Keywords: Amplifier, Linearization, IMD3 Suppression, Adaptive Source Harmonic Termination

Transcription

1 The Institution of Engineering & Technology Hong Kong Younger Members Exhibition & Conference 2010 Power Amplifier Linearization by Source Harmonic Termination Optimization WANG, Dian City University of Hong Kong, ABSTRACT Nowadays, different methods with non-constant envelop signal are implemented to meet limited channel bandwidth. As a consequence, the final stage of power amplifier must behave linearly over the whole dynamic range with desirable efficiency. Traditional methods based on 3rd order intermodulation distortion (IMD3) sweet spot use biasing adaption. However, they are limited by narrow dynamic range and might fail at higher output power level. In this paper IMD3 sweet spot was studied. Biasing and harmonic termination effects on sweet spot were investigated. A novel adaptive source termination was proposed. ADS simulation, which uses a single state BJT amplifier working at 2.4GHz, has shown more than 20dB improvement for IMD3 at peak output power with a wider dynamic range. In the two tone test experiment, the same amplifier with 12.86dB gain was used. Its IMD3 can be well controlled below -44dBm throughout the whole output dynamic range by optimizing source harmonic termination at each input power level. Comparing with adaptive biasing method, 45dB improvement can be achieved at peak power. In addition, less than -50dBc adjacent channel power ratio (ACPR) over the whole dynamic range was got from two tone excitation. Under -40dBc and -50dBc ACPR requirement, boosted upper dynamic range with enhanced efficiency was realized in comparison with adaptive biasing method. The result demonstrated good linearity and dynamic range. With those desired features and simple design, the method, thus, is applicable. Keywords: Amplifier, Linearization, IMD3 Suppression, Adaptive Source Harmonic Termination 1. INTRODUCTION With the continuing growth of modern wireless communication, high spectrum efficiency is critical for transmission of data within the limited channel bandwidth. Different methods with non-constant envelop signal, such as OFDM and QAM, are implemented to improve the channel capacity. As a consequence, the final stage of power amplifier in a communication system must behave linearly over the whole dynamic range including the peak power condition. On the other hand, as power amplifiers consume a significant portion of battery power, higher power-added efficiency (PAE) is another crucial factor to be considered for a longer operationtime. However, this high efficiency is usually achieved near saturation operation which has high distortion level. It is desirable to push power amplifier to high compression region with large linearity distortion. Thus many researches focus on how to improve linearity at high output power region. Traditional methods applied for linearization are either complex (e.g. feed-forward [1]-[3], digital pre-distortion [4]) or limited in their performance (e.g. analog pre-distortion [5]-[6]). Due to the special physical characteristic of transistors and the networking, an unexpected IMD minimum, which referred as sweet spot, may appear in

2 IMD pattern. This small and large signal IMD minimum has already been discovered and studied for years [7]- [8]. Different methods have also been proposed to utilize dynamic range and sweet spot control. Biasing control is one of the efficient methods to improve IMD and many techniques based on it have already been proposed[9]. However, this method is inevitably limited by harmonic termination conditions [7]. Studies on harmonic termination are much tougher and its detail effects are still unclear. However, it is known that the baseband and second harmonic termination have more essential influence. Thus, we extended our research to dynamic IMD3 sweet spot control by source harmonic terminations and designed a compact structure which can provide with the tuneable source termination. 2. SWEET SPORT CONTROL BY HARMONIC TERMINATION Discussed by Nuno Borges de Carvalho and Jos e Carlos Pedro[7], different amplifier classes result in different IMD3 patterns and some are with or without sweet spot. Harmonic termination seems to have strong impact on the sweet spot and IMD3 pattern. Sometime it directly influences the presence of sweet spot. As studied by previous researchers, both baseband signal and second harmonic termination impedance are important. In the following different terminations are discussed with given simulation results using a power amplifier based on BFP 450 transistor. The following figures provide with different IMD3 patterns under different source harmonic terminations. Fig.1. (a) Source baseband termination and sweet spot. (b) Harmonic termination control on IMD3 sweet spot. In Fig.1 (a) evidently, an open source baseband termination gives no sweet spot. On the contrary, the shot source baseband termination leads to an IMD3 pattern with sweet spot. Apart from the baseband termination, 2 nd harmonic termination (Z S/L_2nd ) is also studied with the following conditions listed in the table. Condition Z S BB (Ω) Z S 2nd (Ω) Z L BB (Ω) A 0-22j 0 B j 0 C j 0 Table. 1. Source Hermonic Termination conditions. The simulation results were plotted in Fig.1 (b). It is found that the harmonic termination demonstrates both dynamic control with regard to P in and optimization on the sweet spot. This phenomenon provides us a very useful direction for IMD3 suppression and linearization. This paper, which is based on discoveries discussed above, only focuses on source harmonic terminations.

3 3. DESIGN AND SIMULATION 3.1 Adaptive harmonic termination design The following figures show how the adaptive source harmonic termination is designed. Fig.2. (a) Source adaptive harmonic termination design. (b) Equivalent circuits at different frequency. In Fig.2 (a), the λ/4 short circuit transmission line at fundamental frequency is used for 2 nd harmonic termination while varactor diode is used for tuning. Fundamental signal can fully pass through the resonator. Since the source, which biases variactor diode also presents an impendence, it is connected to the λ/4 line which acts like a RF choke at fundamental frequency and a short circuit for DC. The resistor is chosen to be large which can be traded as an open circuit at high frequency. Thanks to these two configurations, the biasing network for variactor diode will only be effective at DC and has no influencing on the high frequency signals. Fig.2 (b) demonstrates the equivalent circuit at different frequencies where f o represents fundamental frequency. At fundamental frequency, if the resonance frequency is perfectly located, merely a lossless line is seen. However, in reality, no prefect component exists and the resonance frequency may shift a bit. In fact, the bandwidth for this resonator is very wide, thus frequency shift may not cause a problem. At 2 nd harmonic frequency, both diode path and resonator path are important. The limiting capacitor acts a very important part as for tuning rage controlling. Typically, the reversed biased varactor diode has merely 0.1pA current. A 1kΩ resistor will only have 1nV voltage drop which can definitely be ignored. In this sense, the DC biasing is directly applied to the diode and hence can control its capacitance. It should also be pointed out that there is hardly any DC power consumption on this tuneable part which is also desirable. 3.2 Simulation The following figure shows simulation result on how the adaptive termination can be tuned by tuning the variactor s biasing. Both Fig.3 (a) and (b) are obtained when the varactor is series with a 0.7pF capacitor. However, Fig.3 (b) was simulated by using non-ideal components: muratas. Obviously, the tuning range has been shifted. More seriously, for non-ideal simulation, it is found that the fundamental termination wanders around 50Ω which is not desirable. It is this effect which causes the mismatch and thus will leading to the drop of gain. According to simulation a worst 1.8dB mismatching loss can result. With this comparison, it is clear that the limiting capacitors are very important. In real circuit, the tuning range can be shifted and with much difference as simulation and different temperature and conditions also cause unpredictable errors. Limiting capacitors, thus, should be selected practically for optimizing the range and compensating these differences.

4 Fig.3. Source harmonic termination tuning (a) with ideal components (b) with non-ideal component. The following is a simple testing amplifier circuit using BFP 640 transistor and 1SV277 varactor diode (the basic experimental circuit structure is also this one). The fundamental frequency is set to be 2.4GHz for Bluetooth and Wi-Fi applications. Fig.4. Adaptive source harmonic termination power amplifier schematic. Based on the circuit given above, simulation was carried out. In the simulation a circuit without source adaptive harmonic termination was used for reference. The simulation was carried out by two tone test under 2.4GHz with 200KHz tone spacing. Notice that two different circuits are simulated in the same biasing condition. It is also worth to mention that the gain for this amplifier in the simulation is around 18.8dB and the 1 db compression point P in1db =-5.5dBm. The output power saturates around 15.8dBm. Because our objective is to focus on high output power region where compression is high while efficiency is high, P in =5dBm was selected in simulation for optimization where the gain compression is about -8.4dB and the output power is 15.4dBm. As Fig.5 (a) implies, not much improvement can be achieved by tuning the bias in this region. This has already been discussed by previous researchers [9]. Fig.5. (a) Biasing tuning and IMD3 at P in =5dBm. (b) Harmonic termination tuning and IMD3 at P in =5dBm.

5 Different capacitance values were used for the varactor diode and their resulted IMD3 patterns are plotted below against the reference design in Fig.6 (a). Fig.6. (a) Different IMD3 patterns for different C Vara. (b)simulation result of IMD3 suppression at high output level using adaptive source harmonic termination. It is not hard to found that as the capacitance of the varactor increases, the sweet spot shifts toward right to a high output power level which demonstrated good improvement at high output power level. If the termination is adaptively tuned the result in Fig.12 (b) can be got. More than 20dB improvement has been achieved at high output power level with a relatively wider dynamic range. The simulation result demonstrated that harmonic termination can be used for linearization at high output power level where adaptive biasing can hardly get any improvement. 4. EXPERIMENT AND DISCUSSION With simulation and discussion in last section, it is known that the circuit can be very sensitive to components and environment. The experiment result may also be different from simulation. Thus simulation results only give directions of operations but not accurate prediction of performance. For the experimental circuit demonstrated in Fig.7, different limiting capacitors were tried, and finally a suitable series capacitor (1.2pF) was found. It should also be remarked here that cables and connections in the experiment results in 0.5dB loss which has not been added to the experiment result. This is because the loss is linear which will not cause differences in IMD3 and output power pattern. However this loss does affect the efficiency. So it is included in efficiency calculations. Again, all the following experiments were done under two tone excitation at 2.4GHz center frequency with tone spacing of 200KHz. It is also important to mention that in the experiment, focus was on IMD3 performance. No matching and load-pull network were included. Thus, the gain and saturation power are relatively low but have no adverse influence on demonstration the proposed method. Fig.7. Adaptive source harmonic termination power amplifier circuits.

6 In the following, the experimental results from adaptive source harmonic termination are comparing with adaptive biasing method [9] unless otherwise stated. The figure below shows the spectra under two tone excitation at peak output power of 13dBm (10dBm each tone) where the amplifier was almost saturate. Fig.8. Spectruaat peak output power. It can be seen that adaptive source harmonic termination method can knock down IMD3 power to less than -50dBm while adaptive biasing method hardly has any suppression. The improvement at peak power is as high as 45dB which is quite desirable. Nevertheless, this improvement is meaningless unless a linearity requirement is set which will be discussed later in this section. From the spectra, it can be seen that the other two sideband which referred as IMD5 hasn t been suppressed. This is mainly due to the single adaptive harmonic termination. It is hard to achieve both IMDs suppression with a single adaptive termination. IMD5 suppression by adaptive termination will be studied in future work. The continuous IMD3 pattern presented below shows how adaptive source harmonic termination can provide sweet spot control at high output level. Fig.9. IMD3 suppression by adaptive source harmonic termination. In Fig.9 it is clear that more than 20dB IMD3 suppression was achieved which is quite desirable at high input power. This experimental result confirmed that the prediction that harmonic termination control can push the sweet spot to higher input power level. At lower input level the performance of adaptive harmonic termination is not as good as adaptive biasing, but still within a relatively small value which is less than -40dBm. For both methods, the output power can reach as high as 13dBm. However, as predicted in simulation in high power region, the gain of adaptive harmonic termination method is less than reference which is basically due to the

7 mismatching at fundamental frequency. The following figure shows the gain and how the varactor voltage should be tuned with respect to input power. Gain (db) Gain(Adaptive harmonic Z S ) Gain(Adaptive biasing) V Vara (Adaptive harmonic Z S ) P in (dbm) Fig.10. Gain and varactor voltage for adaptive source harmonic termination. In lower P in level the gain of the two methods are approximately the same. However, starting from P in =-4dBm they start to deviate from each other. But also note that the voltage biasing for the varactor diode before P in =- 6dBm is high where the capacitance as well as the series resistor are low. But this small resistance may not lead to such a gain drop. From simulation (Fig.3) it can be seen that the fundamental is well matched when V Vara is high. However, as the voltage of the varactor diode drops, the fundamental matching firstly shifts away and then back. It results a maximum 1.8dB mismatching loss. Although this is only simulation result which may not give exact prediction, it is still a valid explanation for the gain difference. Suffering from this mismatch factor, when the voltage drops from 7V to 1V, the gain of the amplifier with adaptive source harmonic termination is approximately 1.6dB lower than reference. Before further comparing, it is also worth to look at DC power consumption for both cases Voltage (V) DC power (mw) 44 DC power (adaptive harmonic Z S ) DC power (adaptive biasing ) P in (dbm) Fig.11. DC power consumption comparison. When P in is larger than -4dBm, DC power consumed by adaptive biasing method is essentially larger than adaptive source harmonic termination method. That is to say, after this point, adaptive biased power amplifier can handle more power. This is also the reason that its gain suffers less compression in Fig.10. For this reason, it is not fair to compare their PAE which involves gain and power handling. Nevertheless, efficiency, which only depends on output power and DC power, is of more interest. When linearity requirement is set, the improvement is clearer. Currently, 3G mobile device requires -33dBc ACPR while base stations have tighter ACPR restrictions around lower than -40dBc. However, for future

8 signals, even tighter restrictions will be set. Thus -40dBc and -50dBc ACPR standards are used for discussion below. Fig.12. ACPR versus power efficiency. In the above figure a conventional power amplifier with merely one sweet spot, which was biased approximately the same as adaptive source harmonic termination, is also presented for reference. Clearly enough, the proposed method provides with the whole dynamic range under both requirements. Thus, the peak efficiency can be utilized. Conventional method and adaptive biasing method may have comparable or even better efficiency at high output power level, but at this high efficiency output power level, the linearity does not meet the requirement. The solution is to backup output power to work in a relatively low region. The upper dynamic range reduces, which causes dramatically drop on efficiency. According to the figure, the upper dynamic range and efficiency for different restrictions are summarized in the following table. Upper dynamic Linearity requirement Method applied Peak efficiency range (dbm) Conventional % Lower than Adaptive biasing % -40dBc ACPR Adaptive source harmonic % Lower than -50dBc ACPR Lower than -40dBc ACPR termination Conventional % Adaptive biasing % Adaptive source harmonic termination % Table. 2. Comparison on upper dynamic range and peak efficiency. Peak efficiency in the table is referred as the highest efficiency under upper dynamic range. Table.2 clearly demonstrates the improvement of adaptive source harmonic termination. Under -50dBc ACPR requirement, it boosts the upper dynamic range to almost twice that of the adaptive biasing method. At the same time, 27% improvement from adaptive biasing method is achieved in efficiency. In additional, in today s mobile device application (such as Wi-Fi, Bluetooth, and mobile phone), where non-constant output power is needed, average power efficiency is more meaningful. Listed in Table.3 adaptive source harmonic termination method demonstrates 10% improvement on average power efficiency in comparison with adaptive biasing method. Linearity requirement Method applied average power efficiency Conventional 16.3% Adaptive biasing 26.7% Adaptive source harmonic termination 36.4% Table.3. Comparison on average power efficiency.

9 5. APPLICATION For hand-set applications where linearity restriction is relatively low, our method can proved improved average power efficiency and boosted upper dynamic range. In future communication systems like 3G or 4G, this allows devices to sustain for longer time and to be capable for high peak to average signal which has higher transmission rate. For base station or highly reliable communication systems, having higher ACPR restriction, the method can provide with enhanced upper dynamic range with peak efficiency. This is critically important for sustainable development and green system. Moreover, due to the simplicity of this method, it can also be applied together with other linearization methods to further enhance performance, such as combining both adaptive biasing and harmonic termination method for a higher gain and efficiency. 6. CONCLUSION Within this paper, a power amplifier with adaptive source harmonic termination was designed. Adaptive biasing method shows good control on IMD3 sweet spot when input level is relatively low. However, this method is limited by its performance and has little control on the sweet spot in high output power range. Thus, adaptive source harmonic termination was designed and applied. Simulation results demonstrated the effect of this method. In the experiment, the proposed method has -45dB IMD3 suppression at saturation output power when compared with adaptive biasing method. Moreover, under -40 and -50dBc ACPR requirement, the efficiency was improved by 19.7% and 27%, respectively. The upper dynamic range is also boosted from 11.1dBm and 10dBm, respectively, to 13dBm. Through discussion and results, we have demonstrated a novel linearization method with simple and compact design. 6. REFERENCE [1] Jing, D., Chan, W.S., Li, S.M. and Li, C.W New linearization method using interstage second harmonic enhancement. IEEE Microwave Guided Wave Lett., vol. 8, pp [2] Kang, S.G., Lee, I..K. and Yoo, K.S Analysis and design of feedforward power amplifier. IEEE MTT- S Int. Microwave Symp. Dig., pp [3] Hau, Y.K.G., Postoyalko, V. and Richardson, J.R Design and characteristics of a microwave feedforward amplifier with improved wide-band distortion cancellation. IEEE Trans. Microwave Theory Tech., vol. 49, pp [4] Kim, J. and Konstantinou, K Digital predistortion of wideband signals based on power amplifier model with memory Electron. Lett., Volume 37, Issue 23, p [5] Hau, G., Nishimura, T. and Iwata, N A highly efficient linearized wide-band CDMA handset power amplifier based on predistortion under various bias conditions. IEEE Trans. Microwave Theory Tech., vol. 49, pp [6] Yu, C.S., Chan, W.S. and Chan, W.L GHz low loss varactor diode pre-distorter. Electron Lett., vol. 35, no. 20, pp [7] Nuno Borges de Carvalho and Jos e Carlos Pedro, Large- and small-signal IMD behavior of microwave power amplifiers. IEEE Transactions on Microwave Theory and Techniques, vol. 47, no. 12. [8] Teeter, D.A., East, J.R. and Haddad, G.I Use of self bias to improve power saturation and intermodulation distortion in CW Class B HBT operation. IEEE Microwave and Guided Wave Letters, vol. 2, no. 5, pp [9] Lau, K.W Self-adaptive biasing technique: linearity and efficiency improvements for microwave power amplifiers. PhD Thesis. City University of Hong Kong.

10 [10] Cripps, Steve C RF power amplifiers for wireless communications. Second Edition. Canton Street, Norwood, MA. [11] Novis, S.R. and Pelletier, L IMD parameters describe LDMOS device performance. Microwaves RF, vol. 37, no. 7, pp ACKNOWLEDGEMENT Here I would like to express my sincerest gratitude to those people who have helped me in this project. First and foremost, I would like to thank my supervisor, Prof. Chi-Hou Chan and Dr. Xue, Quan for their continuous guidance and support. In addition, I owe my heartfelt thanks to Prof. Franke in University of Illinois at Urbana- Champaign. I d also like to thank Roy and King for their support in simulation software and experiment.