In a message dated 8/5/2003, [email protected] writes:
We have products that meet all the Surge requirements of NEBS GR-1089, FCC-68 and EMC 4-5. But, the same products are continuously failing in the field due to real-world lightning strikes. I have spoken to four other manufacturers who make similiar interfaces (DSL) and they all have the same problem : they meet the standards, but fail in the real world. Hi Anil: I am familiar with the problem that you describe. In general, I believe that the GR-1089 requirements are fairly appropriate, although the FCC-68 and EMC 4-5 are not really adequate for good performance in the field. In my experience, if you are passing GR-1089 and are still experiencing field failures, the cause is likely to be one of the following: 1) Many test labs do not perform the GR-1089 tests properly. For example, some labs only test the equipment in an unpowered state, rather than testing all operating states. It is sometimes difficult to apply surges with the equipment held in a legitimate operating mode with all the proper conditions (DC battery, etc.). Another thing that some test labs skip is testing the "sneak-under" surge tolerance per Note 5 of Table 4-2 of GR-1089, Issue 3. 2) I have seen some DSL interface designs that failed only when used in combination with primary protectors. For example, certain capacitors within the DSL interface can charge up on the rising edge of the surge, but will discharge rapidly when the primary protector fires. Sometimes this discharge path damages the line driver IC. If the test is performed with no primary protector, everything works fine. The requirements in the 2000 editions of ITU K.20 and K.21 seem more directed at uncovering this type of problem than the coordination requirements in GR-1089. 3) All the standards that you reference, including GR-1089, *assume* that there is a primary protector in place that will limit differential surges to about 1000 volts peak and common mode surges to about 2500 volts peak. This is probably an acceptable assumption for central office equipment that is generally well maintained, but for customer premises equipment the primary protector is sometimes inoperative. A common problem is that the ground wire for the primary protector has been inadvertently disconnected. Regarding problem #3 above, the situation is not as dire as it first appears. Even with no primary protector installed, differential (metallic) surge voltages will tend to be limited by the secondary protection in the DSL interface. The short-circuit currents typically do not exceed 100 amps. For common mode (longitudinal) surges, the open-circuit voltage will be limited only by the internal breakdown threshold in the telco cable and infrastructure, unless the DSL interface also contains common mode protection to ground. I have only been able to locate limited field studies of what the upper bound of such surges might be, but the consensus view seems to be that it is in the range of 4000 to 6000 volts with no primary protector. If any list members are aware of published studies on this parameter, please let me know. Where possible, I prefer to implement a 5000 volt isolation barrier that has inherent resistance to common mode surges. This is generally easy to do in transformer-coupled DSL circuits, especially in customer premises equipment. In some central office equipment, protection to earth ground is required for other reasons, so the focus turns to short-circuit current rather than open-circuit voltage. To comment further on possible causes of the specific failures that you are experiencing, I would need to see a schematic of your present interface and a description of the failure mode. However, I think there is a good chance that the cause of the discrepancy (pass GR-1089, fail in the field) is related to one of the three items listed above. The good news is that it is possible to achieve low failure rates in the field without spending a lot on interface protection. The first thing you need to do is carefully compile a list of appropriate tests. I think that GR-1089 (properly performed) is a good starting point, but you may want to add supplemental tests to cover items 2 and 3 above. Once you have identified the tests that you want to pass, some careful attention to the architecture of your protection circuit will usually yield a suitable circuit at minimal increase in cost and/or board area. Joe Randolph Telecom Design Consultant Randolph Telecom, Inc. 781-721-2848 [email protected] http://www.randolph-telecom.com
