY IC circuit determine the load of the discharge current path and therefore are the important factors that determine the CDE current waveform.
C. The ESD events during CDE
The Ethernet cable connection includes several metal to metal contacts, which can lead to multiple ESD events during connection. For the case of plugging a shielded Ethernet cable into a shielded Ethernet connector, the first ESD event is the discharge between the shielding of the cable to the shielding of Ethernet connector as this is the first contact point. Normally this ESD is not likely to cause any failures if the DUT is shielded. For the case of an unshielded cable or an unshielded Ethernet connector the current must flow in the wires of the UTP.
When an Ethernet cable is plugged into a connector, many possible contact sequences can occur. Theoretically there could be eight separate ESD events, one for each pin. However, each contact will minimize the voltage between the other pins and the connector. This is caused by the large mutual capacitance between the wires, and by charging the Bob Smith Termination.
To illustrate the charging of the capacitor in the Bob Smith circuit, a 100 m long shielded cable was charged to 100 V between the shield and its wires and inserted into an Ethernet device. A result from this experiment is shown in Fig 3. Because the capacitance between the wires and the shield is much larger than the 1000 pF capacitor, the voltage reaches over 90 V immediately after first contact, thereby lowering the voltage difference between other connecting pins. For the next seven contact events the voltage difference across the contacts is drastically reduced.
The voltage waveform shown in Fig 3 gradually diminishes due to the 100 MW resistive voltage probe used in the measurement. Without this drain path the voltage on the cable will remain high for a very long time after the connection is made. If the other end of this cable is plugged into an Ethernet device another cable discharge event may occur.
Because the differential pairs of the cable have a welldefined 100 W discharge impedance, after the first wire has made contact and the high voltage capacitor has a low impedance path to ground, such that the second wire also experiences the discharge as it is making contact, resulting in a differential ESD event. This differential ESD current can easily transfer through the magnetics to the isolated PHY circuit, which is hazardous for the PHY chip.
III. THE ETHERNET CDE TEST SYSTEM CONCEPT
A good Ethernet CDE test setup should be able to control as many of the parameters related to the cable discharge event as possible, and provide a repeatable test. To do this the setup must have these three main components; a Controller, a Charge Module, and a Failure Test Monitor.
The Controller must have the capability to control cable charge voltage and polarity, charge and discharge different types of Ethernet cable arrangements, and separately control the charging, floating, grounding, or discharge of each wire. In addition, the test system should maintain the electrical characteristic of the entire discharge current path as close to real world CDE cases as possible. In particular, it should control common and differential mode impedance of each pair, and the contact sequence delay time between each discharge to the same order of magnitude of a real cable as it is plugged into a device.
The Charge Module represents the Ethernet cable used as the discharge source for the CDE event. This consists of different lengths of various CAT5 cable types. For research purposes, different types of cables and how they are arranged must be studied. For industry testing purposes, a good charge module will provide a worst case, real world CDE source.
ill provide a worst case, real world CDE source. The Failure Test Monitor provides a means for verifying the Ethernet performance. It will check if a failure has occurred, or simply if a degradation in communication speed has occurred. It is important this is automated due to the vast array of tests that can be performed to check all discharge sequences. A general block diagram of the CDE test system is shown in Fig 4.
A. CDE Test Controller
The main CDE test controller consists of high voltage supplies (dual polarity), relays to separately control the connection of each Ethernet cable wire, including whether it is charging, floating, or grounded. The relays are bounce free and provide a clean discharge for each wire. The transmission lines on the board maintain the controlled 100 W differential pair structure of the Ethernet cable, and current probes are embedded into each wire to monitor CDE discharge current. The system also has remote control capability integrated into the upper level system as a part of the automatic test equipment (ATE). A simplified diagram of the main CDE controller is shown in Fig 5.
A common test procedure may consist of the following four steps, as illustrated by the block diagram in Fig 6.
1. Charge cable status control
The controller separately controls the voltage (level and polarity) for each wire in the cable bundle. This includes the status of each wire indicating if it is charged, floating, grounded, or a through path.
2. Pre-discharge preparation
The controller will disconnect the charging path from each wire in the Ethernet cable bundle and disconnect the DUT discharge path between the DUT connector and ground. This will leave all wires floating before the discharge step.
3. Control wire discharge while monitoring ESD Current
The controller will close the relays allowing the wires coming from the charge line to discharge into the DUT or termination load. The sequence of the relay closure and the delay between relay closures can be set by the user.
4. Preparation after a discharge and check DUT status
The controller will open all cable discharge relays such that the cable can be charged again. To charge the wires the high voltage charging relays are closed. In additional all remaining charges on the DUT need to be removed. This is achieved by closing the DUT discharge relays.
The controller will open all cable discharge relays, then close all high voltage charging relays and DUT connector discharge relays to prepare for next CDE test.
B. Charge Module
In real world Ethernet cable installations there are many different possibilities in regard to cable type and arrangements, leading to unlimited test scenarios. Some common real world arrangements are:
A. Cable hanging on the celling (relatively far from ground)
B. Cable on the floor (very close to the ground)
C. A spool of new cable
D. Cable pulled through conduit (very close to the ground)
C. Failure Test Module
Besides the setup for CDE test control, it is also important to test for normal Ethernet performance and functionality. It is important to understand the effects of applied CDE on the DUT, and as in the block diagram of Fig 4, an Ethernet traffic test system should not only monitor the status of DUT during CDE test, but also check the performance of the CDE applied ported after each test level.
D. Calibration of the Ethernet CDE Test System
A CDE Tester (Model ES631-LAN) with two different types of Ethernet Charge Modules (shielded and unshielded) was built. The analysis of real discharge waveforms is difficult. For that reason a proper calibration method with well-defined test loads is important to verify the functionality, to understand the current paths and the output parameters of the CDE test setup. A set of calibration loads including short, open, and 100 W differential load were built using 10 cm CAT5E UTP cables, and are shown in Fig 8.
The EIA 568B cable standard was used for all Ethernet wiring, Fig 9.
In the measurement setup the oscilloscope was set to capture the waveforms: Channel 1-voltage on line 5, Channel 2-current on line 5, Channel 3-voltage on line 4, and Channel 4-current on line 4. The voltage probes were 1010:1 (5 kW/50 W resistor bridge = 101:1, plus 20 dB attenuator). And the current probes were 20:1 (5 V/A plus 40 dB attenuator). The CDE Controller in all tests was set to 500 V, and depending on the test configuration the Charge Module and DUT may have been grounded.
Setup 1 consisted of a short connected to CDE Controller output, one differential pair with one side grounded and preconnected to the load, and a 100 meter Charge Module using S/UTP cable with the chassis not grounded. The setup and measurement waveforms are shown in Fig 10. From the simplified circuit, after the relay is closed in line 5, we expect a large current flowing from line 5 to line 4 over the differential 100 W transmission line structure and the peak current should be 5 A (500 V/100 W). From the waveforms it can be confirmed that the current waveform starts with 5 A peak. Then the waveform shows significant cable loss as the waveforms rapidly decay. After the current reaches the cable ends, reflections occur due to mismatches and several reflections happen until the total signal approaches zero. The voltage is a RC discharge waveform through the voltage measurement path.
Setup 2 is similar to setup 1 except the charge module chassis is connected to ground. The setup and measurement waveforms are shown in Fig 11. As can be seen, the current waveforms are the same as the first setup, but the voltage waveform RC time constant is much larger due to the grounding of the Charge Module chassis.
Setup 3 consisted of an open load connected to the CDE Controller output, one differential pair with one side grounded and pre-connected to the load, and a 100 meter Charge Module using S/UTP cable and chassis grounded. The setup and measurement waveforms are shown in Fig 12. There is little displacement current as the load is an open, and the voltage on line 5 is 494 V peak, which is very close to the charge voltage.
Setup 4 consisted of a 100 W load (with center tap between two 50 W resistors grounded) connected to the CDE Controller output, one differential pair with one side grounded and preconnected to the load, and a 100 meter Charge Module using S/UTP cable and chassis grounded. The setup and measurement waveforms are shown in Fig 13. With the 100 W load in place the current waveform magnitude drops to half of that measured with the short as in setups 1 and 2, and there are no reflections.
Except for the shape and magnitude of the measured calibration waveforms, the time delay of the measurement is also important to understand, in particular, which part of the waveform is due to the 100 W transmission line extension, and which part is due to the DUT. The waveforms shown in Fig 14 are with the CDE Controller connected to a CAT5E UTP cable only. The current waveform shows the 100 W transmission line extension first, then the open end. The real measurement of the DUT starts at the time where the open is shown.
E. Test of the CDE System on Ethernet Systems
Two 10/100Mbps Ethernet DUTs were tested with the CDE test system as shown in Fig 15. The test parameters are: CDE source voltage of 1 kV, a 100 meter Charge Module using S/UTP cable, and oscilloscope channels are scaled to 1 A/div. The first pin discharge waveform is captured and compared between unit A and unit B. Unit B has a much smaller current peak and higher current duration. This is mainly because the common mode impedance of unit B is much smaller than unit A.
Most systems using the Bob Smith Termination are almost fully charged after the initial discharge, so the current magnitude of the 2 nd – 8 th discharges are too small compared to the first one. This is not discussed here.
A Power over Ethernet (POE) system, having more complex termination was also tested. The measurement waveforms are shown in Fig 16. When the first pin contact DUT, first discharge current is generated in common mode through this pin. Then when the other wire in the same twisted pair contacts DUT, another discharge current is generated in differential mode through this pair. Both discharge signals could be transferred from termination to PHY circuit leading to damage in the Ethernet system.
IV.CONCLUSION
In this article, several important physics for CDEs have been discussed, a general CDE test and measurement solution developed, and some test results with well-defined structures and a few real world Ethernet systems are presented and discussed.
The next steps for CDE related test and measurement would be modeling the overall test system with different terminations, and running more tests with Ethernet systems to understand the design principle for CDE.
ACKNOWLEDGMENT
We would like to thank the EMC Laboratory, Missouri University of Science and Technology for the partnership of CDE related study and research and Cisco for the partnership in CDE related tests and measurements.
