Process and system for reducing leakage of local oscillator

[0002] The present invention relates to a process and a system for reducing leakage of LO to the RF-port (LLO) in a real system that converts the frequency of a first signal, i.e. the LO signal, by mixing the first signal with a second supplied signal SI to a third signal SO of a frequency that is different from the frequency of the first and second signal. More specifically the invention relates to a process and a system that are applied to suppress LLO in real image reject balanced passive mode transistor mixers. Further, the invention also relates to a balanced passive mode transistor mixer and an image reject balanced mixer to be used in said process and system.

[0003] In modern radio communication there is used a conversion from a first carrier wave frequency to a second frequency. For this conversion a frequency converter/mixer is employed. If the second signal is the IF and the third signal is the RF, the mixer is called an upconverter. Otherwise, if the second signal is the RF and the third signal is the IF, the mixer is called a downconverter. The mixer may be realised equally for these two modus of the mixer. This is the case for the kind of mixer discussed herein. For the upconverter, the IF is supplied and the RF is produced, while for the downconverter the RF is supplied and the IF is produced. In the following description, the upconverter configuration is used, but the present invention is not restricted to this embodiment.

[0004] Ideally the mixer functions so that it receives the signal which is to be converted—IF—and a local oscillator signal—LO—and emits only one signal—RF—which has a frequency equal to the sum or the difference of the frequencies for IF and LO. However, practical limitations cause several other undesired signals also to be present together with RF. A so-called “image”—IM—will be present (both “the sum signal” and “the difference signal” will come to the RF-gate—the one is desired, while the other is IM and consequently undesired). LO will leak out to the RF-port (Even if the mixer is used as a downconverter, the LLO is the LO leaking to the RF-port) In addition a series of other undesired signals named spurious signals, will arise in the mixer. The important ones can be made small and insignificant by employing as far as possible a linear mixing element and by allowing LO to be strong and dominating relative to IF and RF. The level of leaked LO—LLO—is proportional to the level of the supplied LO—SLO. The two dominating spurious signals will therefore be IM and LLO.

[0005] The prior art describes some approaches for reducing the LLO. A short review of these approaches is given in U.S. Pat. No. 6,343,211, as incorporated herein by reference.

[0006] U.S. Pat. No. 6,343,211 also describes a system and a process where two balanced mixers are combined in an image reject mixer configuration. The varying LOL from each of the balanced mixers are 90 degrees out of phase, and these two leakages (named LOL1 and LOL2 in the indicated application) are added before the RF-port, and the resulting leakage will be cancelled.

[0007] Applicants U.S. patent application Ser. No. 09/218,880 also describes a system and a process where two balanced mixers are combined in a image reject mixer configuration. The varying LOL from each of the balanced mixers are 90 degrees out of phase, and these two leakages (named LOL1 and LOL2 in the indicated application) are added before the RF-port, and the resulting leakage will be cancelled.

[0008] Yngve Thodesen, A resistive HEMT image reject upconverter with tunable LO supression. 30^thARMMS Conference Digest, 1999 describes two related approaches. The first approach, also disclosed in U.S. patent application Ser. No. 09/218,880, is a resitive HEMT mixer with no drain bias, and is cold. A gate bias Vg is usually applied. The LO is supplied to the gate, while the IF and RF exist at the drain. The LOL then depends on how the LO leaks through the HEMT This application is easy to implement and suitable for MMIC design. However, it is not very robust.

[0009] The first approach is limited by the fact that the LO-power needed to drive the mixer element into a wel-behaved modus simultaneously limits the dynamic in the regulations. The approach to this problem is to get parameters usable for the regulation that are not overlapping the ones used to control performance. These new parameters are provided by four PIN-diodes. This second approach is more robust, but the four PIN diodes make the solution expensive and not suitable for MMIC design.

[0010] The aim of the present invention is thus to develop a solution that combines robustness, simplicity and MMIC compatibility.

[0011] The purpose of the present invention is to minimise the limitations of prior art solutions and systems, by providing a process, system and mixers where passive mode transistors are utilised in mixers, and where these are arranged together with reflecting elements so that the unwanted LLO is reduced considerably. The system and the process according to the present invention can thus be employed where limitations in the components produce non-ideal conditions.

[0012] In accordance with the present invention there is provided a process for reducing leakage of LO in a non-ideal system that converts the frequency of a first signal, i.e. the LO signal, by mixing the first signal with a second supplied signal SI to a third signal SO of a frequency that is different from the frequency of the first and second signal, wherein

[0013] the first signal is divided via a number of couplers into a number of part-signals, preferably 4, which are phase-displaced relative to each other, and

[0014] that each such part-signal is led to its respective transistor there being partly absorbed, partly leaking through and partly reflected, and

[0015] that the part being absorbed by each transistor are mixed with the second signal SI and frequency-shifted, and

[0016] that the reflected signal from each transistor, while being adjusted by some regulating means, are transferred back to each transistor where a part of it is leaking through, and

[0017] that the frequency-shifted part-signals are added to a third signal via a number of couplers after the signals are phase-displaced relative to each other, and

[0018] that the regulating means being adjusted so that the LO-leakage signal caused by the reflection and the regulating means is cancelling the direct LO-leakage signal so that the total LO-leakage is reduced.

[0019] A preferred embodiment of the present invention relates to a process, wherein the IF-lines are arranged so that none of these crosses, wherein the signals from the two couplers are fed to the respective decoupling capacitors to make possible for the two outer filters to be placed in the same branches as the two inner filters respectively, so that an IF-structure with couplers may be placed together inside the mixer, while being connected to the outside of the mixer through filters, while utilising couplers that do not attenuate the IF-signal, and that a decoupling capasitor is placed after the last coupler to isolate the IF-signal from leaking to the RF-port.

[0020] Further, the invention also relates to a system for reducing the leakage of LO in a non-ideal system that converts the frequency of a first signal by mixing the first signal with a second supplied signal SI to a third signal SO of a frequency that is different from the frequency for the first and second signal, wherein the system comprises:

[0021] coupler means wherein the first signal is divided into a number of part-signals, preferably 4, which are phase-displaced relative to each other, and

[0022] transistor means whereto each signal is fed, there being partly absorbed, partly leaking through and partly reflected, and that the part being absorbed by each transistor are mixed with the second signal SI and frequency-shifted, and

[0023] regulation means wherein the reflected signal from each transistor being adjusted and transferred back to each transistor where a part of it is leaking through, and

[0024] coupler means wherein the frequency-shifted part-signals are added to a third signal, after the signals are phase-displaced relative to each other, and

[0025] that the regulating means being adjusted so that the LO-leakage signal caused by the reflection and the regulating means is cancelling the direct LO-leakage signal so that the total LO-leakage is reduced.

[0026] A preferred embodiment of this system is characterized in that the IF-lines are arranged so that none of these crosses, wherein decoupling capacitors are placed after the couplers to make possible for the two outer filters to be placed in the same branches as the two inner filters respectively, so that an IF-structure with couplers may be placed together inside the mixer, while being connected to the outside mixer through filters, while utilising couplers that do not attenuate the IF-signal, and that a decoupling capacitor is placed after the last coupler to isolate the IF-signal from leaking to the RF-port.

[0027] Further, the invention also relates to a balanced passive mode transistor mixer and an image reject balanced mixer.

[0028] Further, the invention also relates to a balanced passive mode transistor mixer employing a PIN-diode or a drain-side of a passive mode transistor as a tunable reflecting load and an image reject balanced mixer wherein the couplers are 90° couplers preferred embodiments of the invention are given in the subclaims 2-12, 14-21, 23 and

[0029] The invention will now be further explained with reference to the accompanying figures and claims, wherein:

[0030] FIGS. 1aand 1bshow a balanced passive mode transistor mixer according to the present invention.

[0031] FIG. 2ashows with symbols an embodiment of a system according to the present invention where there are utilised two passive mode transistor mixers of the type as illustrated in FIG. 1bcoupled together in an image reject coupling.

[0032] FIG. 2bshows a construction of a system according to the present invention where there are utilised two passive mode transistor mixers of the type as illustrated in FIG. 1acoupled together in an image reject coupling.

[0033] FIG. 2cshows a preferred embodiment of the present invention that prevent crossing IF-lines.

[0034] FIG. 3 shows a preferred embodiment of the present invention that prevents crossing I.F.-lines.

[0035] FIG. 4 shows the reflection-coefficient S₁₁at the PIN-diode in a phase-diagram.

[0036] FIG. 5 shows the DLO from the mixer shown in FIG. 3 together with the limits of the ILO from the “phase-shifter” implemented in the same circuit, all displayed in a phase-diagram.

[0037] FIG. 6 shows measured values for the mixer shown in FIG. 3 without the use of the “phase-shifter”.

[0038] FIG. 7 shows measured values for the mixer shown in FIG. 3 with the “phase-shifter” activated to phase out the DLO.

[0039] FIG. 8 shows the reflection coefficient S₁₁at the LO-input.

[0040] FIG. 9 shows the reflection coefficient S₂₂at the RF-output.

[0041] FIGS. 10 and 11 show how the LO is transported through the mixer.

[0042] Table 1 shows measured performance of the mixer shown in FIG. 3.

[0043] In the FIGS. 1-11 the coupling points (terminals) are indicated as—0—. In order to make the Figures surveyable, we have chosen not to insert reference numerals at all these coupling points.

[0044] The following abbreviations are employed:

[0045] FIG. 1ashows a modified version of a balanced passive mode transistor mixer 90. In this circuit, 90° couplers 102 and 128 are applied. Mixers with this configuration have quite good input and output reflection-loss (reflection-coefficients S₁₁and S₂₂close to zero). The reason is that most of the two LO-signal parts reflected from the transistor-inputs and the two RF-signal parts reflected from the transistor-outputs are terminated in the terminations coupled to or included in the 90° couplers 102 and 128, respectively (K. W. Chang, et al, “High Performance Resistive EHF Mixers Using InGaAs HEMTs”, 1992 IEEE MTT-S Digest, pp. 1409-1412). In FIG. 1aa specific 90° coupler is chosen as the LO-coupler 102. This coupler has two ports 101,100 at the input and two ports 104,106 at the output. To form a 90° coupler one of the ports at the input is used as the input-port, i.e. 101, while the other is terminated, i.e. 100. If the two transistors 116 and 118 at the output of the coupler 102 are quite similar, most of the two LO-signal parts reflected from the two inputs will be added in phase at the terminated port 100, and nearly nothing will reflect back to the input-port 101. For more details, see Rajesh Mongia, RF and Microwave Coupled-Line Circuits, Artech House, 1999. In the process and system according to the present invention, a tunable reflecting load 98 is applied instead of the usual termination. This may be a PIN-diode, the drain-side of a passive mode transistor or another tunable load. By making the load higher than the system impedance, a part of the inserted signal is reflected back against the inputs of the transistors with a zero change in phase; by making it lower, a 180° phase-shifted signal is reflected. The level of this reflected signal is dependent on how far the load is regulated from the system impedance (see FIG. 4).

[0046] One portion of an inserted LO-signal will always leak through a mixer. In the present invention, this part may be viewed as the sum of two signals: one signal directly related to the LO-input at 101, that is DLO, and another related to the reflected signal from the adjustable load 98, that is ILO. The sum of these two signals, LLO, will exist at the RF-port 130. There is one major difference in the way the two signals are handled by the balanced circuit: DLO is suppressed by the balanced circuit (B dB), while ILO is not. However, ILO is dependent on the reflection-loss at the inputs of the transistors (R dB) and at the load (Dn dB). This will now be described in more detail.

[0047] The LO-signal is fed to the input-port 101 of the coupler 102 and divided into two parts with a phase displacement of 90° before being fed to the transistors 116 and 118. Some of these signals leak through the transistors 116,118 and are added in a coupler 128 with a phase displacement of 90°. A total phase displacement of 180° (90°+90°) implies that this signal, the DLO, is suppressed by the balanced circuit. The reflected signal from the tunable load is fed to the usually terminated port 100 of the coupler 102 and divided in two with a phase displacement of −90° before fed to the transistors 116,118. Some of these signals leak through the transistors 116,118 and are added in a coupler 128 with a phase displacement of 90°. A total phase displacement of 0° (−90°+90°) implies that this signal, the ILO, is not suppressed by the balanced circuit. The difference in dB between the |ILO| and the |DLO| is described by eq. 1. $\begin{array}{cc}\left|\mathrm{ILO}\right|-\left|\mathrm{DLO}\right|=B-R-\mathrm{Dn}\bigvee n=\{\begin{array}{c}1\\ 2\end{array}& \mathrm{Eq}.1\end{array}$

[0048] R and Dn are the decibel-value (dB) of the return-lose at the inputs of the transistors 116, 118 and at the load 98, respectively. B is the decibel-value of the suppression of the DLO caused by the balancing-technique.

[0049] FIG. 1bshows a symbol of the mixer 90 shown in FIG. 1a.

[0050] If the mixer 90 in FIG. 1 had been ideal, the DLO would have been zero. Then there would have been no need for ILO, which of course had to be made zero by adjusting the load equal to the system-impedance, which implies that it function as a termination. This corresponds to the point of origin in the phase-diagram given in FIG. 5. Since the mixer is not ideal, the DLO is not zero and may be displayed as a vector in the phase-diagram, where the distance from the origin determines the amplitude, and the direction from the origin determines the phase.

[0051] The present invention aims at mounting two mixers of the type shown in FIG. 1 in an image reject coupling. For the embodiments that are shown in FIGS. 2aand 2b, the first signal is an LO-signal, the second signal an IF-signal and the third signal an RF-signal. 2aand 2billustrate how the LO-signal is divided into two equal portions, i.e. SLO_aand SLO_b(see FIGS. 10 and 11). In the most preferred embodiment these signals are phase-displaced AQ1 (0° to 90°) relative to each other in a coupler 84 in the mixer 91. SLO_aand SLO_bare led out through terminals 86 and 88, respectively, and fed to each of its mixers, 90 and 92 respectively. The signal that is led out through terminal 86 is further divided in coupler 102 in the mixer 90 into two portions SLO_a1and SLO_a2that in the most preferred embodiment are phase-displaced about 90° relative to each other, and the signals are led out through the terminals 104 and 106, respectively. The signal that comes from the terminal 104 is led to a transistor 116 where SLO_a1is mixed with IF(0), which is fed through the low pass filter 120, and the signal that comes from terminal 106 is led to a transistor 118 where SLO_a2is mixed with IF(180), which is fed through the low pass filter 122. V₁and V₂can be set to a value optimised for best performance of the mixer 90. The signals are added in the coupler 128, after they are, in the most preferred embodiment, phase-displaced about 90° relative to each other, and are led out through the terminal 130. The RF and IM signals are each added in phase. A part of SLO_a1will leak through the transistor 116, that is DLO_a1, and a part, i.e. RLO_a1, of SLO_a1will be reflected back against the coupler 102. Likewise, a part of SLO_a2will leak through the transistor 118, that is DLO_a2, and a part, i.e. RLO_a2, of SLO_a2will be reflected back against the coupler 102. DLO_a1and DLO_a2are added in the coupler 128, after being phase-displaced relative to each other, preferably by 90°. The result is that DLO_a1and DLO_a2are added out of phase and a suppressed DLO_ais led out through the terminal 130. The two reflected signals are quite similar in amplitude, and phase-displaced 90° relative to each other. Therefore most of these signals will add (RLO_a1+RLO_a2=RLO_a) to be transferred to port 100. Usually this port is terminated, but in the present invention a tunable load 98 is mounted to port 100. Dependent on the size of the load, a part of the reflected LO-signal RLO_ais reflected back against the transistors. The level of this adjusted signal (ARLO) is dependent on how far the load is from the system-impedance, and the phase of this ARLO signal is changed 0° or 180° dependent on the load being larger or lower than the system impedance, respectively. The ARLO signal is divided in coupler 102 in the mixer 90 into two portions that are phase-displaced −90° relative to each other, and the signals are led out through the terminals 104 and 106. The signal, which comes from the terminal 104, is led to a transistor 116 where a part leaks through, i.e. ILO_a1. The signal, which comes from the terminal 106, is led to a transistor 118 where a part leaks through, i.e. ILO_a2. The signals are added in the coupler 128, after they are phase-displaced 90° relative to each other. Therefore ILO_a1and ILO_a2are added in phase, and the ILO_athat is not suppressed by the balancing technique is led out through the terminal 130.

[0052] Correspondingly, the signal that is led out through terminal 88 is further divided in coupler 140 in the mixer 92 into two portions SLO_b1and SLO_b2that in the most preferred embodiment are phase-displaced about 90° relative to each other, and the signals are led out through the terminals 142 and 144. The signal that comes from the terminal 144 is led to a transistor 156 where SLO_b1is mixed with IF(90), which is fed through the low pass filter 160, and the signal that comes from terminal 142 is led to a transistor 154 where SLO_b2is mixed with IF(270) which is fed through the low pass filter 158. V₃and V₄can be set to a value optimised for best performance of the mixer 92. The signals are added in the coupler 166, after they are in the most preferred embodiment phase-displaced about 90° relative to each other, and are led out through the terminal 168. The RF- and IM-signals are each added in phase. A part of SLO_b1will leak through the transistor 156, that is DLO_b1, and a part, i.e. RLO_b1, of SLO_b1will be reflected back against the coupler 140. Likewise, a part of SLO_b2will leak through the transistor 154, that is DLO_b2, and a part, i.e. RLO_b2, of SLO_b2will be reflected back against the coupler 140. DLO_b1and DLO_b2are added in the coupler 166, after they are, in the most preferred embodiment, phase-displaced 90° relative to each other. The result is that DLO_b1and DLO_b2are added out of phase and a suppressed DLO_bis led out through the terminal 168. The two reflected signals are quite similar in amplitude, and phase-displaced 90° relative to each other. Therefore most of these signals will add (RLO_b1+RLO_b2=RLO_b) to be transferred to port 138. Usually this port is terminated, but in the method according to the present invention a tunable load 136 is mounted to input 138. Dependent on the size of the load, a part of the reflected LO-signal RLO_ais reflected back against the transistors. The level of this adjusted signal (ARLO) is dependent on how far the load is from the system-impedance, and the phase of this ARLO signal is changed 0° or 180° dependent on the load being larger or lower than the system impedance, respectively. This signal is divided in coupler 140 in the mixer 92 into two portions phase-displaced −90° relative to each other, and the signals are led out through the terminals 142 and 144. The signal that comes from the terminal 144 is led to a transistor 156 where a part leaks through, i.e. ILO_b1. The signal that comes from the terminal 142 is led to a transistor 154 where a part leaks through, i.e. ILO_b2. The signals are added in the coupler 166, after they are phase-displaced 90° relative to each other. Therefore ILO_b1and ILO_b2are added in phase, and the ILO_bthat is not suppressed by the balancing technique is led out through the terminal 168.

[0053] The favourable features obtained by the system and process according to the present invention are that a circuit may operate at the same time both as an image reject mixer, which is capable of suppressing IM, and as a phase-shifter, which is capable of suppressing LLO by adjustment. The LLO is reduced to approximately equal 0 (or the origin in FIG. 5) by correct adjustment of parameters X and Y used to control the “phase-shifter”. These parameters usually are electrical signals.

[0054] The SLO, fed to the LO-input 80, is in the coupler 84 split into two parts, SLO_aand SLO_b, that are phase-displaced AQ1 compared to each other. Some part of them leaks through each balanced mixers 90 and 92, respectively, that is DLO_aand DLO_b, before they are phase-displaced AQ2 compared to each other and added in coupler 170. It is most preferred that AQ1+AQ2=90°, which implies that the IM is suppressed, but also that DLO_aand DLO_bis phase-displaced 90° compared to each other before they are added in coupler 170. This implies that also ILO_aand ILO_bare phase-displaced 90° in relation to each other. ILO_aand ILO_bmay each be adjusted between +ILO_m1and −ILO_m2, where ILO_m1is the maximum limit for 0° phase-shift in the loads and ILO_m2is the maximum limit for 180° phase-shift. These limits are shown in FIG. 5, measured for the realised device shown in FIG. 3. The result is that the present invention may operate as an image reject balanced mixer and as a phase-shifter at the same time. The phase of DLO is, however, unpredictable, which implies that DLO_aand DLO_bmay add to a signal with a phase that equals the phase of ILO_aor ILO_b. Then the other must be adjusted to zero, which implies that ILO_m1and ILO_m2have to be at least 3 dB larger than the level of DLO_a(and DLO_b) to ensure that ILO is capable of phasing out DLO. This implies that eq. 2 has to be fulfilled. $\begin{array}{cc}\mathrm{Dn}\le B-R-3\bigvee \mathrm{PDn}=\{\begin{array}{c}0°\\ 180°\end{array}\bigvee n=\{\begin{array}{c}1\\ 2\end{array}& \mathrm{Eq}.2\end{array}$

[0055] PDn is the phase of Dn, which may be 0° or 180°. If R=2 and B=15, the return-loss of the load has to be tunable down to 10 dB for each phase (0° or 180°). If the inputs of the passive mode mixers are matched to be 10 dB, the load has to be tunable down to 2 dB. Making the balance better, say 20 dB, the diodes only need to reach 7 dB.

[0056] The adjusting of the ILO may imply that the return-losses at the LO-input ports 86 and 88 of the balanced mixers 90 and 92 are degraded. If the couplers 102 and 140 were ideal, the balanced mixers 90 and 92 would have return-losses |S₁₁|₁and |S₁₁|₂at the LO-input ports 101 and 139, respectively, given by eq. 3. $\begin{array}{cc}{\left|S_{11}\right|}_n=2R+\mathrm{Dn}\bigvee n=\{\begin{array}{c}1\\ 2\end{array}& \mathrm{Eq}.3\end{array}$

[0057] When the LO-tuning is activated, eq. 4 is fulfilled. $\begin{array}{cc}\mathrm{Dn}⩾B-R-3\bigvee n=\{\begin{array}{c}1\\ 2\end{array}& \mathrm{Eq}.4\end{array}$

[0058] This implies that the return-loss at the input 101, 139 of the balanced mixers 90, 92 is given by eq. 5. $\begin{array}{cc}{\left|S_{11}\right|}_n\ge R+B-3\bigvee n=\{\begin{array}{c}1\\ 2\end{array}& \mathrm{Eq}.5\end{array}$

[0059] If R=2 and B=15, the return-loss at the LO-port of the balanced mixer would be better than 14 dB. The return-loss might be a little degraded by real couplers, however, if AQ1 is set to 90° in the coupler 84, there will be a certain improvement, and the return-loss of the LO-port 80 of the total mixer will then probably be described well enough by eq. 5. The return-loss at the output 172 should be very little affected by the adjustments.

[0060] In comparison with the known prior art solutions the following advantages are obtained with the present invention;

[0061] 1. The method and the system are robust and ensure that LLO is sufficiently suppressed it X and Y are adjusted to correct values.

[0062] 2. The method and the system are not dependent upon a extern phase shifter. The same circuit that suppresses IM also suppresses LLO.

[0063] 3. The method and the system are independent upon the power of TLO, which imply that this power may be set to a level optimised for best performance of the mixer.

[0064] 4. A passive transistor mixer is much more linear than a diode mixer.

[0065] With the novel method a very linear mixer with negligible LLO can be obtained.

[0066] Best Mode

[0067] Below is described, a presently most preferred embodiment of the present invention that is specifically developed for reducing the costs of production, avoiding undesired resonance effects, and together providing a relative good balance in the mixer. This solution comprises a circuit solution that avoids crossing IF lines, as shown in FIG. 3. The adjustable loads are realised as PIN-diodes 98′ and 136′, together with virtual grounds 94′ and 132′, where dc-currents may be connected through low-pass filters 96′ and 134′, respectively. The dc-currents are looped through each coupler 102′ and 140′, respectively, and through a common coupler 84′ and low-pass filter 82′ to dc-ground. The parameters X and Y applied by the “phase-shifter” are realised as do voltages Vd1 and Vd2, which control the diode-currents.

[0068] Decoupling capacitors, which are arranged in order to prevent the signals from flowing between the terminals of the couplers and thus leaking away or short-circuiting each other, are moved in order to achieve the aforementioned advantages. The coupler 84′ is chosen to have AQ1=90°, while the coupler 170′ is chosen to have AQ2=0°. The couplers 128′, 166′ and 170′ may easily be designed to have isolation between the adding-ports at the LO-frequency, but not at the IF-frequency. Thus the IF-signals will each reach their respective transistors without loosing the balance in level. The desired phase difference between the four IF-signals is maintained by taking care that the physical distances for the signals are equal. This is not shown in FIG. 3.

[0069] In FIG. 2bthe IF structure with 180° and 90° couplings are not shown. In the solution of FIG. 3 this is included in order to show clearly a complete realisation.

[0070] FIG. 3 thus shows a system where transistors are utilised in the passive mode. The explanation of the Figure is as for FIG. 2b, but with the difference and clear advantage that the connections for IF are on the outside of the circuit and no lines cross each other as they are obliged to do in the solution of FIG. 2bwhen the IF structure is included. The Lo-signal is divided into two equal portions in a coupler 84′ that most preferably are phase-displaced about 90° relative to each other. The one portion of the signal is further divided into two signals via coupler 102′, most preferably mutually phase-displaced about 90°, before the two signals are led out through the terminals 104′ and 106′. The decoupling capacitors 108′ and 110′ separate the two input-biases V1 and V2 from each other and the coupler 102′. After the transistors the decoupling capacitor 126′ separates the connections for the two IF-signals, which now are arranged the same transistor branch. The decoupling capacitor 124′ now separates IF(O) from leaking to the RF-output.

[0071] Correspondingly for the portion of the signal that is led from coupling 84′ via terminal 88′ to the coupler 140′, the signal is further divided into two portions, most preferably mutually phase-displaced about 90°, before the two signals are led through the terminals 142′ and 144′. The decoupling capacitors 146′ and 148′ separate the two input-biases V₃and V₄from each other and the coupler 140′. After the transistors, the decoupling capacitor 162′ separates the connections for the two IF-signals, which are now arranged on the same transistor branch. The decoupling capacitor 164′ now separates IF(90) from leaking to the RF-output.

[0072] In accordance with this most preferred embodiment of the invention, all the IF-signals are met on the inside of the mixer structure, and the IF-structure can be arranged there. This is done with two couplers 121′ and 159′, most preferable 180°, and a coupler 123′, most preferable 90°. Choosing AQ3=90°, the upper sideband is chosen as the RF; choosing AQ3=−90°, the lower sideband will be the RF. In order to feed the IF-signal into the IF-structure, two low-pass filters 163′ and 125′ are employed together with the fact that the coupler 170′ easily may be designed to not isolate at the IF-frequency. The result is that the IF-structure does not affect the balance in the image reject coupling. The IF-signal is thus fed at terminal 169′, is fed to the RF-line via low-pass filter 125′, is led unimpeded through coupler 170′ and low-pass filter 163′, then divided into two equal portions in a coupler 123′ that are most preferably phase-displaced about ±90° relative to each other, before led further to coupler 121′ and 159′, where the signals are split to IF(0), IF(180), IF(90) and IF(270). The phases of these are most preferable about 0°, 180°, 90°, 270°, respectively. It clearly emerges from FIG. 2cthat no IF-crossings are necessary.

[0073] A mixer based on the present invention includes four packaged HEMTs—NE325S01—as the mixing-elements. The couplers 84′, 102′ and 140′ are realised as branch-couplers, while the couplers 128′, 166′ and 170′ are realised as Wilkerson-couplers. The 90° phase displacement in couplers 128′ and 166′, are realised as differences in physical length of the RF-lines coupled to the Wilkerson-hybrids. To avoid crossing lines, a Wilk-hybrid with negligible isolation for the IF-frequency is utilised. In this way, most of the IF-network may be concentrated in the middle of the mixer, producing a very compact, easily producible and reliable circuit. The decoupling capacitors 124′, 164′ and 171′ are changed to be realised as coupled microstrip filters. These give a better isolation for the IF-signal than the small decoupling capacitors utilised for 108′, 110′, 146′, 148′, 126′ and 162′, but are physically larger. However, this is utilised for naturally displacing the IF-connections further out. Various undesired reflections in the IF-structure can thus be avoided. Couplers in the IF structure are diversely packed circuit elements. A variable resistance plus two variable capacitors have also been used in order to tune phase-difference and amplitude balance in the four internal IF connections. An adjustment of the suppression of IM is now possible. Two packaged PIN-diodes—MPP4203-MMSM—are applied, which implies that the mixer may be mounted by an automatic ensemble (pick and place).

[0074] FIG. 4 shows the reflection-coefficient S₁₁at one of the PIN-diodes in a linear phase-diagram. The diode-current is swept between 0 and 3 mA, being 0.23 mA where the S₁₁is closest to 0 (the return-loss is highest).

[0075] In FIG. 5, the DLO at three frequencies in a chosen frequency-band—the lowest, the middle and the highest—are shown. In the same diagram, the limits of the ILO from the “phase-shifter” are shown for the same frequencies, when the diode-currents are limited to between 0 and 1.4 mA. The LO-power is 15 dBm. It has been verified that the novel method in accordance with the present invention is quite independent of the LO-power.

[0076] FIGS. 6 and 7 show the RF, LLO and IM for the mixer in FIG. 3, given in an xy-diagram, where the x-axes is the frequency in GHz and the y-axes is the amplitude in dBm. In FIG. 6 the “phase-shifter” is not activated; in FIG. 7 it is.

[0077] FIG. 8 shows the normal and the worst-case reflection-coefficient S₁₁at the LO-input of the mixer in FIG. 3, when the “phase-shifter” is not activated and when it is activated, respectively. The return-loss at the LO-input is measured to be at least 13 dB in the selected frequency-band for all adjustments in table 1. In fact, the return-loss would for this example always be much better because the frequency for the worst return-loss always was on a different frequency than the frequency applied.

[0078] FIG. 9 shows the reflection-coefficient S₂₂at the RF-output of the mixer in FIG. 3. The return-loss at the output was better than 17 dB.

[0079] FIGS. 10 and 11 show how the LO is transported through the mixer (see FIG. 3) being partly reflected (RLO), partly leaking through (DLO) and partly absorbed for the mixing process in the transistors. In FIG. 10 the reflected and adjusted signal (ARLO) is shown at the adjustable loads. FIG. 11 shows how the ARLO is transported through the mixer being partly reflected, partly leaking through (ILO) and partly absorbed.

[0080] Finally, the performance of the mixer, shown in FIG. 3, is measured over a selected frequency-band as shown in table 1 below, when LO-power is 15 dBm. The gates of the HEMTs are not matched to system-impedance, which implies R=2. The V₁, V₂, V₃and V₄are set by self-bias through each 10 KΩ resistance to ground. If the circuit was realised in MMIC technology, the good performance could be kept over a broader bandwidth. 90° hybrids are quite suitable to be realised in MMIC technology. The structure described by this novel method and system ensures that the PIN-diodes may easily be mounted outside the chip. If the diodes where replaced by FETs, the complete circuit may be realised by MMIC technology.

[0081] The solution of FIGS. 2band 2ccan be compared with a corresponding diode mixer solution. In M. J. Schindler et al: “A Comparison of GaAs Transistors as Passive Mode Mixers” it is evident that diode mixers typically have OIP3 (“output third order intercept point”—defines how linear they are) of −6 dB to −3 dB relative to the LO-power. The solution according to the present invention has however an OIP3 of over +4 dB relative to the LO-power. It implies that IMD3 (output third order intermodulation) is improved more than 14 dB.