GB2255256A

GB2255256A - Method of and apparatus for reducing vibrations

Info

Publication number: GB2255256A
Application number: GB9107765A
Authority: GB
Inventors: Malcolm Alexander Swinbanks
Original assignee: W S ATKINS ENGINEERING SCIENCE
Current assignee: W S ATKINS ENGINEERING SCIENCE
Priority date: 1991-04-12
Filing date: 1991-04-12
Publication date: 1992-10-28
Anticipated expiration: 2011-04-12
Also published as: GB2255256B; FR2675296A1; GB9107765D0; DE4212176A1; AU1474992A

Description

2 2 -, c - -,:"

METHOD OF AND APPARATUS FOR REDUCING VIBRATIONS The invention relates to a method of reducing vibrations and to an apparatus for carrying out the method. The invention is particularly but not exclusively applicable to noise reduction.

Various systems which seek to cancel or attenuate vibrations have previously been proposed. UK Patent No 1304329 discloses a method of reducing the noise of rotors, utilizing a wave generator mounted adjacent a rotor axis to produce a wave having a front which spreads radially from the generator and with a frequency effectively the same as but in antiphase to the noise field, so as to destructively interfere therewith at locations distant from the rotor.

A further method for the active attenuation of recurring vibrations is described in UK Patent No 1577322. Here, a primary repetitive vibration is at least partly cancelled by a secondary vibration which is a synthesized waveform generated by a waveform generator fed with a triggered signal which is derived from the source of primary vibration. The waveform generator can be arranged to adapt its output on the basis of success achieved in cancelling the vibrations from the source. The adaptive technique employs a microphone located in the region in which cancelling is required, a simple memory being used to determine whether, following a change in secondary vibrations, the cancellation action has been improved or not.

A further method and apparatus for cancelling vibrations is disclosed in UK Patent No 2107960. The method disclosed therein employs a technique in which residual vibrations resulting from interference between primary vibrations from a source of vibrations, and secondary vibrations from a driven actuator are transformed from the time domain into the frequency domain to form a plurality of independent pairs of components which define the amplitudes and phases of the residual vibrations in the frequency domain, each synchronized to the repeat period of the primary vibrations. These frequency domain components are separately modified, and transformed back into the time domain and are used to provide drive signals for the actuator. The control of the separate modifications of the pairs of components is carried out so as to reduce the amplitude of the residual vibrations.

The present invention is particularly concerned with reducing vibrations which are of a generally repetitive nature, and which comprise a plurality of discrete frequencies each of which may be tire varying in amplitude.

A particular case in which this type of vibration is a nuisance is in an aircraft cabin where the noise produced from aircraft propellers consists mainly of a plurality of repetitive signals at discrete frequencies.

According to the present invention there is provided a method of reducing unwanted vibrat44.ons in a zone wherein such unwanted vibrations arise from a source of vibration which generates a signal having at least one vibrational component at a substantially constant frequency over a detection period, the method comprising:

within the detection period deriving a tracking signal at said constant frequency; providing in-phase and quadrature components of said tracking signal; monitoring vibrations at each of a plurality of predetermined locations within the zone to provide a set of observation signals; 3 multiplying together each of said in-phase and quadrature components of said tracking signal with each of said observation signals and filtering each resultant to provide a set of output values, each of which represents the extent to which the respective observation signal contains vibrations at said constant frequency; and utilising the set of output values and the in-phase and quadrature components of the tracking signals to produce a plurality of drive signals which when supplied respectively to each of a plurality of actuators cause vibrations to be produced within the zone which tend to cancel those generated by the vibration source.

The method can also be applied where the source of vibration generates a plurality of signals at respective different but substantially constant frequencies, in which case there are generated a plurality of tracking signals at each of said constant frequencies, and in-phase and quadrature components are provided for each tracking signal.

The invention also provides an apparatus for reducing unwanted vibrations in a zone wherein such unwanted vibrations arise from a source of vibration which generates a signal having at least one vibrational component at a substantially constant frequency over a detection period the apparatus comprising:

a tracking processor for providing in-phase and quadrature components of tracking signals at said constant frequency; a set of vibration monitors arranged to monitor vibrations at each of a plurality of predetermined locations within the zone to provide a set of observation signals, an observation processor coupled to said tracking processor and to said vibration monitors and operable to multiply together each of said inphase and quadrature components of said tracking signal with each of said observation signals and to filter each resultant to provide a set of output values each of which represents the extent to which the respective observation signal contains the constant frequency; a plurality of actuators which, when driven, supply vibrational signals to the zone; and output processing means coupled to said tracking processor and to said observation processor and capable of utilising the set of output values and the in-phase and quadrature components of the tracking signal to produce a plurality of drive signals which when supplied respectively to each of the plurality of actuators cause vibrations to be produced which tend to cancel those generated by the vibration source.

For use in an environment where the source of unwanted vibration generates a plurality of signals at respectively different but substantially constant frequencies, the tracking processor can be arranged to provide in-phase and quadrature components of tracking signals at each of said frequencies.

In one embodiment, the in-phase and quadrature components of the or each tracking signal are derived from signals from one or more sensors positioned adjacent the vibration generating source.

The zero crossings of the signals from the sensors are detected to determine the frequency of said signals. A look up table comprises a plurality of memory locations at which are stored respective values which, when output sequentially, generate a single cycle of a sine wave which is used to constitute the in-phase and quadrature signals. The values are output sequentially at a rate dependent on the frequency determined by the zero crossing detections.

In another embodiment the in-phase and quadrature components of the or each tracking signal are derived from one or more reference sinusoid generated by a waveform generator. The frequency of the reference sinusoid is identified and the signal is filtered to clean it up. An accurate 90 0 phase shift is introduced by a delay circuit whose delay is obtained from a look up table in dependence on the identified frequency.

Preferably, the step of filtering of the resultant of multiplication of said in-phase and quadrature components of the tracking signals together each of said observation signals is controlled by selecting a filter from a plurality of available filters of different response characteristics. The function of the filter is to average the resultant with respect to time. The response characteristic of the filter affects the accuracy of the average in relation to the time taken to produce it. The ability to select filters enables the dynamic response of the vibration reduction apparatus to be adapted to the requirement of the particular environment.

A separate filter may be used for filtering each resultant of multiplication associated with each tracking signal of a different constant frequency, the separate filters having different response characteristics.

It is well known in noise reduction systems that the monitoring of noise at predetermined locations in a zone and injection of vibration signals based on the result of such monitoring, can, while reducing the noise at those locations, result in increased noise elsewhere within the zone and perhaps even in increased overall noise average. Techniques for avoiding this are known, as described for example in the paper "Proceedings of the Institute of Acoustics;" Vol 8; Part 1 (1986) of M.A. Swinbanks, and are termed in the art "optimisation calculations".

Preferably therefore, the set of output values are subjected to an optimisation calculation which modifies the set of output values to produce sets of modified output values associated with each of the plurality of actuators. The optimisation calculation is carried out so that subsequent multiplication and summation gives appropriate actuator drive signals to provide substantial overall vibrational cancellation in the zone. The optimisation calculation may be performed on the basis of stored system transfer values, for example as is dqscribed in the papers "Proceedings of the Institute of Acoustics"; Vol 8; Part 1 (1986) and Vol 6; Part 4 (1984) of M.A. Swinbanks.

The modified output values, and the in-phase and quadrature components of the tracking signals may be processed in accordance with the following calculation to give actuator drive output signals Ym:

> N_ (Anm Cn + Bnm Dn) = Ym n=1 where Ann. and B= represent the modified output values associated with m th actuator and n th tracking signal frequency for in-phase and quadrature components respectively, and Cn and Dn represent respectively the in-phase and quadrature components of the n th tracking signal.

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings in which:

Figure 1 shows schematically the inputs and outputs of a system for reducing vibrations; Figure 2 shows schematically the architecture of the vibration reducing system; Figure 3 shows schematically the operation carried out by observation and output processors of the architecture of Figure 2; Figure 4 shows in block diagram form the operation of an output processor; Figure 5(a) shows the circuitry of a tracking processor which uses input reference sinusoids; Figure 5(b) shows the circuitry of a tracking processor which uses inputs from sensors positioned adjacent to the vibration-producing source; and Figures 6(a) and (b) illustrate by way of computer simulation the effect of the use of various types of filter; Figures 7(a) and (b) illustrate the effect of the use of various types of filter on a higher frequency signal than that of Figures 6(a) and (b); Figures 8(a) and (b) illustrate the effect of the use of various filter types on a signal including a random noise component; Figure 9 shows schematically the architecture of the system for reducing vibrations.

In Figure 1 of the drawings,-the inputs and outputs for a vibration reduction system are shown. The vibration reduction system is arranged to supply to a zone suffering from unwanted vibrations from a vibration source vibratory signals arranged to cancel or significantly reduce these unwanted vibrations. The vibration reduction system is particularly applicable to reducing noise in aircraft cabins arising from the propellor rotation. In the following, the preferred embodiment of the invention is described in this context, but it will be appreciated that the invention can be applied in many other situations to reduce unwanted vibrations. A processor system 2 receives a plurality of tracking inputs 4 which represent primary vibrations produced by the vibration source, and a plurality of observation inputs 6 which are derived from monitoring the zone where vibrations are to be reduced. The processor system 2 utilises the tracking and observation inputs 4,6 to produce a plurality of output drive signals 8 which control actuators to produce the vibratory signals intended to cancel the unwanted vibrations in the zone.

The present invention utilises the realisation that in many cases of unwanted vibration, the primary vibrations are at a plurality of discrete frequencies, each of which may be time varying in amplitude. In many cases knowledge of these frequencies is available from the physical configuration of the vibration generating source. For example, in the case of an aero engine propeller, a fundamental frequency of vibration is given by the product of the frequency of the propeller shaft rotation and the number of propeller blades. Typically, the frequency of rotation may be of the order of 20 Hz, and the propeller may have four blades; thus the fundamental frequency is 80 Hz, with harmonics at multiples of this frequency.

Figure 2 shows in more detail the components forming the processor system 2. The processor system 2 includes a tracking processor 10, an observation processor 12, and output processing means 13 comprising an output processor 14 and a host computer 16.

The tracking inputs 4 selected to represent the primary vibrations produced at the vibration source are input to the tracking processor 10. In the case of an aircraft having two engines the frequencies to be controlled might be selected as the fundamental frequency and three of its harmonics for each engine. The tracking processor lo is arranged to produce sinusoidal tracking signals at each of the discrete frequencies generated by the vibration source which it is desired to control, and to provide for each of these tracking signals cosine and sine wave components (referred to herein as in-phase and quadrature components I and Q respectively) phase-locked to the tracking inputs 4. In the above mentioned example of an aircraft, where it is desired to control a fundamental frequency and three harmonics for each of two engines, the tracking processor 10 outputs eight signals (four I or cosine signals, and four Q or sine signals) associated with each engine, i.e. a total of sixteen signals.

It is important that for each discrete frequency to be controlled, accurately generated I and Q components of the tracking signals which are phase-locked to the appropriate tracking inputs should be generated. Two possible methods by which the tracking processor 10 can generate the I and Q components of the tracking signals are described herein. It will be appreciated however that other methods may be utilised.

Figure 5(a) shows a tracking processor arranged to produce in-phase I and quadrature Q components of the tracking signals. Reference numeral-18 denotes a unit for receiving eight reference sinusoids produced by frequency generators. Complementary cosine components are generated from these input sinusoids by effecting a delay to the input sinusoids which is equivalent to a 90 0 phase shift, in a manner described below.

For each input sinusoid there is provided a DC filter 20 for filtering the sinusoid. The output of the filter 20 is passed to a finite impulse response (FIR) 2nd order filter 22. In essence, this filter identifies the frequency of each sinusoid 18. FIR filters are known, but the present tracking processor uses the known FIR filter in a new way. That is the FIR filter is implemented with only three filter coefficients: 1,2A,1; where 2a is a variable filter coefficient, where:

a = cos 2 Tr f A t and A t is the sampling time (for example 1/1500 s).

(Eqn 1); The output of the FIR filter 22 is passed to an LMS (least mean square) update unit 23 the output of which is fed back to the FIR filter 22 as an error signal to alter the coefficient 2a. An error control loop is thus defined, the gain constant of which is controlled by the LMS update unit 23. The output of the DC filter 20 is also passed to an infinite impulse response (IIR) filter 24 which reduces signal noise and distortion. The IIR filter 24 acts as a narrow band filter which is locked on to the specific frequency of signal to be processed under the control of the output of the LMS update unit 23. The output from the IIR filter 24 is normalised to unit amplitude in an automatic gain control filter 26, the output of which is a cleaned-up ncwmalised cosine wave of known frequency. The output of the LMS update unit 23, representing the FIR coefficient value, is also passed to a look-up table 30 which stores values of filter coefficients required for a Q-filter 28 against values of a from the FIR filter 22. The Q-filter coefficients generated for a particular value of a are those required to effect an appropriate time delay to the cosine signal to give the complementary sine wave.

The method of operation of the Q-filter is as follows. Given an accurate normalised cosine tracking wave, a corresponding sine tracking wave can be obtained by delaying the cosine wave by exactly 90 degrees. In practice, there may not be available a continuously adjustable delay, but rather a set of signal values sampled at preset delay intervals corresponding to the sampling rate (eg integer multiples of 0.67 milliseconds at a sampling rate of 1500Hz). If that integer value of delay which is the closest approximation to the required value is selected, the output signal can be regarded as a sum of the ideal 90- degree shifted wave, together with an unwanted error which is, by definition, in-phase with the input cosine.

Since this error component is in-phase with the input cosine, it can easily be eliminated by subtracting an appropriate proportion of the (undelayed) input. The net output then consists only of the required 90degree phase-shifted component.

This can be achieved using a two-coefficient FIR delay line filter, with coefficients w-and P for the zero delay and "nearest 90-degreell delay respectively.

The integer delay sample number giving closest to 90-degree phase shift is:

DLY = INT (l/4 X 1 + 0.5) f A t The Q-filter is then set up vith coefficients (0) and (DLY). The real and imaginary outputs of this filter become:

Re (In Phase) = tK + f cos 2 -K f (DLY x.6 T) Im (out of Phase) = @ sin 2-wf (DLY x 6 T) If DLY were exact, 2 ITf (DLY x AT) would equal -rr/2, so taking cK = 0 and 0 = 1, Re = 0, and Im = 1. In practice, DLY will not be exact, but if:

P = 1/ sin 2 ir f (DLY x T) and ce- = - g cos 2 -ff f (DLY x T) a result of Re = 0 and Im = 1 would be obtained.

For the required range of frequency values, the corresponding values of DLY, c4 and C can be precalculated and stored in the look-up table 30 (referred to as a Q-filter table). The output -a of the FIR filter 22 can then be used as an address pointer to this table to select the appropriate values corresponding to the specific frequency of the input signal. As the frequency changes, so the coefficients and delay value are changed accordingly, and the filter maintains an output which is always of unity gain and 90 degrees out-of-phase with its input.

The outputs from the tracking processor 10 are therefore the required inphase I and quadrature Q components of the tracking signals, and these are generated for each of the eight discrete frequencies.

The values of from the FIR filter 22 a are also fed to the host computer 16 which calculates ':'&rom Eqn 1) the precise frequencies which these values represent, for use in subsequent process steps.

Figure 5(b) illustrates a tracking processor for implementing another method of providing the tracking signals. In this method detectors are synchronized directly with the vibration producing source, or transducers are located close to the vibration producing source. In essence, the I and Q components are derived directly from values stored in look-up tables in a processor memory by moving through these tables at a rate derived from the input signal.

Each signal 31 (analogue or digital) from a transducer on or adjacent an engine propeller is sampled at a fixed high-speed sample rate (for example 60kHz) in a waveform acquisition unit 32 and passed through a D.C. filter 33. The zero crossings of the input signal are detected in a zero crossing detection unit 34 and used to derive the time for one rotation of the propellor shaft, from which the rotation rate can be derived using a look-up table 36. Information as to the frequencies which it is desired to control such as the harmonics of the basic propeller shaft rotation rate (which are dictated by the number of propeller blades) is previously stored in the processor memory as the harmonic number, which represents the frequency multiple of the basic propeller shaft vibration rate which is to be generated. A look-up table 40 comprises a memory having a plurality of addressable locations, termed herein elements. There is stored over a predetermined number of elements a single-cycle of a sine wave. That is, each element stores a value between 0 and 1 depending on its relative position in the addressing sequence so that sequential addressing of all elements generates a sine wave. The number of elements required to be addressed to recreate one cycle of the sine wave is the table size. In the described embodiment the table size is 8192.

In an address calculation unit memory addresses in the look-up table are derived from an interpolation calculation:

cos address = (time into cycle x rotation rate x harmonic number x table size) + offset sine address = (time into cycle x rotation rate x harmonic number x table size) where the offset is the address shift required to represent the appropriate 90 0 phase shift between cosine and sine signals.

In the described example, the offset is 2048 elements.

It will be appreciated that it is not essential for the look-up table to have stored an entire cycle of sine wave; the table could be arranged to include one quarter or one half of a sine wave, the address calculation unit being arranged to select the appropriate addresses to generate the entire cycle.

The values output from the look-up table 40 generate the I and Q components of the tracking signals, the rate at which the values are output being dependent on the rotation rate and harmonic number. It will be apparent that should the input signal temporarily be lost, then those I and Q components will continue to be generated based on the lastavailable information detected by the zero crossing detection unit 34. It will also be apparent that this tracking processor can operate on any type of periodic input signal, provided only that the signal has zero crossings once any d.c. offset has been removed.

Where the vibration-producing sources are two engines, the processing sequence is operated for input signals 31 derived from each engine. Where it is desired to control four frequencies for each engine, as is suggested in the case of a twin-engine aircraft, eight sets of I and Q signals i.e. a total of sixteen signals are output. The I and Q components of the tracking signals are led to a serial dataport. The engine rotation times are additionally input to the host computer 16 to enable it to derive the frequency information required for subsequent processing steps.

The observation processor 12 receives the observation inputs 6 which are derived from transducers located in the zone of unwanted vibrations and arranged to pick up the vibratory levels in that zone as an indication of the performance of the vibration reduction system. In the case of an aircraft, these transducers can conveniently comprise a set of microphones distributed over the cabin area. Preferably the microphones are arranged at locations where unwanted noise is at its most annoying, for example by the headrests of individual seats or just below overhead lockers.

The observation processor 12 also receives the in-phase I and quadrature Q components of the tracking signals and derives from these and the observation inputs 6 measurements of the in-phase and quadrature levels in each observation input 6, at each discrete frequency to be controlled. From this information, the corresponding 1 and Q levels for the output processor 14 are derived, these being processed by the output processor 14, along with the phase-locked I and Q components of the tracking signals, to produce drive outputs 8 for actuators which produce the vibrations which tend to cancel the primary vibrations.

Figure 3 shows in its left-hand half, circuitry for implementing the processing operation of signals carried out by the observation processor 12. The observation processor 12 takes the I and Q components of the tracking signals from the tracking processor 10, and multiplies each observation input 6 together with each of the I-and Q components, and filters the resultant. The inventor has realised that the multiplication of signals representative of the expected discrete frequencies causing unwanted vibrations in a zone together with signals representing the actual conditions in that zone followed by filtering results in a signal which renders it particularly useful for further analysis yet contains all the vital information about the effect of the vibration reduction system on the zone. The result of multiplication and filtering is a set of output values Iout, Qout representing the extent to which the respective observation signals 6 contain each constant frequency. This can be seen from the following:

An observation signal 6 can be represented as:

A. sin( wt + j6) + n(t) where A.(t), 9(t) are the unknown and time varying amplitude and phase respectively, w(t) is also time varying but can be tracked, and n(t) is a noise component (possibly containing similar signals at other frequencies).

This can be written:

A. sin (wt + 95) = A cos Wt + B sinwt where A = A. sin 0, B = A. cos; It is desired to determine A(t) & B(t), which are the level of in-phase and quadrature components of frequencycu in the observation signal.

If accurate in-phase and quadrature components of the tracking signals representative of the expected frequency w.(t)(= 2-fff.(t) are available, namely cos tLbt & sin u.,,t, the identities:

COS W t COS w,,t = 1/ 2 COS ( L&>+ U.o) t + cos w t sin Wt cos wt = 1/2 sin( up+ t4b)t + sin t,0)t coswt sin wp= 1/2 sin t,;b)t - sin A,)t sin wt sin ui = 1/2 -cos u:),) t + cos ti - u--b) t mean that multiplying together the observation signal and the components of the tracking signals gives rise to resultants having frequencies which are the sum and difference frequencies (o + o.) and (u)- o,,).

Under ideal tracking conditions, lo= t4(t); in practice there may be small tracking phase errors so that w = u,(t) + E(t).

Then 4> + u:)0 = 12. L#-),> + C, t4_-) - w,, = a. The f ormer represent components at essentially twice the tracking frequency, while the latter represent low frequency components, which tend to DC as to tends to u-% (the ideal). So if the result of such a multiplication process is passed through a low-pass (DC) filter which essentially carries out an averaging process, the sum components 2uj. + E are immediately filtered out, leaving only the difference components. Thus, the resultant of the multiplication and filtering, Iout and Qout are, denoting the filter by an overbar:

lout fA, sin(u>t +0 + n(t)cos L..,4,t = tl/2 A cos et + 1/2 B sin Et3 Q0uk = + n (t) cos u:b t [A, sin(u>t + n(t)5 sin -C,!:6t- = {1/2 B cos L7t 1/2 A sin Etj + n(t) sin tcht Examining each of the components of the right hand side of the expression in turn, the tighter the tracking accuracy, the more closely cos at ---:o 1 and sin it---p0, so that the expressions fl/2 A cose t + 1/2 B sin 7t and - - Otj L1/2 B cos F_t 1/2 A sin gtj tend to 1/2 A and 1/2 B respectively. The residual components n(t) cos t,%t, and n(t) sin tgb contain similar sum and difference frequencies, but since these do not match the tracking frequency, such difference frequencies always represent fluctuating AC components. The bandwidth of the DC filter then defines the extent to which such components are rejected. So, Iout and Qout are representative of the extent to which the respective observation signals contain each constant frequency.

It is noted that two considerations define the accuracy with which the components A & B can be extracted: (1 The tracking accuracy of the reference tracking frequency The nature

and extent of the DC-filtering process (ie DC-filter bandwidth and sideband roll-off).

Once (i) has been adequately satisfied, (defined by the source of the input tracking signals and precision of the cos/sin tracking system), the precision of measuring A & B depends entirely on (ii). The trade-off becomes one of speed of response versus ultimate accuracy; this performance is controlled by the choice of DC-filter characteristic as discussed in more detail below.

The set of Iout and Qout values are fed to the host computer 16 which performs an optimisation calculation 44 to provide optimised values Ilout and Qiout. As explained hereinbefore, the optimisation calculation seeks to provide a set of modified values which will have the effect of reducing the noise overall.

It will be appreciated that various standard techniques are available for carrying out the optimisation calculation. one method described in the paper "Proceedings of the Institute of Acoustics"; Vol. 6, Part 4, -(1984) and Vol. 8; Part 1 (1986) of M.A. Swinbanks involves deriving the overall set of transfer functions of the environment in which the vibration is to be controlled (e.g. in the case of an aircraft, specifically the cabin environment), by appropriate known techniques, and storing these to form a reference database.

The optimisation calculation effects a multi-degree-of-freedom statistical manipulation of the lout and Qout values. What is ultimately required by the vibration counteraction system is a set of output drive signals 6 which control actuators to produce vibratory signals which tend to cancel the unwanted vibrations in the zone of interest. The optimisation calculation effects a modification of the Iout and Qout values representative of the extent to which respective observation signals 6 contain each constant frequency, such that the modified values Ilout and Q1out, when multiplied by corresponding I and Q components of the tracking signals and appropriately summed (as discussed below) produce the required drive signals 8 for respective actuators.

This technique is standard procedure for characterisation of an active control system (see for example the discussion in Active Noise Control: Review Lecture DAGA 1985 M.A. Swinbanks). This database is then used to form the basis for an optimisation strategy to define the required output drive levels, having observed the corresponding monitoring levels. The calculation procedure may be according to the technology disclosed in Proceedings of the Institute of Acoustics Vol. 18 Part 1 (1986) of M.A. Swinbanks, referred to above. The procedures for deriving the output parameters may be based on a 'lone shot" calculation, but under conditions where the reference discrete frequencies are varying would preferentially be implemented as a continuous process of re-calculation and update.

The method described in "Proceedings of the Institute of Acoustics" Vol. 8; Part 1 (1986) of M.A. Swinbanks can be implemented in the present invention to achieve a vector of optimised output values F given by:

F = (a 0 T) (TO T)-1 where ao represents the vector of in-phase and quadrature components of the tracking signals, T represents the matrix of sy stem transfer function parameters which can be derived by prior testing in a manner well known to persons skilled in the art, and T and TO represent respectively the complex conjugate and transpose of the matrix T.

The set of optimised values Ilout and QOout are fed to the output processor 14 which multiplies the Ilout and QOout values by the I and Q components of the tracking signals at the relevant frequency, and forms a summation over all the reference frequencies. This results in active drive output signals which are supplied to actuators in the aircraft cabin to cause them to produce vibrations which tend to cancel the unwanted noise in the cabin.

The operation of deriving the actuator drive output signals is shown schematically in Figure 4. The set of optimised values Ilout and Q'out are fed from the host computer 16 to the output processor 14, indicated diagrammatically by a broken line in Figure 4. The set of optimised values Ilout and Q'out are arranged in respective matrices A and B having components Amn and Bmn, where A is the matrix of Ilout values, B is the matrix of Q'out values, m is the number of the actuator and n represents the number of the discrete frequency which it is intended to control, with the total number of frequencies denoted by N. In the preferred embodiment of the aircraft described previously there are Ilout and Q'out values associated with eight frequencies (N = 8), and associated with 16 loudspeakers. The output processor also receives the I and Q components of the tracking signals via a serial link from the tracking processor 10, represented in Figure 4 as vectors C (I component) and D (Q component) having components Cn and Dn respectively. The output processor 14 performs the following multiplications and summation:

N (Anin Cn + B= Dn) = Ym where Ym. are components of a vector Y of the actuator drive output signals. The actuator drive output signals Ym are passed through a digital to analogue convertor 46 before being amplified and fed to the individual loudspeakers.

The matrices A, B are continually updated as new Ilout and Q1out values are generated.

Returning now to the operation of DC filtering utilised in the multiplication process carried out in the observation processor 12, as noted previously it is the DC filter (represented by numeral 42 in Figure 3) which essentially controls the dynamical response of the entire active system. The function of the filter is essentially to average out the resultant of the multiplication.

It is important to note that the speed of the process of converting the I and Q components of the tracking signals to the corresponding V out and Q1 out values is dependent only on the speed of the optimisation calculation. With powerful, modern processors, this calculation could be performed within the space of a single sample period (typically 0.7 milliseconds). Under such circumstances, the only component in the entire system which imposes any significant input-output delay is the D.C. filter. Thus, this single component can ultimately control the entire dynamical response of the active system. Choice of a long-time constant (narrow bandwidth) filter will give high precision, but a slow response to dynamical changes. Conversely, choice of a wide bandwidth enables fast varying components to be controlled, at the expense of reduced accuracy under "noisy" signal conditions.

Various D.C. filters can be utilised. The simplest form of D.C. filter is a single-pole recursive digital filter. This choice is the easiest and most straightforward to implement an active system with minimal calculation complexity. However, there are many alternative choices which might be made with advantage.

Specifically, the D.C. filter could be preceded with a "comb-filter" according to (for example) the description in UK Patent No 2054999B. The advantage of this approach is that the infinite impulse response of the recursive filter, with exponential decay, can be converted to a linear decay over a well-defined time interval. This ensures that after this specific time interval has elapsed, the effect of any change to the system can be guaranteed to have been accurately registered.

Alternatively, higher order D.C. filters, either of infinite impulseresponse time, or finite impulse-response time, could be incorporated for example by cascading two or more comb-filter/recursive filter combinations).

The advantage of such techniques is that the response time of the system can be optimised, while maximising the rejection of contributions from other frequencies present in the system, by appropriate choice of the bandwidth and sideband characteristic of this filter.

The effect of the use of various filters is illustrated with reference to Figures 6 to 8 which give the results of computer simulations of the response of output values Iout,Qout in response to a step change in frequency of an observation signal. specifically, the figures illustrate the following points:

M (ii) The transient performance of three different types of DC filter for a "clean" input signal which is subject to a step change in amplitude, frequency and phase.

The similarity of this response for a different set of input frequencies.

(iii) The effect of "noise" on the signal, and the resultant trade-off between rapid response and precision.

Figure 6(a) shows a simulated observation signal, namely a 50Hz sinusoid sampled at 1500Hz, which suddenly changes its frequency to 100Hz. It is initially described by A cos 2 Ift + B sin 2 1 ft where f =50, A = 1 and B =0. At the frequency change, the amplitude increases and the phase changes by 20 degrees, so that f = 100Hz, A = 2 and B = 0.73.

A corresponding tracking signal is shown (the sine component); a similar cosine tracking component is assumed, as described in the text.

Figure 6(b) shows these step changes in A and B, together with the output (Iout or Qout) of the multiplication and filtering process for three choices of DC filter. It can be seen that each filter has an associated transient response, defining the speed and accuracy with which the changes in A & B are registered. The following should be noted:

(1) For any particular choice of filter, the transient response time is identical fpr both A & B components, regardless of their initial final amplitudes.

(2) The simplest DC-filter is a single pole recursive (exponential) filter, but to achieve adequate accuracy, a comparatively long timeconstant is required, thus slowing the overall response of the system.

(3) A constant-weighting "moving average" filter (which can be constructed either by averaging samples explicitly, or, more efficiently, by preceeding the single-pole recursive filter with a comb-filter) gives a straight-line response to the parameter change. Response time and accuracy is dictated by the number of samples chosen for the average. This filter has the advantage that the exact time required for any change to register completely is well-defined.

(4) A two-pole recursive filter. This filter gives better rejection of high-frequency components, so that its time-constant can be correspondingly reduced. This can provide faster response than either of the previous filters.

Figures 7(a) and 7(b) show a repeat simulation, but for a higher frequency signal (150Hz doubling to 300Hz). The residual high frequency "ripple" is less, but otherwise the basic response times are the same as for the previous example.

Figures 8(a) and 8(b) show the effect of introducing additional "noise" into the measurement. Figure 8(a) shows exactly the same input signal as for 6(a), but now with a component of random noise superposed. corresponding DC filter response.

Figure 8(b) shows the It can be seen that under these circumstances the wider bandwidth, fast- responding 2-pole DC filter is more sensitive to this random noise; more apcurate long-term response is actually obtained using the narrow bandwidth single-pole filter. The "moving average" filter probably represents the best compromise of the three, giving a useful trade-off between speed of response and precision.

The specific filters chosen are intended to be purely illustrative; by using higher-order DC filters, one can undoubtedly improve the transient response and noise rejection. It should be apparent that the precise choice of this element completely defines the ultimate speed and accuracy of the entire vibration reduction system.

Specifically, identical filters could be implemented for every discrete frequency present in the system. Under such circumstances, the response time of the entire system can be varied by changing the parameters of these filters in common, for example by varying the feedback gain of a single-stage recursive filter. Alternatively, under circumstances where different frequencies are varying on different time-scales (e.g. an aircraft installation where one engine is accelerating and another is held constant) one can choose dissimilar time-constants for the appropriate analysis filters.

A specific advantage can occur in the context of controlling time-varying periodic noise. Under such circumstances, the higher harmonic components automatically vary through a wider frequency range than the lower frequency components. It may therefore be advantageous to choose longer time-constant (narrower bandwidth) filters for the lower harmonics, and short time-constant (wide bandwidth) filters for the faster varying, high frequency components.

One specific circumstance merits further discussion. In the case of multiengined propeller driven aircraft it is not uncommon to encounter "beating" from closely spaced, but not synchronous, discretes. The present processor can cope with this situation by adopting the following strategy:

1. For well separated discretes (e.g. > 1Hz separation), long timeconstant D.C. filters are selected, and the two tonal components are separated out and controlled independently.

2. For closely spaced discretes (e.g. < 1Hz separation) an apparent "single discrete" is observed with time varying amplitude and phase. Under such circumstances, a broad bandwidth DC. filter with rapid response time is selected and the amplitude-phase variations are tracked as for a single time-varying discrete.

A "decision frequency separation" can be selected to define the transition between broad and narrow bandwidth observations. The choice of 1Hz suggested above is simply by way of example.

By way of summary, Figure 9 shows schematically architecture for the entire vibration reduction system, in which like reference numerals are used to indicate like parts already referred to. The system includes a tracking processor 10, observation processor 12 and output processor 14 arranged respectively on separate processor boards, each coupled to the host computer 16. The host computer effects upload of the signal information from the processor boards, download of the transfer function information, and programmes for control of the processors. Tracking inputs 4 undergo signal conditioning in unit 48, and analogue to digital conversion in the ADC (analogue to digital convertor) 50, for example a 32 input ADC, before input to the tracking processor 10. Observation inputs 6 from the plurality of transducers also undergo signal conditioning in unit 52, and analogue to digital conversion in

ADC 54, before input to the observation processor 12. The actuator drive output signals from the output processor 14 are fed to the digital to analogue convertor 46 before being amplified and fed to respective loudspeakers.

Claims

1. A method of reducing unwanted vibrations in a zone wherein such unwanted vibrations arise from a source of vibration which generates a signal having at least one vibrational component at a substantially constant frequency over a detection period, the method comprising:

within the detection period deriving a tracking signal at said constant frequency; providing in-phase and quadrature components of said tracking signal; monitoring vibrations at each of a plurality of predetermined locations within the zone to provide a set of observation signals; multiplying together each of said in-phase and quadrature components of said tracking signal with each of said observation signals and filtering each resultant to provide a set of output values, each of which represents the extent to which the respective observation signal contains vibrations at said constant frequency; and utilising the set of output values and the in-phase and quadrature components of the tracking signals to produce a plurality of drive signals which when supplied respectively to each of a plurality of actuators cause vibrations to be produced within the zone which tend to cancel those generated by the vibration source.

2. A method of reducing unwanted vibrations for use where the vibration source generates a plurality of vibrational components at respective different but substantially constant frequencies in which a plurality of tracking signals are generated in the detection period at each of said frequencies.

3. A method of reducing unwanted vibrations according to claim 1 or 2 wherein the in-phase and quadrature components of the or each tracking signal are derived from the vibration generating source.

4. A method of reducing unwanted vibrations according to claim 3 wherein the in-phase and quadrature components of the or each tracking signal are derived from values stored in respective memory locations which when sequentially addressed generate a single cycle sine wave the rate of addressing being dependent on the frequency of one or more signal derived from the vibration generating zone.

5. A method of reducing unwanted vibrations according to claim 1 or 2 wherein the in-phase and quadrature components of the or each tracking signal are derived from at least one reference sinusoid generated by a waveform generator.

6. A method of reducing unwanted vibrations according to claim 5 wherein the or each reference sinusoid is fed to a finite impulse response filter having filter coefficients 1, 2a, 1; where 2a is a variable filter coefficient, where a = cos 2 1 f A t, and where A t is the sampling time and f is the frequency of the sinusoid.

7. A method of reducing unwanted vibrations according to claim 6 wherein the or each reference sinusoid is also fed to an infinite impulse response filter which is coupled to receive the value of A from the finite impulse response filter, to lock the infinite impulse response filter onto the appropriate reference sinusoid frequency.

8. A method of reducing unwanted vibrations according to claim 7 wherein the value of a from the finite impulse response filter is used to determine the values of delay coefficients for a delay means, and wherein the output from the infinite impulse response filter provides directly one of the in-phase or quadrature components of the tracking signals and wherein the other component is obtained by passing the output from the infinite impulse response filter through the delay means.

9. A method of reducing unwanted vibrations according to any preceding claim, wherein the step of filtering each resultant of multiplication of said in-phase and quadrature components of the tracking signals by each of said observation signals comprises selecting a filter from a plurality of filters of different response characteristics.

10. A method of reducing unwanted vibrations according to claim 9 wherein for tracking signals of different frequencies filters of respectively different response characteristics are selected.

11. A method of reducing unwanted vibrations according to any preceding claim wherein the set of output values are modified to produce sets of modified output values which when supplied to said actuators tend to cause a reduction in the mean level of vibration in the zone.

12. A method of reducing unwanted vibrations according to claim 11 wherein the set of modified output values, and the in-phase and quadrature components of the tracking signals are processed in accordance with the following equation, to give actuator drive output signals Ym:

(Anm Cn + B= Dn) = Ym n where Anm and Bnm represent -the modified output values associated with the m th actuator and n th tracking signal frequency for in-phase and quadrature components respectively, and Cn and Dn represent respectively the in-phase and quadrature components of the n th tracking signal.

13. A method of reducing unwanted vibrations as claimed in any preceding claim when used to cancel noise produced by aero-engine propellers at selected locations in an aircraft cabin.

14. A method of reducing unwanted vibrations substantially as hereinbefore described with reference to the accompanying drawings.

15. An apparatus for reducing unwanted vibrations in a zone wherein such unwanted vibrations arise from a source of vibration which generates a signal having at least one vibrational component at a substantially constant frequency over a detection period, the apparatus comprising:

a tracking processor for providing in-phase and quadrature components of tracking signals at each of said constant frequency; a set of vibration monitors arranged to monitor vibrations at each of a plurality of predetermined locations within the zone to provide a set of observation signals, an observation processor coupled to said tracking processor and to said vibration monitors and operable to multiply together each of said in- phase and quadrature components of said tracking signal with each of said observation signals and to filter each resultant to provide a set of output values each of which represents the extent to which the respective observation signal contains the constant frequency, a plurality of actuators which when driven, supply vibrational signals to the zone, and output processing means coupled to said tracking processor and to said observation processor and capable of utilising the set of output values and the in-phase and quadrature components of the tracking signal to produce a plurality of drive signals which when supplied respectively to each of the plurality of actuators cause vibrations to be produced which tend to cancel those generated by the vibration source.

16. An apparatus for reducing unwanted vibrations according to claim 15 for use where the vibration source generates a plurality of vibrational components at respective different but substantially constant frequencies, the tracking processor providing in-phase and quadrature components of the tracking signals at each of said frequencies.

17. An apparatus for reducing unwanted vibrations according to claim 15 or 16 wherein the tracking processor is adapted to receive signals from one or more sensors positioned adjacent the vibration generating source and to process these to provide the in-phase and quadrature components of the or each tracking signal.

18. An apparatus for reducing unwanted vibrations according to claim 17 wherein the tracking processor comprises a memory having a plurality of memory locations at which are stored values which when output sequentially generate a single cycle of a sine wave, and is adapted to generate the in-phase and quadrature components of the or each tracking signal by addressing said memory locations at a rate dependent on the frequency of the signal(s) from the or each sensors.

19. An apparatus for reducing unwanted vibrations according to claim 15 wherein the tracking processor comprises circuitry operable to receive at least one reference sinusoid and to process the or each reference sinusoid to provide the in-phase and quadrature components of the tracking signals.

20. An apparatus for reducing unwanted vibrations according to claim 19 wherein the tracking processor includes a finite impulse response filter coupled to said circuitry and arranged to receive the or each reference sinusoid, the finite impulse response filter having filter coefficients 1, 2a, 1; where 2a is a variable filter coefficient, where a = cos 2 ITf At where A t is the sampling time and f is the frequency of the reference sinusoid.

21. An apparatus for reducing unwanted vibrations according to claim 20 wherein the tracking processor also includes an infinite impulse response filter arranged to receive the or each reference sinusoid, and which utilises the value of a from the finite impulse response filter to lock on to the specific frequency of the reference sinusoid.

22. An apparatus for reducing unwanted vibrations according to claim 21, wherein the tracking processor further includes a delay means and a memory wherein are stored values of delay coefficients selectable in dependence on the value of a of the finite impulse response filter wherein the output from the infinite impulse response filter provides directly one of the in-phase and quadrature components of the tracking signals and is led via the delay means to provide the other of the in- phase and quadrature components.

23. An apparatus for reducing unwanted vibrations according to any one of claims 15 to 22 comprising a plurality of filters having different response characteristics, and selection means for selecting appropriate one of said filters.

24. An apparatus for reducing unwanted vibrations according to claim 23 wherein said selection means are adapted to select a particular one of said filters according to the frequency of the tracking signal.

25. An apparatus for reducing unwanted vibrations according to any one of claims 15 to 24 wherein the output processor is arranged to modify the set of output values to produce a set of modified output values which when supplied to said actuators tend to cause a reduction in the mean level of vibration in the zone.

26. An apparatus for reducing unwanted vibrations according to claim 25 wherein the output processor is further arranged to process the sets of modified output values, with the in-phase and quadrature components of the tracking signals according to the following calculation, to give actuator drive output signals Ym:

(Anm Cn + Bnn Dn) = Ym n=l where Anm and B= represents the modified output values associated with the m th actuator and n th tracking signal frequency for in-phase and quadrature components respectively, and Cn and Dn represent respectively the in-phase and quadrature components of the n th tracking signal.

27. An apparatus for reducing unwanted vibrations substantially as hereinbefore described with reference to or as shown in the accompanying drawings.

Amendments to the claims have been filed as follows 4 CLAIMS 1. A method of reducing unwanted vibrations in a zone wherein such unwanted vibrations arise from a source of vibration which generates a signal having at least one vibrational component at a substantially constant frequency over a detection period, the method comprising:

2. A method of reducing unwanted vibrations according to claim 1 for use where the vibration source generates a plurality of vibrational components at respective different 36 4 but substantially constant frequencies in which a plurality of tracking signals are generated in the detection period at each of said frequencies.

6. A method of reducing unwanted vibrations according to claim 5 wherein the or each reference sinusoid is fed to a finite impulse response filter having filter coefficients 1, 2a, 1; where 2a is a variable filter coefficient, where a = cos 2 liT f t, and where Cú t is the sampling time and f is the frequency of the sinusoid.