GB2336216A - Detecting the shape of the frequency response of a filter - Google Patents

Abstract

Electrical filter circuits 4 are tested by connecting to the filter inputs without the need to connect to the filter outputs or to disconnect the outputs from a load 5. A signal generator 2 of known source resistance applies a.c. signals successively over a range of frequencies to the filter inputs, and a voltmeter 3 monitors the voltage across the filter inputs. Different types of filter have different characteristic shapes for the voltage/frequency curve, and processing is applied to the measured results in a computer to determine the location of inflections in the curve and other characteristics of the curve. Methods are disclosed for determining the values of the individual sub-components of the filter. Where the filter is an L-C filter, an interactive process is applied to successively improve the accuracy of the component value determinations.

Description

2336216 1 METHOD AND APPARATUS FOR DETECTING THE SHAPE OF THE FREQUENCY

RESPONSE OF A FILTER 1 -- 2--

Background of the invention:

This invention relates to methods and apparatus for testing frequencydependent electrical circuits or circuit elements, particularly for example electrical filter circuits or circuit elements.

Electrical filter circuits often need to be tested, for instance for the purpose of verifying whether they are serv, iceable or not, i.e. to check their operational integrity, or for the purpose of assessing their electrical characteristics, or both. The usual method of testing such circuits is to apply a.c. (alternating current) input signals of predetermined voltage, e.g. of measured and/or fixed standard voltage, from a low-impedance source, over a range of frecruencies, to the input terminals of a filter circuit under test, and to measure the corresponding voltage delivered at the output terminals.

Most methods for determining the freauency response of an electrical circuit rely on measurement of both magnittide and phase information at the inputs and outputs of the circuit in order to derive the frequency and phase response of the circuit. Once input and output measurements have been made it has been possible to estimate the insertion loss of the filter, i.e. ratio of output to input voltages under the test conditions, usually expressed in decibels, and generally as a function of frequency. The insertion loss can give desired information about the integrity or otherwise, and the characteristics, of the filter under test.

However, increasingly there are situations in which it would be desirable to test frequency-dependent electrical circuits, for example filters, but in which the usual method of testing indicated above is not feasible. Tn modern electronic systems, especially where packaging densities are high, electrical filters are often incorporated in devices. aircraft for example, the increasing number of pilot and - 2 3 -! navigator aids, provided by electrical equipment, means that t'-e integration of ecuiDment is advanced, and there is little ava-"ab--'-', itv for test -ooints w-.h-n eauioment. Furthermore, once the circuits 'nave been incorporated into a larger piece of equipment it may be advantageous to test the circuit for Situations where it is not possible to have access both to input and to output connections, as would be required _n order to carry out the traditiona test methods are increasingly commonplace. Often the output side is inaccessible.

The present inventor's International Patent Application No. W095/04935 discloses a method and apparatus for testing filters and the like applied to the input measurements are made terminals and/or the the integrity of the characteristic of the in which a.c. input test signals are o- the filter under test, and of the voltage across the input current through the input terminals, and filter is dezermined by identifying a resultant measurements. The measurements can simply be compared with those obtained from a sample filter which is known to be good. Alternative! and y preferably the measurements obtained are processed to detect exiDec-.e-,characteristics of the measurements. This gives an indication as to whether the filter comes up to specification or not.

That system proves to be very effective, but can only identify a faulty unit. We have appreciated that there would be advantages in being able to determine why the unit has failed. in particular, it would be advantageous tc determine wh--'c'- of the subcomponents of the filter under test are giving rise to the problem. This would make it possible to repair the unit, and/or would provide the manufacturers with _nformation enabling them to improve the quality of their production by removing frequently appearing faults.

hus, we have appreciated that it may be desirable to be able to determine the values of selected subcomponents of the filter under test. It may also be desirable to be able to deterr-.--ne the values of character istics of the filter such as the -'nsertion loss ef the filter.

Summary of the Invention

The invention in its various aspects is defined in the Jndependent claims below, to which reference should now be made. Advantageous features are set forth in the appendant claims.

1 = 33 3 z Preferred embodiments of the invention are described below with reference to the drawings. in these embodiments of the invention electrical filter circuits are tested by connecting to the filter inputs without the need to connect to the filter outputs or to disconnect the outputs from a load. A signal generator of known source resistance applies a.c. signals successively over a range of frequencies to the filter inputs, and a voltmeter monitors the voltage across the filter inputs. Different types of filter have different characteristic shapes for the voltage/frequency curve, and processing is applied to the measured results in a computer to determine the location of inflections in the curve and other characteristics of the curve. Methods are disclosed for determining the values of the individual sub-components of the filter. Where the filter is an L-C filter, an interactive process is applied to successively improve the accuracy of the component value determinations. Using the techniques described enables the insertion loss of the filter also to be readily calculated by the computer.

Brief Description of the Drawings

Preferred embodiments of the invention will now be described in more detail, by way of example, with reference to the drawings, in which:

Figure 1 is a schematic diagram of an arrangement of a computercontrolled test apparatus for testing frequency-dependent electrical circuits or circuit elements, particularly for example electrical filter circuits or circuit elements, according to an embodiment of the invention; Figure 2A shows one embodiment of test set-up required to test a parallel capacitive, or C-section, low-pass filter; j 1 1 Figure 2B shows the frequency response obtained when servceable and unserviceable examples of parallel capaciior filter circuits or circuit elements are tested using the apparatus of Figure 2A; Figure 3A shows the test set-up -for a T-section filter; Figure 3B shows the test set-up) for an L-section fil.:er is a special case of the T-section filter with L-=0; Figure 3C shows the frequency response characteristic of serviceable and unservIceable examples of low-pass T-section or L-section filter circuits or circuit elements when tested using the apparatus of Figure Figure 4A shows the test set-up required for testing a p-'-section filter; Figure 4B shows the frequency -response characteristic of serviceable and unserviceable examples of low-pass pi-section --- ter circuits or circuit elements are tested; Figure 5 shows a -flowchart for testing a pi-section low-pass filter using the method of the invention; and Figure 6 shows a flowchart for calculating the total capactance of a pi- section -filter.

Description of the Preferred Embodiments of the Invention.

General Description

A general description of apparatus embodying the _nvention and its method of operation will first be given, and then the preferred embodiments will be described more specifically in more detail as to their construction and operation.

one particularly desirable method of obtaining the rec--iire--measurement -'s to apply each of a series of a.c.

-nn,,t signals from a signal generator, over a predetermined frequency range, successively to a circuit or circuit element under test. The signal generator fcrms part of a signal source which is of standard or calibra.ed or otherwise determinate voltace (at and which also has a known and fixed source impedance,preferably a purely resistive impedance) associated with it. The resulting voltage developed, at each of the series of frequencies, between the terminals or connections made to the circuit or circuit element under test is then measured. Alternat-4vei-y, the open circuit voltage could be measured using the voltmeter reauired for measuring the test voltage by switching the filter out of the test circuit.

The source impedance should be effectively constant over the frequency range used for the testing.

For example, a suitable signal source can comprise a variable frequency signal generator of low output impedance, of standard output voltage (or provided with means for measuring or standardising its output voltage), and a fixed impedance, e.g. a resistor, in series with its signa7 output. A suitable measurement device can comprise a high-impedance digit.al voltmeter connected across the connections made -.c the circuit or circuit element under test.

In practice the source impedance, usually pure resistance as far as this can be realised in practice, '.-.as a value chosen to make an appreciable difference between the resulting voltage at the connections made to the circuit or circuit element under test, on the one hand, and the opencrcuit voltage of the signal source on the other hand. For example (but without limitation) a 50-ohm source resistance can be used for filters which have input impedances, at one or more of their corner frequencies or characteristic frequencies, in the range of a few ohms (e.g. 1-1-0 ohm) to a few hundred ohms (e.g. 100 ohm to 1 kohm).

The a.c. input signals are chosen so that they cover a frequency range which ranges well to each side of each corner or characteristic frequency of the selected type of circuit or circuit element which is under test, e. g. a range that includes pass and stop bands. In the case of a single "lel para -capacitative filter, or other low-pass fJiter, the needed range extends well both below and above the corner-frequency of the configuration which corresponds to the time constant of the capacitor in conjunction with the associated source and/or load resistance.

7he value of the voltages used is chosen so as to produce minimum active effect in the circuits connected t the output of the filter being tested. These circuits mav 'nclude active components such as transistors, and thus the voltage applied to the filter is less than 0.7 volts. Preferably the voltage is in the range 0.2 to 0.5 volts, and may conveniently be nominally set at 0.3 volts.

1 2 -- in cractice, a useful test method and apparatus catering a wide variety of important types of filters can utilise a sianal generator with a range from about 1-0 kHz to about 50 MHZ, which can sweep the desired frequency range in suitably chosen linear or logarizhmic steps to demonstrate the characteristic under test. Use of a lower freauency at the better-. of the range may be desirable if filters incorporating large capacitance values are to be tested, but the indicated range has been found suitable for most filters encountered in pract4ce.

-he data-processing arrangement and identification steps of the method and apparatus embodying the invention can be arranged or -performed in any of a number of ways. Examples of such method and apparatus for use with L-C 'inductancecapacitance', -filters are described below.

Embodiments of the invention can take the form of a computer-controlled test rig comprising a digital computer and a variable-frequency a.c. signal generator, of standard or calibratable output voltage, under program control by the diailtal comouter, and arranged to feed its a.c. signal.zhrouch a source resistance to a single pair of connections made to a circuit or circult element to be tested. A digital voltmeter is arranged to measure the voltage developed across the single pair of connections made to the circuit or circuit element to be tested and to deliver data representing the results of such measurement as input data to the digital computer so as to cooperate with the program control. The comuter is arranged under program control to cause a steDw-se succession of a.c. sianals, of a stepwise succession. of freouencies, to be applied to the circuit or circuit element of selected tvoe to be tested, and for the data 3-- delivered by the digital voltmeter to be processed so as to provide for the user of the test rig an indication related to the integrity or performance of the circuit or c--'rcu--,-: element which is under test.

The diagnostic features of the test data show a dependence upon the type of circuit or circuit element under test, and accordingly it will often be necessary to carry out a different form of comparison or other dataprocessing for each of several such types. Several representative types of circuit or circuit element, and the characteristic forms of data given by serviceable and unserviceable examples of each, are described below.

Preferred erdDodiments of the invention are described in further detail below, but without limitation on the scope of the invention, and with reference to the accompanying drawinas.

General Apparatus for Testing Filters Referring now to the drawings, Figure 1 shows the arrangement of test apparatus controlled by a central n=cessor unit (CPU) 1 in accordance with one embodiment of.he invention. CPU 1 may conveniently be an industry-standard microcomputer such as a personal computer based on a 80386, 80486 or Pentium (Trade Marks) processor and equipped with hardware expansion slots and/or --'/o (input/output) ports such as RS-232 ports. Either bv hardware expansion cards or by cable links via --'/o 1 is functionally linked to control a programmable variable-frequency a.c. signal generator 2 and to receive data from a programmable digital voltmeter 3, preferably under control of menu-driven software.arranged to control computer to carry out the functions described herein.

The connections to the signal generator 2 from the CPU 1, and from the voltmeter 3 to the CPU 1, may be made via an interface (not shown) if necessary. Alternatively an interface sub-routine may be used to allow the CPU to control the signal generator and voltmeter. The signal generator is connected to the two input terminals of the filter 4 to be ports, CPU -he 7 -:

tested, oreferablv with a source resistor R. ar-ranged in series w-t'h the signal generator to provide an accurately known and stable source resistance. Alternatively the signal generator may be provided with a suitable internal resistance as standard which may be used allowing the separate external resistor R. to be omitted from the test circuit.

The voltmeter is also connected to the two Inputs of a filter 4 under test. The filter (shown as a functional block) has two input terminals iN and has its output terminals connected to an output or load 5. The load 5 has a load impedance Z,., assumed to comprise a resistor R. and capaciter C. in parallel, shown as connected across the outputs of the filter.

If one connector serves several -filters, each of these filters may be tested in turn using different pins of the connector. A mechanism for switching the signal's between. different pin connections enables all -he filters to be tested -without having to reconnect the test equipment. Preferably routing relays, controlled by the CPU, are used.

in a practical example, the filter may be the filter in a combined filterconnector, that -'s a connector which incorporates a filter circuit or circuits.

The signal generator 2 can be an industry-standard unit chosen (e.g. from the Hewlett-Packard rance) to give an outcut level in this examn-le of 0.5V RMS (--5'6) over the rance !")kHz to 50MHz, programmable so that the signal frequency is also 5%, relative to nominal programmed frequency. The generator is chosen to-give an output which is sinusoidal with good purity indicated by low total harmonic distortion preferably not exceeding -40dB (--ecibels), and good linearity over a load impedance range from abcut 5 ohm to about lO kohm. The generator is of a type able to drive any impedance value from open-circuit (O/C) to short-circuit (S/C',, and whether resistive or reactive, without damage. i the iz)re-ferred embodiment, the source impedance of the generator is arranged, using either the series, or internal, -es' (as purely resistive as stor R,,, to be effectively 50 ohm can be realised).

n - 9 -- C Z.Ji - The programmable digital voltmeter 3 is an industry-standard unit chosen (e.g. from the Hewlett-Packard range) to give RMS (root mean square) voltage detection and measurement over the frequency range 1OkHz to 50MHz, over a range of levels from 2mV RMS to 0.5V RMS, with an accuracy preferably from about 1% of measured voltage at maximum level which may gradually increase to 10% of measured voltage at minimum level. A high voltmeter input impedance is desired, preferably at least about 10Okohm shunted by no more than lOoF capacitance over the usable frequency range. It is preferred that the measurement accuracy is maintained when about 100 different frequency settings are programmed within a period of about 10 seconds.

The arrangements described in this embodiment can conveniently be used for testing filters with capacitive component values up to about 0.5 microfarads (pF). if larger cacactance values are likely to be encountered, a lower bottom -frequency may be used and the signal generator and voltmeter chosen accordingly.

in Figure 1 the generator 2 and voltmeter 3 are dlagrammatically shown connected to the filter 4 under test. Filter 4 may have any load impedance Z- connected across i-s outputs, as described above, or the filter may be unloaded and Z, may be absent. In many practical situations, the output load may be assumed to consist of a capacitor and load resistor arranged in parallel, as shown. This assu-npiion is exploited by the method for determining the filter component values. The connections of a filter 4 under test to both the signal generator 2 and voltmeter 3 are made via matched coaxia-l cables 6,7,8,9 connected as illustrated to the two input terminals of filter 4, via a conventional four-terminal measurement connector configuration (not shown). In practice cables 6 and 9 are constituted by one coaxial cable and cables 7 and 8 by another.

It may be convenient to make these connections switchable, e.g. by a bank of progra.mmab-"e relays, to any one cf a number, e.g. conveniently 120 in one example, of plugi'socket connectors suitable for receiving filters to be -1 Such relays can preferably be subject- to operation under software control from CPU 1 in manner known per se.

snec-al-"v where sue- routing relays are used, it can be helpful to take null-signal measurements from voltmeter 3 when all routing relays are left open, and to apply any such null-signal va,-.,ies tc comoensate the measurements taken when a relay -'s closed and a circuit or circuit element is under test.

A problem of equalisation tends to occur in connection with the transmission of high-frequency measurement signals alona such links. Especially where an example of the system is optImised for cables, filters and/or other components having characteristic impedance of e.g. 50 ohm, the measurement conditions can be controlled or compensated for ieauall--'ties in -frequency response by carrying out a frequency sweep of the system, ',e.g. separately for al! relay positions if a bank-switched system _Js in use), us-nc a standard signal generator level and a pure resistance of e.g. 50 ohm in the place normallv occupied by the (or each', filter to be tested, and storing calibration data or compensaton data corresponding to the resultant.neas,-,rements. Any departure from a 'flat' frequency response can then be applied as a factor when data- processing the measurement data obtained under test conditions, to c)m-- ensa.e for any inequality of frequency response of the measurement system.

7-n the preferred embodiment CPU l is connected via an output port 10 to a control input of the signal generator 2. The CPU is provided with the necessary interface technique, either an interface sub-routine er a compatible signal eene-rator, to enable the CPU to control the frequency of the outr)ui tes--) signal applied to the filter 4. Alternatively, the signal generator may be provided by an add-on board within a standard computer, in which case the inputs to the -filter 4 are connected directly from the relevant output of the computer.

The digital voltmeter 3 is connected to the input 11 of the C7-U, allowina the CPU to read the RMS voltage measured by 'I 3 5 the voltmeter. Alternatively, exception processing and timers, provided within the computer, may be used to provide RMS detection with suitable software, in which case the filter inputs are connected to the input of the computer via a buffer if necessary.

The CPU 1 is provided with Read Only Memory (ROM) and programmable memory (not shown). The ROM mav contain the operating instructions to the test procedure including different algorithms encompassing all the different types of filter to be supported. Alternatively, the algorithms may be stored on a write protected floppy disk, or magnetic tape, and the CPU provided with a suitable disk drive. It is now preferred, however, that the program is held in a PCMCIA card and the CPU provided with an appropriate card reader. The test results can also be held on the card.

Test Procedure On initialisation of the test procedure, the user is required to input, for example via a keypad (not shown) associated with the CPU, data commensurate with the type of filter to be tested. This data may be the model number of the filter given in the equipment instruction manual- or on the equipment, or alternatively the user may select the filter type from a menu provided on a display (not shown) also associated with the CPU. The CPU then controls which one, or series, of a number of test algorithms, or test sub-routines, are executed in accordance with the filter type.

Regardless of filter type, the CPU initiates a procedure which causes the signal generator to apply a known, and incrementible, frequency sinusoid to the filter. In the embodiments described in detail below, a voltage measurement from the unloaded signal generator is required. Preferably, the voltmeter 3 has been chosen to give a known open-circuit voltage across the test frequency range. If the open circuit voltage of the signal generator variles with frequency, the connections to the filter from the signal generator and vol.imeter may be switched to allow the open to be measured.

If routing relays are used, the CPU then causes the required relays to close, connectina the selected filter into,-he test circuit, and the test voltage, at a first frequency f, is measured by the voltmeter 3, read by the CPU and stored _n programmable memory. The test voltage, frequency of the test signa- applied, and open-circuit voltage if necessary, are stored as associated data forming one entry in a data set, or table. The processor then sends a signal to the signal generator to increase the test signal frequency.

At this stace the test may proceed in one of two ways. The processor may commence data processing according to the 'fied filter type processing on the data as it becomes specavailable, or a complete frequency sweep can be implemented and al! the data stored for subsequen. processing. In either case the i3rocessing principle -'s the same.

The data is processed to determine whether the voltages measured indicate that the correct frequency response has been observed. If the correct frequency response has indeed been measured, then further data processing, to determine the filter comnonent values circuit, takes place. differ from the correct failed and some limited failure may be provided printout.

The results of the test may be displayed in a number of ways. A visual output display of the -frequency response may be nrovided in the case of filter operational failure. In addition, or alternativelv, the display may indicate the result by displaying a text message. If the ---i-'ter is operational, then the component values and insertion loss will be displayed in numerical form on the display, and/or by a crint-out.

The type of data processing depends on the filter under!est. in 90 - 95% of svstems employing pre-filtering, or filterinj f the input signal, the fi4l-:iers requ-'red are c j - circuit voltage and the -'nsertion loss of the If the frequency response is found to frequency response, the filter has in.ferences regarding the type of for the user '--v a display and/or 1 2 _5 low-pass filters in order to remove high frequency noise. The description below examines the test procedure of the four common types of passive low- pass filters, C-section, L-section, T-section, and pi- (i.e. n) section filters. In principle the method and apparatus nay be used to test other types of filter, for example, high-pass or band-pass filters. In these cases the software controlling the data processing would need to be adapted to the specifics of the filter under test.

In use, the processor is provided with input effectively consisting of program software P to perform the functions described herein, and a setting S set by the operator, to correspond to the type of filter to be tested, e.g. one of the types described in connection with Figures 2 to 4 below. Setting S then determines which of several variants of the data processing are to be carried out in connection with the test of the filter 4. The computer 1 provides results R in anv convenient form, provided by software in manner known per se.

Low-band-pass L-C filters Figure 2B, Figure 3C and Figure 4B show the frequency responses of the filter types shown in Figure 2A, Figures 3A and 3B, and Figure 4A respectively when the above described apparatus is used to test each type of filter. The forms of the frequency response for these types of filter are well known. The frequency response of the filter is verified by the system and used, as described below, to ascertain the filter component values and the insertion loss of the filter, under the assumption that the load may be modelled as a parallel combination of capacitance and resistance. In practice, this assumption is often valid.

As noted above, the user must input information to the processor which will enable the processor to ascertain the type of filter, for example a filter model number displayed on the equipment may be entered which the processor can identify from a look-up table.

The system uses the recognition of the shape of the frequency response to determine whether the fi'ter 's operational. If the oneration of the filter is verified, the component values and 'finally the insertion loss of the filter are estimated. If the filter has failed, the processor can determine -from the frequency response that the filter has failed, and provide some information regarding the type of failure, that is, as to which components in the filt-er have become o::)en-circuit or shortcircuit.

The raw RMS voltage -readings are first converted by calculating the ratio of the test voltage to the open circuit voltage, and expressing the result in decibels. The converted voltage reading is associated with the logarithm of the test- frequency, thus forming a data set. The data sets may be used to plot a frequency response on a loaarithm-Jc-logar-Jthrr,ic scale of voltage ratio 'in decibe-'s', versus frequency. However, the verification of the frequency response, and subsequent data processing rewuireci to estimate the filter comi,onents and the -'nsertion loss of te filter, Js performed by mathematical processing cf the data set by a -o.m:3uter using an iterative method to estimate the filter components. The test apparatus may be used to zest capacitive and LC filters which on their output sides are either unloaded, or 'Loaded where the load may be modelled a resistor and capacitor arranged in parallel. The overall iir.iDedance is denoted Z_ However, the main features of the frequency response, such as curve 1-1a in Figure 2B, define--' bv the overall shape of --he freauency response, are substantiallv unaffected over a range Of load conditions. lt Js, therefore, generally unnecessary to have prior knowledge of the nature or magnitude of any such load impedance.

If the load comprises active components, should be tested with the loadunpowered. It -'s an aim of the invention that the test may be performed in situ. -n practce, therefore, the environment should preferably be as noise-free as)oss-ible during testinc and surrounding e'ectric:a-' equ_pment should normally be unpowered.

1 z Characteristic Frequency response of a C-section filter.

Figure 2A is a circuit diagram of the system when the filter being tested is a C-section or parallel capacitive filter. The capacitance is shown as C.in filter 4 in Figure 2A.

Figure 2B shows the frequency response obtained when serviceable and unserviceable examples of a parallel capacitive filter, or C-section filter, are tested using the apparatus of Figure 1. The function:

12 0 x loglo (Zj (Z-+R) (ordinate) against the logarithm of frequency (abscissa, with high frequencies to the right in the diagram of -Figure 2B) is plotted to give the frequency response of the filter, where Zthe parallel impedance of C,, C. and R- (as shown in Figure 2A) at any frequency, in series with the source resistance R_ Curve 1-1a of Figure 2B shows the characteristic frequency response for an operational C-section filter. if capacitor C- is faulty and open- circulit, then the response of the filter under test is independent of frequency and the -p 'ling portion of the normal characteristic, i.e. curve fa portion la, is replaced by curve portion 2. If, on the other hand, the capacitor is faulty and short-circuit, the response is again frequency independent but the overall resistance of the filter is lower, causing curve 1 - la to be replaced by curve 3.

In order to determine the serviceabilitv of the filter the data processing is arranged to discriminate between curves of the types 1, 2 and 3, and thereby to provide an outnut which is indicative of the state of the component under test, whether it is serviceable, open-circuit or short-circuit.

Testing C-section low-pass filters: Measurement The data processing software, when set to examine a filter of this type, is arranged to analyse a data set of nurnber-pairs each representing a point on one of the curves 1, 2 or 3 of Figure 2B. These number-pairs result from - 5 3!app"ying a.c. signals of a range of frequencies, at standard voltage, to the C-section -filter under test. As noted above, each number-pair consists of a digital numerical representation of one of the several test frequency values, pa-'red with a digital representation of the corresponding voltage measurement made by voltmeter 3 across the input of the filter 4 when the signal of that frequency is applied.

With the filter 4 electrically isolated from the signal generator 2, ie switched out of the circuit for example using routing relays, the processor reads zhe value of the voltmeter 3 using a suitable interface detector sub-routine, and this value is stored by the processor as the open circuit vo"tage. If the open circuit voltage alters with frequency, either measurements must be taken at each test frequency or the manufacturer's handbook must be consulted in order to obtain open circuit voltages at the test frequencies. The -filter is then connected into the circuit using an interface relay s-b-routi-'ne controlled by the processor.

The test voltage across the input terminals of the filter 4 is then measured using the voltmeter 3. The CPU 1 reads the RMS voltage across the input terminals of the filter from the voltmeter and stores the result in association with the freauency, and if necessarv with the open- circuit signal voltage at that test signal frequency. The CPU 1 then sends a control signal, generated in accordance with appropriate interface generator sub-routines, to the signal generator 21 to increase the frequency of the cutout sinusoidal waveform, and the new test voltage is m.easured, and the value stored. In practice, tne test Itages at each of a range of test frequencies are usually stored in an arrav in ascending frewuency order, each vc-"tage being iir.i:)licit-v linked to a test frequency.

The test may proceed in one ef two ways; either a complete frequency sweep is made, or the frequency sweep may be limited by the results of earlier data processing. If a co-r,plete frequency sweep is to be made then the above measurements are reoeated at increasing frequencies and the entire set of results is obtained)rior to anv data processing. In one example the increasing test frequencies may be the logarithmic progression 31.1602kHz, 10OkHz, 316.2kHz, 1MHz, 3.16MHz, 10MHz and 31.52MHz. However, limited data may provide enough information to determine, and verify, the shape of the filter, and it may prove more efficient to introduce an element of data processing during the testing procedure. In either case, further measurements, tailored to the specific type of filter, are required.

Estimation of the cut-off frequency The basic testing and processing procedure for finding the cut-off frequency of the filter is the same for C-, L-, T- and pi-section low- pass filters. Subsequent processing, to verify the characteristic frequency response for L-, T-, and pi-section filters is required prior to estimation of the filter components.

requency of the test signal of 1OkHz. In practice, this datum frequency is chosen with regard to the signal generator. Frequency steps are also chosen with regard to the signal generator. It is a simple step to adapt the method to accommodate any test frequencies. The example below is illustrated with a limited set of readings determined at frequencies of 1OkHz, 31.62kHz, 10OkHz, 316.2kHz, 1MHz, 3.162MHz, 10MHz and 31.62MHz. In the presently preferred embodiment, however, 131 frequencv settings are used at closer spacings over the frequency band of 1OkHz to 50MHz. This allows the data to be smoothed, us'ng known windowing techniques, and improves the detection of the turning points for the pi-section filter using the differencing method described below. The frequencies are chosen to be substantially equally spaced on a logarithmic scale.

Assuming the more-limited frequency set is used, once readings of the test voltages, and if necessary the open circuit voltages, at frequencies of 1OkHz and 31.62kHz have been obtained, the ratio of the test voltage V. to the open circuit voltage V,, each frequency is estimated in dB, ie the ratio 201oa._-(V.,/V,,,) dB is estimated. If the dif ference The method below assumes a nominal datum - iB - I z between the ratios obtained at 1OkHz and at 31.62kHz is greater than 3dB, then it is assumed that the cut-off frequency has been exceeded, and an estimate of the actual cut-off frequency can be made by interpolation between the measured data seis. If a sufficiently large number of steps are used, interpolation may not be necessary. if the difference is less than 3dB, the cut-off frequency has not been reached and the processor sends a signal to the signal generator to increase the frequency of the test signal to say 10OkHz. The new open circuit and test voltages are measured and stored. The ratio between the test voltage and the open circuit voltage at --0OkHz is now calculated, and compared with the 1OkHz datum value.

The test freauency is increased in this way in predetermined steps, until the difference between the latest voltage ratio and the 1OkHz (or datum) voltage ratio is greater than 3dB. In practice it- mav be that none of the test frequencies chosen precisely coincide w-it"- the cut-off frequencv because of the finite step sizes. The CPU, therefore, interpolates between the data points to -find the cu.-off frequency. Denoting the lowest test frequency at which the voltaQe ratio is more than 3dB less than the datum -1-eauency voltage dBm, the cut-off - frequency, equation --):

Fm, and the veltage ratio at Fm as F,:1z, is found in accordance with r 3 a' 2 = FM where y = antilog,11 - [Equ 1 - ', 1 Failure Modes for C-section filters If, at any test -frequency, the measured test vo ' tage below 1OmV (RMS), or the ratio of test voltage to open circuit voltage is estimated to be greater than 34dB, then the test is deemed to have fa--'led. There are three types of which could cause this result: the filter load -esstan,ce is below 1 ohm and is outside the specification of the irstrument, the capacitance of the fi-'izer is greater than

17pF and is outside the specification of the instrument, or the capacitor has gone short circuit. In any of these circumstances, the frequency response Is not of the characteristic form. If the frequency response of the filter is not of the expected form, calculation of the filter components and insertion loss are impossible. However, the user is alerted to the failure of the unit, allowing it to be replaced.

- 5 Limitations of test ecluipment Since a signal generator generally has a maximum frequency of 50MHz, the detection of the capacitance values are limited by the filter load impedances. As a general guide, for a maximum frequency of 50MHz and a filter load impedance greater than 50 ohms, the minimum filter capacitance that can be detected is 150p-F; for of 10 ohms the minimum filter capacitance that detected is 50OPF; and for a load resistance of minimum f.-'-iter capacitance that can be detected a filter load can be 1 ohm the _s 5. 0 n 1 Estimation of the filter components for C-section filters. Tn order -Eo estimate the filter components, an estimate c-' the load resistance R. must first be made. In the preferred embodiment the source resistance R. is chosen to be 50 ohms. in practice, the source resistance must be known but need not be 50 ohms. For a low-nass filter at a freQue-cv of 1OkHz, any capacitance of the load C, of the filter C, is negligible, and the resistive component R, of the load mav be related using Ohm's Law to the open circuit voltage, the test voltage and the source resistance, which are all either known or measured. The re-"a--ions'.-ip between the open- ci4rcu-'t voltage of the generator V_ and the voltage V,,h- at the filter inputs is given by:

V LOAD R L V_ R, + R- [Equ(L','i Alternatively, the result can be es.ima-Led from the frequency response.-, s-'nc the voltage ratio in dB at l.kHz:

dB = 201og., By rearrangement:

RL RS + RT X R- dB where X = anz-log - 1H 1 - X', 2 0--- [ Eciu(3 1,] [Ecru (4) and R, can be estimated' by the processor.

At low fre=encies, regardless the filter type, the resistance R. of the capacitor C- is negligible, and the effective resistance R, of the -.1,ier ' is related tc the source and load resistances R- and R- as follows:

R, R R, L 1 (R, + RS The load resistance R- has been calculated, the source resistance R, is known, and, hence, 'i'-e effective resistance 211 R of the filter may be calculated.

For a C-section filter, the value of the filter capacitance C. can be calculated from the cut-off frequency and the effective resistance.

a, n addition, on curve 1 - la, consider point B which is exactiv 3 dB lower than c1B10,,,. (or the lowest measurement At this point:

(Cl + C L) 2 FT F 3dB Rr C where: (CI + Cl) = total circuit capacJ tance C- F-dg the frequency at point B R effective circuit resis--ance R,, - R L R S (RL + R S and hence the total circuit capacitance can be calculated.

2 1, [Equ (5) 1 Estimation of the insertion loss for a C-section filter The insertion loss is frequency dependent and in accordance with an aspect of this invention can be deduced an estimate of the magnitude of the attenuation A of the filter at any frequency. By substituting the known values, the attenuation may be expressed in terms of a frequency response. The insertion loss, I,, may then be estimated. The data processing software, when set to examine a filter of this type, is arranged to calculate the insertion loss from the curve in Figure 2B. The magnitude of attenuation AdB at any frequency f can be deduced from:

A = 201oglo 2 2 C 2 +is + W RS T 1 [Equ (6) where w = 2.1I.f. The insertion loss I, at any frequency is given by:

R A-2'1)1og, dB [Equ (71'/'] where R- is known, and R- C- and A have been calculated _from eaua---cns (41,, (5) and respectively. Writing equations and---1) as a single equation gives:

201og,, R- 2 C + + (,) R S 7 R, 1 - 201og., RL SI 1.

RI + R it --an be useful to arrange that the -frequency response _s --isn-aved visually on a plot providing a useful way for -'.-e onerator to judge the result of the test procedure.

L and T-section Filter Test Procedure Figure 3A shows the test set-up for a ---section filter.

--section filter has two inductors L. and L- and one caDac-'tor C_ The inductors are connected in series between o-e terminal and one output terminal of the filter, and the canacitor I's connected in parallel between the junction bet.,,Teen the inductors, on the one hand, and the other of each .f:'-e input and outuut terminals of the filter, on the other. The ---section filter is a special case of the T-secticn filter and corresponds to inductance L, having a nec-igib-'e value and being replaced by a short circuit as shown in -Figure 3B.

3C shows the frequency response obtained when serviceahle and unserviceable examples of L-section and T-sec-icn low-pass L-C filter circuits or circuit elements are tested using the apparatus of Figures 3A or 3B. The voltages measured for a serviceable example of L-, or T-, section filter result in a freauencv res-conse of the form of sclid curve I (continued at la and Ib' shown in Figure 3C. 7-he pronounced dip in the log-log plot is characteristic of z.,ese i,,:)es of filters and can be used to identify t-a-. the 1 I- filter is 'in serviceable condition, -free from short-circuit and opencircuit faults.

If the caiDacitance C. is open-circuit, then the -frequency response follows curve 1 initially but rises along curve 2a instead of falling along curve la, and the characteristic dip is absent. If the inductance Lis open crcuit (i.e. in this connection is of substantially zero inductance, e.g. because the ferrite bead that should provide it has split and fallen off its associated wire), the frequency response initially follows curve -- and la but fails to turn u)wards along curve lb, following instead curve 2b. If the capacitance C, is short circuit then the response is frequency independent and the characteristic response is replaced by curve 3.

By comparing the measured frequency response with the characteristic frequency response for this type of filter, the operational integrity of the filter can be assessed. In practice, the response may not be as pronounced or clean as the characteristic response. The CPU should therefore desirably be able to detect whether the response is analogous to the characteristic response from imperfect data. it follows from the shape of the frequency response that both 1and T-section filters have cut- off frequencies which can be estimated in the same way as for the --- section filter.

3-- Estimation of the cut-off frequency The test procedure initially follows that of the C-section filter, described above, with open-circuit and test voltages being measured for a range of frequencies. Once the cut-off frequency of the filter has been established, the crocessor must also establish that the frequency response subsequently rises at higher frequencies in order to verify the characteristic frequency response and operational integrity of the filter.

If the response does not fall by 3dB or more below the 50MHz measurement, then the capacitance is open circuit and the filter is defective. In this case, the filter components and insertion loss --anno. be estimated. In practice, this would enable an inoperat--'ve filter to be replaced.

- I- Verification of the frequency response for L- and T-section filters.

In order ±-c establish that the frequency response of the -is correct, the programmable signal generator 2 should then be set to provide a test signal of the cut-off frequency of the filter. At the cut-off frequency, the test voltage is measured. The processor first checks the ratio of test --c open-circu-Ji voltage to ensure that it is 3dB below -he datum value as expected. If the test fails, the processor recalculates the cut-off frequency and retests at the new cut-off frequency until the cut-of,' frequency is est-ahlished. on a logarithmic-logarithmic scale, the frequency response of the filter will falL at 20dB/decade until the effect of the inductors causes the freauency -esccnse to rse. Once the cut-off frequency has been established, the shape of the frequency response can be verified by checking --hat the response subsequently rises.

-f the resoonse rises the difference in dB beiween the datum reading and the current frequency reading will be _ess than the difference between the datum frequency and the previous est freauency reading. Given t'-a-z the response dro-Qs at 20dB/decade, a threshold can be calculated for any frequency above the cut-off frequency which will determine whether the frequency -response has begun to rise.

example below uses test freQuencies of 3.162 x 13 ------ 1 31.62 x and 100 x The processor sets the programmable generator to the freauency given in the first row of Table 1 appended tc this i.e. 3.162 x The,:)rocessor checks that t-'-e voltage in dB at this higher frequency is less than lOdB 0.5dB lower than the value at the datum frequency, the threshold shown in the second column of Table 1. If the test conalition is not met, the proQrammable aenerator is set to t'--- next frequency, i.e. 10 x measuring the test -,To-'tage and converting i": to dE. The new reading is compared with the corresponding threshold. For so long as the condition is not met, the frequency response has not begun to r'se and the shape of the filter is unverified. The frequency must be increased and the next threshold used. if different frequencies are used, the thresholds need to be recalculated using the fact that the fall-off is 20dB/decade to estimate the new thresholds. The test is continued at higher frequencies until the test condition is met. At this frequency the voltage ratio has begun to rise along curve lb as shown by the characteristic frequency response in Figure 3C and the shape of the L- or T-section filter has been verified.

Another method of determining the minimum frequency is described below in relation to pi-filters.

2'- 3, If a generator frequency of 50MHz is reached i.e. the maximum frequency, and the test condition has not been met, then the filter is faulty, and inductance L. of the T network is open-circuit. On the other hand, if the operation of the T -filter is verified, the filter components can then be calculated. Note that it is not possible using this procedure to establish whether inductance L- is satisfactory.

Estimation of the filter component values for L- and T-section filters. in order to estimate the filter component values, the filter load resistance, RLI must first be estimated. For Land T-section filters, over the pass-band the resistive component of the inductors and capacitors will be negligible. The load impedance, ZLI is 'assumed to comprise a parallel combination of resistance RL and capacitance C.. The induc.ance of the T filter has two components due to the two 'nductors L. and L_ The inductors have associated loss resistances R,- and Rel which are frequency variable. At 1OkHz, these associated loss resistances are negiigible, but at frequencies of F3,, and above the resistances are significant. The value of R, can be calculated as for the --section filter, e.g. from equation (4).

- 1 1 L S - The zotal caDacitance C- c--- the circuit -'s a function of.he effect-'ve resistance R, and can be calculated from equation (18):

C- = cl 1 2!1 E-, d3 RP where:

- ' + CL RS (RL + Rel + R e2 RS + RL + RIi + R e2 [Equ (8)] [Equ (8)a] There are too many unknown variables to determine the value of --- at this stage. In order to proceed the values of R,. and R,- are set to zero, -'.e. R,,=R,-0, in order to estimate an interir. value for R-. Further -information is available from the -frequency response by considering the minimumin the freauency response, at frequency f,,,. The minimum of the frecuencv response -'s estimated using an inflection alacrithm. The inflection alaorithm is described below for t'-e pi-section filter. m-he minimum in the frequency response caused -'-v the resonance of the inductcr L- and capacitor C-. At this point L- and C. are related by equation:

1 -) 2,, 2 n), ' f.,ir c., [Equ (9) In general for a T filter, C. -will be small compared wIth C, i- which case C- = C, and C. is given by equation (8) A-'so the inductors L- and L- take the same value, making R_ = R- -ipractice. R., mav be related to the m--'niTu,-n of the frequency response:

D R- Rl = (1- D" where:

[Equ (10)] D = antilogl,) ( dB at fmin [Equ (11) 1 Given that the above assumptions hold, a value for Rel = R,-, can be used to estimate a new improved value for C. in Equation (8) and hence new values for R,, and R,,2. This iterative loop is repeated until successive values of C, are within a predetermined amount such as 1% of each other, and this is then taken to be the value of component C.. Finally, the values of L, = L2 can be calculated from equation (9).

Estimation of the insertion loss for T- and L-section filters The insertion loss IL of the T filter can be found by calculating:

I- = 201og,, L 3 Z-R, + Z 2 Z 1 + Z 2 Z 3 + Z LRS + Z IZI + ZLZ3 + Z,Rs + Z 3 Z J - 201ogI, ZL Z, + R S] where:

Z1 R el + jwL 1 Z 2 R e2 j coL 2 Z = 1 3 j WC 1 ZL = RL The insertion loss L of the L filter is the same but with L, set equal to zero. Thus the insertion loss of the L filter can be found by calculating:

2 Q log,, L 3 IL = 1 ZLRS + ZLZ1 + Z LZ3 + Z 3 RS + Z,Z - 201ogl. ZL Z, + R, - 1 - _1 - 71 1 Pi-section filter test procedure -7igure A shows the test set-up for a pi-secticn _ow-pass filter. The pi-section filter 4 has one inductor L and two capacItors C.- and Capacitor C- -'s conne=ed across the two in-out terminals and capacitor C- is connected across the two output terminals. Inductor L, couples one input terminal to one outi:-ut terminal, the other of eac" of the input and output terminals being connected together. The characteristic frequency response is shown in Figure 4B. The tained for serviceable and unserviceable r crcu-ts or by a serviceable example of such a filter creates a frequency response of the of solid curve 1 starting at la and continued at lb, lc and ld in Figure 4B, with a noticeable S-shape in this log-log plot. This corresponds to an example of the filter in serviceable condition, free from shor-,-circui.: and open-c-'rcuit faults. The types such a filter have little effect frequency response ob examples of a pi-section low-pass L-C filte circuit elements are shown. The data given of load normally applIed = on the overall shaoe of the frequency response, shifting the whole response hor-izon.:al-',v or vertically depending on the reactance of the load.

7 47 remote capacitance C- is ooen circui-, then -he response corresponds to a curve made up of iDort-ions la, 'La and ld. If the caQacitance C, is open-circuit then the response approaches that of an L-section filter i.e. with a dip and a rise, following curve la, lb,!c, and 2b. If the inductance, L or L, Is open circuit, the- the frequency response corresponds tc curve portions!a, lb and 1 L. C. If either capacitance C. or C- is short circuit, then the curve is replaced by curve 3. An inductor provided by a ferrite bead threaded around a wire does not generally develop a short-circuit fault, though its inductance may largely disappear if the bead is damaged or missing.

As shown, curve portion lc has:)ositi7.re slope on Figure B. Depending on the component values, however, it r,,,av not be such a sharp in-Election and may have a negative slope, which is nevertheless less than the negative slope of curve cor.i-4cns lb- and ld, which have substantially the same slope.

- 29 2 2 0 1 z j - The system will however detect a less-pronounced inflection of such a tvDe.

The steps in verifying the frequency response, and es,zin,ating the filter component values and the insertion loss of the filter, are shown in the flowchart of Figure 5. This flowchart will be self-explanatory to those skilled in the art in conjunction with the above description and a verbal descriction is not therefore repeated here.

Estimation of the cut-off frequency The characteristic response for the pi-section filter s simlar to that of the L- or T-section filters at low and differs in that the response rises and then falls, being effectively characterised by an S-shape. Using the method described above with reference to the C-sectien filter, the cut- off frequency of the filter is first determined.

Verification of the frequency response for a pi-section filter.

!he process for determining that the frequency response rises -'s the same as for the L- and T-section filters. Once the rise of the frequency response has been detected the fall-off of the response must be verified. Denoting the freQuencv at which the response begins to rise as F,,,,, the signal generator frequency-is increased by a factor of 3.162, for example if F,1,,=10 x F,:.,.ff, then the signal generator _requency is set to 31A2 x F,,.t.ff, At this new frequency, the difference in dB between the previous value and the current value s'-ould be greater than lOdB, with the current value being more than 10dB lower than the previous value. -If this condition is met, the filter is functional. If the new value falls short of being 10dB higher than the previous value, capacitor C, is opencircuit.

Once the frequency response has been verified, i.e. the shape of the response has been shown mathematically to correspond to the characteristic shape of the particular f-'---er, the com:)onent values of the filter rr.av be estimated.

z -j. - Estimation of the filter component values for a p!-section filter.

The filter load impedance is assumed to comprise a parallel cc.mbinat-'on of resistance R- and capacitance C_ The of the pi filter comprises an inductor L., and associated loss resistance R,, which is frequency dependent and which is negligible at 10 kHz, but which has a significant value at F-, and above.

The value of R- is estimated in the same manner as for -section filters using the 1OkHz datum frequency L and T test data and Equation The total capacitance of the circuit is given by:

1 C- = - 27 F 3dB RP were:

R- (R + R L e R = - RS + R L + R e (-- = C C- L [Equ (12' j [Equ --2a 1 Note that, at this stage of compu--at--'on, R. at F_ is unknown and a value of R, = 0 is used to obtain an interim value of R_ Estimation of the inflection points for a pi-section filter. Further information is available in the inflection --oints of the frequency response. A curve inflection derIvation method -'s used which is based on the difference in the voltage ratio between the open-circuit and test voltages dB at different frequencies. In practice, 131 test f-equencies, equally spaced on a logarithmic scale, are used _n the measurement of the test voltage and subsequent calculation of the dB level at the -zest voltage. The results are placed In an a--ray in frequency order and then smocthed using a window length o 5 samples. The __rst and last two 3 1 1 - 3 1 samples remain unchanged in the array due to the window length. Once smoothed, the inflection points of the array must be estimated. For regularly spaced frequency readings, the slope of the curve is given by the difference in dB between any two of the readings. In practice, a smoothina window length of 7 samples is used, i.e. samples n and (n - 7) are used, hence:

dy/dx = (dB[n] - dB[n+71)/7.

A new array is used to store the difference values, with the first and 'Last three entries being null due to the length of the difference window chosen. point C, which as shown in the minimum in the curve, is estimated bv determining the lowest frequency at which the difference va"ue falls below the average of all the difference values.

The minimum frequency of an L- or T-section filter frequency response mav also be calculated in this manner. inflection point D, which as shown in the subseauent peak, is then estimated as the first entry in thedifference array after point C which is greater than the average value. At point C denote the dB level = dB(fmin)r and frequency = f,l:_ At point D denote the dB level = dB(fmax) and the freauency Using the fact that the frequency response drops at 20dB/decade for a C- section filter, the test data mav now be extrapolated to estimate a point F. The value in dB at point D is used to establish a point E on curve portion ld of Figure 4B which has a value 3dB lower than point D:

dB, = (dBm.,. - 3) where the frequency at this point E is denoted f,.

The position of the point F, corresponding to a on curve portion 2c of Figure B with the same value in dB as point E, has a frequency f., the attenuated reauencv, where:

f F,,,, x A :Equ 13)l where:

A = antilog,, {( dB,O,,; - dB,)/20}. The processor then estimates the extrapolated frequency ratio k where:

- -2 - k = f h a [Equ (14) and calculates the value of the first capacitance C. from:

c 2 5 [Eau In general, for pi filters, C. = C,, that is both internal capacitances are equal. In this case:

[Equ At -the first minimum after F,,,, that is at f,.., the inductance L is in resonance with C- + CL) Hence, the effective inductance of the circuit is aven bv:

L - 2 ' C, + C, (2 -1 ') f:ri., 1 [Equ (17', Also at this freauency it is possible -io derive an equaton -for R,:

R e where:

W =121 11-1 E B R - 17 C-2 R S n-n -p antilog,,, [dBs at -n -- n I 20 [Equ (18', 1 [Equ (19', The crocessor now replaces -he in-itial value for R. in equation.;'2a, -- c give a new value for C- i- equaton (12 and hence a new value for C in equation (15) and a new value for R- in equatien (18). This interactive loop is repeated until successive values of C. are within a predetermned ainount such as 1' o.: eac'- other. Figure 6 shows a flowchart corresponding tc the estir-riation of C- as descr--'be.d above. As with Figure 5, a verbal description of this F1gure unnecessary.

Estimation of the insertion loss of a pi-section filter. The insertion loss IL Of the Pi filter is given by:

I, = 201oglc Z L (Z i 2Z 3 + Z 1 z 3 2 + Z 1 z 2 z 3 p where:

P = R [Z2Z, + 2Z,Z Z + Z 2z + Z z 2 + Z z 21 S - - - 2 3 2 3 1 3 2 3 - R Z- FZ,2 + 2Z Z + 2Z Z + z 2 + 2Z,Z + Z 2 S 1 2 1 3 2 - 3 3 z L z 2 z + Z z 2 + 2Z Z Z + z 2 z + Z Z3 2 L " 1 2 1 2 1 2 3 1 3 1 - I- - Z 1 2z 2 z 3 + Z 1 z 2 2 z 3 + Z 1 z 2 z 3 [ Equ (2 0) j and where:

Z, = 1 j W C.

Z, = R + jiL 1 e e Z3 = -: 1 j W (CL + C 2) Z- = R_ Extension to other filter characteristics the case of high-pass L-C filters that respectJvely corresnond in configuration to the low-pass filters of Figures 'L to 4 after changing L components to -- components and vice versa, the forms of the log-log characteristic curves are for many purposes substantially similar except for the reversal of the log- frequency scale, i.e. in the case of the high-pass filters the low- frequency end is to the right of each plot instead of to the left.

Several other types of filter show other characteristic da.:a patterns and differences between serviceable and faulty - 34 examples, as may be ascertained by the reader skilled in the art by using apparatus embodying the present invention, e.g. that of Figure 1, to obtain data and curves corresponding te those of Figures 2B, 3C and 4B in the case of each of severafilter types that may be of interest, using serviceable and faulty examples.

In addition to the use of the test system and method as described to detect fault conditions such as short circuits and open circuits, and component values, the system and method can also be used tc detect other characteristics of filters or circuit elements under test. For example, also detectable are changes, e.g. by way of deterioration, with time, and/or with use or wear, of the component va-ues in such filters or circuit elements, or changes in value of the load impedances connected to such elements.

Where alternative measurement circuit arrangements are used, e.g. current measurement or use of a high-impedance constant-curren' source, corresponding changes generally need to be made in handling the forms of data that result, as will be appreciated by the skilled reader.

The present invention is susceptible to many other modifications and variations and the present disclosure extends to other combinations and subcombinaticns of the features mentioned above and illustrated in the draw-ngs.

1 - Table 1

1 1 Test Frequency Test voltage ratio 1OkHz (datum frequency) Vfilter f:ut-c" VIiter 3dB f -,- x 3.162 V,Ite, lOdB f 2,4t-", X 10 v_--- 20dB f n- x 31.62 Vfi-ter 30dB f X 100 V, 40dB f f x 316.2 V'--ter - 5OdB f X 1000 Vfi--er - 60dB 36

Claims (7)

1. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs, comprising the steps of: (a) providing a low-pass filter to be tested, the filter having inputs; applying successive a.c. test signals over a predetermined range of different frequencies to the incuts of the filter; measuring for each of the frequencies the test voltage developed across the inputs of the filter; obtaining estimates of the open circuit voltage at one representative frequency or over the predetermined range of frequencies; estimating the voltage ratio in dB of the test voltage to the open circuit voltage at each of the frequencies; associating the ratios in dB with the particular frequency of the 'Lest signal and ordering the associated data by frequency; estimating the frequency at whic1, the ratio is 3dB less than the estimated voltage ratio at the lowest frequency; comparing the associated data at the 3dB frequency with the next highest frequency data to determine whether the voltage ratio has decreased by an amount equivalent to less than 20dB per decade; (i'i repeating step (h) with successive sets of associated data until the ratio has decreased by an amount equivalent to less than 20dB per decade or the last test -frequency is reached; continuing to step through the associated data until the ratio increases by an amount equivalent to less than 20dB per decade or the last test frequency is reached; differencing the associated data and subtracting the mean value to form a set of d. Lfferenced data; (b) (c) (d) (e) (f) (g) h) 2 =.

3, z (k) (1) stepping through the differenced data and noting the first negative value and if applicable the next subsequent positive value as the turning points of the frequency response.

2. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs according to claim 1, including the step of using the turning points of the frequency response to estimate the component values of the filter.

3. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs according to claim 1, in which steps (a) to (k) are fully automated using a microprocessor.

4. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs according to claim 1, including the step of interpolating between the associated data to estImate the frequency at which the ratio is 3dB less than the estimated voltage ratio at the lowest frequency.

5. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs according to claim 1, wherein the a.c. test signal has a voltage less than 0.7 volts.

6. A method of automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs according to claim 1,wherein the step of differencing the data comprises determining the lowest frequency at which the difference value falls below the average of all the difference values.

3C

7. A method of automatically verifying the characteristic -Prequency response of a low-pass filter having inputs and output connections according to claim 1, in which - E the associated data is differenced using a windowing technique.

(a) (b) i c 2 E 3 C (c) (d) (e) (f) 3 Z, ' i) Apparatus for automatically verifying the characteristic frequency response of a low-pass filter having inputs and outputs, comprising: a signal source for applying successive a.c. test signals over a predetermined range of different frequencies to the inputs of a filter; a meter for measuring for each of the frequencies the test voltage developed across the inputs of the filter; means for obtaining estimates of the open circuit voltage at one representative frequency or over the predetermined range of frequencies; means for estimating the voltage ratio in dB of the test voltage to the open circuit voltage at each of the frequencies; means for associating the ratios in dB with the particular frequency of the test signal and order-ing t'ne associated data by frequency; means for estimating the frequency at which the ratio is 3dB less than the estimaied voltage ratio at the lowest freauencv; (g) means for comparing the associated data at the 3dB frequency with the next highest frequency data to determine whether the voltage ratio has decreased by an amount equivalent to less than 20dB per decade and for repeating the foregoing wizh successive sets of associated data until the ratio has decreased by an amount equivalent to less than 20dB per decade or the last test frequency is reached; (h) means for continuing to step through the associated da.:a until the ratio increases bv an amount equivalent to less than 20dB per decade or the last test frequency is reached; means for differencing the associated data and subtractina the mean value to form a set of differenced data; and (j) means for stepping through the differenced data and noting the first negative value and if applicable the next subsequent positive value as the turning points of the frequency response.