Adaptive Second Harmonic Active Load For Pulsed-IV/RF Class-B Operation

Transcription

1 Adaptive Second Harmonic Active Load For Pulsed-IV/RF Class-B Operation Seok Joo Doo, Patrick Roblin, Venkatesh Balasubramanian, Richard Taylor, Krishnanshu Dandu, Gregg H. Jessen, and Roberto Rojas Electrical and Computer Engineering, The Ohio State University, Columbus, OH 431, USA Telephone: ( Texas Instruments Corporation, Sunnyvale, CA 9489, USA Air Force Research Laboratory, WPAFB, Dayton, OH 45433, USA Abstract To provide a proper load impedance termination under class-b operation, for pulsed-iv biasing and pulsed-rf (pulsed-iv/rf excitations, an adaptive harmonic active load is presented. Pulsed-IV/RF measurement is the technique of choice for characterization and modeling of RF devices as it guarantees that the transistor is not afflicted by low-frequency memory effects. It is also a promising technique to assist with the design of efficient pulsed power-amplifiers for TDMA or radar applications. In this work the multi-harmonic active loadpull technique used for CW excitations is extended to pulsed- IV pulsed RF excitations. The harmonic load was implemented with an active load and a large signal network analyzer (LSNA both under computer control for adaptive tuning. The technique is demonstrated on both a GaN HEMT and a 65 nm NMOS transistor operating in class-b to quantify the impact of memory effects in pulsed-iv/pulsed-rf operation compared to CW operation. Index Terms Pulsed-IV, pulsed-rf, memory effects, large signal network analyzer (LSNA, class-b. I. INTRODUCTION A pulsed-iv pulsed-rf measurement system is useful to characterize transistors while avoiding low-frequency memory effects such as traps and thermal effects. Since low-frequency memory effects have indeed a slow time response, the fast pulsed-iv pulsed-rf (pulsed-iv/rf measurements will maintain a constant temperature or trap state in FETs making it possible to obtain accurate RF characteristics from devices exhibiting low-frequency memory effects [1], []. A pulsed-iv/rf measurement system with improved dynamic range can be realized using a large signal network analyzer (LSNA [3]. For a pulsed-rf signal with a.3% duty rate, the dynamic range in a conventional single-tone network analyzer is reduced by 5dB due to the pulse desensitization log(duty Cycle. On the other hand only 1dB of desensitization results in the LSNA pulsed-iv/rf measurement system when the amplitude and phase of the tones in the main lobe of the sinc spectrum are acquired. An active load-pull integrated to a pulsed VNA under pulsed biasing also has been reported to measure RF power variations and RF carrier shift within the pulse under class-ab and class- C operations [4]. Only the fundamental frequency was considered. To maximize the device performance, however, it is also desirable to control the load impedance of each harmonic. For instance in class-b power amplifiers, only 5% efficiency could be obtained by optimizing the fundamental frequency, whereas 78.5% efficiency can be achieved when all harmonics are optimized together [5] (class-f. For this purpose, a multi-harmonic active load-pull system provides an efficient approach as was reported in [6]-[7]. However these systems have been so far limited to continuous wave (CW active loads. In this paper, an active pulsed RF-load for the adaptive tuning of harmonic output-impedance is presented and demonstrated for the automatic characterization of devices under class-b operation for pulsed-iv/pulsed-rf excitations. II. SYSTEM IMPLEMENTATION WITH A LSNA Fig. 1 shows the measured pulsed-rf signal which typically has a.33µs pulse width and.33% pulse duty cycle. Consider an FET with the pulsed-rf signal excitation at the gate with simultaneously pulsed bias voltages applied to the gate and drain of the device as shown in Fig.. The pulse biasing with a 1µs pulse width and 1% duty rate is typically fast enough to avoid slow memory effects. Using the source and load tuners in conjunction with the LSNA, we are able to gain access to the dynamic RF response of the transistor for modulated large signals. In general both tuners contribute to provide the desired source and load impedances at the fundamental frequency. However the default 5Ω load termination at the fundamental frequency is used in this initial work to maintain sufficient RF power levels required for the active harmonic load-pull operation. A stable open-loop pulsed-rf active-load for the n th harmonic load impedance can be achieved by injecting an additional pulsed-rf signal at the n th harmonic on the drain side. In this work focus is placed on the dominant nd harmonic which must be shorted for class-b operation. This can be accomplished by using a CW signal generator with a pulse modulator and a pulse generator as illustrated in Fig.. The pulse generator provides the pulse timing for the pulse modulator. It is essential to make sure that the pulsed-rf generators for and are synchronized together so that the pulsed-rf active-load is applied in the same time window.

2 Large Signal Network Analyzer Port1 Tuner DUT Tuner Port ( Pulsed RF generator Current sensor Oscilloscope DC Bias supply Current sensor (a synchronization Gate Pulser Pulse generator synchronization Drain Pulser Diplexer Couplers ( Pulse modulator Pulsed RF generator ( ( Isolator HPF Fig.. Pulsed-IV pulsed-rf measurement setup with an adaptive second harmonic load. (b Fig. 1. Pulsed-RF measurements using the LSNA. (a Sinc spectrum (white: magnitude, red: phase. Only the main lobe (61 tones of the spectrum is used for the analysis. (b Pulsed-RF signal in time. The pulsed-rf excitation uses the pulse width of.33µs and pulse repetition frequency of 1kHz. An alternative way to implement the nd harmonic active load is to inject a CW signal instead of a pulsed-rf signal for the nd harmonic. For this purpose, the pulsed-rf signal generator ( block in Fig. can be replaced by a CW RF signal generator. Since the transistor is biased using pulsed-iv, the transistor will still avoid slow-memory effects. The diplexer provides a path to deliver the harmonic pulsed- RF signals to the device output terminal while providing a good impedance matching at the fundamental frequency. A circulator is also used to isolate the harmonic RF sources from the output signal originating from the device. III. ADAPTIVE SECOND HARMONIC ACTIVE LOAD FOR PULSED-RF EXCITATION The incident and reflected pulsed-rf waves exhibit a sinc spectrum centered on each harmonic. The sinc spectrum data (61 tones in the main lobe acquired by the LSNA are used to reconstruct the pulsed-rf time-dependent a (t, and b (t, waveforms at the center of the pulse [3]. It has been demonstrated that this reconstruction scheme has the advantage of removing most of the desensitization problem present in conventional single-tone pulsed-rf measurement systems. An automatic tuning of the active load was implemented for the second harmonic via successive phase and magnitude corrections to obtain a pulsed-rf signal at the port verifying a (t, = b (t, at the center of the pulse. Fig. 3. Adaptive tuning of the pulsed-rf active load for a SHORT at. The consecutive magnitude and phase corrections are represented by plain (blue and dashed (red lines, respectively. The automatic adaptive tuning is monitored by the user on a graphical display as shown in Fig. 3. This permits the setting of the required harmonic load condition with improved speed and accuracy. This practical implementation is made possible thanks to the measurement capability of magnitude and phase in pulsed modulation mode of the LSNA. The minimized v (t, is shown in Fig. 4 for pulsed-iv/rf class-b operation. When a discrete set of tones in the frequency domain with M the total number of single-side band (SSB of the sinc is considered, the output power at the pulse center (t = t c can be obtained by applying P out (, t c = 1 M M (v p iq ej(p q tc, p= M q= M where v p = v ( + p and i q = i ( + q. The PAE

3 .5 a ( t, v ( t,. b ( t,.15.1 Fig. 4. Waveforms at the center of the pulse when the nd harmonic impedance is shorted. The minimized v (t, is obtained when a (t, = b (t,. In the plot, red is for v (t,, yellow for a (t,, and green for b (t,. TABLE I FOUR TYPES OF LOAD-LINE MEASUREMENTS Method Bias Gate excitation Γ L ( Active load #1 [8] DC-IV CW -1 using CW # [9] Pulsed-IV Pulsed-RF (no control #3 Pulsed-IV Pulsed-RF -1 using CW #4 Pulsed-IV Pulsed-RF -1 using pulsed-rf at t = t c then can be defined as P AE(, t c = P out(, t c P in (, t c. P dc (t c IV. EXPERIMENTAL RESULTS In order to demonstrate the capabilities of the developed approach, a GaN HEMT on SiC was first investigated to obtain pulsed-rf load-lines at GHz under class-b operation. Fig. 5 shows the complete set of acquired pulsed-rf class-b dynamic load-line when the condition a (t, = b (t, is satisfied at the center of the pulse-rf signal. With this condition, the output power of.13 W and the PAE of 68.7% was obtained with Γ L ( =. Notice that there is a shift of the biasing point due to the applied pulsed- RF stimulus even thought the device is biased at cut-off for class-b operation. To quantify the impact of memory effects, four types of measurement conditions were considered based on different biasing conditions, gate RF excitations and the methods for obtaining an active nd harmonic load. This is summarized in Table I. The Γ L ( was set to zero (5Ω for all cases. The same experiments were repeated with MOSFETs (Si NMOS 65 nm to compare dynamic load-lines. Since the MOSFETs usually exhibit very small memory effects, their measurements are useful for validating the proposed pulsed- IV/RF class-b measurement methods. A. GaN HEMTs Fig. 6 compares the obtained RF dynamic load-lines for the GaN HEMT based on the measurement types in Table I. The (a Gate load-line (b Drain load-line Fig. 5. Dynamic pulsed-iv/rf class-b load-line for the GaN HEMT. In (b, the shifted self-biasing points is marked as a triangle. For IV characteristics, the is varied from -4 to V in steps of 1 V. same RF excitations (v 1 =.9 V were used for the various load-lines as shown in Fig. 6(a. Method #1 [8] which is based on a normal DC-IV biasing and a CW active load at yields a class-b load-line. However it is expected to be afflicted by memory-effects. Method # [9] uses a pulsed-iv biasing and pulsed-rf excitation for the fundamental frequency with a nd harmonic termination of 5Ω instead of a SHORT. It does not exhibit a class- B load-line, and this confirms the importance of harmonic terminations in getting proper class-b operation. Methods #3 and #4 rely on the previously described pulsed-iv/rf schemes with CW and pulsed-rf injections for the nd harmonic respectively. The acquired class-b dynamic load-lines are found to be in good agreement between methods #3 and #4. Note that the larger drain current in method #1 (DC- IV/CW compared to methods #3 and #4 (pulsed-iv/rf is found to arise from the reduction of the threshold voltage by

4 #1. DCIV CW(, #. PIV PRF( #3. PIV PRF( CW( #4. PIV PRF(, =8mA 5 #1. DCIV CW(, #. PIV PRF( #3. PIV PRF( CW( #4. PIV PRF(, =1mA (ma same V T =7mV (a Gate load-line (a Gate load-line #1. DCIV CW(, #. PIV PRF( #3. PIV PRF( CW( #4. PIV PRF(, 5 #1. DCIV CW(, #. PIV PRF( #3. PIV PRF( CW( #4. PIV PRF(, (ma (b Drain load-line (b Drain load-line Fig. 6. Comparison of dynamic load-lines for the GaN HEMT. In (a, the difference of maximum drain currents between methods #1 and #3 (or #4 is approximately 8 ma, which is corresponding to 7 mv threshold voltage reduction. In (b, the triangles indicate the obtained DC-IV and pulsed-iv biasing points under class-b operation. approximately 7 mv which is presumably due to self-heating arising under CW operation. B. MOSFETs Fig. 7 shows the comparison of load-lines which are acquired from a NMOS transistor. As confirmed with the GaN HEMT device, method # does not demonstrate a proper class- B load-line because of inappropriate harmonic load terminations. However, the class-b load-lines from method #1 (DC- IV/CW are pretty similar to ones from #3 and #4 (Pulsed- IV/RF except for a slight reduction in drain current. Since it is believed that the used NMOS transistor is not much affected by low-frequency memory effects, the obtained dynamic loadlines should indeed share the same RF response for a CW or pulse-rf excitation. Note that the small drain current reduction of 1 ma presumably comes from a small self-heating effect Fig. 7. Comparison of dynamic load-lines for the NMOS transistor. For IV characteristics in (b, the is swept from to 1. V in steps of.1 V. The similar biasing points marked as triangles were obtained for all cases. or memory effects associated with the body voltage. V. CONCLUSION In this paper, a computer-controlled adaptive secondharmonic active-load for pulsed-iv/rf class-b operation was implemented with a LSNA. The adaptation was realized by automatic successive magnitude and phase corrections. This system was used for the comparison of dynamic load-lines obtained from GaN HEMTs and NMOS transistors. The major advantage of the proposed system is to maintain iso-thermal and iso-trapping conditions during device characterization while attaining a desirable harmonic load termination which is essential for the design of high efficiency power amplifiers. In order to maximize the PAE or output power of class-b power amplifiers, however, it is also necessary to provide an optimum load termination at the fundamental frequency. This

5 can be achieved by applying load-pull measurements at combined with the proposed adaptive active load at. Therefore the adaptive harmonic active load presented here should prove beneficial to characterize transistors for the design of high-efficiency class-b or class-c power amplifiers for pulsed-rf applications. ACKNOWLEDGMENT This work was sponsored in part by a fellowship from the Automatic RF Techniques Group (ARFTG. Additional support was provided by the National Science Foundation (NSF under grant ECS-63 and by Texas Instruments Corporation. REFERENCES [1] J. Scott, J. G. Rathmell, A. Parker, and M. Sayed, Pulsed device measurements and applications, IEEE Trans. Microwave Theory and Techniques, vol. 44, pp , Dec [] P. Roblin and H. Rohdin, High-speed Heterostructure Devices. Cambridge University Press,. [3] S. J. Doo, P. Roblin, S. Lee, D. Chaillot, and M. V. Bossche, Pulsed- IV pulsed-rf measurements using a large signal network analyzer, in ARFTG 65th Conf. Dig., (Long Beach, CA, June 5. [4] C. Arnaud, D. Basataud, J. Nebus, J.-P. Teyssier, J. Villotte, and D. Floriot, An active pulsed RF and pulsed DC load-pull system for the characterization of HBT power amplifiers used in coherent radar and communication systems, IEEE Trans. Microwave Theory and Techniques, vol. 48, pp , Dec.. [5] S. C. Cripps, RF power amplifiers for wireless communications - nd ed. Artech House, 6. [6] J. Benedikt, R. Gaddi, P. J. Tasker, and M. Goss, High-power timedomain measurement system with active harmonic load-pull for highefficiency base-station amplifier design, IEEE Trans. Microwave Theory and Techniques, vol. 48, pp , Dec.. [7] X. Cui, S. J. Doo, P. Roblin, G. H. Jessen, R. G. Rojas, and J. Strahler, Real-time active load-pull of the nd & 3rd harmonics for interactive design of non-linear power amplifiers, in ARFTG 68th Conference, (Colorado, Nov. 6. [8] P. McGovern, J. Benedikt, P. J. Tasker, J. Powell, K. P. Hilton, J. L. Glasper, R. S. Balmer, T. Martin, and M. J. Uren, Analysis of DC-RF dispersion in AlGaN/GaN HFET s using pulsed I-V and time-domain waveform measurements, in IEEE MTT-S Int. Dig., (Long Beach, CA, June 5. [9] S. J. Doo, P. Roblin, G. H. Jessen, R. C. Fitch, J. K. Gillespie, N. A. Moser, A. Crespo, G. Simpson, and J. King, Effective suppression of IV knee walk-out in AlGaN/GaN HEMTs for pulsed-iv pulsed-rf with a large signal network analyzer, IEEE Microwave and Wireless Components Letters, vol. 16, pp , Dec. 6.