This paper intends to put to rest a longstanding EMC controversy. Thus far an interrelated set of unresolved problems existed for power feed line filters. One never knew whether or not they would behave as predicted; that is, would they attenuate in the stopband, as desired, and/or ring in the passband, as not desired? Successful testing, according to accepted standards, does not guarantee satisfactory performance in the real world. This problematique is solved by combining extreme-value theory (newly applied) of mismatch with application of a properly categorized statistical data base (now established) on interface impedances. The delineation of the surprisingly transparent conditions for filter dysfunction permits establishment of practical measures for design and testing of filters that must often work under indeterminate and fluctuating interface conditions.

Assuredly Effective Filters

D.O.D. Directive 3222.3 with its clearly stated objectives (Section IV) of operational and built-in design compatibility necessitates a fresh approach to certain inadequate MIL STD's. One of them is in particular need of revision. MIL STD 220A, with its 50? reference, is based on assumptions contrary to the facts which are: 1) Interfacial impedances change with frequency, time, and location and can be nearly anywhere on the Smith Chart; 2) Interfacial resonances in the stopband, and eigen and interactive resonances in the passband, render the conventionally predicted performance of the filter illusory. Hence, new, more realistic criteria for filter testing are necessary: I. Compatibility: The test method must be a qualification test and must be able to permit selection of the more insensitive filter: insensitive meaning that in the broad area about the cutoff frequency the filter is relatively unaffected by interface changes. II. Predictability: The filter test method must permit a selection of a filter to be effective as stipulated minimally by the control plan prior to installation of the system and independent of changes in the system. III. Reliability: The filter must not be destroyed by insertion gain in the so-called passband. IV. Practicability: Filters meeting the specs must be economical and practical. V. Simplicity: The test method should be reasonably simple and brief. From the various solutions which have been proposed, two try to overcome fundamentally the limitations inherent in MIL STD 220A: (a) the in situ measurement (b) the worst case measurement.

The Controversial MIL STD 220A

This paper describes a technique to measure insertion loss of filters in a 5 ohm system over an extended frequency range and under full load from 14 KHz to 10 GHz. This technique required development of extended range buffer networks and electromagnetic shielding requirements in excess of 120 db. Techniques have been developed for measuring high power, high attenuation, rf powerline filters in impedances that sometimes approximate actual operating impedances. Present specifications and standards require updated and improved measurement techniques. The frequency range of present measurements under full-rated load is limited to 100 KHz through 20 MHz and the circuit impedance is kept at 50 ohms. This provides inadequate and misleading information about the performance of the filter under actual operating conditions. By use of techniques described in this paper, meaningful measurements that simulate actual operating conditions can be achieved. It further describes insertion loss measurements that were performed on a filter measured in a 5 ohm system, under full load conditions, over the extended frequency range from 14 KHz to 10 GHz by employing extended frequency range buffer networks.

Insertion Loss Measurements in a 5 Ohm System

The most important requirement of a power-line filter is its insertion loss under load. This paper discusses methods of extending and improving the technique of measuring insertion loss as specified in MIL-STD-220A. It discusses present equipment limitations in MIL-STD-220A that restricts full-load tests to frequencies above 100 kHz and limits maximum current to 100 amperes. However, it is realized that MIL-STD-220A is the only available specification at present. Therefore, the paper discusses methods of extending the MIL-STD-220A technique for measuring the insertion loss of filters for frequencies as low as 14 kHz and under loads as high as 200 amperes. A new buffer network that allows such measurements is described. In addition, a method is discussed for improving voltage regulation and reducing distortion of the power-line waveshape caused by power filters. These are two parameters which have been totally ignored when specifying power filters and which have been found to cause considerable electrical problems after installation.

Jules Milton

Papers

Improving the Specifications for Power-Line Filters