D.Hammou,E.Moldovan,and S.O.Tatu

(Institut National de la Recherche Scientifique-Énergie Matériaux et Télécommunications,Montréal,Québec H5A 1K6,Canada)

Abstract:This paper presents a new implementation of a millimeter-wave heterodyne receiver based on six-port technology.The six-port model is implemented in Advanced Design System(ADS)using S-parameter measurements for realistic advanced simulation of a short-range 60 GHz wireless link.Millimeter-wave frequency conversion is performed using a six-port down-converter.The second frequency conversion is performed using conventional means because of low IF.A comparison between the proposed receiver and a conventional balanced millimeter-wave mixer shows that the proposed receiver improves conversion loss and I/Q phase stability over the local oscillator(LO)and RFpower ranges.The results of demodulating a V-band quadrature phase-shift keying(QPSK)signal at a high data rate of 100 Mb/s-1 Gb/s are discussed.The results of a bit error rate(BER)and error vector magnitude(EVM)analysis prove that the proposed architecture can be successfully used for wireless link transmission up to 10 m.

Keyw ords:millimeter wave;six-port;frequency conversion;heterodyne;front-end;wireless LAN

1 Introduction

R ecent dramatic growth in wireless communication systems has caused microwave frequency bands to become overcrowded.

The millimeter-wave spectrum provides multi-Gb/s data transmission[1],[2]and is a possible solution to this problem.

Millimeter-wave frequencies enable the design of compact,low-cost wireless millimeter-wave communication front-ends.These can offer convenient terminal mobility and high-capacity channels.

This paper describes a millimeter-wave heterodyne receiver based on a new six-port architecture dedicated to high-speed wireless communication systems.Six-port technology is widely used in microwave applications for low-cost circuit characterizations[3]and has also been used as an unconventional technique for performing frequency down-conversion[4],[8].

2 Millimeter-Wave Six-Port Circuit

Anumber of six-port architectures have been described in literature.All of these architectures are functions of fabrication technologies and specific applications,such as down-converter,direct modulator,and reflection coefficient measurement.

The best-known six-port architecture for communication purposes comprises a 3 d BWilkinson power divider and three 90°hybrid couplers.At high frequencies,the two branches of the Wilkinson power divider must be placed very close to each other to be connected to the 100Ω resistor.Astrong,undesirable mutual coupling between the output lines occurs when using a conventional Wilkinson power divider[9],[10].

Therefore,a second six-port is built with four 90°hybrid couplers to avoid the Wilkinson power divider.The power divider is implemented using one of the 90°hybrid couplers and a transmission line of 90°.This deteriorates the desired amplitude and phase-split ratio over a wide band and increases phase and amplitude imbalance.The S-parameter measurements for this architecture show that,for the coupled ports being considered,phase imbalance is less than 10%over the entire bandwidth[11].However,the amplitude imbalance reaches 1.5 d B at higher frequencies.

In addition,S-parameter measurements for the ring power divider developed in[12]show that the circuit has a quasi-null amplitude imbalance of 0.25 d B and additional insertion loss of＜0.5 d Bfrom 60 GHz to 70 GHz.

In light of recent progress in power divider design,a new six-port architecture is proposed.This circuit is constructed using three hybrid couplers(H-90°)and a new ring power divider[13].

▲Figure 1.Microphotograph of the proposed six-port circuit.

For full S-parameter characterization of a six-port circuit,15 circuits prepared for two port measurements are needed.To reduce the cost of circuit fabrication,and regarding the symmetry of the six-port architecture,our approach is to evaluate the most important S-parameters from port 5.These parameters include transmission,output phase balance,and RFinput isolation.

Five six-port circuits(5-1,5-2,5-3,5-4,and 5-6),prepared for two port measurements,are integrated on a TPS superstrate with relative permittivity of 9.9 and thickness of 127μm.Fig.1 shows this circuit in S54configuration.

S-parameter measurements are taken in the 60-65 GHz band because of the characteristics of the measuring equipment.Measurements show that the RFinputs at ports 5 and 6 are well matched;that is,results are better than-15 d Bacross the whole frequency band.Isolation between the two RFinput ports is at least 15 d B and reaches 30 d B at the central frequency.High isolation is needed to avoid local oscillator(LO)leakage to the antenna.

The S-parameter measurements for transmission show＜0.5 d B of additional loss,and the relative amplitude imbalance between requested ports does not exceed 8.33%across 60-65 GHz.

The S-parameter measurements for phase performance between two requested ports show a relative phase imbalance of only 7.6%,related to the expected quadratic reference of 90°across 60-65 GHz.

3 Mixer Topologies

Here,we describe a six-port millimeter-wave mixer(SPMM)that performs down-conversion and can improve mixing signals in terms of conversion loss versus RFsignal and LO power sweep.

The proposed down-converter comprises the new six-port circuit as wellas two pairs of parallel Schottky diodes that are connected to two differentialamplifiers and act as power detectors(Fig.2).

A conventional I/Q mixer uses couplers and two anti-parallel pairs of Schottky diodes acting as LO-driven switches(Fig.3).For comparison with the proposed architecture,a six-port circuit is used instead of conventional mixer couplers.The intermediate frequency(IF)amplifiers have the same gain as the IFdifferential amplifiers of the proposed mixer.The block diagram for this architecture is shown in Fig.3.

To estimate the conversion loss for both mixers and the stability of the I/Q phase,harmonic balance simulations are performed using Advanced Design System(ADS).Zero-bias Schottky diode(Spice model HSCH-9161)is used as a non-linear element in both topologies.First,we analyze the behaviour of the two topologies in relation to the LO drive in the range-15-15 d Bm.RFpower is fixed at-25 d Bm,and the amplifier gain is fixed at 10 d Bfor both mixers.

▲Figure 2.SPMMusing paralleldiodes.

▲Figure 3.Mixer using anti-parallel diodes.

Figure 4.▶Simulation results:conversion loss vs.LO.

Over the entire LO power range,the output I/Q power of the proposed mixer is better than that of a conventional mixer by 42 d Bm(Fig.4).The conversion loss of the proposed mixer is improved by 42 d Bm.In addition,the LO driving power of the quasi-conventionalmixer should be at least 10 d Bm to move it into saturation area.

◀Figure 5.Simulation results:I/QIF phase balance vs.LO.

The phase performances of both mixers are quasi-similar and close to the quadratic reference of 90°until LO power of 5 d Bm,at which point I/Q phase performance of the conventional mixer deteriorates dramatically(Fig.5).At an LO power of 6 d Bm,phase performance is 10°.The proposed mixer has a phase disparity of only 0.5°over the entire LO drive range.

Second,we analyze the two topologies in relation to RFdrive in the range-50-0 d Bm.LO power is-10 d Bm for the proposed mixer,and the sweep is-10-5 d Bm for the quasi-conventionalmixer.The amplifier gain is 10 d Bfor both mixers.

Fig.6 shows the results of HB analysis in terms of I/Q IFpower versus RFdrive.The proposed mixer gives excellent results.When RFpower is-50 d Bm,I/Q IFpower is approximately-60 d Bm for the proposed mixer and-80 d Bm for the quasi-conventionalmixer with 5 d Bm of LO power.

This represents a 20 d Bm improvement in conversion loss with 30 times less LO power.

Keeping the same parameters for the mixers,we analyze I/Q phase(Fig.7).

The I/Q phase of the proposed mixer is almost the same as the quadratic reference of 90°from-50-0 d Bm.The I/Q phase of the quasi-conventional mixer has quasi-similar performance for RFpower over a limited range of-36-10 d Bm.

Low-cost receivers need reduced LO power and high I/Q phase stability.So the proposed mixer is an excellent candidate for the applications dealt with in the next section.

4 Receiver Architecture and Operating Principle

Heterodyne architecture is widely used in wireless communications because it has many well-known advantages.However,the main problem in the millimeter-wave domain is the design of a high-quality low-cost mixer.To overcome this problem,the receiver uses the new SPMM described in the previous section instead of a conventional mixer to perform millimeter-wave down-conversion.Fig.8 shows the proposed heterodyne receiver architecture.

In this section,we show that the first down-conversion from RFto IF,shown in Fig.8(a),can be performed using specific properties of the SPMM.

The SPMM inputs are connected to a low-noise amplifier(LNA)and to the first(millimeter-wave)local oscillator(LO1),respectively.The intermediary frequency module(IFM)allows quadrature IFsignals to be obtained using detected output six-port signals and differential amplifiers.The second local oscillator(LO2),two mixers,and two low-pass filters(LPF)are used togenerate the primary baseband signals.The baseband module(BBM)amplifies these signals,and sample and hold circuits(SHCs)operating at the bit-rate frequency generate improved demodulated output signals.Signal processing techniques compensate for rotation of the demodulated constellation caused by instability of the LOs,especially LO1.

Figure 6.▶Simulation results:conversion loss vs.RF.

Figure 7.▶Simulation results:I/Q IFphase balance vs.RF.

◀Figure 8.Block diagram of the heterodyne six-port receiver.

The operating principle of the six-port direct-conversion receiver is described in[14].In this paper,we demonstrate that a six-port circuit can down-convert millimeter-wave modulated signals to intermediary frequency IF.The second frequency conversion,from IFto baseband,can be easily obtained by conventional means related to relative low frequency operation.

Generally,the output signals(bi)of a multiport can be expressed with the dispersion parameters Sijas

The scattering matrix of the proposed six-port circuit can be obtained using the scattering matrix of a 90°hybrid coupler and the power divider in Fig.1.

It can be assumed there are two normalized wave inputs:a5from the LO and a6from the RFinput(Fig.8).These two normalized input waves have an amplitude ratioα,phase difference Δφ(t)=φ6(t)-φ5,and frequency shift Δω=ω-ω0.These are expressed as

Assuming a perfect match a1=a2=a3=a4=0,the four normalized output waves can be calculated using(1)and the six-port scattering matrix(2):

To obtain the IFoutput signals,four power detectors are connected to the multiport outputs.The output voltage of an idealpower detector is proportional to the square magnitude of the RFinput signal:

Supposing identical power detectors are used,that is,Ki=K,for i=1 to 4,the output voltages are

The output voltages at ports 1 and 3,and 2 and 4 are phase opposites.Therefore,the quadrature output signals can be obtained using two differential amplifiers at the outputs of the SPMM stage:

After the second frequency conversion and low-pass filtering,the output I/Q signals are obtained:

Equations(3)-(13)depend on the amplitude ratioα,phase difference Δφ(t),and frequency shiftΔω.The proposed heterodyne receiver based on the new six-port architecture can demodulate arbitrary phase-shift keying(PSK),QPSK,M-ary phase-shift keying(MPSK),and M-ary quadrature amplitude modulation(M-QAM)schemes.

5 Demodulation Results for the Heterodyne Six-Port Receiver

To perform realistic simulations using ADS,and to evaluate demodulation performances,we used our V-band six-port circuit model,validated by S-parameter measurements in[13].Envelope simulations were performed using pseudorandom QPSK-modulated signals.The transmitter uses a vector modulator and two pseudorandom sources(Iand Q input signals)with two different voltage levels of±1 V.The carrier consists of a 61 GHz V-band signalthat is modulated in the range 100 Mb/s-1 Gb/s.The propagation path simulation is performed using the Friis model,and the receiver incorporates the configuration in Fig.1.

▲Figure 9.Typicalspectrum of a quadrature IFsignal(Ior Q)for six-port receiver.

▲Figure 10.Typicalspectrum of a baseband quadrature signal(Ior Q)for six-port receiver.

According to the Friis equation[15],the free space line-of-sight(LOS)attenuation is 88 d B for d=10 m.In this system analysis,the antenna gains are set at 10 d Bi.The LNAgain and noise figure(NF)are 21 d B and 3.8 d B,respectively.These are common values for today's 60 GHz integrated amplifiers.Limiters are used in the baseband stage of the receiver to obtain data output squared waves,and consequently,perfect demodulated constellations.In keeping with future high-speed requirements of the IEEE 802.15.3.c wireless standard,IFof 900 MHz was chosen.The LO is perfectly synchronized,an important advantage of this architecture compared to direct conversion.Perfect synchronization can be obtained by controlling the lower frequency LO from the digital processing block to IFM block(LO2).

Fig.9 and Fig.10 show typical spectrums of a quadrature IFsignal(Ior Q)centered at 900 MHz and a baseband quadrature signal(Ior Q)obtained after the second down-conversion.The results show the performance of a six-port receiver designed to replace a conventional millimeter-wave mixer.The signals are pseudorandom QPSK-modulated at 100 Mb/s.

Fig.11 shows the demodulation results of a pseudorandom QPSK bit sequence of 700 ns.The demodulated signal shapes are the same as the I/Q signals generated by the transmitter.

The gray line represents the baseband signal before it reaches the SHC,which improves the demodulated signal shape.

To evaluate the wireless link quality for Gb/s pseudorandom QPSK-modulated signals,a BER analysis is performed for the pseudorandom QPSKmodulation data rate 100 Mb/s-1 Gb/s.BERresults are presented as a function of Eb/No,where Ebis the average energy of a modulated bit,and Nois the noise power spectral density(Fig.11).

Simulations show that BER performance is less than 10-6for a typical Eb/Noratio of 14 d B in all cases.This is an excellent result if we consider that it corresponds to a millimeter-wave LO stability of 10-3,compared to 10-6/°C of commercial oscillators

Fig.12 shows only 2.2 d B additional shift in Eb/No.This is an excellent result if we consider that it corresponds to the tenfold increase in data rate from 100 Mb/s-1 Gb/s.To reach a BERof 10-9(required for uncoded HDTV transmission),Eb/Nomust increase a further 2.8 d B to reach 16 d B.

To calculate error vector,we consider the demodulation results for a pseudorandom QPSKbit sequence of 700 ns in Fig.11.A comparison is made for data rate of 100 Mb/s-1 Gb/s.

Figure 11.▶Demodulation resultsof 100 Mb/s pseudorandom bit sequence.

Fig.13 shows the idealand QPSK demodulation constellation diagrams at 100 Mb/s.The four clusters are very individualized,the output signals are square waves,and the output constellations are situated at the corners of a square that is centered at the origin.

◀Figure 12.BERperformance for a QPSKmodulation data rate of 100 Mb/s-1 Gb/s.

▲Figure 13.Pseudorandom QPSKdemodulation constellation diagrams.

However,constellation normalization must be enabled in order to effectively calculate the error vector magnitude(EVM)[16].

In the complex plane,we define the baseband signalГin,nandГout,n as

These vectors represent the transmitted and ideal baseband signals.Constellation normalization is achieved by affecting the output constellationГout,nwith a K coefficient.

The symbols in the two diagrams in Fig.13 become correlated[16]

Thus,when symbols have been normalized,EVM is defined as the root-mean-square(RMS)value of the difference between a simulated symbols and idealsymbols:

where N is the number of symbols.

Fig.14 shows the EVM-per-symbol calculations for a pseudorandom QPSK bit sequence of 700 ns at 1 Gb/s.The instantaneous error vector magnitude is＜1.5 over 700 ns.

The effective EVM for each QPSK data rate from 100 Mb/s-1 Gb/s corresponds to the mean of the EVM-per-symbol with the bit sequence length(number of symbols).

These results are shown in Fig.15.As expected from the EVM-per-symbol calculations,the effective EVM is less than 2%for 100,200,500,800 Mb/s and 1 Gb/s QPSK data rates.

These calculations confirm the BER results in the previous paragraph and give an indication of the capability of the six-port heterodyne receiver for the millimeter-wave wireless high data-rate communications systems.

6 Conclusion

A new millimeter-wave heterodyne six-port receiver is presented in this paper.To perform millimeter-wave frequency conversion,this receiver uses the specific properties of the new six-port circuit in Fig.1.This avoids the need for a conventional I/Q mixer with anti-parallel diodes and conventional couplers or a costly active mixer.Realistic system simulations for a short-range 60 GHz wireless link were performed using a measurement-based six-port model and a QPSK-modulated signal of 1 Gb/s.The demodulation results are validated by BERand EVM calculations.The proposed six-port heterodyne architecture enables the design of compact,low-cost wireless millimeter-wave receivers for future high-speed wireless communication systems.

Figure 14.▶EVM-per-symbol for a transmitted sequence at 1Gb/s.

◀Figure 15.Effective EVMvs.QPSKdata rates at 0.1-1 Gb/s.

Acknowledgement

The financial support of the National Science Engineering Research Council(NSERC)of Canada is gratefully accepted.The authors would like to express their gratitude to chief technologist,Jules Gauthier,Poly-Grames Research Center,École Polytechnique de Montréal,for his technicalassistance.