In many fields, there is a general desire to be able to generate two signals with a mutual 90° phase difference, both substantially identical to one input signal. Examples of fields where such signal combination is desirable are for instance: television tuners, mobile telephones (GSM, NMT), wireless telephones (DECT), etc. As is commonly known, quadrature signals are used in, for instance, television tuners for, inter alia, mirror rejection.

For generating such signal combination as mentioned above, several techniques are already available. Each of those available techniques suffers from some disadvantages.

One example of such known technique is to use frequency division with gate circuits, flip flops, etc. A disadvantage of this known technique is that the frequency of the input signal must be chosen at least twice as high as the desired frequency of the 0° and 90° signals.

Another technique is to use a separate oscillator which generates two output signals having 90° phase difference, the oscillator being coupled to the original input signal by means of a phase locked loop. A disadvantage of this technique is that it involves a relatively large amount of electronic circuitry and a relatively large amount of energy dissipation.

A further technique that is known per se is to use a phase-shifting network.

Phase-shifting networks generate one output signal identical to the input signal, with a fixed phase delay that can be set to 90°. However, a disadvantage of phase-shifting networks to date is that they require the input signal to be sine-shaped. More particularly, phase-shifting networks to date are not capable of generating two output signals having a mutual 90° phase difference and both at least substantially identical to an input signal having a square wave shape, although most local oscillators in the above-mentioned examples generate a square wave signal.

The present invention aims to provide an approach which allows the use of a phase-shifting network to generate two output signals having a mutual 90° phase difference, without the restriction that the input signal needs to be sine-shaped. To this end, the invention provides a method and device for generating two signals having a mutual phase difference of

90° as defined in the independent claims. The dependent claims define advantageous embodiments. In a preferred embodiment, the present invention provides a phase-shifting network capable of receiving a square wave signal and of outputting two square wave signals having a mutual 90° phase difference and both at least substantially identical to the input signal.

The invention is based on the insight that a square wave signal such as generated by a local oscillator can be approximated as a Fourier series of a fundamental wave and a limited number of odd harmonic waves, each of said waves being sine-shaped. In practice, the number of odd harmonic waves that need to be taken into account depends on the frequency of the fundamental wave: the higher the frequency of the fundamental wave, the lower the number of odd harmonic waves that play a significant role.

Therefore, based on this insight, the present invention proposes a phase- shifting network which operates on the fundamental Fourier component and at least the third harmonic Fourier component, and preferably also the fifth harmonic Fourier component, of an input signal in such a way that these components are shifted over the same amount of time equal to one fourth of the period of the fundamental Fourier component.

In a preferred embodiment, the phase-shifting network of the invention comprises a passive polyphase filter. Such filter has an intrinsic property of shifting all frequency components within a pass band over the same angle. Assuming that the desired shift angle is equal to 90°, said intrinsic property would be correct for the fifth harmonic Fourier component but would be wrong for the third harmonic Fourier component, which should be shifted over 270°, which is equivalent to a shift over-90°. The invention is based on the further insight that this is equivalent to a +90° shift of the third harmonic Fourier component with negative frequency. Therefore, according to the invention in this preferred embodiment, the phase-shifting network has a frequency characteristic that passes the third harmonic Fourier component with negative frequency and suppresses the third harmonic Fourier component with positive frequency.

These and other aspects, characteristics and advantages of the present invention will be further clarified by the following description of exemplary embodiments of a network in accordance with the invention.

In the drawings, in which same reference numerals indicate equal or similar parts, Fig. 1 schematically shows a polyphase filter; Figs. 2A-C illustrate how a polyphase filter can be used to generate a two- phase output signal with positive frequency on the basis of a one-phase input signal; Figs. 3A-C illustrate the shifting of Fourier components; Fig. 4 is a block diagram schematically illustrating an embodiment of a device according to the invention; Figs. 5A-C illustrate how a polyphase filter can be used to generate a two- phase output signal with negative frequency on the basis of a one-phase input signal; Fig. 6 shows the frequency characteristic of a polyphase filter according to the present invention; and Fig. 7 schematically illustrates a simple embodiment of a polyphase filter according to the present invention.

Fig. 1 schematically shows a polyphase filter 10 having a first input 11, a second input 12, a first output 13 and a second output 14. More particularly, this polyphase filter 10 is a two-phase filter. Polyphase filters are known per se. For instance, reference is made to the article"A Fully Integrated 900 MHz CMOS Double Quadrature Downconvertor" by J. Crols et al. in 1995 ISSCC Digest of Technical Papers, Vol. 38, IEEE press, 1995, p. 136-137. Therefore, an elaborate explanation of the design and operation of the polyphase filter 10 is not necessary here. However, in order to introduce some symbols and expressions, some aspects of the operation of the polyphase filter 10 (two-phase filter) are described here.

Assume that two input signals Xn ( (o) and X12(#) are applied to the two inputs 11 and 12, respectively, the two input signals X)) (co) and X12 (w) being sine-shaped and having the same frequency c3, but having a phase difference of 90°. This can be written as ##11-#12# = 90° [mod 360°], wherein #11 is the phase of the first input signal X11(#) applied to the first input 11, while +12 is the phase of the second input signal X12 (#) applied to the second input 12. Two situations can be distinguished: 1) X t (m) is leading, i. e. 0 1-#12 = +90°

2) X12 (w) is leading, i. e. + 12 =-90° A sine-shaped signal may be represented in complex notation as X ( (o) = X#ej#t, keeping in mind that the actual physical signal is the real part of the complex expression.

Then, both the above-mentioned cases of #11-#12 = +90° and #11-#12 = -90° can be written as xi 1(#) = X#ej#t and X12(#) = jX#ej#t Using X = #X##ej# yields : Re (XI I) = #X#cos(#t+#) and Re (XI2) = #X#cos(#t+# + 7c/2) so that wiz corresponds to the case of f 1-#12 = +90°, whereas co>O corresponds to the case of #11-#12 = -90°.

At its outputs 13 and 14, the polyphase filter 10 generates sine-shaped output signals Y13(#) and Yl4 (c3), respectively, having the same frequency o as the two input signals Xi 1(#) and X12(#).

It is further assumed that the polyphase filter 10 has a transfer characteristic H (co) that can be described as H(#) = Y13/X11 = Y14/X12 (1) in the case that X12(#) = jX11(#).

If the (normalized) transfer characteristic H (#) of the polyphase filter 10 is such that, for a certain positive frequency sox the following equation (2) is valid: H(#x) = 1 and H (-o) x) = 0 (2) then the polyphase filter 10 can be used for generating two output signals Y13 and Y, = jY13 on the basis of only one input signal Xi 1, as will be explained with reference to Figs. 2A-C.

In Fig. 2A, a first input signal X 1A = X(#x) is applied to the first input 11, and a second input signal X12A = jX11A = jX(#x) is applied to the second input 12. It follows from the above equations (1) and (2) that the polyphase filter 10 then generates a first output signal

,) and a second input signal Y14A = jY13A = jY(#x) at its two outputs 13 and 14, respectively.

In Fig. 2B, a first input signal XI IB = X (#x) is applied to the first input 11 and a second input signal X12B = -jX11B = -jX(#x) is applied to the second input 12. It follows from the above equations (1) and (2) that the polyphase filter 10 then generates zero output signals Y13B = 0 and Y14B = 0 at its two outputs 13 and 14, respectively.

The polyphase filter 10 is a linear filter, which means that, if two input signals are added, the corresponding output signals are also added. In Fig. 2C, the two input signals used in Fig. 2A and in Fig. 2B, respectively, are added.

In other words, the first input 11 receives XI 1C = X11A + X11B = 2X (OX) 9 while the second input 12 (not connected) receives X12C = X12A + X12B = jX(#x) + ( -jX(#x)) = 0.

Then, also the two output signals of Fig. 2A and Fig. 2B are added; therefore, the polyphase filter 10 then generates a first output signal Y13C = Y(#x) and a second output signal Y14C = jY(#x) at its two outputs 13 and 14, respectively.

In other words, if the first input 11 receives a real input signal X (#x) while the second input 12 is zero, the polyphase filter 10 then generates first and second output signals Y13(#x) = 1/2#X(#x) and Y14(#x) = 1/2#jX(#x). These are two real output signals with 90° phase difference, also indicated as a sine-shaped two-phase output signal having positive frequency. In the above, any possible phase difference between Y13(#x) and X (#x) is neglected.

It is noted that the above explanation would also apply in case the sign of Yl4A would be reversed. In that case, the two-phase output signal may be indicated as having negative frequency. However, it is also possible to consider output 14 as the"first"output and to consider output 13 as the"second"output.

Thus, using a polyphase filter, it is possible to generate two signals Y13 (o) x) and Yl4 (#x) with 90° phase difference with respect to each other, on the basis of one input signal X (wu).

In the above, two assumptions have been made. One assumption is that the input signal X(#x) is sine-shaped. The second assumption is that the frequency cox of the input signal X (#x) lies in a frequency region where equation (2) is valid. Such frequency region will hereinafter also be indicated as opposite sign rejecting pass region, abbreviated as OSR pass region, indicating that frequencies within the region are passed whereas identical frequencies with opposite sign are rejected or at least suppressed.

To date, broadband polyphase filters exist where the OSR pass region ranges from 0 to very high frequencies, approximating infinity, at least for all practical purposes.

Also, polyphase filters have been designed for cooperation with a specific local oscillator operating in a specific frequency band with central frequency COLO and bandwidth BWLO ; those polyphase filters have an OSR band-pass region coinciding with the operational frequency band of the local oscillator, the transfer function H (c3) being zero for all other frequencies.

With such prior art polyphase filters it is, however, not possible to employ the same simple approach in generating two signals with 90° phase difference with respect to each other, on the basis of one input signal, if the input signal is a binary signal and the output signals are required to be binary signals, too. More specifically, in many applications, the local oscillator generates a square wave signal with 50% duty cycle; the technique described in the above can not be used in such cases. This will be explained in the following.

Assume that the local oscillator generates an output signal A being a square wave signal with 50% duty cycle, having a period Tl, as illustrated in Fig. 3A. As is well known, such square wave signal A can be developed into sine-shaped signal components (Fourier series). These sine-shaped signal components comprise a fundamental wave with the fundamental frequency (ol = l/TI, which will be indicated as Al (col), as illustrated in Fig. 3B.

The Fourier series further comprises odd harmonic waves A3 (#3), A5(#5), ......, A2n+t ( (02n+1), wherein n = 1,2,3....... Herein, the frequency component A2n+l (o) 2n+l) is the (2n+I)-th harmonic wave with respect to the fundamental wave Ai, having a frequency o) 2n+i being equal to (2n+1) times the fundamental frequency mj. Fig. 3B also shows a part of the 3rd and 5th harmonic waves.

It is noted that, although in general the Fourier series is an infinite series having an infinite number of frequency components, in most practical circumstances the output signal A from the local oscillator can be approximated very well by a-limited number of Fourier terms, for instance five.

If a polyphase filter would be used to generate quadrature signals on the basis of such square wave local oscillator signal A, the polyphase filter will operate on each of said sine-shaped Fourier components in the way described above. Thus, if the polyphase filter would have a relatively narrow OSR band pass characteristic, only accommodating the fundamental frequency c31 of the local oscillator, the polyphase filter will only generate two sine-shaped output signals Y13 (co1) and Yl4 (l) with a mutual phase difference of 90°.

Although it is possible to construct square wave signals on the basis of such sine-shaped output signals, for instance by using an amplifier with large gain so that the signals will clip, this will necessitate further circuitry, while further small deviations in the sine-shaped output signals may lead to important timing deviations in the constructed square wave signal. In order to improve the accuracy in the zero-crossings, more Fourier terms should be taken into account.

If, on the other hand, the polyphase filter would have a relatively wide OSR pass characteristic, also accommodating the harmonic frequencies #3, #5, #7, etc, the polyphase filter would not generate the 90° output signal in a correct way, as will be explained hereinafter with reference to Fig. 3C.

First, it is pointed out that the required 90° phase difference between the two output signals Yl3 and Y14 relates to the fundamental frequency wl. Thus, in fact, it is required that the two output signals Yl3 and Y14 are identical, yet shifted in time over a time distance Tl/4 [mod TI] with respect to each other.

Second, it is pointed out that, in order to meet the objective that the second output signal Yl4 is equal to the first output signal Yl3 yet shifted over Tl/4 [mod Tl], it will be necessary that, when creating the second output signal Y14, each frequency component A2n+l (oz2n+l) of the input signal A is shifted over a time distance T ;/4 [mod Tl]. However, as mentioned above, the polyphase filter 10 has the intrinsic property of shifting all frequency components A2n+1(#2n+1) within its pass band over a phase angle of 90°, this phase shift of

90° being always measured with respect to the corresponding frequency C32n+1 of the signal component in question. Such a phase shift does not correspond to the required time shift for all frequency components.

For instance, for the fundamental wave Al and the fifth harmonic As, the ninth harmonic A9, etc., a time shift oft/4 corresponds respectively to a phase shift of 90°, 450°, 810°, etc., which is all equivalent to a respective phase shift of +90° [mod 360°] ; in other words: these are"matching"shifts.

However, for the third [seventh] {eleventh} harmonic wave A3 [A7] {An}, etc., the required time shift of Tri/4 corresponds to a required phase shift of 270° [630°] {990°}, respectively, in each case being equivalent to a required phase shift of-90° [mod 360°]. As mentioned earlier, the conventional polyphase filter can not deliver such phase shift. In fact, if a conventional polyphase filter would have a relatively wide OSR pass characteristic, also accommodating the harmonic frequencies c33, 037, cal l, etc, these harmonic waves are likewise shifted over +90°, i. e. 180° wrong.

According to an important aspect of the present invention, this problem is overcome if an additional phase shift of 180° is exerted on the third [seventh] {eleventh} harmonic wave A3 [A7] {A11}, etc. This can, for instance, be done by individually selecting the separate harmonic waves, for instance by means of appropriate band pass filters, then individually processing each harmonic wave such that for each individual harmonic wave a +90° or a-90° [mod 360°] shift, as required, is obtained, and then combining the shifted harmonic waves.

Fig. 4 shows an embodiment where the harmonic waves that need to be shifted over +90° are combined, and one single broadband polyphase filter I OA is used to perform the +90° shift for all those harmonic waves in common, whereas the harmonic waves that need to be shifted over-90° are also combined, and one single broadband polyphase filter 1 OB is used to perform the-90° shift for all those harmonic waves in common. Fig. 4 shows a circuit 1 having an input 3 for receiving an output signal A from a local oscillator 2, and two outputs 8 and 9 for generating an in-phase output signal I and a quadrature output signal Q, both being identical to the input signal A. As described above, the local oscillator signal A is

a square wave signal with 50% duty cycle, having a period Tl. As mentioned above, the local oscillator signal A can be developed into sine-shaped signal components AI (#1), A3(#3), As , ......, A2n+1(#2n+1), wherein n = 0, 1,2,3....... The circuit 1 of Fig. 4 is designed to process four Fourier components, and is suitable for a situation where the output signal A from the local oscillator can be approximated by four Fourier components. From the following description, it will be clear to a person skilled in the art how this embodiment is to be supplemented in order to take higher order Fourier components into account.

The circuit 1 comprises a first Fourier component selection section 4, selecting those Fourier components which need to be shifted over +90° in the quadrature output signal Q. These are the Fourier components AI ( A5(#5), A9(#9), ......, A2n+1(#2n+1), wherein n = 0,2,4,6....... For each of those Fourier components, the first Fourier component selection section 4 comprises a corresponding band pass filter 602n+1. In the present embodiment, the first Fourier component selection section 4 is intended to select the fundamental wave AI (col) and the fifth harmonic A5 (e3s). Thus, the first Fourier component selection section 4 comprises a first band pass filter 601 having a pass band 611 in the frequency range between #1-BW1/2 and #1+BW1/2, and a second band pass filter 605 having a pass band 615 in the frequency range between ws-BW5/2 and (05+BW5/2.

The input terminals of these band pass filters are connected to the input 3, while the output terminals of these band pass filters are coupled to a first adder 71, the output of which is coupled to the first input 11A of a first polyphase filter 1 OA. Thus, the first input 11A of this first polyphase filter l0A receives an input signal XIIA = Al (cal) + A5 (os). The second input 12A of this first polyphase filter I OA receives a zero signal.

Similarly, the circuit 1 comprises further a second Fourier component selection section 5, selecting those Fourier components which need to be shifted over-90° in the quadrature output signal Q. These are the Fourier components A3 (#3), A7(#7), A11(#11), ...., A2n+1(#2n+1), wherein n = 1, 3,5....... For each of those Fourier components, the second Fourier component selection section 5 comprises corresponding band pass filters 602n+1. In the present embodiment, the second Fourier component selection section 5 is intended to select the third harmonic A3 (cl3) and the seventh harmonic A7(#7). Thus, the second Fourier component selection section 5 comprises a third band pass filter 603 having a

pass band 613 in the frequency range between (03-BW3/2 and O3+BW3/2, and a fourth band pass filter 607 having a pass band 617 in the frequency range between (07-BW7/2 and (07+BW7/2.

The input terminals of these band pass filters are connected to the input 3, while the output terminals of these band pass filters are coupled to a second adder 72, the output of which is coupled to the second input 12B of a second polyphase filter 10B. Thus, the second input 12B of this second polyphase filter 1 OB receives an input signal X12B = A3(#3) + A7 ( (07). The first input 11B of this second polyphase filter 10B receives a zero signal.

The two polyphase filters 1 OA and 1 OB are broadband polyphase filters having, at least for realistic frequencies, (normalized) transfer characteristics H (m) in accordance with the following equation (3): H (#) = 1 for (o > 0 and H (#) = 0 for m < 0 (3) In fact, the two polyphase filters 10A and 10B may be identical.

As will be clear from the above explanation of the operation of the polyphase filters, the first polyphase filter 10A provides at its first output 13A a first output signal Y13A according to 2Y13A = XI IA = A1 (col) + A5(#5), and at its second output 14A a second output signal Y) 4A according to Y14A = jY13A. It may be that the first output signal YA has a certain time delay ATA with respect to the input signal X11A.

Further, as will also be clear from the above explanation of the operation of the polyphase filters, the second polyphase filter 10B provides at its second output 14B a third output signal Y14B according to 2YI4B = X12B = A3 (#3) + A7 (m7), and at its first output 13B a fourth output signal Y13B according to Y13B = -jY14B = -jX12B. It may be that the third output signal Y) 4B has a certain time delay ATB with respect to the input signal Xl2B ; the two polyphase filters I OA and 10B should be matched such that said two time delays ATA and ATB are equal.

The first output 13A of the first polyphase filter 10A and the second output 14B of the second polyphase filter 10B are coupled to a third adder 73, the output of which is coupled to the first output terminal 8 of the circuit 1 to provide the first output signal I = Y13A + Y14B = (A1(#1) +A3 (#3)+A5(#5)+A7(#7))/2

Similarly, the second output 14A of the first polyphase filter I OA and the first output 13B of the second polyphase filter 1 OB are coupled to a fourth adder 74, the output of which is coupled to the second output terminal 9 of the circuit 1 to provide the second output signal Q = Y14A + Yl3B = j(A1(#1)-A3(#3)+A5(#5)-A7(#7))/2 The circuit proposed in Fig. 4 functions satisfactorily. However, it needs two polyphase filters. Preferably, the step of shifting the subsequent harmonic waves over alternatively +90° and-90° is performed by one single polyphase filter that receives all harmonic waves at one input.

According to the present invention, such a design is possible because an additional phase shift of 180° is equivalent to using a Fourier component with negative frequency -#3, -#7, -#11, etc. in the polyphase filter, as will be explained with reference to Figs. 5A-5C.

Assume that, for a certain positive frequency o3x, the transfer characteristic H (co) of the polyphase filter 10 obeys the following equation (4): H ((ou) = 0 and H (-wax) = 1 (4) In Fig. 5A, a first input signal XI IA = X (#x) is applied to the first input 11, and a second input signal X12A = jX11A = jX(#x) is applied to the second input 12. It follows from the above equations (1) and (4) that the polyphase filter 10 then generates zero output signals Y13(#x)=0 and Y14(#x)=0 at its two outputs 13 and 14, respectively.

In Fig. 5B, a first input signal XI IB = X (#x) is applied to the first input 11 and a second input signal X12B = -jX11B = -jX(#x) is applied to the second input 12. It follows from the above equations (1) and (4) that the polyphase filter 10 then generates a first output signal Yl3B = Y (#x) and a second output signal Y14B = -jY13B = -jY(#x) at its two outputs 13 and 14, respectively.

In Fig. 5C, the two input signals used in Fig. 5A and in Fig. 5B, respectively, are added. In other words, the first input 11 receives X11C = X11A + X11B = 2X(#x), while the second input 12 receives Xl2c = Xl2A + X12B = jX(#x) + (-jX (#x))=0. Then, also the two output signals of Fig. 5A and Fig. 5B are added; therefore, the polyphase filter 10 then

generates a first output signal Y13C = jY(#x) and a second output signal Y14C = -jY13C = -jY(#x) at its two outputs 13 and 14, respectively.

In other words, for such frequencies (ox for which equation (4) applies, if the first input 11 receives a real input signal X (#x) while the second input 12 is zero, then the polyphase filter 10 generates first and second output signals Y13 (Cox) = 1/2-X (<Dx) and Y14(cox) = -1/2 @ jX(#x) = -jY13(#x). These are two real output signals with-90° phase difference, also indicated as a sine-shaped two-phase output signal having negative frequency.

Based on this insight, the present invention proposes a polyphase filter 10 with a novel frequency characteristic 40 as shown in Fig. 6, adapted for use with a local oscillator having a central oscillator frequency caLo and a bandwidth BWLO. This frequency characteristic 40 has a first OSR band pass region 41 with a positive central frequency #1 = #LO and a bandwidth BWI substantially equal to the bandwidth BWLO of the local oscillator. More specifically, it is indicated in Fig. 6 that H (#) = 0 in the rejection range 51 between-c3l-BWI/2 and -#1+BW1/2, i. e. frequencies around-c31 are effectively suppressed.

The frequency characteristic 40 further has a second OSR band pass region 42 with a negative central frequency O42 =-3 (0 ; and a bandwidth BW42 substantially equal to 3 times the bandwidth BWs of the first OSR band pass region 41. More specifically, it is indicated in Fig. 6 that H (o) = 0 in the rejection range 52 between 3#1-3BW1/2 and 3#1+3BW1/2, i. e. frequencies around +3#1 are effectively suppressed. In view of this second OSR band pass region 42 (and the corresponding rejection region 52), the polyphase filter 10 can correctly process the third harmonic wave A3(#3).

Preferably, the polyphase filter 10 is also designed for correctly processing the fifth harmonic wave A5(#5). To that end, the frequency characteristic 40 further has a third OSR band pass region 43 with a positive central frequency #43 = +5#1 and a bandwidth BW43 substantially equal to 5 times the bandwidth BW of the first OSR band pass region 41.

More specifically, it is indicated in Fig. 6 that H (e3) = 0 in the rejection range 53 between -5#1-5BW1/2 and -5#1+5BW1/2, i. e. frequencies around-5 (0) are effectively suppressed.

Preferably, the polyphase filter 10 is also designed for correctly processing the seventh harmonic wave A7 (c37). To that end, the frequency characteristic 40 further has a fourth OSR band pass region 44 with a negative central frequency (o44 =-7e31 and a bandwidth BW44 substantially equal to 7 times the bandwidth BWl of the first OSR band pass region 41. More specifically, it is indicated in Fig. 6 that H (co) = 0 in the rejection range 54 between +7c)}-7BW)/2 and +7#1+7BWE1/2, i. e. frequencies around +7#1 are effectively suppressed.

In general terms, the polyphase filter 10 has N OSR band pass regions, each having a central frequency according to On = (-1)(n+1)#(2n-1)##1 for n = 1,2,3,4,...... N, and a bandwidth BWn substantially equal to (2n-1) times the bandwidth BWI of the first OSR band pass region 41.

In practice, it may be sufficient if N = 2, although preferably N is at least equal to 3. More preferably, N > 5.

It is observed that, for optimal operational quality, N should be as large as possible. However, in view of the fact that the bandwidth of the successive band pass regions increases, N can not be chosen infinitively high; the limit of the possibilities is reached if neighboring band pass and rejection regions touch, which will be the case if N = c3l/BWI + 1.

Fig. 7 shows a relatively simple embodiment of a polyphase filter 110 having the frequency characteristic 40 as described above, wherein N=2, i. e. the polyphase filter 110 of Fig. 7 operates on a fundamental wave and a third harmonic wave. The polyphase filter 110 has four input terminals 111, 112,113,114, and four corresponding output terminals 121, 122,123,124. The first (second) [third] {fourth} output terminal 121 (122) [123] {124} is connected to the first (second) [third] {fourth} input terminal 111 (112) [113] {114} through a first (second) [third] {fourth} transfer channel comprising a series connection of a first resistor R11 (R12) [R13] {R14} and a second resistor R2, (R22) [R23] {R24}, all first resistors Rl i being substantially equal to each other, and all second resistors R2i being substantially equal to each other.

The first (second) [third] {fourth} node between the first resistor RI 1 (R12) [R13] {R14} and the second resistor R2l (RZ2) [R23] {R24} of the first (second) [third] {fourth} transfer channel is indicated as N1 (N2) [N3] {N4}.

The first (second) [third] {fourth} input terminal 111 (112) [113] {114} is coupled to the second (third) [fourth] {first} node N2 (N3) [N4] {N1} through a first (second) [third] {fourth} first stage capacitor C12 (C23) [C34] {C41}, while the second (third) [fourth] {first} node N2 (N3) [N4] {N1} is coupled to the first (second) [third] {fourth} output terminal 121 (122) [123] {124} through a first (second) [third] {fourth} second stage capacitor C21 (C32) [C43] {C14}.

The first and third input terminals 111 and 113 define a first signal input. In Fig. 7, an input signal A from a local oscillator 102 is received at the first signal input 111, 113 in a balanced way.

The second and fourth input terminals 112 and 114 define a second signal input. In Fig. 7, these terminals are not connected to any signal source or voltage source, i. e. they are floating. Alternatively, they may be connected to zero.

The first and third output terminals 121 and 123 define a first signal output. In Fig. 7, the in-phase output signal I is taken from these two output terminals 121 and 123.

The second and fourth output terminals 122 and 124 define a second signal output. In Fig. 7, the quadrature output signal Q is taken from these two output terminals 122 and 124.

The polyphase filter 110 of Fig. 7 may be designed for a fundamental wave of cs l equals about 910 MHz, as used in GSM, for instance, with a narrow bandwidth of about 40 MHz, while (03 is about 2730 MHz, by selecting the parameter values approximately as follow: R11 (R12) [R13] {R14} = 22 # R21 (R22) [R23] {R24} = 210. 26 Q C12 (C23) [C34] {C41} = 2. 568 pF C21 (C32) [C43] {C14} = 820 pF It will be clear to a person skilled in the art of designing polyphase filters how these values are to be amended for obtaining different values for ol and 0) 3. Further, it will be clear to a person skilled in the art of designing polyphase filters how the circuit of Fig. 7 is to be expanded for processing a fifth harmonic wave, a seventh harmonic wave, etc.

Thus, the present invention succeeds in providing a method and device for generating two output signals I; Q each substantially identical to a square-wave input signal A from a local oscillator 2, wherein the first output signal I may have a certain time shift with respect to the input signal A, and wherein the second output signal Q is shifted over T1/4 [mod TI] with respect to the first output signal I, T, being the period of the input signal A.

To generate the first output signal I, Fourier components SI (#1), S3(#3), S5(#5), S7(#7), S9(#9), S11(#11) etc of the input signal A are combined.

To generate the second output signal Q, Fourier components Sl (#1), S5(#5), Sg (#9) etc of the input signal A are phase shifted over +90'while Fourier components S3 (03), S7 ((D7), Sn (on) etc of the input signal A are phase shifted over-90°, and the thus shifted Fourier components of the input signal A are combined.

It should be clear to a person skilled in the art that the scope of the present invention is not limited to the examples discussed in the above, but that several amendments and modifications are possible without departing from the scope of the invention as defined in the appending claims. For instance, in the embodiment of Fig. 4, the second polyphase filter 1 OB having a broadband normalized transfer characteristic H (co) = 1 for ##0 and H (#) = 0 for wiz can be replaced by a polyphase filter having a broadband normalized transfer characteristic H (co) = 1 for o<0 and H (o) = 0 for t9>0, in which case the output of the second adder 72 will be fed to the first input 11B, the second input 12B will receive a zero signal, the first output 13B will be coupled to the first combiner 73, and the second output 14B will be coupled to the second combiner 74.

Further, it is also possible that, in order to generate the second output signal Q, Fourier components S (#1), S5(#5), S9(#9) etc of the input signal A are phase shifted over - 90° while Fourier components S3 (#3), S7(#7), S11(#11) etc of the input signal A are phase shifted over +90°, and the thus shifted Fourier components of the input signal A are combined.

Further, it is observed that in the above explanation the characteristic 40 of the polyphase filter is described only in relation to the passbands or rejection bands around ~#1,

(03, N5, (07, etc, these bands having a bandwidth of BWI, 3BWI, 5BWs, 7BWI, respectively, BWI being the expected bandwidth of the local oscillator. For frequencies between said bands, the characteristic of the polyphase filter has not been defined. It is noted that, in principle, the characteristic of the polyphase filter outside said bands is not critical.

After all, no frequency components are expected in the frequency regions between said bands.

Thus, said bands may in fact be wider than mentioned; the indicated bandwidths are to be considered as minimum widths. Further, although in Fig. 6 the characteristic 40 is illustrated as a combination of pass bands 41,42,43,44 with zero transfer function in between, the desired functioning may also be obtained by a characteristic that can be described as a broadband pass characteristic having a certain number of reject regions 51,52,53,54.

In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word"comprising"does not exclude the presence of elements or steps other than those listed in a claim. The word"a"or"an"preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the device claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.