Application Note 5457

Transcription

1 MGA-633P8 GaAs MMIC LNA Enables 900 MHz BTS Amplifier with Industry Best Noise Figure and Linearity Application Note 5457 Introduction The weakest signal that a wireless receiver can recover is defined by its sensitivity and is calculated by Equation 1, [1] : Rx_sen (dbm) = log BW + SNR + F Eq. 1 BW is the bandwidth in Hz, SNR the required signal to noise ratio, and F the system noise figure. A low noise amplifier (LNA), as its name implies, improves the receiver sensitivity by reducing the cascade noise figure. The Friis formula, Equation, is used to calculate the total noise figure of a cascade of stages, each with its own noise factor and gain. As Equation 1 shows, the noise figure, F 1, of the 1st amplifying stage in the receiver chain the LNA dominates the noise performance, and the noise factor of subsequent stages, F, F 3 and so on are less important. F 1 F 3 1 Ftotal = F Eq. G 1 G 1 G G n is the gain of the nth stage in the receive chain. Cellular base-stations (BTS) and microwave relays have detached LNA stages located in the aerial tower to decrease NF degradation from pre-lna cable loss. In the typical BTS architecture, the LNA stage is preceded by a Transmit-Receive (Tx-Rx) diplexer for duplexing a common aerial and an interference filter for preventing out-of-band blocking or desensitization. However, both the duplexer and filter have losses that must be minimized as they occur before amplification []. A LNA with extra noise performance margin will relax the duplexer-filter s loss requirement and potentially lower overall system cost. Other critical LNA performance parameters include high gain to overcome loss in the long cable connecting the towermounted LNA and the ground-level radio shack and high linearity as the RF spectrum can be very crowded due to site sharing with other wireless transmitters. The Avago MGA-633P8, which is optimized for the 400 MHz to 1.5 GHz range, is part of a new active-bias LNA family that provides a common printed circuit board (PCB) footprint for the 400 MHz to 4 GHz range. It is ideal for GSM and CDMA cellular infrastructure applications. Besides the tower mounted LNA application, other uses are found in BTS radio cards, repeaters, remote radio heads, consumer premises equipment (CPE) and access points (AP). For optimum performance at frequencies from 1.5 GHz to 4.0 GHz the MGA-635P8 LNA is recommended.

2 Device and Package Technologies The MGA-633P8 is a microwave monolithic integrated circuit (MMIC) built on a new process optimized for low noise. The Avago GaAs enhancement-mode pseudo-morphic high electron mobility transistors (ephemt) process has a 0.5 m feature size. The MGA-633P8 is a common-source amplifier with an active bias regulator. Quiescent current (Ids) can be adjusted by varying either R BIAS, which connects the supply voltage Vdd to the shunt regulator, or by the external V BIAS control voltage. The regulator s drive bias current is a low 1 ma maximum, which makes it compatible with most CMOS control circuits. The adjustable bias enables better linearity but with higher power. Available in an 8-pin Quad Flat No-lead (QFN) package measuring.0 x.0 x 0.75 mm, the MGA-633P8 is suitable for compact applications where space and volume are limited. Vbias [1] RFin [] Bias Cct [8] NU [7] RFout NU [3] [6] NU NU [4] [5] NU Vb & Rbias vs. Vdd=5 V mv Vbias Rbias 450 Figure 1. MMIC internal structures (top) and characteristic of the adjustable internal bias regulator (bottom) V BIAS, R BIAS vs. Idd with Vdd = 5 V kω Centre: ma Id 4.5 Span: ma

3 Application Demonstration To evaluate RF performance, a 900 MHz LNA for a cellular base-station was designed around the MGA-633P8. External components C1-L1 and C-L provide the matching and biasing functions. In addition to the DC blocking and RF choking functions, C1-L1 also rolls-off undesirable gain below the operating frequency. Both L1 and L should be operated below their self-resonant frequency (SRF) for effective choking. The insertion loss, A(f), of a single transmission resonator at an evaluated frequency (f) is given by Equation 3, [3] : Q l 0 A() = 10 log Eq. 3 Q l 1 Qu f 0 is the resonant frequency; Q l is the loaded Q of the matching network; and Qu is the components unloaded Q, usually the Qu of the inductor as it is lower than that of the capacitor. At the centre of the resonator s pass-band, f 0 can be substituted for f, hence A = 0 log Q u Q 1 Q u Eq. 3(a) The loaded Q l of the matching network is determined by the ratio of input and source impedances, R HI /R LO [4] : R HI Q L = 1 Eq. 4 R LO Substituting for Q l gives: R HI Q u 1 R LO A = 0 log Eq. 5 Q u When the R HI /R LO ratio approaches 1, insertion loss, A, is approximately 0 db. The lowest loss is achieved when the source and input impedances are equal. To keep the R HI /R LO close to unity, the Avago proprietary ephemt process noise was optimized by scaling the transistor size and bias current for an input impedance of approximately 50. Capacitors C3-C5 decouple RF from the bias lines. By selecting C4 so that its reactance is approximately 6 at f 0, it also works with R1 to rolloff the gain below f 0. When it is required to switch the LNA via a logic-level control signal applied to pin 1 (Vbias), C3 should be reduced to the lowest capacitance value that will still provide a low reactance at f 0, for example 7 pf at 900 MHz. This low C3 value will speed up the switching time that is limited by the R BIAS x C3 time constant. Turn-on time is typically several microseconds for small C3 values in the pf range. 3

4 in C3 C1 L1 R bias [1] [] [3] [4] Vdd R1 C5 C4 L [8] C [7] [6] [5] Centre tab out The printed circuit board (PCB), which measures 1.5 mm x 18 mm x 1.4 mm, is made from 10-mil Rogers RO4350 microstrip with a coplanor ground and 1. mm FR4 material glued to the bottom as stiffener. RF connections were made through edge-launch SMA-to-microstrip transitions (Emerson Network Power/Johnson Component P/N ). The DC supply was connected via a -pin straight PCB header. The LNA demonstration board is powered from a single 5 V power supply. At the nominal Idd of 50 ma, approximately 0.5 V is dropped across R1, giving a Vds of approximately 4.5 V. The component values in Table 1 were used for the demonstration board and circuit but may be slightly different if either a different layout or PCB material is used. Table 1. MGA-633P8 Demonstration Board Part List Position Value Description C1, C 100 pf Murata GRM15 L1, L 33 nh Toko LL1005 C4 7 pf Murata GRM15 C3, C5 4.7 F Murata GRM15 R1 9.1 Note: All components are 040 size unless otherwise noted Figure. LNA circuit schematic (top), part placement diagram (middle), and photograph of assembled demonstration board (bottom) The demonstration LNA was simulated in Avago Technologies' ADS006A software in order to obtain the matching network components initial values. The MGA-633P8 was modelled using Touchstone formatted scattering (S-parameter) and noise parameter files supplied by Avago Technologies. To reduce the circuit complexity and simulation time, only first-order parasitics for the passive chip components (RLC) were considered. The inductor model used typical Q UL values at the nearest frequency (800 MHz) published by the vendor [5]. Inductor parasitic capacitances, Cpst, were calculated from their published typical SRF values but with an extra 0.1 pf added to account for pad-to-pad capacitance. The parasitic inductance, Lpst, in the capacitor model was also obtained from the vendor [6]. The -pin header and its associated pads were excluded because they were found to have little impact on the simulated results. The SMA to microstrip transitions were modelled using an ADS006A parameterized component model for a coaxial line with the parameter values obtained from the manufacturer; however, for simplicity, the discontinuity effects at the transitions were ignored. 4

5 P1 Num=1 TL1 L=5. mm C1 R=Rpst Ω C=100 pf TL Taper1 W1=W mm W=Wpad mm 1 Ref Taper W1=W mm W=Wpad mm TL3 C R=Rpst Ω C=100 pf TL4 L=5. mm TL6 P Num= CF4 Rbias R=6.8 kω C=Cpst pf TL5 A=1.5 mm Ri=4.6 mm Ro=5.3 mm L=9.375 mm T= mm Cond1=4.1E7 Cond=1.5E7 Er=.08 TanD=0.001 CF5 TL9 L=1.5 mm TL10 L=.6 mm Corn1 L1 L=33 nh Ql=8 Fl=800 MHz C=Cpst pf Corn TL8 L=6.5 mm TL11 L=0.7 mm CF6 TL7 L=0.6 mm C3 C=4.7 uf Bend Angle=45 M=0 TL19 W=0.3 mm L= mm SNP1 File="MGA633P8 5V50mA.sp" Bend1 Angle=45 M=0 TL0 L=1. mm Figure 3. Avago Technologies' ADS006A simulation circuit of a MGA-633P8 based 900 MHz LNA CF1 C4 C=7 pf C5 C=4.7 μf TL1 L L=33 nh Ql=8 Fl=800 MHz C=Cpst pf TL14 L=0.6 mm TL13 L=1.9 mm CF R1 R=9.1 Ω CF3 Var Eqn VAR1 W=0.5 Wpad=0.8 Lpst=0.5 Cpst=0.3 Rpst=0.5 Var Eqn VAR CF=0. MSub MSub1 H=0.5 mm Er=3.48 Cond=4.7E+7 T=0.0 mm TanD=0.004 Results and Discussion The general measurement conditions were f 0 = 900 MHz, Vdd = 5 V, and Idd = 50 ma. Key specifications for BTS design engineers are low noise in conjunction with good return loss (RL). This is because diplexers and filters are detuned by reflective terminations. At 900 MHz the demonstration board LNA achieved a noise figure, F, of 0.35 db, a gain, G, of 18 db and both IRL and ORL better than -15 db. The wide bandwidth of the input and output match ( GHz at the -10 db point) is favourable from the system standpoint as it prevents detuning of the input/output filters out-of-band frequency response. For both the G and RL parameters there is good agreement between simulated and measured results. The simulated noise figure is 0.15 db higher than the measured result at f 0. This small difference is not unusual and has been ascribed to a system repeatability issues. An ATN source-pull system was used for device noise characterization, and an Agilent 8970S noise figure meter was used for the final measurement. There are also different sample and day-to-day variations that contribute to repeatability [7]. Although this LNA targets a fixed-frequency cellular BST application, the good noise figure and return loss results from UHF to S-band show that the MGA-633P8 will be useful in wideband/multiband applications, such as Cable/Satellite TV distribution infrastructure, scanners, and military and multi-service radios. The measured frequency response exhibited out-of-band gain peaks at 10, 13.5 and 18 GHz. But they are well below the unity gain level and are not expected to cause instability. The Rollett stability factor, k, is greater than 1 when evaluated from HF to approximately 0 times the operating frequency. This means the LNA will be unconditionally stable with any termination having a positive real part. With a 50 ma quiescent bias, the gain dropped 1 db from nominal at an RF input drive of 5.3 dbm. This corresponds to an output 1-dB compression point (P1dB) of.3 dbm. 5

6 IRL, ORL and G vs. Frequency S11 mes S mes S11 sim S sim S1 mes db S1 sim Centre: MHz #, 5 V 53 ma Span: 1.00 GHz Figure 4. Simulated and measured IRL, ORL and G vs. frequency F vs. frequency mes sim Centre: MHz 09-Dec-09 4:14:41 PM sn0 5 V 50 m Figure 5. Simulated and measured noise figure vs. frequency Span: 1.00 GHz HP8970 Wideband gain and Rollett stability factor db S1 30 k Start: MHz #, 5 V 53 ma Stop: 0.00 GHz 0 Figure 6. Wideband G and k vs. frequency Output power, Gain and Drain Current vs. Pi at 900 MHz db Po G Id ma Centre: 3.00 dbm Pi 30 Span: dbm Figure 7. Po, G and Idd vs. Pi dbm 0 Fundamental, IMD3 and OIP3 vs. Pi (dm01, Vdd=5 V) Po imd OIP3 dbm Start: dbm Pi Stop: 0 dbm 8 Figure 8. Fundamental, third-order intermodulation product, and, OIP3 vs. Pi Oip3 vs. Ids (Vdd=5 V, f1 and f=900 & 901 MHz) dbm G P1dB Oip3 Figure 9. OIP3 vs. Ids (f1 = 900 MHz, f = 901 MHz, Pi = -10 dbm and Vdd = 5 V) dbm Start: 5.00 ma Idd Stop: ma 31 6

7 The point where the fundamental signal power (P fund ) and the third order intermodulation distortion theoretically intersect is called the third-order intercept point, OIP 3, and it is a measure of amplifier linearity. Equation 6 shows how OIP 3 is calculated: OIP 3 = P fund + ΔIM Eq. 6 IM is the difference between the fundamental and the intermodulation product power in db. Two input tones at 900 MHz and 901 MHz were used for this measurement, but other frequency spacing is not expected to change the results significantly. In the linear operating region (Pi < -10 dbm), OIP3 is slightly above 37 dbm at standard biasing. If either higher or lower linearity is required, the OIP3 can be varied as much as 10 db by operating at different Idd values in the 5 ma to 75 ma range. Gain and P1dB are minimally affected (G and P1dB ~0.5 db) when Idd is varied. Summary An MGA-633P8 LNA for a cellular BTS application has been designed and prototyped. Measured and simulated results show that the Avago MGA-633P8 has excellent NF, G, and ORL performance with minimal external component count. References and Resources 1. Agilent Technologies application note, AN57-1 Fundamentals of RF and Microwave Noise Figure Measurements, [Online] Available: I. Hunter, R. Ranson, A. Guyette, and A. Abunjaileh, Microwave Filter Design from a Systems Perspective, IEEE Microwave Magazine, pp , Oct K. V. Puglia, A General Design Procedure for Bandpass Filters Derived From Low Pass Prototype Elements: Part I, Microwave Journal, pp. 8, Dec C. Bovick, RF Circuit Design. Carmel, IN: Howard W. Sams, 198, pp Toko Inc. product specification, LL1005FHL Multilayer Chip Inductor, [Online] Available: 6. Murata Manufacturing Co. software, Murata Chip S-Parameter & Impedance Library Version 3.6.0, C. A Morales-Silva, L. Dunleavy, and R. Connick, Noise Parameter Measurement Verification by Means of Benchmark Transistors, High Frequency Electronics, Feb Avago Web: 9. Avago Flash Video: Avago Wireless Semiconductor Solutions for RF and Microwave Communications: collateral/wireless_semi_solns_sg.pdf 11. Avago Semiconductor Wireless Applications and Selection Guides: Semi_wireless_apps_SG.pdf For product information and a complete list of distributors, please go to our web site: Avago, Avago Technologies, and the A logo are trademarks of Avago Technologies in the United States and other countries. Data subject to change. Copyright Avago Technologies. All rights reserved. AV0-344EN - November 14, 011