BACKGROUND OF THE INVENTION Control devices exercise a restraining or directing influence over a controllable device. During its operation an automatic control device typically receives one or more input parameters and in response to the received input parameters outputs one or more control functions. Control functions are further input to a controllable device and they bring about an effect on the operation of the controllable device. One control device may control several controllable devices and one controllable device may be controlled by one or more control devices. The input parameters of the control device may originate from a proc- ess of which the controllable device is a part, or they may be fed into the con- trol process from outside, for example as control commands or as output pa- rameters from another process.

In automatic control devices the input parameters are typically de- rived from a series of sampled data. The document NUMERICAL RECIPES IN C : THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) copyright C) 1988-1992 by Cambridge University Press, pp. 496-510, provides the basic principles of Fast Fourier Transform and the referred pages are incorporated herein by reference.

The computation is based on the perception that a physical process can be described in the time domain by the values of some quantity h as a function of time t, or in the frequency domain where the process is specified by giving its amplitude H as a function of frequency f. These two representations can be related to another by means of the Fourier transform equations: In the most typical situations, function h (t) is sampled at evenly spaced intervals in time, so that the sequence of n sampled values hn is hn = h(n#) n =...,-3,-2,-1,0,1, 2, 3,... (2) where A is the sampling rate. The integral of equation (1) can be approximated by a discrete sum

In practical solutions the computing related to Discrete Fourier Transform is significantly reduced by utilising Fast Fourier Transform (FFT) algorithm. In the first section of FFT, the data is first sorted into bit-reversed order. The second section of FFT has an outer loop that is executed og2N times and calculates, in turn, transform of length 2, 4,..., N. For each stage of the process the two nested inner loops range over the sub-transforms already computed and the elements of each transform, implementing the Danielson- Lanczos Lemma. This variant of the FFT is called a decimation-in-time or Coo- ley-Tukey FFT algorithm. The decimation-in-frequency (Sande-Tukey) FFT algorithm first goes through a set of 1092N iterations on the input data, and then rearranges the output values into bit-reverse order.

In another class of FFT variants the initial data set of N is sub- divided down to some small power of 2, for example N=4 (base-4 FFT) or N=8 (base-8 FFT). These small transforms are done by small sections of optimized coding that utilize special symmetries of the particular N. Assuming, for exam- ple, that N=4, leads to the trigonometric sines and cosines being 0, +1 or-1, which by far eliminates multiplications, and leaves additions and subtractions.

Base-4 or base-8 FFTs have been considered to decrease the computing of the order to 20 to 30 percent.

One example of automatic control devices is an electric protection device. The general purpose of electric protection devices is to prevent or limit

damage to the protected device or to protect service of the protected device from interruption. In electric circuits the current flows are rapidly varied and the power may surge instantaneously. Therefore the power feed of an electrically driven device is always provided with one or more protection functions that are capable of isolating the device from harmful electrical transients from the power feed. In more elementary devices the protection devices are typically relay arrangements that are triggered when present current or voltage levels exceed a pre-determined threshold. In more complex systems the current and voltage levels are rigorously monitored and protection functions responsive to various parameters acquired through monitoring are utilized to safeguard the system from harmful effects in the power system.

In general electrical devices are designed to function with sinusoidal voltage, and for accurate functionality highly developed systems also require a faultless voltage to function correctly. One of the main harmful effects deterio- rating the quality of electricity in the power systems are the harmonics. Har- monics are divided into different components by their properties (ranking num- ber). The component class shows the rotational direction of the phase phasor of the harmonic with respect to the fundamental frequency.

When the curve form of the voltage or of the current deviates from the sinusoidal, it can be considered formed of several sinusoidal signals of dif- ferent frequency. The function is typically broken down into its components by using the above Fourier analysis, on which the mathematical treatment of har- monics is based.

In complex protection devices the method of determining a protec- tion function based on said input parameters generally comprises balancing between the expenses of the device and the response time from receiving the input parameters to outputting the determined protection function. In known protection devices a digital signal processor (DSP) is considered the most powerful tool for handling defined input signals and for determining a protec- tion function thereupon. However, the component costs related to digital signal processors are considerable, and in a complex system the costs due to the number of digital signal processors need to be optimised.

On the other hand, the embedded systems facilitate parallel usage of a variety of programmable algorithms. A number of variable algorithms are available and component costs are not directly increased with the number of operable protection functions. However, going through the number of calcula-

tions takes time and the accumulated response times become longer than what is actually required in order to implement the protection functions appro- priately. There are a number of cases where the acceptable response times are only slightly exceeded, but for reliability reasons the costly DSP very often end up being preferred over the programmable algorithms.

Consequently, one of the disadvantages associated with the known electric protection devices is that the available means for determining protec- tion functions in response to various input values do not facilitate adequate optimization that takes into consideration both the technical and economical factors related to operable protection functions.

BRIEF DESCRIPTION [DISCLOSURE] OF THE INVENTION An object of the present invention is thus to provide a method and an apparatus for implementing the method so as to alleviate the above disad- vantage. The objects of the invention are achieved by a method and a control device which are characterized by what is stated in the independent claims.

The preferred embodiments of the invention are disclosed in the dependent claims.

The invention is based on the idea of appropriately fixing the num- ber of sampled current and/or voltage data and optimizing the discrete Fourier transform algorithm based on operations on the fixed coefficients.

An advantage of the method and arrangement of the invention is that it provides significantly shorter response time than the earlier general pro- grammed solutions without increasing the related costs as much as the con- ventional digital signal processors, and therefore enables more flexible optimi- zation between technical and economical factors in an electric protection de- vice.

BRIEF DESCRIPTION OF THE DRAWINGS In the following the invention will be described in greater detail by means of preferred embodiments with reference to the attached drawings, in which Figure 1 is a block chart illustrating the operating environment of a control device according to the present invention; Figure 2 illustrates schematically the basic functional structure of a control device 1 according to the first embodiment of the present invention il- lustrated in Figure 1;

Figure 3 is a flow chart illustrating a method of automatic control ac- cording to the present invention; Figure 4 illustrates another embodiment of the invention.

DETAILED DESCRIPTION OF THE INVENTION As an embodiment of a control device according to the present in- vention, the invention is described in the following by means of an electric pro- tection device of an electric motor. Naturally, it is obvious to a person skilled in the art that said solution is applicable to any control device arranged to monitor the quality of power supply on a basis of series of measured values of current and/or voltage, and on the basis of the measured values to implement a con- trol function to control the operation of a device connected to said power sup- ply.

Figure 1 is a block chart illustrating the operating environment of a control device according to the present invention. The control device 1 is con- nected to an electric line 2 feeding an electrically operated device 3, whose purpose is to perform an operation functionality F (x) 4. The operation function- ality 4 may be a continuous operation that is continuously controlled by control function f (x) 5 input through a control input 6, or the operation functionality 4 may be a step function triggered by sequential control functions f (x) 5 input through said control input 6, or a combination of both. It should be noted that only essential elements for illustrating the operating environment of the control device according to the invention are shown in Figure 1. For a person skilled in the art it is clear, for example, that a complex system typically comprises a plu- rality of inputs and outputs like the one shown in Figure 1.

During its operation the control device 1 typically receives input data (x) 7 and in response to the received data 7 outputs one or more control func- tions f (x) 5. The input data 7 that is input to the control device 1 may originate from a process of which the controllable device 3 is a part, or it may be fed into the control process from outside, for example as control commands or as pa- rameters output from another process. In this first embodiment the input data x 7 comprises sequentially measured values of current in a power feed of an electric motor, hereinafter called as sampled current values.

In computing the basic algorithm of a discrete Fourier transformation can be expressed as:

void dft (double *x, double *re, double *im, int N) { int k, n; double a ; for (k=0 ; k<N ; k++) { re [k] =0 ; im [k] = 0 ; for (n 0 ; n <N ; n++) { a=k*n*PI2/N ; re [k] += x [n] * cos (a); im [k] -= x [n] * sin (a); <BR> <BR> <BR> <BR> <BR> <BR> }<BR> } where sampled data re= real values of transformation im=imaginary values of transformation N=number of samples P12=2*pi The current levels of the power feed are measured sequentially and the measured values are arranged into an array of N samples. The sample rate R is adjusted to the nominal frequency f of the power feed so that R=fN.

In this first embodiment of the present invention the number of samples is ad- justed to 32, and with the nominal frequency of 50 Hz, the sample rate is 1,6 kHz.

By adjusting the number of samples N to a value in power of 2, the sine and cosine functions have a corresponding range of N/4 possible non- zero absolute values. In this first embodiment of the present invention the sample rate of 32 leads to eight different non-zero values for the sine and co- sine functions. r [O] =cos (l*pi/16) r [l] = cos (2 * pi/16) r [2] =cos (3*pi/16)

In increasing computing speeds integer values are typically pre- ferred, and therefore computed values can be scaled into an integer form with a fast shift-operation. In this first embodiment the sine and cosine functions are scaled into integer form using factor 214 (16384), whereby an array of r [O] = cos (l pi 16) 16384 r [l] =cos (2*pi/16) *16384 r [2] =cos (3*pi/16) *16384 r [7] =cos (7*pi/16) *16384 is formed. Thereafter the basic algorithm is optimized on the basis of comprehensive consideration of the coefficients r. Said optimization com- prises eliminating the time-consuming calculations for the trigonometric func- tions, where the operations comprising multiplication by zero are discarded, and multiplications by-1 or +1 are eliminated by merely changing or retaining the sign of the sampled values. Said optimization preferably comprises also locating individual equations including a common factor r and consolidating the equations into multiplication by combined sums. The rows of the basic algo- rithm are optimally reduced and consolidated to the extent that identical opera- tions need to be performed only once.

In this first embodiment of the present invention the optimized Fou- rier transformation of the array of 32 x-values can be expressed in C language format as: #define SHIFT 14 void ft32 (int *x) { unsigned char i, j ; int k [7] = {16069, 15137, 13623, 11585,9102, 6270,3196} ; int a [16], b [16], c [8], d [8], e [8], f [8], g [57]; for (i = 0 ; i < 16; i++) { j=i+16 ; a [i] = x [i] + xo] ;

b[i] = x[i] - x[j]; } c[0] = a[0] + a[8]; d [0] = a[0] - a[8]; for(i = 1;i < 8; i++) { j = 16 - i; c [i] = a[i] + a[j] ; d [i] = a[i] = a[j] ; e [i] = b [i] + b [j] ; f[i] = b[i] - b[j]; } g[0] = c[0] + c[4]; g [l] = c [2] + c [6]; g [2] = c [l] + c [3] + c [5] + c [7]; g [3] = k [0] * f[1] # SHIFT; g [4] = k [0] * e [7] » SHIFT; g [5] = k [0] * f >> SHIFT; g [6] = k [0] * e [3] # SHIFT; g [7] = k [0] * f [3] # SHIFT; g [8] =k [0] *e [5] » SHIFT; g [9] = k [0] * f [7] >> SHIFT; g [10] = k [0] * e [l] >> SHIFT; g [11] = k [l] * f [2] >> SHIFT; g [12] = k [l] * e [6] >> SHIFT; g [13] = k [l] * f [6] # SHIFT; g [14] = k[1] * e[2] # SHIFT; <BR> <BR> <BR> g [15] = k [l] *(c[1] - c [7] ) » SHIFT; g [16] = k[1] * (d [3] + d [5]) » SHIFT; g [17] = k[1] * (c [3] -c [5]) # SHIFT; g [18] = k[1] * (d[1] + d [7]) # SHIFT; g [19] = k [2] * f [3] # SHIFT; g [20] = k [2] * e [5] # SHIFT; g [21] = k [2] * [l] » SHIFT; g [22] = k [2] * e [7] # SHIFT; g [23] = k [2] * f [7] # SHIFT;

g [29] = k [2] * e [l] » SHIFT; g [25] = k [2] * f [5] » SHIFT ; g [26] = k [2] * e [3] » SHIFT; g [27] = k [3] * f [4] » SHIFT; g [28] = k [3] * e [4] » SHIFT; g [29] = k [3] * (c [2] -c [6]) >> SHIFT; g [30] = k [3] * (d [2] + d [6]) » SHIFT; g [31] = k [3] * (c [l]-c [3] -c [5] + c [7]) » SHIFT ; <BR> <BR> g [32] = k [3] * (d [l] + d [3]-d [5]-d [7] ) >> SHIFT ; g [33] = k [4] * f [5] » SHIFT; g [34] = k [4] * e [3] » SHIFT; g [35] = k [4] * f[7] # SHIFT; g [36] = k [4] * e [l] » SHIFT; g [37] = k [4] * f [l] » SHIFT; g [38] = k [4] * e [7] » SHIFT; g [39] = k [4] * f [3] » SHIFT; g [40] = k [4] * e [5] » SHIFT; g [41] = k [5] * f [6] » SHIFT; g [42] = k [5] * e [2] » SHIFT; g [43] = k [5] * f [2] » SHIFT; g [44] = k [5] * e [6] » SHIFT; g [45] = k [5] * (c [3] -c [5]) » SHIFT; g [46] = k [5] *(d[1] + d [7]) » SHIFT; g [47] = k [5] * (c [l]-c [7]) » SHIFT; g [48] = k [5] * (d [3] + d [5]) » SHIFT; g [49] = k [6] * ft7] » SHIFT; g [50] = k [6] * e [l] » SHIFT; g [51] = k [6] * f[3] # SHIFT; g [52] = k [6] * e [5] » SHIFT; g [53] = k [6] * f[5] # SHIFT; g [54] = k [6] * e [3] # SHIFT; g [55] = k [6] * ftl] » SHIFT ; g[56] = k[6] * e[7] # SHIFT; <BR> <BR> <BR> <BR> x[0] = g[0] + g[1] + g[2];<BR> x[1] = O;

x[2] = b[0] + g[3] + g[11] + g[19] + g[27] + g[33] + g[41] + g[49]; <BR> <BR> <BR> <BR> x[3] = -b[8] - g[50] - g[42] - g[34] - g[28] - g[20] - g[12] - g[4];<BR> x[4] = d[0] + g[15] + g[29] + g[45]; x[5] = -d[4] - g[46] - g[30] - g[16]; x [6] = b [0] + g [21] + g [43] -g [51] -g [27] -g [5]-g [13]-g [35]; x [7] = b [8] -g [36]-g [14]-g [6] -g [28] -g [52] + g [44] + g [22]; x [8] = c [0]-c [4] + g [31] ; x [9] = d [6] -d [2] -g [32]; x [10] = b [0] + g [37]-g [43] -g [7] -g [27] + g [53] + g [13] + g [23]; x [11] = g [8] -b [8] -g [24]-g [14]-g [54] + g [28] + g [44]-g [38] ; x [12] = d[0] + g [47] -g [29]-g [17] ; x [13] = d [4] -g [18]-g [30] + g[48] ; x [14] = b [0] + g [55]-g [l 1]-g [39] + g [27] + g [25] -g [41] -g [9]; x [15] = b [8]-g [10]-g [42] + g [26] + g [28] -g [40]-g [12] + g [56] ; x [16] = g[0] - g[1] ; x [17] = d [3] -d [l]-d [5] + d [7]; x [18] = b [0]-g [55] -g [l 1] + g [39] + g [27] -g [25] -g [41] + g [9]; x [19] = g [12]-b [8] -g [10] + g [42] + g [26] -g [28] -g [40] + g [56] ; x [20] = d [0]-g [47] -g [29] + g[17] ; x [21] = g [48] -d [4]-g [18] + g[30] ; x [22] = b [0]-g [37] -g [43] + g [7]-g [27] -g [53] + g [13]-g [23]; x [23] = b [8] -g [24] + g [14]-g [54] -g [28] + g [8] -g [44] -g [38]; x [24] = c [0]-c [4] -g [31] ; x [25] = d [2]-d [6] -g [32]; x [26] = b [0]-g [21] + g [43] + g [51] -g [27] + g [5]-g [13] + g [35]; x [27] = g [14]-b [8]-g [36]-g [6] + g [28]-g [52]-g [44] + g [22] ; x [28] = d[0] - g[15] + g[29] - g[45]; x [29] = d [4] -g [46] + g [30] -g [16] ; x [30] = b [0]-g [3] + g [l 1]-g [19] + g [27] -g [33] + g [41]-g [49] ; x[31] = b[8] - g[50] + g[42] - g[34] + g[28] - g[20] + g[12] - g[4]; }

The embodiment shows further that the results of the optimized Fourier transformation are stored into the original array of x-values as follows : x [0] = lore (DC) x [1] = Ioim (=0) x [2] = Ilre x [3] = Ilim x [4] = Ivre x [5] = I2im x [30] = Ilsre x [31] = Iisim thereby comprising the DC-component and real and imaginary values of 15 harmonic components. This optimizes the use of memory, because no new arrays need to be created to store the real and imaginary values of the har- monic components. Furthermore, the choice of using 32 samples per cycle has been detected to be an optimal choice because this way it provides reasonable accuracy in computation of up to 15 harmonic components with reasonable amount of computable samples.

The real values of harmonic components 117... 131 are equal to the real values of harmonic components 115... 11, correspondingly, and the imagi- nary values of harmonic components 117... 131 are opposite to the imaginary val- ues of harmonic components 115... 11, correspondingly, whereby the absolute values of harmonic components 117... bi are equal to the absolute values of harmonic components 115... 11, correspondingly. For further optimization, calcu- lation of real and imaginary values of harmonic components 117... 131 of the ba- sic algorithm can therefore be omitted.

With the above optimized algorithm the computation of the harmonic components is significantly faster and causes considerably less load to the processor. In the above description the calculation of real and imaginary val- ues of harmonic components of current are shown, but the same algorithm can be applied to the measured values of voltage as well. The computed real and imaginary values of harmonic components of current and voltage can be used to calculate a variety of important parameters to which the control functions of the control device may be made responsive. In this embodiment of the present invention the root-mean-square current can be calculated from the computed values by

where n is the number of harmonic components.

Correspondingly the power factor can be calculated from the com- puted values of the first harmonic by <BR> <BR> <BR> <BR> <BR> <BR> I1*U1 where 11 i and Ui denote the lengths of the current and voltage vectors, corre- spondingly.

By calculating the sign from sign = I1re*U1im-I1im*Ur1re (8) enables determining the phase angle of the power feed, and thereby whether the coupling is capacitive or inductive. If the sign is positive (sign > 0), the phase angle is lagging, and the connection is inductive. If the sign is negative (sign < 0), the phase angle is leading, and a capacitive connection is involved.

Furthermore, the distortion can be calculated from the computed values by The apparent power of three phases can be derived from S-I1RMS * U1RMS + IO2RMS * U2RMS 9 I3RMS * U3RMS <BR> <BR> <BR> <BR> <BR> (10)<BR> P=#cos(phi)#S where 11 RMS is the root mean square current of phase 1 12RMS is the root mean square current of phase 2 13RMS is the root mean square current of phase 3 U 1 RMS is the root mean square voltage of phase 1 U2RMS is the root mean square voltage of phase 2 U3RMS is the root mean square voltage of phase 3 and the earth fault current can be calculated from Figure 2 illustrates schematically the basic functional structure of a control device 1 according to the first embodiment of the present invention il- lustrated in Figure 1. The control device includes processing means 21, an element that comprises an arithmetic logic unit, a number of special registers and control circuits. Connected to the processing means there are memory means 22, data medium where computer-readable data or programs, and/or static or dynamic data can be stored. The memory means typically comprise memory units that allow both reading and writing (RAM), and memory whose content can only be read (ROM). The unit also comprises an interface block 23 with input means 24 for inputting data for internal processing in the unit, and output means 25 for outputting data from the internal processes of the unit.

Examples of said input means comprise a plug-in unit acting as a gateway for the information delivered to its external connection points, a keypad, or a touch screen, a microphone, or the like. Examples of said output means comprise a plug-in unit feeding information to the lines connected to its external connec- tion points, a screen, a touch screen, a loudspeaker, or the like. The process- ing means 21, memory means 22, and interface block 23 are electrically inter- connected for performing systematic execution of operations on received and/or stored data according to the predefined, essentially programmed proc- esses of the unit.

In the first embodiment the predefined processes comprise a pro- grammed algorithm that on the basis of the sequentially sampled current levels computes a plurality of parameters representative of the quality of the power feed to the electric motor. Said processes also comprise a number of protec- tion functions that, responsive to the computed values, initiate, adjust and/or terminate a variety of control operations. The algorithm is optimized for im- proved response time and saved processor capacity. The input means 24 comprise an input to sensors that with a defined sample rate measure current and voltage values of the power feed, the signals from the sensors being first processed from analog to digital format with an analog-to-digital converter.

The input means 24 typically comprise also an interface to an on-site user ter- minal and/or to the fieldbus for on-site and/or remote controlling of the device.

The output means 25 comprise primarily an output to a relay through which the electric motor can be isolated from the power feed whenever one of the moni- tored parameters passes through a predefined threshold level. The output means 25 typically comprise also an interface to an on-site user terminal and/or to the fieldbus for on-site and/or remote processing of the information on the processes in the device.

Figure 3 is a flow chart illustrating a method of automatic control ac- cording to the first embodiment of the present invention. In step 30 a counter for counting the number of data samples is reset to zero, and in step 32 the measured values of current and voltage in digital form are collected. In step 33 the counter is incremented and the current value of the counter is compared with the predefined number of data samples N in step 33. If the counter indi- cates that there are less samples than the predefined number of data samples N, the process returns to step 31 where next values of current and voltage are read. Otherwise the predefined number of data samples have been read and will be analyzed with the algorithm optimized as described above. In step 34 a predefined number of real and imaginary values of harmonic components are calculated from the sampled values of current and voltage, and relevant char- acteristics of the power feed, including for example root-mean-square current, the power factor, distortion, or earth fault current are determined therefrom. In step 35 the determined values are compared against corresponding threshold conditions. If none of the threshold conditions are met (step 36), the process may continue automatically and no control function needs to be triggered. If a threshold condition is met, the procedure will move to step 37 where a control function f is triggered. The control function itself is not relevant to the invention and will not be described in more detail. The procedure thus returns to step 30 where the counter is reset to zero in order to collect a new set of samples. It should be noted that the current and voltage levels are measured continu- ously, and therefore the procedure typically contains buffering of values to fa- cilitate an appropriate reciprocity between the reading and computing opera- tions. However, due to the optimized algorithm the delay from computing is minimized.

In the above described first embodiment of the present invention the control function is a protection function whereby the controllable device is iso- lated from the power feed whenever one of the defined electrical characteris- tics of the power feed exceeds a predefined level. The electrical characteristics

of the power feed are determined from sequences of measured values of cur- rent and voltage using the optimized computation algorithm. It is clear for a person skilled in the art that said control function can be any restraining or di- recting operation performed to affect the controllable device. The control func- tion may comprise, for example, sending an alarm signal to a central control device through the fieldbus, or providing a feedback signal to adjust the opera- tion of the power source.

Figure 4 illustrates another embodiment of the invention where the controllable device 40 is the power feed itself, for example an electric genera- tor. Electric generators are known to persons skilled in the art and as such will not be described here in more detail. The operation of the power source is con- trolled by a control device 41 connected to it. The control device 41 is an as- sembly of variable control functions, ranging from simple mechanical switches, functions performed with on-site user interface terminals, automated control functions performed through the fieldbus system, to automatic control opera- tions performed with powerful digital signal processors. The control device 41 according to the present invention is connected to the power line 42 to receive information on the generated power F (x) as a sequence of measured current and voltage levels, and on the basis of the sampled data to compute a group of defined parameters representing the characteristics of the quality of the power feed. In case said parameters indicate an unwanted effect, a corrective control function is initiated.

The optimization of the algorithm is based on primarily fixing the number of data samples per cycle to be processed with the Discrete Fourier Transform. The optimized algorithm thus corresponds to a defined sample rate, and thus to a defined nominal frequency. In case the nominal frequency is changed, the sample rate needs to be adjusted to provide the predefined number of samples within the cycle of the new nominal frequency.

In the above first embodiment the DC-component and real and imaginary values of 15 harmonic components are computed. The number of harmonic components to be calculated can be adjusted depending on the level of required accuracy. The smaller the number of components, the less time will be needed for computing, but at the same time the accuracy of the result will reduce.

Therefore, an automatic control device comprises optimally a num- ber of variable control devices chosen and adjusted according to the parame-

ters relevant to each individual control operation. The control device according to the present invention provides a solution where the response time of an economical configuration is improved, and can even be adjusted, and which therefore provides for optimization that takes into consideration both the tech- nical and economical factors related to operable protection functions. The op- timal arrangement of individual control devices can thus be configured accord- ing to the current operating environment to provide the necessary actions in appropriate times with economical efficiency.

It will be obvious to a person skilled in the art that, as technology advances, the inventive concept can be implemented in various ways. The in- vention and its embodiments are not limited to the examples described above but may vary within the scope of the claims.