Indian Journal of Engineering & Materials Sciences Vol. 18, June 2011, pp. 175-182 Design of harmonic noise frequency filtering down-converter with asymmetrical inductance tank in InGaP/GaAs HBT technology Cong Wang & Nam Young Kim* RFIC Laboratory, Kwangwoon University, 447-1 Wolgye-dong, Nowon-ku, Seoul 139-701, Korea Received 31 May 2010; accepted 13 May 2011 A harmonic noise frequency filtering (HNFF) down-converter is fabricated using asymmetric inductance tank (AIT) in InGaP/GaAs heterojunction bipolar transistor (HBT) monolithic microwave integrated circuit (MMIC) technology. In the voltage-controlled oscillator (VCO) design, internal inductance is found to lower the phase noise which is based on an analytic understanding of phase noise. This VCO directly drives the on-chip double balanced mixer (DBM) to convert radio frequency (RF) carrier to intermediate frequency (IF) through local oscillator (LO). Furthermore, final power performance is improved by output amplifier. This paper presents the design for a 1.721 GHz enhanced LC VCO, high-power DBM, and output amplifier that have been designed to optimize phase noise and output power. The proposed VCO exhibited a phase noise of -133.96 dBc/Hz at 1 MHz offset and a tuning range from 1.46 GHz to 1.721 GHz. In measurement, on-chip down-converter shows a third-order input intercept point (IIP3) of 12.55 dBm, a third-order output intercept point (OIP3) of 21.45 dBm, an RF return loss of -31 dB, and an IF return loss of -26 dB. The RF-IF isolation is -57 dB. Also, the conversion gain is 8.9 dB through an output buffer. The total on-chip down-converter is implanted in 2.56 × 1.07 mm2 of chip area. Keywords: Harmonic noise frequency filtering (HNFF), Asymmetric inductance tank (AIT), Double balanced mixer (DBM), Phase noise, Voltage-controlled oscillator (VCO) In the tremendous growth of wireless handheld devices, high performances become a major consideration in MMIC designs. This paper explores high-performance on-chip down-converter MMIC design for HBT VCO and DBM co-design through their applications in an RF front-end transceiver. Integrated VCOs are important blocks in communication systems, which are used in virtually all types of up-conversion, down-conversion, frequency modulation, and high-frequency generation systems. While most VCOs are designed to generate a sinusoidal voltage waveform, their use and required performance vary widely1,2. An integrated circuit reduces the size of the devices such as on-chip inductors and capacitors; thereby it can improve the spectral purity of VCOs. Also, the varactor diode which is used for frequency tuning of VCOs can also be realized on the same chip. A further advantage is that 1/f noise is reduced by using the ledge function of the HBT process3. Many VCOs constructed with the InGaP/GaAs HBT technology have been developed to achieve the characteristics of low power consumption, low phase noise, and high operation frequency4,5. —————— *Corresponding author (E-mail: nykim@kw.ac.kr) In this paper, the uses of AIT structure, highquality tank, and HNFF technique are described to design the proposed VCO. In order to improve the phase noise performance of the integrated LC VCO, the core structure of the VCO is based on a crosscoupled differential configuration. This topology has been widely adapted in low gigahertz regimes due to its simplicity and differential operation5,6. One advantage of this architecture is its high loop gain which makes it as a popular solution for differential VCO designs in RFIC/MMIC. The frequency conversion process is necessary for the quality of the signal as well as the wider communication area using a small antenna. Therefore, the mixer is a critical block for this process to perform the frequency translation. The most important elements to consider in the mixer design are high gain, low noise, and low power consumption. The main topology of the proposed DBM is double balanced Gilbert cell topology. This topology is a fundamental building block which is used in a wide array of IC applications including modulators, de-modulators, and RF mixers. This RF mixer can produce positive conversion gain and multi-decade bandwidth operation. It requires low LO power with compact size. Furthermore, its baseband and multi-decade bandwidth performance make INDIAN J. ENG. MATER. SCI., JUNE 2011 176 it attractive for wideband instrumentation and RF communications7. Recently, active Gilbert cell mixers with high conversion gain are reported8,9. Enhanced LC VCO and DBM Co-Design Asymmetrical inductance tank structure The proposed VCO consists of a LC tank, a pair of emitter follower buffers, and a negative resistance block. A cross-coupled differential structure with capacitive coupling feedback is used to reduce the 1/f noise. To realize a high quality factor, the core of the VCO with AIT and symmetrical inductance tank (SIT) structure will be concerned which is shown in Fig. 1a and 1b, respectively. To obtain low phase noise performance, the high Q resonator is optimized; this is one of the most important aspects of oscillator design10. As we know, the quality factor can be expressed as the phase of the open-loop transfer function Φ (w) at the resonance of the LC tank in an oscillator QTank = R ω0 dφ = P 2 dω ω = ω ω 0 L 0 where RP is the parallel resistance of the LC tank at the resonant frequency (ωo), and L is the inductance of the tank11. If the asymmetrical inductor is used in tank, the RP will be two times better than the symmetrical tank as shown in Fig. 2. In the view of the layout, the area of inductor in AIT is about 1.5 times bigger than SIT. The inductance of AIT should be 1.5 times bigger than SIT, because the inductance is proportional to the area of the inductors. ωo depends on the center carrier frequency, therefore it will be almost the same in both the AIT and the SIT. In order to acquire a high Q value, inductors with low parasitic series resistance are chosen because these inductors have high Q values and provide low phase noise in the oscillator. To this end, the Q-factor of the AIT is definitely superior to the SIT, which is shown in Fig. 3. The proposed VCO can be modeled simply as a lossy LC tank combined with a noiseless energy restorer that keeps constant oscillation amplitude. The phase noise for this model can be analytically calculated as: (a) (a) Asymmetrical inductance tank structure (b) Symmetrical inductance tank structure Fig. 1—Brief circuits of the LC VCOs: (a) AIT; (b) SIT … (1) (b) Fig. 2—Equivalent circuits of the LC tank: (a) AIT-VCO; (b) SIT-VCO Fig. 3—Q values of the asymmetrical and symmetrical inductor tank WANG & KIM: DESIGN OF HARMONIC NOISE FREQUENCY FILTERING DOWN-CONVERTER 2kT L(∆ω ) = 10 log Psig ωo ∆ 2 Q ω tan k 2 … (2) where k is Boltzmann’s constant, T is the temperature, ωo is the oscillation frequency, Psig is the signal power, Qtank is the Q factor of the LC tank, and ∆ω is the offset form the oscillation frequency. It tells us that phase noise improves as both the carrier power and Q factor increase. Increasing the signal power improves the ratio simply because the thermal noise is fixed, while increasing Q improves the quadratic ratio because the tank’s impedance falls off as 1/Qtank∆ω12. A comparison of the phase noise simulation performance as a function of offset frequency between the AIT-VCO and the SIT-VCO and the fabricated VCOs are shown in Fig. 4, respectively. The most important point to note here is that the phase noise of the AIT-VCO is greater than 3.4 dB at 1 MHz offset than SIT-VCO. As a result, the AIT structure can be optimized with high RP and Q factor when compared with the SIT. But using the AIT, the wave shape of the core stage has a little distortion in the headroom, which came from the asymmetric capacitor in the resonator part. It can be eliminated with symmetric capacitance in the AIT. Harmonic noise frequency filtering technique In the differential LC oscillator, the current source is required to fix the bias current that supplies high impedance with the switching active devices of the differential pairs in series combination. In the case of perfectly balanced circuit, odd harmonic circulate Fig. 4—Phase noise simulation results of the AIT-VCO and SITVCO (the fabricated AIT-VCO and SIT-VCO are shown on the upper right) 177 with no current flowing through the current source in out-of-phase-operation. But even harmonics flow in a common-mode path through the active devices, LC resonator circuit and current source are in the in-phase operation. The low noise frequency of the current source is initially up-converted to high frequency noise around the fundamental frequency due to the mixing effect of non-linear transconductance and intrinsic capacitance of the active oscillators. Because of very small level of the third and higher order harmonics in the VCO, the effect of the second harmonic can be considered. That means it is necessary to create condition of current source bypassing for the second harmonic. It is necessary to use the special filtering techniques for the second harmonic suppression to get pure spectral purity and it also can overcome the problem of noise in VCOs13. The VCO includes a feedback oscillation signal path and a frequency control path. The phase noise of the oscillator is primarily generated by these two paths, into which the noise is injected. The proposed HNFF shorts the noise signals, including the supply voltage and feedback noise, to ground. The capacitance is chosen to down the noise frequency in parallel with whatever capacitance is present at the feedback source of the differential pair as shown in Fig. 5. There are numerous literatures regarding noise reduction techniques in VCOs13-15. However, these techniques required the use of two or more inductors to reduce noise in the VCO which will have a very huge size. If VCOs are used by the HNFF, only filtering capacitor (C1) is needed because the proposed HNFF technique includes the internal inductance of the tank Fig. 5—Schematic circuit of the differential LC VCO using HNFF technique 178 INDIAN J. ENG. MATER. SCI., JUNE 2011 and negative resistance cell. Small value C1 (2.722 pF) and inductance (almost under 0.5 nH) act like low-pass filter until 5th order harmonics. Therefore, the harmonic noises can be removed. In Fig. 5, arrow (1) shows the flow of the 2nd harmonics in the loop. Before the 2nd harmonics flow out of the output buffer, the noises of the harmonics short to the ground through arrow (2). Figure 6 shows the phase noise variation as a function of control voltage with and without HNFF technique and the fabricated two VCOs, respectively. The phase noise at 1 MHz offset increased from -124.52 dBc/Hz for the VCO without harmonic noise frequency filtering capacitor to -129.7 dBc/Hz for HNFF-VCO. Therefore HNFF technique can be adapted to reduce phase noise due to the bypassing for the 2nd harmonic. The cross-coupled transistors (Q1 and Q2) form a positive feedback loop in the VCO, and two buffer circuits (Q3 and Q4) can be isolated to the VCO core by means of the load impedance. In addition to contributing noise, Q1 and Q2 saturate heavily if the peak swing at tank node exceeds. In order to solve this issue, a capacitive divider (C2) can be inserted in the feedback path allowing greater swings and higher common-mode level at tank nodes than at the bases of Q1 and Q2. Figure 7 shows the schematic circuit of the designed high-power DBM. The circuit is designed in such a way that the transistor in LO input part acts as a switch (Q1, Q2, Q3, and Q4) and can be operated in the cutoff area by applying minimum bias. The bias points of Q5 and Q6 are in the saturation region and those of Q1-Q4 are within a linear region. Besides, C1 is the signal path for the balanced input signal. By itself, the RF input differential pair is not linear enough for useful applications, so it must be linearized. A common technique to achieve it is known as resistive degeneration using resistors R1 and R216,17. Furthermore, linearity is improved by inserting a broadband resistor (R5) at the RF port, but it increases the noise figure. Both the LO differential input ports have on-chip broadband 50 ohm resistive terminations Vdc=5 V I=72.6 mA IF + IF - Vb2 LO + R6 R7 R8 R3 Q1 Q2 Q3 Q4 LO R4 Vb1 Design of the high-power DBM The DBM utilizes a Gilbert cell core in this paper because of its low DC power consumption, high linearity, and good spurious output rejection. Also, it can provide a high gain and very low noise figure. Strict symmetry is maintained for the balanced structure to minimize the amplitude and phase errors. Fig. 6— Phase noise simulation results of the HNFF-VCO and the VCO without harmonic noise frequency filtering capacitors (the fabricated VCOs are shown on the upper right) R9 RF R10 Q5 R5 R1 Q6 R2 C1 Ibias Fig.—7 Schematic circuit of the DBM core Fig.—8 Schematic circuit of the integrated down-converter WANG & KIM: DESIGN OF HARMONIC NOISE FREQUENCY FILTERING DOWN-CONVERTER (R3 and R4), resulting in excellent broadband return losses and superior inter-modulation performance. Co-design of the enhanced LC VCO and DBM Figure 8 shows a complete schematic circuit of the integrated down-converter consisting of the presented VCO, DBM, and output buffer. The output buffer is a cascode differential pair with inductive emitter degeneration represented by L1 and L2. This configuration is chosen to reduce the Miller capacitance and increase the power gain. Off-chip pull-up inductors (L3 and L4) enable a high output voltage swing. AC coupling capacitors (C1) help improving the power matching to 50 ohm. Results and Discussion A microphotograph of the fabricated downconverter with a die size of 2.56 × 1.07 mm2 is depicted in Fig. 9. It is fabricated using a commercial InGaP/GaAs HBT process in which the HBT devices had a cutoff frequency (fT) of 48 GHz and a maximum oscillation frequency (fMAX) of 80 GHz. The fabricated HNFF down-converter is measured using RF on-wafer probes. The chip is mounted on a printed circuit board (PCB) with silver epoxy. In this chip, the DC contact is created using golden wire (length: 10 mm; diameter: 1 mm) to provide a stable DC supply. Figure 10 shows a PCB for chip 179 measurement and probe tips contacting on the chip. Key parameters are measured using various equipments, such as probe station, spectrum analyzer, signal generator, power supply, and vector network analyzer. HNFF VCO with AIT Figure 11 shows the phase noise and output spectrum characteristics of the proposed VCO. It produces an output power of -10.251 dBm (including the cable loss) and has superior phase noise of -117.3 dBc/Hz and -133.93 dBc/Hz at 100 kHz and 1 MHz offset frequencies respectively from the carrier frequency of 1.721 GHz. The oscillation frequency and output power variation as a function of control voltage sweep are depicted in Fig. 12. The oscillation Fig. 11—Phase noise of the proposed VCO measured at the lowest tuning voltage (output spectrum of the VCO is shown on the lower left) Fig. 9—Microphotograph of the fabricated down-converter Fig.—10 PCB and probe tips contacting on the chip Fig. 12—Variation of oscillation frequency and output power as a function of tuning voltage INDIAN J. ENG. MATER. SCI., JUNE 2011 180 frequency is varied from 1.46 GHz to 1.721 GHz with a control voltage from 0 to 3 V. Using the HNFF capacitors produces low phase noise performance since the noises in both the feedback oscillation signal path and the frequency control path are reduced. This is also due to an increase in the parallel resistance and the Q factor of the tank. Therefore, the proposed VCO shows low phase noise performance. But the output power is not good enough, which mainly depends on the active device, bias condition, and matching network. There are several possible explanations for this result. Some of this loss can be attributed to test setup losses (around ~ 1.5 dB) which are not calibrated out of these measurements. Second may be a problem with the buffer amplifier biasing because the output power is decreased from the simulation value. It is believed that the buffer amplifier design is responsible for the majority of the loss since the buffer amplifier did not bias to the expected current in the proposed VCO. Finally, the reason of the low output power may be the mismatching of the matching network at the output. A smaller output device could have been used to match 50 ohm since the 1/gm is equaled to 50 ohm. But the design is current limited in driving the output. More current is thus run in the source follower stage to drive the output and it is deliberately chosen to not match the output to minimize the power loss. Therefore, before this buffer amplifier and matching circuit design are used in future iterations, the reason behind the faulty biasing and matching must be determined. A widely used figure-of-merit (FoM) is defined18 as DBM with the differential cascode output buffer Integrated matching resistors provide broadband return loss without matching networks. Figure 13 shows the return loss of the RF port is under -20 dB in the frequency range from 400 MHz to 12 GHz, while, the return loss of the LO port is under -10 dB in the frequency range of 400 MHz to 8 GHz. Also, the flat conversion gain is obtained as +2.6 dB from 400 MHz to 10 GHz. Fig. 13—Conversion gain of the DBM as a function of RF frequency and RF, LO return loss f P FOM = L(∆f m ) − 20 log 0 + 10 log diss … (3) 1mW fm In this work, we achieve FoM of -181.3 dBc/Hz. It shows a much better performance than the other VCOs in different frequencies and processes. Table 1 presents the VCO performances from recent literatures with this work. Fig. 14—The measured IMD3 of the fabricated DBM Table 1– Summary of the previously published VCO performances Reference Frequency Tuning Range Phase Noise FoM Process [19] [20] [21] [21] [22] [23] [24] Present work (GHz) 0.915 1.22 12.2 13.5 5.6 2.18 0.78 1.721 (MHz) 160 341 600 800 500 511 107 261 (dBc/Hz) -119 @ 1MHz -120 @ 1MHz -113.8 @ 1MHz -111.8 @ 1MHz -117 @ 1MHz -123 @ 1MHz -120 @ 1MHz -133.96 @ 1MHz -174.1 -173.5 -180.7 -176.6 -180.7 -180 -180 -181.3 BiCMOS 0.18 µm CMOS InGaP/GaAs HBT InGaP/GaAs HBT BiCMOS 0.18 µm CMOS 0.18 µm CMOS InGaP/GaAs HBT WANG & KIM: DESIGN OF HARMONIC NOISE FREQUENCY FILTERING DOWN-CONVERTER 181 In order to measure the inter-modulation effect, especially the third order inter-modulation, two signals with a 1 MHz offset are applied to the input port of the mixer. Figure 14 shows when the applied input power is -10 dBm, then the third order intermodulation distortion (IMD3) is 53.7 dBc. The inputreferred third order intercept point (IIP3) and the output-referred third order intercept point (OIP3) are calculated by using these results, IIP3 is found to be +16.78 dBm and OIP3 is +19.38 dBm. The integrated differential output buffer generated a high outputreferred 1 dB compression point (P1dB, out) of 10 dBm, which is shown in Fig.15. Fig. 15—The measured P1dB compression point Down-converter The current consumption of the fabricated downconverter is 94 mA at DC supply voltage of 5 V. Figure 16 shows the return loss of the RF port is under -20 dB in the frequency range of 100 MHz to 4.1 GHz and -31 dB at the center frequency. Return Fig. 16—Simulation and measurement results of RF return loss (the results of IF return loss is shown on the lower right) Fig. 18—Measured P1dB compression point Fig. 17—The measured IMD3 of the proposed down-converter (RF power: -10 dBm; LO power: -10 dBm) Fig. 19—Fundamental and 3rd order harmonics versus RF power of the two-tone test 182 INDIAN J. ENG. MATER. SCI., JUNE 2011 loss of IF port is about -26 dB. The leakage of RF-toIF ports measured at -30 dBm (RF power) is -57 dB. Figure 17 shows when the applied input power is -10 dBm, the IMD3 is 48.86 dBc. The measurement results of the one-tone test are shown in Fig. 18. The conversion gain is about 8.9 dB to compensate for cable loss at an RF power level of -30 dBm and an LO power level of -10 dBm. IIP3 is found to be 12.55 dBm and OIP3 is 21.45 dBm. Figure 19 shows the IIP3 and the OIP3 point, respectively. References Conclusions The LC-tuned VCO and down-conversion DBM are two major building blocks of a high-frequency radio front-end receiver. In this paper, both circuits have been examined and designed to combine each module for down-converter system on chip (SoC) applications. However, total current consumption is greater than in the CMOS design which is an inherent characteristic of the technology. The VCO has superior phase noise performance due to the HNFF technique and AIT structure. The minimum phase noise improvement of 8.9 dB at 1 MHz offset is reported which can be adapted to other topologies for reducing the phase noise. The designed DBM has a cascode output buffer using pull-up inductors. Integrated matching resistors provided excellent broadband return loss. Resistive and inductive emitter degeneration improves the linearity of the mixer and the cascode output buffer. In conclusion, the successfully superior operation of the L-band downconverter shows the appropriateness of the proposed design technique. 7 1 2 3 4 5 6 8 9 10 11 12 13 14 15 16 17 18 19 20 Acknowledgment This research was supported by the MKE (The Ministry of Knowledge Economy), Korea, under the ITRC (Information Technology Research Center) support program supervised by the NIPA (National IT Industry Promotion Agency) (NIPA-2011-C10901111-0002). This work was also supported by the Research Grant of Kwangwoon University in 2011. 21 22 23 24 Zhang L, Wang Z H & Chen H Y, Chin J Electron, 16 (2007) 66-68. Ahn H T & Burm J, IEICE Trans Electron, E93-C (2010) 420-422. Shrestha B, Kennedy G P & Kim N Y, Chin J Electron, 16 (2007) 236-238. Kurachi S & Yoshimasu T, IEICE Trans Electron, E88-C (2005) 678-682. Eo Y S, Kim K C & Oh B D, IEEE Microwave Wireless Components Lett, 13 (2003) 259-261. Zirath H, Kozhuharov R & Ferndahl M, IEEE J Solid-State Circuits, 40 (2005) 2077-2086. Kobayashi K W, IEEE GaAs Integrated Circuit Symp, (2003) 251-254.. Ahn D, Kim D W & Hong D C, IEEE Microwave Wireless Components Lett, 19 (2009) 227-229. Seo J B, Kim J H, Sun H & Yun T Y, IEEE Microwave Wireless Components Lett, 18 (2008) 803-805. Cha C Y & Lee S G, IEEE Radio Frequency Integrated Circuits Symp Dig, (2004) 583-586. Hamel J S, LC Tank Voltage Controlled Oscillator Tutorial (Waterloo, Ontario, Canada), 2005. Lee T H & Hajimiri A, IEEE J Solid-State Circuits, 35 (2000) 326-336. Hegazi E, Sjoland H & Abidi A A, IEEE J Solid-State Circuits, 36 (2001) 1921-1930. Yang F L, Zhang Z F, Li B Q & Min H, Int Conf Solid-State and Integrated Circuits Technology Proc, (2004) 1315-1320. Andreani P & Sjoland H, IEEE J Solid-State Circuits, 37 (2002) 342-348. Li L & Tenhunen H, 47th Midwest Sympium on Circuits and Systems, (2004) 193-196. Babcock G, Proc 2001 Bipolar/BiCMOS Circuits and Technology Meeting, (2001), 90-93. Westra J R, Verhoeven C J M & van Roermund A H M, Classification, anaylysis and synthesis, (Kluwer Academic Publishers), 1999. Teichmann P, Peter J & Laute A, IEEE 7th European Conference on Wireless Technology, (2004) 201-204. Kim H R, Cha C Y, Oh S M, Yang M S, & Lee S G, J SolidState Circuits, 39 (2004) 952-955. Baek D H, Ko S S, Kim J G, Kim D W & Hong S C, IEEE Trans Microwave Theory Tech, 52 (2004) 1353-1359. Astis G D, Cordeau D, Paillot J & Dascalescu L, IEEE J Solid-State Circuits, 40 (2005) 2087-2091. Leung L L K & Luong H C, IEEE Trans Microwave Theory Tech, 56 (2008), 39-48. Park K G, Jeong C Y, Park J W, Lee J W, Jo J G & Yoo C, IEEE Microwave Wireless Components Lett, 18 (2008) 413-415.