We present in this paper a wideband RF demodulator using a five-port correlator and a power detector for channel sounding applications. The demodulator has been fabricated using microstrip components. The correlator receives from the five-port qualities that allow it to be low-cost and less sensitive to the phase and amplitude imbalances. A calibration procedure is proposed to find the complex envelope of the RF signal applied at the input of the five-port correlator. Simulation with Advanced Design System software and measurement results have been conducted to demonstrate its capabilities as a RF signal demodulator operating in a wideband around 2.4 GHz frequency.

Wireless communication systems development and optimization require a good knowledge of the radio propagation channel. The wideband channel sounder has been already discussed by several studies in the time-Frequency domain [1-3], but a few are carried out in the spatial domain [4,5]. An experimental characterization of the complex wideband propagation channel is necessary in order to propose models for the development of such systems. Single Input Multiple Output (SIMO) or Multiple Input Multiple Output (MIMO) wideband systems use this complexity to increase the channel capacity [6,7]. To optimize the performance of these techniques, it is important to quantify the propagation channel parameters in time and space domain [8,9]. Sounders are the tools used to characterize a radio channel. Several wideband sounders are based on Single Input Single Output (SISO) system using a Vector Network Analyzer (VNA) for frequency domain characterization, and digital oscilloscope for time domain characterization [

The object of this paper is to present a new homodyne demodulator based on a five-port correlator and power detectors in order to design a low-cost/high performance circuit. Simulation study was validated by measurement performed on the fabricate five-port. The paper is organized into six sections. The operating principle of the fiveport correlator and its architecture will be presented in Sections 2 and 3, respectively. The calibration method will be detailed in Section 4; and in Section 5, measurement results of the prototype will be exposed. Section 6 will conclude this study.

The five-port correlator is described in _{1} (RF signal) and a_{2 }(LO signal), the power detectors measure the power of these three interferometric mixtures.

The five-port correlator is modeled by the following equations:

Complex parameters A_{k} and B_{k} depend on the S parameters of the RF linear circuit at frequency f_{LO} = f_{RF}. a_{LO} and a_{RF} are the RF and LO input signals, respectively. bk is the transmitted pseudo power wave at ports 3, 4 and 5 of the ring.

We can then write:

A_{RF} and A_{LO} are the amplitudes of the radiofrequency and local oscillator signals. I and Q represent the inphase and quadrature signals. If the two signals are synchronized, it is assumed that. The three real RF voltages are obtained with the real part of Equation (1):

α_{k} and β_{k} are the modulus of A_{k} and B_{k}, respectively; φ_{k} and θ_{k} are their respective arguments. The last term contains the main information to recover.

In practical, the three output demodulator voltages must be measured and then I, Q parameters will be deduced after a calibration procedure. There are several calibration methods like for example the five-port self calibration [

The circular passive five-arm ring is chosen as the interferometric circuit. It is realized using microstrip technology with an epoxy substrate. The ring operates at 2.4 GHz frequency with a bandwidth of 1 GHz. The impedance characteristic of the ring was set to 44.7 Ω. All accesses have an impedance of 50 Ω, which gives symmetry of the device.

The following three figures present the simulation results of the five-arm ring obtained with Agilent Advanced Design System (ADS) environment.

_{ii} for i = 1 to 5. All these curves are confounded and shows that the system is well matched since |S_{ii}| < −10 dB for a wide frequency band around 2.4 GHz, and especially at the operating frequency for which S_{ii} = −24.19 dB. Reference [

Vectors representing the three output voltages, and are rotating according to a Φ rotation. Therefore, there will be a relative phase shift ±Φ between the outputs. However, it is possible to correct this phase shift error by using a calibration procedure.

Figures 4 and 5 show the variations in modulus and phase of S_{1J} parameters depending on the frequency. First note that as well for the magnitude curves as for the phase curves, S_{12} and S_{15} (curves in green and red) on the one hand; and S_{13} and S_{14} (curves in pink and blue) on the other hand are confounded. As expected, the Five-arm ring operates at 2.4 GHz as a power divider, ensuring an equal distribution of almost half of power (0.503 for S_{13} or S_{14}, and 0.461 for S_{12} or S_{15}) to each accesses with a phase shift of +/−120˚ approximately, which will enable us to find both signals I and Q and to obtain the complex envelope of the baseband signal.

As shown in

where V_{T} = 26 mV and I_{s} represents the saturation current of the diode (I_{s} = 3 μA for HSMS2850 of Agilent). If the amplitude of the RF input voltage is lower than V_{T}, Equation (5) can then be approximated as follows:

This approximation is valid for low input power [_{V} represents the video resistance of the power detector [

The voltage is proportional to the baseband term of, which corresponds to the even-order term of Equation (6). The three output voltages are:

with.

Equation (7) represents the quadratic law of a power detector. By using Equations (4) and (7) and assuming that f_{LO} = f_{RF}, we obtain Equation (8) representing the output voltages for k = 3 to 5.

with.

We see that each voltage is composed of three terms:

• corresponds to the DC-Offset of the LO self-mixing signal;

• corresponds to the RF auto mixture. It is the second order inter-modulation term or IM2;

• corre- sponds to the mixing signal between LO and RF signals.

The last term is the desired signal I/Q, it carries the information and contains the complex envelope of the RF signal. In the classical receiver, the baseband I/Q signals are directly obtained by two mixings. So, a signal processing must be applied on the three output voltages to reject the dc offset and the even-order terms, and regenerate the I/Q signals.

The calibration method of the five-port demodulator is the same as that of the six-port, and the literature related to this subject is numerous [18,19]. When used as a reflectometer, the six-port calibration is both very powerful and complicated because it requires human intervention for the calibration standards (short, open, and matched load).

The calibration procedure consists to determine three complex constants to regenerate the complex envelope from the three output voltages [

In Equation (8), the term can be written as the sum of two terms as follows:

where and represent the dc and time-variant components, respectively. Equation (8) becomes:

where is the dc component of the output voltage. These components can be estimated by averaging, and eliminated by subtraction. We thus obtain:

definies the average.

The expressions of the three output voltages after dcoffset cancellation are:

with: and

The three output voltages are a linear combination of the three time-variant terms and which can be expressed in the following matrix form:

where.

Assuming the matrix P invertible, it is possible to write:

where. (15)

From these two relationships, we obtain the following equations:

Equations (16) and (17) define two relations between I (t), Q(t) and the 3 output voltages, and show the six calibration constants and with k = {3, 4, 5}. Thus, after determination of these constants and measurement of the voltages, it will be possible to regenerate the signals I(t) and Q(t), and obtain the complex envelope of the baseband signal:

with

g_{3}, g_{4} and g_{5} are the three complex calibration constants previously mentioned.

In order to validate this calibration method, measurements were carried out. Two RF generators, Marconi 2031 and HP-4432B, are connected to the input ports 1 and 2 of the five-port correlator, respectively.

A 10 MHz clock synchronizes both generators and a frequency difference Δf = 100 Hz is set between the RF input signals. Let us remind that the five-port calibration is to get the complex calibration constants and thus determine the complex envelope. A simultaneous sampling of the five-port outputs is made by a data acquisition system, the sampling frequency of the acquisition board is set at f_{s} = 10 KHz. The carrier frequency is fixed at 2.4 GHz and the RF generators are set to supply P_{OL} = 0 dBm and P_{RF} = −20 dBm of power. We obtain symbols equally distributed along a circle, so the calibration constants generated are reliable.

Measurement results of the three demodulator output voltages are presented in

The demodulator output voltages are sinusoids resulting from the dephasing between the RF and LO signals. Non-zero phase shift between at least two of the three accesses should be verify to demodulate the signal. Moreover, the redundancy of the five-port correlator architecture allows correcting the interferometer error phases. We note a complementary variation of the three voltages; each one has only one maximum or minimal value for a dephasing between the LO and RF signals ranging from 0 to 360˚. The five-port junction can then be considered as a phase discriminator. Each voltage is 2π periodic and the three minimal voltage values are separated by multiples of 2π/3. This consideration is used to estimate the directional of arrival of multi-path signals by measuring the phase difference of detected signals [20,21] and also, it is used to estimate the time of arrival by measuring the phase difference over two successive frequencies.

In this paper, a new wideband demodulator using a fiveport correlator for channel sounding application has been presented. We focused on three main criteria, which are calibration, phase and demodulation capacity in the whole band. The complex envelope can be evaluated by making power measurements of the five-port wave correlator outputs. In order to accomplish power detection over a larger dynamic, a calibration procedure was used to set aright the five-port correlator imperfections and compensate the phase and amplitude losses in the system. Measurement results were conduced to evaluate the performance of the wideband demodulator, and results obtained with the circuit have shown an excellent agreement between theory, simulations and measurements. In addition of the wideband operating, the five-port correlator has a redundant output allowing it to be more robust against the phase and amplitude unbalance of the local oscillator. Nevertheless, these features will only be exploited if a calibration procedure is applied, which could lead to erroneous results if it is not performed correctly.