Development of 0.34 THz Sub-Harmonic Mixer Combining Two-Stage Reduced Matching Technology with an Improved Active Circuit Model

Abstract

In this paper, a high-performance 0.34 THz sub-harmonic mixer combining two-stage reduced matching technology with an improved active circuit model is established and analyzed. The mixer’s improved active circuit model is realized by decomposing passive functional models into basic transmission line units and their impedance matching and filtering are realized through automatic optimization. The improved active circuit model takes out the RF (radio frequency) transition model separately to set an optimization goal instead of operating directly in the mixing circuit. Compared with traditional active circuit models in the SDM (subdivision design method) and GDM (global design method), it provides a massive optimization space, larger working bandwidth, and better results. In the RF frequency range of 320–360 GHz, the SSB (single sideband) conversion loss of the 0.34 THz sub-harmonic mixer is below 9.5 dB and the RF return loss is less than 12 dB.

1. Introduction

In recent years, the enormous demand for high-speed communication, high-resolution imaging, and high-precision measurements has promoted the development of terahertz devices [1,2]. Among them, mixers play an important role and have become a research hotspot in many fields such as detection, fast imaging, planetary science, and remote sensing [3,4,5]. Over the past few decades, a series of mixers with different functions, structures, and materials have been proposed and fabricated. The mixer based on the GaAs Schottky diode is one of the most researched mixers due to its simple structure, high stability, low noise, and room temperature working conditions [6]. At present, the development of the terahertz sub-harmonic mixer is moving towards high frequency, high harmonic, and high integration [7,8,9].

For practical demands, two active circuit models have been widely used for the design of terahertz mixers. Among them, the mixer’s active circuit model based on the SDM is the most common, which is characterized by its stable performance and simple design [10]. The performance of a mixer based on this model is hindered by its limited optimization variables and small optimization space. As Figure 1 shows, the structures and parameters of a mixer’s RF transition model and LO-IF (local oscillator-intermediate frequency) diplexer model are fixed and nonadjustable in the optimization process. The impedance matching of this model only relies on the suspended microstrip on both sides of diode. As a result, the working bandwidth of a mixer based on this model is usually narrow and the return loss is not satisfactory in most situations. Moreover, if the conversion loss is not optimal, it is time-consuming to re-optimize passive functional units. In [11], a 205–230 GHz sub-harmonic mixer combining single-stage reduced technology with SDM was reported. The measured SSB conversion loss was 8–11 dB when IF frequency was fixed at 1 GHz. The mixer also showed a complicated structure and the circuit size (8.908 mm) was large.

To increase optimization variables and optimization space, an active circuit model based on the GDM was proposed [12]. By utilizing basic transmission line units, passive functional models were broken down into individual units to simulate a mixer, as shown in Figure 2. Compared with the active circuit model based on the SDM, the lengths and impedance of the waveguide and suspended microstrips can be optimized in this model and the circuit structure can be more flexible and changeable. Usually, a mixer based on this model has the characteristics of a simple structure and larger working bandwidth. However, the active model based on the GDM also has disadvantages. As there are many optimization variables, the mixers’ design process needs constant optimization and iterations, which is much more time-consuming. In addition, the lack of initial simulated parameters causes random optimization results, which are uncertain in most cases. Distinguished from the active circuit model based on the SDM, the RF transition model and LO-IF diplexer model are not simulated in this model. In actual applications, the active circuit model based on the GDM show poor stabilities. In [13], a 199–238 GHz sub-harmonic mixer based on the GDM was reported. The measured SSB conversion loss was just 9–12 dB and was not flat, which showed an obvious difference to simulated results.

In the mixer optimization and reconstruction process, some problems such as poor conversion loss, long time consumption, and bad RF return loss cannot be solved well by these models. However, at this moment, the demand for high-performance mixers is huge. For example, in a terahertz communication system, the mixer performance has a close connection with the transmitter output power and the receiver noise figure, which are closely connected with the communication distance. When the mixer’s working frequency reaches the terahertz band, the size of the entire circuit decreases dramatically and the parasitic effect makes it difficult to establish an accurate diode model, which is conducive to the design of the terahertz mixer. When the working frequency exceeds 300 GHz, the performance of the mixer has an obvious drop due to the limitations of Schottky diodes. The mixer’s RF return loss cannot be as good as the low frequency band. Under this circumstance, pursuing a terahertz mixer with low conversion loss and good return loss becomes a challenge. Meanwhile, the demand for mixers with large working bandwidth is becoming huge. Therefore, it is necessary to open up a new road for a mixer with high performance.

In this paper, to obtain a large working bandwidth, low conversion loss, and good return loss, a high-performance 0.34 THz sub-harmonic mixer combining two-stage reduced matching technology with an improved active circuit model is established and analyzed. Section 2 establishes the model of a Schottky diode, discusses the improved active circuit model and gives the design and simulation of the mixer. In Section 3, measurements are carried out to evaluate the mixer’s performance. In the mixer tests, the SSB conversion loss is 8.95–12 dB in the range of 320–360 GHz. In the last section, a conclusion is presented.

2. Mixer Design

2.1. Schottky Diode

As the parasitic effect of the diode will greatly influence the mixer’s performance, a simulation method combining the electric field and circuit was applied to this mixer [14]. A 3D EM model of an anti-parallel Schottky diode with its photograph is depicted in Figure 3a and is used to present its structure and parasitic effect. In the actual application, the parasitic impedance and the parasitic capacitance generated by the diode’s physical structure and package can cause energy loss and affect the coupling of the input signal [15]. The anode junction diameter of this diode is 1 um and the junction capacitance and series resistance are 9 fF and 7.5 ohms, respectively. In the active circuit of the mixers, the diode is simulated by a SPICE model, as Figure 3b shows [16]. The simulated result of the diode will be exported to the circuit in the ADS (advanced design system) for optimization.

2.2. Improved Active Circuit Model

After studying the active circuit model based on the SDM and the GDM, an improved active circuit model was proposed for the design of 0.34 THz terahertz mixer. As Figure 4 shows, an external auxiliary circuit was creatively added to the ADS to simulate together with mixing circuit and was used to manipulate RF insert loss. The mixing circuit includes a DC grounding, suspended microstrip transmission line, RF probe transition, suspended microstrip step structure, hammer-head filter, LO probe transition, CSMRs (compact suspended microstrip resonators) filter, RF/LO waveguide transition, and the shorted circuit end of RF/LO reduced waveguide. The external auxiliary circuit was taken from the RF part of the mixing circuit and had the same parameters as it. The impedance of port impedance termination for S-parameters was extracted from the suspended microstrip transmission line at one side of the diode. Four optimization goals were applied to this model, which include conversion loss, RF/LO return loss, and RF insert loss. In addition, a two-stage reduction process was applied for the waveguide to broaden working bandwidth.

The improved active circuit model has many advantages. Compared with the active circuit model based on the SDM, there are more optimization variables and larger optimization space. It is much easier to obtain optimal results and the working bandwidth can be broader. Compared with the active circuit model based on the GDM, it can be more effective to regulate and control conversion loss and RF return loss due to the existence of the external auxiliary circuit. The simulations of the mixer’s RF transition model and LO-IF diplexer model are necessary for this model as the optimization of initial parameters can save considerable time spent on the optimization process. By utilizing this model, the performance of terahertz mixer can be optimal.

2.3. Filters, RF Transition Model and LO-IF Diplexer Model

In order to reduce the overall size of mixing circuit, a hammer-head filter and a CSMRs filter which have characteristics of compact structure and good stop-band rejection were developed as shown in Figure 5a,b. The CSMRs filter was used as IF LPF (low pass filter) and the length was 0.86 mm. The S21 of the CSMRs filter was less than −40 dB with a range of 162–226 GHz, which can pass IF signal and isolate other signals. The hammer-head filter was used as LO LPF and the length was just 0.38 mm. The S21 of this filter was less than −40 dB with a range of 300 GHz to 420 GHz, which can effectively prevent the leakage of the RF signal.

To obtain initial optimization parameters of the 0.34 THz sub-harmonic mixer, a RF transition model and a LO-IF diplexer model were established, which are shown in Figure 5c,d. The RF signal was put by WR 2.8 waveguide (0.712 mm × 0.356 mm) and the LO signal was put by WR 5.1 waveguide (1.295 mm × 0.648 mm). The thickness of the quartz substrate and the gold strip on its surface was 50μm and 1 μm, respectively. The waveguide and suspended microstrip structure only allow single-mode propagation. As the frequency ranged from 324 GHz to 373 GHz, the return loss and the insert loss of RF transition model was more than 15 dB and less than 0.28 dB, respectively. The return loss and insert loss of the LO-IF diplexer model were more than 20 dB and less than 0.35 dB, respectively, with a range of 163–185 GHz LO frequency. After getting the desired results, the RF transition model and LO-IF diplexer model were broken into basic circuit units to form an improved active circuit model, as shown in Figure 4, and their calculation results were used as the initial optimization parameters.

2.4. Simulation

In Figure 6, a passive model of the 0.34 THz sub-harmonic mixer combining two-stage reduced matching technology with an improved active circuit model is illustrated. There are 18 variables involved in the optimization process, including six length variables related to the waveguides, six length variables, and six impedance variables related to the suspend microstrip line. For the optimization of the mixer, the IF frequency was fixed at 1 GHz and the power of the LO signal was 7 dBm. The RF sweep frequency changed from 320 GHz to 360 GHz with an RF power of −10 dBm. The optimization goal named RF insert loss (<0.3 dB) was set in the external auxiliary circuit. The other optimization goals which include SSB conversion loss (<9 dB), RF return loss (>10.5 dB), and LO return loss (>9 dB) were set in the mixing circuit. The simulated results of the mixer’s improved active circuit model are shown in Figure 8.

After acquiring the desired results, the mixer’s overall structure can be established in HFSS (high frequency structure simulator). It is advised that reconstructing the passive model of the terahertz sub-harmonic mixer from the IF filter to the DC grounding and adding the RF/LO waveguide is performed at the end. In addition, it should be noted that after the diode is added to the mixer’s passive model, new optimizations goals are needed in the mixing circuit. Here, three optimization goals, which include SSB conversion loss (<9 dB), RF return loss (>15 dB), and LO return loss (>8 dB) were given to obtain a sub-harmonic mixer with high performance as the restrictions on the matching circuit of the RF part were broken. Finally, a high-performance 0.34 THz sub-harmonic mixer combining two-stage reduced matching technology with an improved active circuit model was completed, as shown in Figure 7. The length of this mixer is 3.061 mm.

The simulated results of mixer’s overall structure are illustrated in detail in Figure 8. In comparison to the simulated results of the improved active circuit model, the final simulated results behave quite well. As for the RF frequency ranging from 320 to 360 GHz, the return loss at the RF part is below −12 dB and the SSB conversion loss is better than 9.5 dB. The return loss at the LO part is less than 9.5 dB. The return loss at the LO part is less than −8 dB. These values are almost optimal for the parameters of a Schottky diode.

3. Results and Analysis

After the final simulation, a high-performance 0.34 THz sub-harmonic mixer was fabricated, assembled, and tested, as shown in Figure 9. The photographs of the 0.34 THz mixers are shown in Figure 9b,c. The mixer’s circuit was integrated on a 50 μm thick quart substrate and was fixed on the copper block with silver glue. In the last part of the mixer, a 50-ohm microstrip line was used to connect the quart substrate and connector. The material for the black substrate was Rogers RT/Duroid 5880. The IF signal was output by SMA connectors.

The performance of the mixer was measured by a down-conversion method as shown in Figure 9a. In the measurement of SSB conversion loss, the mixers’ RF signal was provided by a signal generator (1465 L from Ceyear) with a multiplier module (82401 TA from Ceyear) and was measured by a THz power meter (Erickson’s PM5). To drive the mixers, a × 12 multiplier chain was used to provide sufficient LO power. In terms of the IF signal generated by the diode’s mixing, it was measured by a calibrated spectrum analyzer (4051 P from Ceyear). The power of the RF signal was set at a low level to make sure that the mixer works in their linear range. The return loss of RF signal and the loss of IF output lines were measured by a vector network analyzer. All the measurements were carried out at room temperature.

In respect of the 0.34 THz sub-harmonic mixer, the power of the RF input signal was set at −13 dBm to −15 dBm and the LO driving power was maintained in the range of 5.5–6.5 mW. In the RF frequency range of 320–360 GHz, the measured SSB conversion loss of this mixer was better than 12 dB. When the IF frequency was fixed at 1 GHz and 2 GHz, the measured SSB conversion loss was better than 10 dB in the RF frequency range of 325 to 340 GHz, which is illustrated in Figure 10a,b. In the LO frequency range of 163–175 GHz, the measured SSB conversion loss of this mixer was 8.95–12 dB with large bandwidth. The measured results of this mixer are almost consistent with the simulated results. This mixer also achieves a good return loss at the RF port, as Figure 10e shows. In the frequency range of 310–355 GHz, the measured RF return loss was less than −7.5 dB. In contrast with our previous work, combining single-stage reduced matching technology with GDM, the measured results are much better, as shown in Figure 10f.

The comparison between the performances of different terahertz sub-harmonic mixers is shown in

Table 1. Summary of some published sub-harmonic mixers.

Ref.	Freq (GHz)	Conversion Loss (dB)	LO Power (mW)	RF Return Loss (dB)	Reduced Matching Technology	Method
[13]	200–240	7–12 (SSB) 7.84 (min)	3	N.A.	Two-stage	GDM
[17]	320–360	7.2–24.1 (SSB) 15 (average)	7–11	N.A.	Single-stage	N.A.
[18]	320–360	7–10 (DSB) 6.9 (min)	6	N.A.	Single-stage	GDM
[19]	290–310	9–10 (SSB) 6–8 (DSB)	2.5–3	>10	Single-stage	SDM
This work	320–360	9–12 (SSB) 8.95 (min)	5.5	>12	Two-stage	An improved active circuit model

. The proposed 0.34 THz sub-harmonic mixer is close to the international and domestic levels in terms of conversion loss and optimal LO power. The proposed 0.34 THz mixer combining two-stage reduced technology with an improved active circuit model have the characteristics of simple structure, low conversion loss, high stability, and large working bandwidth.

4. Conclusions

In this paper, a high-performance 0.34 THz sub-harmonic mixer combining two-stage reduced matching technology with an improved active circuit model is designed, fabricated, and measured. For this mixer, the measured SSB conversion loss is almost better than 12 dB and the best value is 8.95 dB. The experimental measurements of this mixer are consistent with the simulated results, which powerfully demonstrate the effectiveness and correctness of this model. We hope that this work will be able to facilitate the development of terahertz sub-harmonic mixers in the future.