30 GHz RF-MEMS Dicke Switch Network and a Wideband LNA in a 0.25 µm SiGe BiCMOS Technology

Advances in Microelectronic Engineering (AIME) Volume 3, 2015 doi: 10.14355/aime.2015.03.001

www.seipub.org/aime

30 GHz RF-MEMS Dicke Switch Network and a Wideband LNA in a 0.25 µm SiGe BiCMOS Technology Shakila Bint Reyaz1,2, Carl Samuelsson3*, Andreas Gustafsson3, Robert Malmqvist1,3, Rolf Jonsson3, Mehmet Kaynak4, Anders Rydberg1 1

Department of Solid-State Electronics, Uppsala University, 751 05 Uppsala, Sweden

2

Departement of Telecommunications, NED University of Engineering and Technology, 75210 Karachi, Pakistan

3

Swedish Defence Research Agency (FOI), 583 30 Linköping, Sweden

4

IHP GmbH, 15236 Frankfurt (Oder), Germany

*Now with SAAB AB, 581 88 Linköping, Sweden shakila.bint.reyaz@angstrom.uu.se; 2carl.samuelsson@sabbgroup.com; 3andreas.gustafsson@foi.se; robert.malmqvist@foi.se; 5rolf.jonsson@foi.se; 6kaynak@ihp-microelectronics.com; 7anders.rydberg@angstrom.uu.se

1 4

Abstract This work presents a novel monolithic integration of a 30 GHz RF-MEMS Dicke switch network and a wideband LNA realised in a 0.25 μm SiGe BiCMOS process. The wideband LNA design has a measured gain of 10-19.9 dB at 2-33 GHz given a DC power consumption (PDC) of 35 mW and a measured noise figure of 5.4-6.3 dB at 14-26.5 GHz when PDC=7.5 mW (the LNA gain is then 10-14.2 dB at 4-26 GHz). The Dicke switch has 3 dB and 22 dB of losses and isolation at 25 GHz. The MEMS switched LNA gain was found to be 10-17 dB lower than anticipated due to some unintentionally missing metal via contacts between the Dicke switch and LNA ground planes. Despite this fact, the MEMS LNA resulted in a measured isolation of 9.0-13.5 dB at 24-31 GHz when the Dicke switch was switched ON and OFF which validates the switching function of the SiGe RF-MEMS wideband LNA design. Such reconfigurable low-power MEMS switched RFICs could be used in highly adaptive broadband receiver front-ends for wireless communication, sensor networks and imaging systems, for example.

Keywords Low Noise Amplifier; Millimeter Wave; RFIC; Receiver; RF-MEMS

Introduction Reconfigurable high-performance integrated circuits are key elements in RF systems for wireless communication, space, defense and security applications at microwave and millimeter-wave frequencies. RF-Micro-ElectroMechanical-Systems (MEMS) switches present low losses/DC power and high linearity/isolation over large bandwidths making them attractive candidates for highly adaptive broadband front-end architecture solutions; e.g. switches used in handsets, and radar sensors [1], [2]. On-chip integration of RF-MEMS and active RF devices is a next step in the development of reconfigurable Monolithic Microwave/RF Integrated Circuits (MMICs/RFICs) as some foundries have integrated MEMS switches on top of III-V and silicon substrates, and it can further improve the system performance [3]–[8]. This is especially important in low-noise receivers where there are the losses before the first amplification stage, i.e. the low-noise amplifier (LNA) has an impact on the over-all noise figure (NF) and receiver sensitivity. RF-MEMS together with active RF circuitry has so far with a few notable exceptions mostly been realised as hybrid circuits and mainly at frequencies below 30 GHz which still leaves room for significant improvements to be made in terms of RF performance, frequency range, functionality as well as to achieve reduced complexity (higher integration) and lower costs. The first known examples of on-chip integration of RF-MEMS in switched LNA and power amplifier circuits up to 30 GHz were demonstrated using the Rockwell Scientific GaAs MMIC process [3], [4]. More recently, some GaAs and silicon based MEMS & switched LNA ICs were presented with an NF of 2-4 dB and 7-8 dB up to 26.5 GHz and 77 GHz, respectively [5]–[8]. The first and only silicon MEMS LNAs operating above 10 GHz that has been reported so far are two narrow-band and dual-band (switchable)

1

www.seipub.org/aime

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

designs made at 60/77 GHz and 24/74 GHz, respectively [7], [8].

In certain mm-wave applications (such as automotive radar, wireless communication/sensor networks and active/passive imaging systems) Silicon-Germanium (SiGe) Bipolar Complementary Metal Oxide Semiconductor (BiCMOS) technology is one of the most cost-effective IC processes because of its high frequency performance and embedded integration of analog/digital circuit functions at a lower cost. Compared with hybrid integration, onchip integration of RF-MEMS devices in SiGe processes can provide shorter connection paths (less parasitic losses), improved RF performances and a higher integration. The SiGe foundry IHP developed capacitive MEMS switches are realised using the Back-End-of-Line (BEOL) metallisation of a BiCMOS process [9].

Figure 1 depicts a receiver front-end architecture used in high-sensitivity (low-noise) radiometers for space and security sensor applications. The sensitivity (or thermal resolution) of such systems is largely influenced by the receiver (LNA) NF, gain and bandwidth and a low-loss Dicke switch network can be used for calibration purposes and compensate for gain variations [10]–[14]. A SiGe LNA and a Dicke switch network with 3 dB/12 dB of losses/isolation at 130 GHz were reported in (Shumakher et al. 2013). References [15]–[17] provide some examples of switched LNAs and receiver modules for satellite communication payloads and radiometer systems at Ka-band (26.5-40 GHz).

FIG. 1 A SINGLE-CHIP SIGE RF-MEMS SWITCHED RECEIVER FRONT-END ARCHITECTURE INTENDED FOR HIGH SENSITIVITY (LOWNOISE) BROADBAND RADIOMETERS FOR SPACE AND SECURITY SENSOR APPLICATIONS.

The main target in this study was to monolithically integrate for the first time a 30 GHz RF-MEMS Dicke switch network and a wideband LNA in IHP’s 0.25 µm SiGe BiCMOS process (fT, fmax=190 GHz, 220 GHz). Some earlier wideband LNA designs made in 0.18 µm/0.25 µm SiGe processes reported an NF of 4-6 dB at 15-35 GHz [18], [19] and one can then expect that a single-chip MEMS switched LNA could reach an NF in the range of 5-8 dB (if realised in such processes and assuming 1-2 dB of switch losses). Such a MEMS LNA may be used in a Dicke switched front-end, for example, if realised with a high gain and isolation (≥10-20 dB) over a broad frequency range.

In this study, we will target an approximately 30% wide instantaneous bandwidth (i.e. 10 GHz for such a circuit design made at 30 GHz). Table 1 shows the target characteristics of a 30 GHz RF-MEMS Dicke switch network and a wideband LNA circuit made in a 0.25 µm SiGe BiCMOS technology.

The wideband SiGe LNA presented in this paper has a measured maximum gain of 10-19.9 dB at 2-33 GHz (the measured LNA NF was 5.4-6.9 dB at 11-26.5 GHz). The MEMS Dicke switch has 3 dB/22 dB of measured min/max (ON/OFF state) transmission losses. Some missing via connections between the switch and LNA ground planes caused the gain of the fabricated SiGe MEMS LNA design to be much lower than expected in the ON state. In spite of this, the MEMS LNA achieved 9-14 dB of measured switched isolation at 24-31 GHz thereby validating the switching function of the Dicke switch used in this design. To our best knowledge, the presented MEMS switch and LNA designs represent the first time; a MEMS Dicke switch network and a wideband LNA have been monolithically integrated in a SiGe process. Furthermore, recent SiGe MEMS switch networks have reported less than 1 dB of losses and 20 dB of isolation [9] which implies that an optimised 30 GHz MEMS switched LNA potentially could reach an NF of 5-6 dB with 20 dB of gain and isolation if realised in the same SiGe process.

TABLE 1 TARGET VALUES OF A 30 GHZ RF-MEMS DICKE SWITCH AND A WIDEBAND LNA MADE IN A 0.25 µM SIGE PROCESS.

Component Switch LNA

2

Gain/Loss (dB) L=1-2 10-20

Isolation (dB) 10-20 N/A

Noise Figure (dB) 1-2 4-6

Return Loss (dB) -10 -10

Band width (GHZ) 10 10

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

www.seipub.org/aime

30 GHz RF-MEMS Dicke Switch and Wideband LNA RFIC Designs The SG25H1 technology used here to fabricate the 30 GHz RF-MEMS Dicke switch network and wideband LNA designs is a 0.25 µm SiGe BiCMOS process which was selected based on its suitability for implementing some Kaband radiometer receivers as depicted in Fig. 1 (see e.g. [20]). High frequency active devices available in this process are based on SiGe:C (npn or pnp) HBTs which are size scalable with a minimum emitter area of 0.21 x 0 .84 μm2 (AE=n x 0.176 µm2 where n is the number of emitter fingers). The back-end includes a stack of five metal layers which can be used for realising high quality on-chip inductors and metal-insulator-metal (MIM) capacitors. Poly silicon based resistors are available in this process.

Figures 2(a-b) show a block schematic and photograph of a 30 GHz SiGe RF-MEMS switched LNA RFIC with chip dimensions of 1.6 x 1.56 mm2 (breakouts of the Dicke switch network and wideband LNA used in this design were also included as separate test structures). The Dicke switch was implemented as a Co-Planar Waveguide (CPW) based SPDT switch network with one RF input/output port. The third port is terminated to ground via an on-chip 50 Ω resistor. Two capacitive shunt MEMS switches are connected to a tee-junction at a quarter wavelength (λ/4) distance to set the input matching at the resonance frequency of the Dicke switch. In the transmission (ON) state, the MEMS switch at port 1 is in up-state while the other switch is in down-state and these results are in a low impedance in the transmission branch (from port 1 to 2) and a high impedance in the other branch, and vice versa for the isolation (OFF) state.

(a)

(b)

FIG. 2 (A) SCHEMATIC AND (B) CHIP PHOTOGRAPH OF A 30 GHZ SIGE RF-MEMS SWITCHED LNA RFIC DESIGN (THE MEMS SWITCHED LNA AND BREAKOUTS CIRCUITS OF THE WIDEBAND LNA/DICKE SWITCH ARE SHOWN TO THE LEFT AND RIGHT, RESPECTIVELY).

SiGe RF-MEMS based SPST Switch Network Figure 3(a) shows a photograph of a SiGe RF-MEMS based SPST (shunt) switch network with circuit dimensions of 540 x 410 μm2 (incl. RF and DC pads). The MEMS switch is in the ON state when the membrane is in the up-state position (unbiased) and this provides a low switch capacitance (25-30 fF) to ground and a small attenuation (loss) to the signal passing between the RF input and output ports. A higher switch capacitance to ground (200-250 fF) exists when the MEMs switch is in the OFF state (the membrane is in the down-state position) which results in a higher attenuation (isolation) between the RF ports. The on and off switching times are around 10 μs [21].The switching reliability of the IHP SiGe MEMS switches has been demonstrated in life cycle tests showing 5-10 billion switch cycles [22].

The SPST circuit was simulated with the electro-magnetic (EM) simulation tool Sonnet using the foundry-provided s-parameters of the SiGe MEMS switch. Small signal testing of the fabricated circuits was first performed at the foundry using a probe station and an HP 8510C network analyser which was calibrated using on-wafer calibration standards. Figure 3(b) shows simulated and measured s-parameters of a SiGe MEMS SPST switch (the effects of the RF pads were removed using open-short de-embedding). The SPST switch has 0.7 dB and 12 dB of measured transmission losses (ON and OFF state) at 30 GHz, respectively (0.4 dB and 15 dB simulated). Compared with the modelled switch data, a slight shift in frequency occurred in the measured transmission due to a variation in the MEMS switch capacitance [23].

3

www.seipub.org/aime

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

(a)

(b)

FIG. 3 A SIGE RF-MEMS BASED SPST SHUNT SWITCH NETWORK: (A) CHIP PHOTOGRAPH AND (B) MEASURED AND SIMULATED SPARAMETERS IN UP-STATE (ON) AND DOWN-STATE (OFF), RESPECTIVELY

A 30 GHz SiGe RF-MEMS Dicke Switch Network Figure 2(b) shows a break-out test structure of a 30 GHz SiGe MEMS Dicke switch (to the right) with a circuit area of 0.77 mm2 incl. RF and DC pads. The Dicke switch was simulated in ADS using the EM simulated s-parameters of the CPW transmission line (TL) structure together with the measured RF-MEMS SPST switch data and a foundry model of the on-chip 50 Ω resistance. A foundry EM model of the SiGe MEMS SPST switch that could be cosimulated with the co-planar structure was not available during the design of the Dicke switch network although such a model is expected to result in a closer agreement between the MEMS switch test results and simulations.

Figure 4 shows the measured and simulated s-parameters of the MEMS Dicke switch (the open-short deembedding was made at 2-50 GHz). Similarly, as for the MEMS SPST switch results shown in Fig. 3(b), a certain shift in frequency occurred between the measured and simulated |s21| data which may be explained by a variation in MEMS switch capacitance [23].

The Dicke switch has 3 dB and 22 dB of measured minimum/maximum transmission losses (ON and OFF) at 25 GHz, respectively (simulated 2.5 dB and 18 dB at 30 GHz, respectively, see Fig. 4)). The measured input matching (|s11|) in the ON state is below -10 dB at 17-50 GHz. The discrepancies between the experimental and simulated MEMS Dicke switch results could be explained by the large center ground area of the CPW TL structure (see Fig. 2(b)) which is connected to the two outer ground planes via the RF probes (air bridges that may be used to connect the different CPW ground planes together were not available in this process). A compact MEMS Dicke switch network with improved RF properties may be realised based on a microstrip circuit design where the ground plane is made in metal 1. Such low-loss microstrip SiGe MEMS SPDT switches are reported in [9] (with < 1 dB of losses and around 20 dB of isolation at 30 GHz).

FIG. 4 MEASURED AND SIMULATED S-PARAMETERS OF A 30 GHZ SIGE RF-MEMS BASED DICKE SWITCH NETWORK IN ON AND OFF STATE, RESPECTIVELY.

4

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

www.seipub.org/aime

A SiGe Wideband LNA Figure 5 shows the circuit schematic of a SiGe wideband LNA design which was combined with a 30 GHz RFMEMS Dicke switch network as shown in Fig. 2(b) (a break-out circuit is shown in the upper right corner). The cascade LNA design used here was optimised for high gain, low NF and wide bandwidth by selecting the appropriate DC bias points and the suitable number of emitter fingers (n = 8 was chosen in the three transistor stages as it resulted in the best possible trade-off between providing both circuit stability with a reasonably high gain, low NF and a wideband matching). The peak values of fT and minimum noise figure (NFmin) are specified in the process design manual to occur at a collector current (IC) of 16 mA and 5 mA when n is equal to eight.

On-chip capacitors and inductors (Lin=Lout=300 pH) are used for DC blocking and impedance matching. The onchip inductors were optimised and simulated in Sonnet. Resistive feedback was used in the third stage to ensure the LNA is unconditional stable (K-factor >1 and stability measure μ > 0). The fabricated SiGe wideband LNA occupies an area of 285 x 114 μm2 without pads (510 x 300 μm2 incl. RF and DC pads).

The layout and simulations of the SiGe LNA design were performed using the Cadence CAD tool. Simulations include post-layout parasitic extraction of the resistance and capacitance values due to the inter-stage metal connections. The parasitic extraction was done for the three stages of amplifier excluding the input/output matching networks as the on-chip inductors were made as full-custom designs. The matching networks’ EMsimualted s-parameter data were later combined with LNA design (3-stages) and simulated in Cadence.

The fabricated wideband LNA design was subjected to several small and large signal tests (incl. s-parameter characterisation, noise figure and intermodulation distortion measurements). Figure 6(a) shows the on-wafer measured and simulated s-parameters of the three-stage LNA when VC1=VC2=VC3=2.5 V and the total collector current (IC) is equal to 14 mA. The small signal gain |s21| is then 19.9 dB at 9 GHz (|s21|≥10 dB at 2-33 GHz). The measured input and output matching (|s11| and |s22|) are below -10 dB from 24 to 39 GHz and 20 to 42 GHz, respectively. The simulated and measured results are in a close agreement except for the measured peak |s21| which is a few dBs lower than expected and the slight shift in frequency occurring around 30 GHz for the measured |s11| and |s22| (the effects of RF the pads were removed by open-short de-embedding at 2-50 GHz). The discrepancies found between the measurements and simulations may be due to process variations and somewhat higher losses than predicted for the on-chip inductors and interconnects.

FIG. 5 SCHEMATIC OF A SIGE WIDEBAND LNA MADE IN A 0.25 µM SIGE BICMOS PROCESS TECHNOLOGY.

The small signal characterisation of the fabricated wideband LNA and MEMS switched circuits was first made onwafer at the foundry. RF ground-signal-ground (GSG) probes with a pitch of 100 μm were used in the measurements and a multi-contact wedge was used for applying the DC bias during the on-wafer testing. For the NF and linearity measurements a diced chip with the LNA design (shown in Fig. 2b) was mounted inside a DC test fixture (0.5-1.5 mm long bond-wires were used to connect to the DC pads). The LNA s-parameters, NF and linearity were measured using an Agilent PNA-X N5245A network analyser connected to a probe station, an N4691-60004 calibration module, an 346C noise source (defined up to 26.5 GHz), an N8487A power sensor and an N1913A power meter.

5

www.seipub.org/aime

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

(a)

(b)

FIG. 6 MEASURED AND SIMULATED S-PARAMETERS OF A SIGE WIDEBAND LNA IN A 0.25µM BICMOS PROCESS TECHNOLOGY: (A) VC1=VC2 =VC3=2.5 V AND IC =14 MA (B) VC1=VC2 =VC3=1.25 V AND IC=6 MA.

Despite the fact that the wideband LNA circuit was designed to be unconditionally stable, when the same DC bias values were applied into the LNA mounted in a test fixture (as were in the previous on-wafer LNA tests) oscillations occurred that may be due the off-chip interconnects between the power supplies and the DC pads. The measurements were instead performed with a reduced DC bias at which the LNA was found stable (VC1=VC2=VC3=1.25 and IC= 6mA). The LNA gain is equal to 14.2 dB at 7.5 GHz with this bias (|s21|≥10 dB at 3-26 GHz) and the return losses are better than -10 dB at 27-40 GHz (see Fig. 6(b)).The s-parameters were in this case measured using a standard impedance calibration substrate (the effects of the RF pads were not de-embedded).

Figures 7(a-b) show the measured and simulated NF and output-referred third order intercept point (OIP3) of the SiGe wideband LNA (VC1=VC2 =VC3=1.25V and IC= 6mA). The measured LNA NF reaches 5.9 dB at 26.5 GHz and is below 7 dB at 13-26.5 GHz. Furthermore, the actual measured NF of the wideband LNA design is estimated to be 0.5 dB less than those values after subtraction of the expected RF input pad losses (based on the characterisation of an on-chip thru standard that shows 1 dB of losses). The measured and simulated values of LNA NF are in a relatively close agreement at higher frequencies (the difference is within 0.5 dB at 24-26.5 GHz). The difference is that several dBs become higher at lower frequencies where the measured |s21| is typically also some dBs lower than simulated. These discrepancies between the experimental and theoretical results may be related to a limited accuracy of the passive and active (noise) models used in this amplifier design (e.g. due to process variations and higher losses than predicted for the passive elements). The measured LNA OIP3 values are in the range of -6 dBm to -3 dBm at 10-26.5 GHz when consuming 7.5 mW from a DC power supply.

(a)

(b)

FIG. 7 MEASURED AND SIMULATED RESULTS OF A SIGE WIDEBAND LNA IN A 0.25 µM BICMOS PROCESS TECHNOLOGY (VC1=VC2 =VC3=1.25 V AND IC =6 MA): (A) NOISE FIGURE (B) OUTPUT THIRD ORDER INTERCEPT POINTS.

A SiGe RF-MEMS Dicke Switched LNA Figures 8(a-b) show theoretical results (VC=2.5 V, IC =19 mA) and a chip photo of the realised SiGe MEMS switched LNA circuit, respectively. The Dicke switch network was simulated in ADS (based on the EM simulated transmission line structures and the SPST switch model) and the wideband LNA was simulated using Cadence and

6

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

www.seipub.org/aime

Sonnet. The MEMS LNA has a predicted gain of 10-18 dB and -5 dB to 10 dB at 25-35 GHz when the Dicke switch is switched on and off, respectively (corresponding to 8-19 dB of switched isolation). The simulated noise figure of the switched LNA is 6.2-7.9 dB at 25-35 GHz which is close to 2 dB higher than the noise figure of the wideband LNA. Figures 9(a-b) show measured and simulated s-parameters of the fabricated SiGe MEMS Dicke switch LNA circuit (VC=2.5 V, IC =19 mA). The measured |s21| is maximum 2 dB when the Dicke switch is in the ON state and this is 16 dB lower than the simulated peak gain values shown in Fig. 8(a). An unexpected design mistake that can explain this anomaly was identified as metal via contacts between the Dicke switch and LNA ground planes which were unintentionally missing in the manufactured layout of the MEMS switched LNA (see Fig. 8(b)). The Dicke switch and LNA ground planes were then not connected to on-chip and although ground connections were provided via the RF probes during the testing; a floating ground occurred as the Dicke switch ground plane to the right became separated and isolated from the other ground planes. In order to verify the measured small signal characteristics of the implemented MEMS switched LNA design, the switched LNA circuit was re-simulated in Sonnet (using the previously derived s-parameters for the passive and active parts) but in this case without the via contacts supposed to connect the Dicke switch and LNA ground planes together. Compared with the simulated MEMS switched LNA results (Fig. 9), there is a some shift in frequency for the measured |s21| which partly also could be explained by the variation in MEMS switch capacitance [23]. A closer agreement between the simulated and experimental switched LNA results is anticipated using a more precise (EM) MEMS switch model that could be co-simulated with the CPW lines in the Dicke switch network. Furthermore, the measured MEMS LNA has 9-13.5 dB of switched isolation at 24-31 GHz which validates the switching function of the SiGe MEMS Dicke switch used in the switched LNA. The presented simulated and experimental results show the potential of a novel monolithic integration of a 30 GHz RF-MEMS Dicke switch network and a wideband LNA realised in a low-cost SiGe BiCMOS process.

(a)

(b)

FIG. 8 A SIGE RF-MEMS DICKE SWITCHED LNA DESIGN MADE IN A 0.25 µM BICMOS PROCESS TECHNOLOGY: (A) SIMULATED SPARAMETERS/NOISE FIGURE AND (B) CHIP PHOTO (THE TWO ELLIPSES INDICATE REGIONS WHERE SOME METAL VIA CONTACTS ARE MISSING).

(a)

(b)

FIG. 9 MEASURED AND SIMULATED S-PARAMETERS OF A SIGE RF-MEMS DICKE SWITCHED LNA WHEN THE DICKE SWITCH AND LNA GROUND PLANES ARE DISCONNECTED (VC1=VC2 =VC3=2.5 V, IC =19 MA): (A) TRANSMISSION (ON AND OFF) (B) INPUT AND OUTPUT MATCHING (ON).

7

www.seipub.org/aime

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

Summary of Results SiGe technology is an attractive candidate for certain emerging applications (e.g. automotive radar, wireless communication, sensor networks and imaging systems). High sensitivity single-chip receiver front-ends (incl. broadband and reconfigurable low noise amplifiers) are key functions in such applications. Table 2 compares the results of the wideband high-gain LNA design presented in this paper with some previously published state-ofthe-art SiGe BiCMOS LNAs. In summary, this work achieves both a high LNA gain of 10-20 dB over a much wider bandwidth (2-33 GHz) and a broadband input and output matching better than -10 dB from 24 to 39 GHz and 20 to 42 GHz, respectively. The measured NF of the proposed LNA design are relatively close to the reported values of some wideband LNAs realised in higher process nodes (see Table 2). Table 3 shows a performance summary of published III-V and silicon based RF-MEMS switched LNA MMIC/RFIC designs within the 10-30 GHz range and beyond. A wideband GaAs RF-MEMS (Dicke) switched LNA design with 3 dB noise figure at 16-26 GHz was presented in [6]. The first and only silicon MEMS LNA ICs operating above 10 GHz that have been reported so far are some narrow-band/dualband (switchable) designs made at 60/77 GHz and 24/74 GHz, respectively [7], [8]. As a result of the unintentionally missing metal via contacts between the Dicke switch and LNA ground planes, the gain of the fabricated MEMS switched LNA design was much lower than expected in the ON state (-4 dB to 2 dB at 24-31 GHz). Despite this fact, the MEMS LNA showed 9-14 dB of measured isolation over a 25% wide RF bandwidth when the Dicke switch was switched on and off, respectively. These experimental MEMS LNA results are relatively close to the anticipated 8-19 dB of in-band isolation which thereby validate the switching function of the Dicke switch network used in this design. The proposed reconfigurable LNA RFIC design is believed to represent a first-time and single-chip implementation of a 30 GHz MEMS Dicke switch network and a wideband LNA made in a 0.25 µm SiGe BiCMOS process. The unexpected floating ground issue that occurred in this design may be easily solved in a future re-designed version of such a MEMS Dicke switched LNA. TABLE 2. PERFORMANCE COMPARISON OF SIGE BICMOS WIDEBAND LOW-NOISE AMPLIFIER CIRCUIT DESIGNS

Reference [24] [25] [26] [18] [19] This work Frequency (GHz) 26-40 3-26 8-18 21-40 10-30 2-33 Gain (dB) 15-23 9 13-16 10-19 10-22 10–20 Input matching ≤ -10 ≤ -10 ≤ -10 ≤ -10 ≤ -10 ≤ -8.5 (dB) (3-30 GHz) (21-23 GHz) (15-19 GHz) (24-39 GHz) 4-6 5.6 Noise Figure 4-5 5-6† 3 5-6 (dB) (2-22 GHz) (35 GHz) (15-28 GHz) (14-26 GHz) Power (mW) 11 33 38 26 28 8-35 OIP3 (dBm) 4 15* 2* 9 13* ≤-3 Technology 0.12 μm 0.13 μm 0.13 μm 0.18 μm 0.25 μm 0.25 μm *estimated based on the given IIP3 data †estimated values based on the measured noise figure data (excl. the effects of the RF input pad) TABLE 3. PERFORMANCE COMPARISON OF RF-MEMS SWITCHED LNA IC DESIGNS REALISED IN DIFFERENT TECHNOLOGIES

Reference Frequency (GHz) Gain (dB) Isolation (dB) Noise Figure (dB) Technologies

[4]

[5]

[6]

[7]

[8]

This work (Measured)

This work (Simulated)

26-30

18–26

16-34

60/77

24/79

24-31

25-36

20-22 N/A

≥ 15 16-20

19–21/21–23 N/A

25/18 N/A

- 4 to 2 9-14

N/A

2–5

7/8

4/8

N/A

GaAs pHEMT

GaAs pHEMT

10-17 20-25 3 (16-26 GHz) GaAs pHEMT

0.25 μm SiGe BiCMOS

0.25 μm SiGe BiCMOS

0.25 μm SiGe BiCMOS

9-18 8-19 6–8 (25-35 GHz) 0.25 μm SiGe BiCMOS

Conclusions We presented a novel monolithic integration of a 30 GHz RF-MEMS Dicke switch network and a wideband LNA made in a 0.25 μm SiGe BiCMOS process. The LNA achieves a high gain (10-20 dB) over a wider bandwidth (2-33 GHz) and a broader input/output matching (≤-10 dB at 24−39/20−42 GHz, respectively) in comparison with some earlier reported SiGe LNAs. Furthermore, the LNA has a measured NF and OIP3 of 5-6 dB and -6 dBm to -3 dBm at

8

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

www.seipub.org/aime

14-26/10-26 GHz, respectively (the DC power consumption is 8-35 mW). The Dicke switch has 3 dB/22 dB of measured in-band losses/isolation. As a result of the unintentionally missing via connections between the Dicke switch and LNA ground planes, the switched LNA gain was 10-17 dB lower than anticipated. The MEMS LNA resulted in 9-14 dB of isolation at 24-31 GHz thereby validating the switching function of the switched LNA design.

The proposed reconfigurable and broadband RFICs represent first-time and single-chip implementations of a 30 GHz MEMS Dicke switch network and a wideband LNA made in a 0.25 µm SiGe BiCMOS process. Such highly integrated low-power and low-cost RFICs may be used in reconfigurable wideband receivers for wireless communication, space and security RF sensors.

REFERENCES

[1]

“http://www.wispry.com.” .

[2]

“http://www.radantmems.com.” .

[3]

M. Kim, J. B. Hacker, R. E. Mihailovich, and J. F. DeNatale, “A monolithic MEMS switched dual-path power amplifier,” IEEE Microw. Wirel. Components Lett., vol. 11, no. 7, pp. 285–286, 2001.

[4]

J. B. Hacker, M. Kim, R. E. Mihailovich, and J. F. DeNatale, “Monolithic GaAs PHEMT MMICs integrated with RF MEMS switches,” Tech. Dig. - IEEE Compd. Semicond. Integr. Circuit Symp. CSIC, pp. 229–232, 2004.

[5]

R. Malmqvist, C. Samuelsson, A. Gustafsson, H. Maher, and T. Vähä-heikkilä, “A K-Band Single-Chip Reconfigurable/Multi-Functional RF-MEMS Switched Dual-LNA MMIC,” in Microwave Symposium Digest (MTT), IEEE MTT-S International, 2012, pp. 10–12.

[6]

A. Gustafsson, C. Samuelsson, R. Malmqvist, S. Seok, M. Fryziel, N. Rolland, B. Grandchamp, and R. Baggen, “A 0-Level Packaged RF-MEMS Switched Wideband GaAs LNA MMIC,” in European Microwave Conference (EuMC), 2013, pp. 1403– 1406.

[7]

A. C. Ulusoy, M. Kaynak, T. Purtova, B. Tillack, and H. Schumacher, “A 60 to 77 GHz Switchable LNA in an RF-MEMS Embedded BiCMOS Technology,” IEEE Microw. Wirel. Compon. Lett., vol. 22, no. 8, pp. 430–432, 2012.

[8]

A. C. Ulusoy, M. Kaynak, T. Purtova, B. Tillack, and H. Schumacher, “24 to 79 GHz frequency band reconfigurable LNA,” Electron. Lett., vol. 48, no. 25, pp. 1598–1600, Dec. 2012.

[9]

M. Kaynak, M. Wietstruck, W. Zhang, J. Drews, R. Barth, D. Knoll, F. Korndörfer, R. Scholz, K. Schulz, C. Wipf, B. Tillack, K. Kaletta, M. Suchodoletz, K. Zoschke, M. Wilke, O. Ehrmann, A. C. Ulusoy, T. Purtova, G. Liu, and H. Schumacher, “Packaged BiCMOS Embedded RF-MEMS Switches with Integrated Inductive Loads,” in Microwave Symposium Digest (IMS), IEEE MTT-S International, 2012, pp. 4–6.

[10] Z. Chen, C. Wang, H. Yao, and P. Heydari, “A BiCMOS W-Band 2×2 Focal-Plane Array With On-Chip Antenna,” IEEE J. Solid-State Circuits, vol. 47, no. 10, pp. 1–17, 2012.

[11] L. Gilreath, V. Jain, and P. Heydari, “Design and Analysis of a W-Band SiGe Direct-Detection-Based Passive Imaging Receiver,” IEEE J. Solid-State Circuits, vol. 46, no. 10, pp. 2240–2252, 2011.

[12] J. W. May and G. M. Rebeiz, “Design and Characterization of W-Band SiGe RFICs for Passive Millimeter-Wave Imaging,” IEEE Trans. Microw. Theory Tech, vol. 58, no. 5, pp. 1420–1430, 2010.

[13] E. Shumakher, J. Elkind, and D. Elad, “Key components of a 130 GHz Dicke-radiometer SiGe RFIC,” in IEEE 13th Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems, 2013, pp. 255–257.

[14] T. Kosugi, H. Sugiyama, H. Matsuzaki, K. Murata, M. Nakamura, H. Satoh, and K. Throngnumchai, “A 140-GHz quadreceivers IC and sub-assembly for compact passive imaging sensors,” in IEEE MTT-S International Microwave Symposium Digest, 2012, pp. 1–3.

[15] R. Wilke, S. Hamid, K. Schraml, R. Khunti, and D. Herberling, “Multi-layer patch antenna array design for Ka-band satellite communication,” in International Microwave & Optoelectronics Conference (IMOC), 2013, pp. 1–4.

[16] C. Miquel, J. C. Cayrou, and J. L. Cazaux, “Flexible Ka-Band Low Noise Amplifier Sub-System for Oncoming Satellite Payloads,” in European Microwave Conference, 2006, no. September, pp. 890–893.

9

www.seipub.org/aime

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

[17] E. Schreiber, S. Anger, and M. Peichl, “Design of an integrated Ka band receiver module for passive microwave imaging systems,” in Semiconductor Conference Dresden (SCD), 2011, pp. 1–4.

[18] P. J. Riemer, B. R. Buhrow, J. D. Coker, B. A. Randall, R. W. Techentin, B. K. Gilbert, and E. S. Daniel, “Ka-Band (35 GHz) 3Stage SiGe HBT Low Noise Amplifier,” in IEEE Microwave Symposium Digest (MTT), 2005, no. C, pp. 1037–1040.

[19] T. Masuda, T. Nakamura, M. Tanabe, N. Shiramizu, S. Wada, T. Hashimoto, and K. Washio, “SiGe HBT based 24-GHz LNA and VCO for Short-Range Ultra-Wideband Radar Systems,” in IEEE Asian Solid-State Circuits Conference, 2005, pp.

425–428.

[20] L. Aluigi, F. Alimenti, and L. Roselli, “Design of a Ka-Band LNA for SoC space-based millimeter-wave radiometers,” in IEEE MTT-S International Microwave Workshop Series on Millimeter Wave Integration Technologies, 2011, pp. 156–159.

[21] M. Kaynak, M. Wietstruck, R. Scholz, J. Drews, R. Barth, K. E. Ehwald, A. Fox, U. Haak, D. Knoll, F. Korndörfer, S. Marschmeyer, K. Schulz, C. Wipf, D. Wolansky, B. Tillack, K. Zoschke, T. Fischer, Y. S. Kim, J. S. Kim, W. Lee, and J. W. Kim, “BiCMOS Embedded RF-MEMS Switch for above 90 GHz Applications using Backside Integration Technique,” in IEEE Electron Devices Meeting (IEDM), 2010, pp. 832–835.

[22] M. Kaynak, F. Korndörfer, M. Wietstruck, D. Knoll, R. Scholz, C. Wipf, C. Krause, and B. Tillack, “Robustness and Reliability of BiCMOS Embedded RF-MEMS Switch,” in IEEE 11th Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems (SiRF), 2011, vol. 1, pp. 177–180.

[23] M. Kaynak, K. E. Ehwald, J. Drews, R. Scholz, F. Korndörfer, D. Knoll, B. Tillack, R. Barth, K. Schulz, Y. M. Sun, D. Wolansky, S. Leidich, S. Kurth, and Y. Gurbuz, “BEOL Embedded RF-MEMS Switch for mm-Wave Applications,” in IEEE Electron Devices Meeting (IEDM), 2009, pp. 797–800.

[24] B. Min and G. M. Rebeiz, “Ka-Band SiGe HBT Low Noise Amplifier Design for Simultaneous Noise and Input Power Matching,” IEEE Microw. Wirel. Components Lett., vol. 17, no. 12, pp. 891–893, 2007.

[25] P. K. Saha, S. Shankar, R. Schmid, R. Mills, J. D. Cressler, and G. Tech, “Analysis and Design of a 3-26 GHz Low-Noise Amplifier in SiGe HBT Technology,” in IEEE Radio and wireless Symposium (RWS), 2012, pp. 203–206.

[26] F. F. Dai, R. C. Jaeger, and J. D. Irwin, “An 8–18 GHz wideband SiGe BiCMOS low noise amplifier,” in IEEE Microwave Symposium Digest (IMS), 2009, pp. 929–932.

Shakila B. Reyaz received the M.Sc. degree in Telecommunications engineering and the Ph.D degree in Engineering sciences with specialization in Microwave from NED University of Engineering and Technology, Karachi, Pakistan and Uppsala University, Uppsala, Sweden in 2007 and 2015, respectively.

From 2007-2010, she has worked as a lecturer in the department of Electronic engineering at NED University of Engineering and Technology, Karachi, Pakistan. From 2010 to February 2015, she has worked as a Ph.D. student in Microwave group at Uppsala University, Uppsala Sweden. Her current research interests include Reconfigurable RF MEMS, MMICs and RFICs for RF and millimeter-wave applications.

Carl Samuelsson received his bachelor degree in electrical engineering from Linköpings institute of technology in 2001 and used to work as a research engineer at the Swedish Defence Research Institute. His main research intrerest includes microwave circuit design and measurements. He is currently employed at SAAB.

Andreas Gustafsson received his MSc in engineering physics at Umeå University in 1996. Since 1997 he has been employed at FOI working with RF front-end technology. In 2004 he received (as first author) the Radar Prize at the European Microwave Week for the design and evaluation of a experimental compact 16channel smart-skin digital beam forming X-band antenna module. He is the author or co-author of more than 20 scientific publications.

10

Advances in Microelectronic Engineering (AIME) Volume 3, 2015

www.seipub.org/aime

Robert Malmqvist received the M.S. degree in Applied Physics and Electrical Engineering and the Ph.D degree in Electronic Devices from Linköping University, Sweden in 1994 and 2001, respectively. He joined FOI (Swedish Defence Research agency) in 2000 where he currently holds a position as Head of the Radar Technology group. He has been involved as project manager at FOI within several nationally funded and European funded research projects. From 2007 to 2015 he had a part-time position as scientist with the Microwave group at Uppsala University, Sweden and he has been the supervisor of one Ph.D. student with a research topic on reconfigurable RF-MEMS, MMICs and RFICs. He has authored or co-authored around 70 scientific papers of which12 in peer-reviewed journals and he has received or co-received three Best Papers Awards in 2004, 2009 and 2012, respectively.

Rolf Jonsson, received the M.Sc. and Ph. Lic. Degrees in Solid state physics from Linköping University, Sweden in 1997 and 2005 respectively. Since 1997 he has been employed as research engineer and scientist at the Swedish Defence Research Agency (FOI), Dept. of microwave technology. He has been engaged in research on wide band-gap microwave amplifiers and high power microwave (HPM) protection. His current research concerns array antenna electronic warfare systems and components for mm‑wave imaging systems.

Mehmet Kaynak received his B.S degree from Electronics and Communication Engineering Department of Istanbul Technical University (ITU) in 2004, took the M.S degree from Microelectronic program of Sabanci University, Istanbul, Turkey in 2006 and received the PhD degree from Technical University of Berlin, Berlin Germany in 2014. He joined the technology group of IHP Microelectronics, Frankfurt (Oder), Germany in 2008. Dr. Kaynak has received the young scientist award of Leibniz institute for the year of 2014. Since 2015, he is acting as the department head of technology group at IHP and network faculty member at Sabanci University.

Anders Rydberg (M’89) received the M.Sc.-degree from Lund University of Technology in 1976. He worked between 1977–83 at the National Defence Research Establishment, ELLEMTEL Development Co., and the Onsala Space Observatory. In 1988 and 1991 he received the Ph.D. degree and was appointed Docent (Associated Professor) respectively, at Chalmers University of Technology. Between 1990 to 1991 he worked as a senior research engineer at Farran Technology Ltd., Ireland. In 1992 he was employed as Senior Lecturer and in 2001 as Professor in Applied Microwave and Millimeterwave Technology at Uppsala University. He is heading the Microwave Group at the Department of Engineering Science, Uppsala University. He is since 2007 also joint-owner of Integrated Antennas AB and WISENET Holding AB, Uppsala, Sweden. Prof. Rydberg has authored or co-authored more than 220 publications in the area of micro- and millimeterwave antennas, sensors, solid state components and circuits and has three patents in the areas. He has graduated 7 Ph.D. students. Prof. Rydberg is a member of the editorial board for the IEEE-MTT, Chairman for Section D of the Swedish Member Committee of URSI (SNRV), and member of the board for the Swedish IEEE MTT/AP Chapter.

11