A KA-BAND RADAR INTERFEROMETER - DEVELOPMENT AND TESTING

Transcription

1 A KA-BAND RADAR INTERFEROMETER - DEVELOPMENT AND TESTING A Thesis Presented by BRUNO GALOBART TOR Submitted to Universitat Politècnica de Catalunya Escola Tècnica Superior d Enginyeria de Telecomunicació de Barcelona in partial fulfillment of the requirements for the degree of ENGINYERIA ELECTRÒNICA May 2013 Electrical and Computer Engineering University of Massachusetts Amherst

2 A KA-BAND RADAR INTERFEROMETER - DEVELOPMENT AND TESTING A Thesis Presented by BRUNO GALOBART TOR Approved as to style and content by: Paul R. Siqueira, Professor Electrical and Computer Engineering

3 To my family

4 ACKNOWLEDGMENTS I would like to express my gratitude to all those who have contributed to the realization of this thesis. First, I would like to thank Professor Paul Siquiera, my advisor, for giving me the opportunity to finish my studies in MIRSL at the University of Massachusetts, and for his assistance during all this period. I would also like to acknowledge Tom Hartley for helping me with some experiments and for his advices when I had problems, and Cristina Llop for her help and guidance, specially at the beginning when I was new here. I would also like to thank the new friends with whom I have shared my stay in Amherst, without them it would have been more difficult. And I don t want to forget my friends in Barcelona, a special thanks to all of them. And last but not least, I would like to thank my parents for their support and encouragement all this years. iv

5 ABSTRACT A KA-BAND RADAR INTERFEROMETER - DEVELOPMENT AND TESTING MAY 2013 BRUNO GALOBART TOR Directed by: Professor Paul R. Siqueira Radar interferometry is a remote sensing technique that consists in transmitting a signal and measuring the phase difference of the reflected signal between two antennas separated by a baseline. This phase difference can be used to estimate the topographic height of the target. The height accuracy is limited by the ability to accurately measure the phase difference. The objective of this thesis is to develop and test a Ka-band dual downconverter and upconverter for an interferometer, specially the phase performance of the downconverter and how it is affected by the temperature. This interferometer will be used in NASA s Surface Water and Ocean Topography (SWOT) mission, that will map the water elevation on Earth. v

6 TABLE OF CONTENTS Page ACKNOWLEDGMENTS iv ABSTRACT v LIST OF TABLES ix LIST OF FIGURES x CHAPTER 1. INTRODUCTION History and Motivation Summary of Chapters RADAR INTERFEROMETRY Introduction Phase Estimation Height Estimation Height Accuracy RADAR HARDWARE DESCRIPTION Interferometer Receiver Ka-band to L-band Downconverter Telemetry L-band to Baseband Downconverter LO Filters L-band Filters vi

7 Baseband Filters Transmitter Ka-band Dual Upconverter Baseband Mixer L-band LO Power Splitter Ka-band Filters RADAR HARDWARE EVALUATION Ka-band to L-band Downconverter Gain Measurement L-band to Baseband Downconverter Gain Measurement Image Rejection Ratio Channel-to-Channel Isolation Phase Performance Ka-band Dual Upconverter Output Power PHASE CHARACTERIZATION Phase Estimation Algorithm Experimental Setup Microcontroller MATLAB Program Phase Stability October 2011 Version May 2012 Version Factors affecting Phase Stability Temperature Control Resistors CONCLUSIONS AND FUTURE WORK Summary of Work Recommendations for Future Work vii

8 APPENDICES A. MATLAB CODE FILES A.1 phase temperature.m A.2 temperature.m A.3 phase temperature plot.m A.4 temperature fix.m A.5 polynomial.m A.6 polynomial best.m B. MICROCONTROLLER CODE FILES B.1 KA2L AtoD Simple.spin BIBLIOGRAPHY viii

9 LIST OF TABLES Table Page 2.1 Interferometry Parameters Interferometer Specifications A-to-D Channel Assignments Measured Parameters of the LO Filter Measured Parameters of the L-band Filter LO Power Splitter Design Parameters Ka-band Filter Design Parameters ix

10 LIST OF FIGURES Figure Page 1.1 SWOT Ka-band Radar Interferometer Interferometric Antenna Geometric Model of an Interferometer Image Bands in the Interferometer Interferometer Block Diagram Block Diagram of the Ka-band to L-band Downconverter Picture of the Ka-band to L-band Downconverter Block Diagram of the L-band to Baseband Downconverter Picture of the L-band to Baseband Downconverter Measured S-Parameters of the Baseband Filter Block Diagram of the Ka-band Dual Upconverter Picture of the Ka-band Dual Upconverter Single-Sideband Mixer HMC497LP Layout of the LO Power Splitter Simulated S-Parameters of the LO Power Splitter Layout of the Ka-band Filter Simulated S-Parameters of the Ka-band Filter (52.5 mil) Simulated S-Parameters of the Ka-band Filter (51 mil) x

11 4.1 Ka-band to L-band Downconverter Gain with different f LO Ka-band to L-band Downconverter Gain (f LO = GHz) L-band to Baseband Downconverter Gain L-band to Baseband Downconverter Image Rejection L-band to Baseband Downconverter Channel Isolation L-band to Baseband Downconverter Phase Difference (f = 240 MHz) Ka-band Dual Upconverter Output Power with different f LO Ka-band Dual Upconverter Output Power (f LO = GHz) Ka-band Dual Upconverter Output Power (f in = 40 MHz) Experimental Setup for measuring the Phase Stability Phase Difference in the Ambient Temperature of the Laboratory Phase Difference in a Thermally Isolated Box Temperature Sensors (ADC 2) and Temperature Control Resistors Phase Difference versus Temperature Phase Difference versus Temperature (with errors) Estimated Frequency, Mean and Amplitude (with errors) Phase Difference versus Temperature (Day 1) Phase Difference versus Temperature (5 days) Phase Difference while moving the Connectors Phase Difference using the Resistors (with errors) Phase Difference using the Resistors xi

12 CHAPTER 1 INTRODUCTION 1.1 History and Motivation The Surface Water and Ocean Topography (SWOT) mission was recommended by the National Research Councils first decadal survey and is being developed by NASA [5]. The main objectives of the mission are (i) measure the ocean level at high resolution and over wide swath to characterize the ocean mesoscale and submesoscale circulation, and (ii) measure the elevation of water on land (rivers, lakes, wetlands) to characterize the storage and discharge of surface waters [4]. This will allow a better study of the climate change and understanding of flood processes. The ocean surface temperature can also be estimated from the different measured heights, because water expands when the temperature increases. The altimetric measurements has been made using radar altimeters, but the spatial resolution of the observations is not enough for this mission. The requirements will be achieved using another measurement technique, radar interferometry. It consists in estimating the topographic height, transmitting a signal and measuring the phase difference of the reflected signal between two antennas separated by a baseline. The main instrument of the SWOT mission will be a Ka-band Radar Interferometer (KaRIN) with a bandwidth of 200 MHz in a spaceborne platform, shown in Figure 1.1. High frequencies, like Ka-band, are used to achieve the required resolution and accuracy, and also allows to reduce the size and weight. A baseline length of 10 m is required for this interferometer, shorter enough for a spaceborne system, and it will permit to reduce the total cost. However, the design, manufacturing and evaluation of the 1

13 Figure 1.1. SWOT Ka-band Radar Interferometer interferometer will be more difficult due to the required accuracy for measuring phase at short wavelengths and the sensitivity to temperature. A Ka-band dual-channel downconverter and Ka-band upconverter are being developed in the Microwave Remote Sensing Laboratory (MIRSL) at the University of Massachusetts Amherst, to use them in the SWOT mission. The objective of this thesis is to continue with the work done in [13, 14, 15]. The upconverter will be redesigned and built to increase the bandwidth to 200 MHz and to change the output frequencies to GHz. The downconverter was previously modified to meet the new requirements [11]. The phase stability of this downconverter is extremely important to measure the phase difference of the interferometer and estimate the topographic height. It will be tested, using a phase estimation algorithm [9], to see how the temperature of the system affects the accuracy of the phase difference. Source: NASA Jet Propulsion Laboratory ( 2

14 1.2 Summary of Chapters Chapter 2 presents radar interferometry and different types of interferometers. The geometry and phase estimation of a single-pass cross-track interferometer are explained to estimate the topographic height. Chapter 3 describes the downconverter and upconverter of the interferometer. A detailed description and some simulations are provided for the modifications made to change the output frequency. Chapter 4 presents the results of the different measurements made to test the downconverter and upconverter. Chapter 5 discusses the performance of the phase difference in the Ka-band to L-band downconverter. It explains the phase estimation algorithm and the experimental setup used to measure the the phase stability and to analyze the effect of temperature. Chapter 6 provides the summary of work and recommendations for future work. 3

15 CHAPTER 2 RADAR INTERFEROMETRY 2.1 Introduction Radar interferometry is a remote sensing technique that consists in measuring the interferences of two different electric fields separated in space or in time. It can be used to measure topography and topographic change, depending on the configuration of the interferometer. A radar interferometer transmits a signal and measures the reflected wave with radar receivers separated in space or in time. The two received signals are used to calculate the phase difference between them. There are different configurations for the interferometers [7]. A single-pass interferometer has two antennas separated by a baseline and takes both measurements simultaneously. A repeat-pass interferometer makes two different measurements separated in time with the same receiver. An along-track interferometer has the baseline oriented in the direction of the antenna motion and the phase difference is related to the velocity of the targets. If the baseline is oriented perpendicular to the motion direction, it is called cross-track interferometer and the phase difference is related to the topography. The interferometer explained below is a single-pass cross-track interferometer. 2.2 Phase Estimation The radar interferometer should measure the range difference ( r) between the incident waveforms in both antennas (A 1 an A 2 ), but the delay between them is too small to measure by pulse time delay only. The range (r) is usually very large 4

16 Δr A 2 B α α r A 1 H Figure 2.1. Interferometric Antenna and using the far-field approximation, it can be assumed that both antennas receive parallel plane waves (see Figure 2.1). Then, the phase difference (φ) between the two antennas can be measured instead of r. The relation between them is φ = 2π r λ. (2.1) The phase difference can be determined by the phase of the cross-correlationn (γ) between the electric field measured at the two antennas [10]. The cross-correlation is γ = E 1 E2 = γ 0 e jφ (2.2) E 1 2 E 2 2 where the brackets indicate averaging over independent samples. The magnitude (γ 0 ) indicates the coherence of the cross-correlation. 2.3 Height Estimation The geometric model of the cross-track interferometer is shown in Figure 2.2 and the description of all the parameters is listed in Table 2.1. The phase difference is 5

17 A 1 B α A 2 r 2 = r + Δr H r 1 = r h Figure 2.2. Geometric Model of an Interferometer measured between the two antennas separated by a baseline B, at a height H from a reference level. The baseline is tilted an angle α relative to the horizontal. The look angle θ produce the range r from the antenna A 1 to the target. The target is at height h, the value to be measured. Table 2.1. Interferometry Parameters Parameter A 1 A 2 B H h r 1 r 2 θ α λ φ Description Antenna 1 position Antenna 2 position Baseline length Reference height Topographic height Range from antenna 1 (r) Range from antenna 2 (r + r) Look angle Baseline angle Radar wavelength Phase difference between channels 6

18 The height of the target can be estimated using Figure 2.2. It is written as, h = H r cos(θ). (2.3) Assuming the far-field approximation and r very large, the next equation can be determined from Figure 2.1, sin(α θ) = r B. (2.4) Joining (2.1) and (2.4), and rearranging terms, the look angle can be expressed as, ( ) λ φ θ = α arcsin. (2.5) 2π B Substituting (2.5) into (2.3), the topographic height can be expressed as, ( ( )) λ φ h = H r cos α arcsin. (2.6) 2π B The accuracy of the topographic height measurement depends on the estimation of the phase difference φ. A 2π phase change cannot be seen in the topographic height. The ambiguity height is the amount of height change that leads to a 2π change in the phase difference [7]. It is given by, h a = 2π h φ = λ r sin(θ) B cos(θ α). (2.7) The ambiguity height decreases when the baseline length increases. Increasing the frequency of the interferometer allows a proportional reduction of the baseline, maintaining all the other variables. The resolution in the cross-track direction is proportional to the bandwidth and high frequencies allows higher bandwidth, achieving better resolution. 7

19 2.4 Height Accuracy To estimate the topographic height accuracy, an absolute error analysis can be done [1]. The height depends on all of the independent variables in the following equations derived from the geometry shown in Figure 2.2, ( ) B 2 + r 2 (r + r) 2 ϑ = arccos 2 B r h = H r cos(γ + ϑ) r = λ φ 2π (2.8a) (2.8b) (2.8c) where the angle γ can be expressed as, π 2 α, and the angle ϑ as, π 2 θ + α. The absolute height error is the sum of the error variances, given by, where, with e 2 h = A 2 r e 2 r + A 2 B e 2 B + A 2 H e 2 H + A 2 γ e 2 γ + A 2 φ e 2 φ + A 2 λ e 2 λ (2.9) A r = h r + h θ θ sin(γ + ϑ) r = A B = h ϑ ϑ B = sin(γ + ϑ) A H = h H = 1 A γ = h γ A φ = h ϑ ϑ r r φ A λ = h ϑ ϑ r r λ ( r 2 2 B r B 2 r ) ( K ) 1 + (r+ r)2 r B 2 K = r sin(γ + ϑ) (2.10d) r+ r sin(γ + ϑ) B = K r+ r sin(γ + ϑ) B = 2π K λ 2π φ + cos(γ + ϑ) (2.10a) (2.10b) (2.10c) (2.10e) (2.10f) ( ) B2 + r K = 1 2 (r + r) 2. (2.11) 2 B r The height accuracy (or the absolute height error) decreases when the baseline length increases, similar to the ambiguity height. 8

20 CHAPTER 3 RADAR HARDWARE DESCRIPTION 3.1 Interferometer The interferometer works from baseband to Ka-band frequency. High frequencies, like Ka-band, allows to reduce the size and weight, and achieves good resolution. This frequency conversion needs two stages of downconversion and upconversion to improve the image rejection, LO isolation and noise bandwidth filtering. The frequency specifications of the interferometer are listed in Table 3.1. Table 3.1. Interferometer Specifications Parameter Signal bandwidth Ka-band frequency range L-band frequency range Baseband frequency range Ka-band LO frequency L-band LO frequency Value 200 MHz GHz MHZ MHz GHz 1300 MHz Using a dual conversion produces three different image bands. Figure 3.1 shows the signal bandwidth, the image bands and the LO frequencies. One of the objectives of the downconverter and the upconverter is to reject these image frequencies. The downconverter and the upconverter are built in printed circuit boards (PCBs) to minimize the size, weight and cost, using the software Allegro PCB Designer from Cadence [3]. The two stages of the two-channel downconverter are designed in different PCBs, the Ka-band to L-band downconverter and the L-band to baseband downconverter. A custom-built two-channel 3 GSamp/sec analog-to-digital conversion board, complemented by a Xilinx FPGA, is being designed and tested that will allow 9

21 33 GHz 34 GHz 35 GHz 36 GHz L-LO Ka-LO L-LO 1.3 GHz GHz 1.3 GHz I.1 I.2 I GHz GHz GHz GHz GHz GHz GHz GHz Figure 3.1. Image Bands in the Interferometer to use only the Ka-band to L-band section [2, 12]. The downconverter has temperature, current and power sensors connected to ADCs that can be read through a separate telemetry port on the board. These telemetry data can be processed to characterize the downconverters. The upconverter is designed in a single PCB. The L-band LO and Ka-band LO are split in the upconverter and each one is sent to the corresponding downconverter. The signal generator is the Tektronix AFG3252 that can generate the baseband inputs, q in and i in, separated by 90 and it has a maximum frequency of 240 MHz. Figure 3.2 shows the hardware blocks of the interferometer. L-band LO 1.3 GHz Ka-band LO GHz Tektronix AFG3252 Arbitrary/Function Generator BB q in MHz BB i in MHz 1.3 GHz Ka-band Dual Upconverter GHz GHz Tx Antenna Analog-to-Digital Converter ( MHz) L-band to Baseband Downconverter Ka-band to L-band Downconverter Rx(+) Antenna Rx( ) Antenna Figure 3.2. Interferometer Block Diagram 10

22 3.2 Receiver The receiver consists of the dual downconverter with two channels. Each stage is in a different PCB and both of them are highly symmetric in order to minimize the differences between the channels to increase the phase stability Ka-band to L-band Downconverter The Ka-band to L-band downconverter was previously designed at MIRSL [11, 12]. Separate boards are used for RF signals, and for telemetry and power. Figure 3.3 shows the block diagram of the RF board. The two channels, Ka-band filters and LO distribution are separated from one another by different cavities of the chassis. Z+ IN 20 db Image: GHz Signal: GHz Signal: GHz BPF GHz LSB: MHz Combiner BPF GHz 22 db LPF 1400 MHz 22 db LSB: MHz 22 db HPF 950 MHz Ka-band LO GHz Z IN 20 db 19 db GHz BPF S GHz BPF Combiner 22 db 1400 MHz LPF LPF 1400 MHz 1400 MHz LPF 22 db 22 db Z+ OUT Z OUT 950 MHz HPF Figure 3.3. Block Diagram of the Ka-band to L-band Downconverter Figure 3.4 shows a picture of the Ka-band to L-band downconverter in the chassis without the lid. The Ka-band part is down and the L-band, up. 11

23 Figure 3.4. Picture of the Ka-band to L-band Downconverter Telemetry The telemetry sensors measure the temperature, current and power of some critical components and at various points along the signal path. Table 3.2 lists all of them. Table 3.2. A-to-D Channel Assignments Channel ADC ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC1 (U73) ADC2 (U76) ADC2 (U76) ADC2 (U76) ADC2 (U76) ADC2 (U76) ADC2 (U76) ADC2 (U76) ADC2 (U76) Pin AIN0 AIN1 AIN2 AIN3 AIN4 AIN5 AIN6 AIN7 AIN0 AIN1 AIN2 AIN3 AIN4 AIN5 AIN6 AIN7 12 Description Temperature LO Amp Temperature LO Filter ( ) Temperature LO Filter (+) Temperature DC Board Power LO Current Sense LO Current Sense Rx (+) Current Sense Rx ( ) Temperature LNA (+) Temperature Mixer (+) Temperature L-band Amp (+) Temperature L-band Out (+) Temperature LNA ( ) Temperature Mixer ( ) Temperature L-band Amp ( ) Temperature L-band Out ( )

24 The output voltage of the 16 sensors is sampled by two analog-to-digital converters. This telemetry data can be read through the serial port of the downconverter. It will be used in Chapter 5 to characterize the phase difference with respect to temperature. Four resistors, two next to each channel line in the Ka-band side can be switched on or off through the serial port to increase the temperature. The effect of these resistors is shown in Section L-band to Baseband Downconverter The previous design of the L-band to baseband downconverter [15] with 100 MHz bandwidth is modified to increase the bandwidth to 200 MHz and some small errors are fixed. The downconverter is a four-layered PCB with standard FR-4 dielectric. The four layers are used for RF, ground, power and telemetry signals. Figure 3.5 shows the block diagram of the board. Z+ IN Signal: MHz Image: MHz Signal: MHz LSB: MHz USB: MHz LSB: MHz Z+ OUT 22 db BPF 1160/230 MHz 22 db 25 db LPF 260 MHz 22 db Power Sensor BPF 1300/12 MHz L-band LO 1.3 GHz 22 db S Power Sensor Power Sensor 1300/12 MHz BPF 22 db 1160/230 MHz BPF 22 db 25 db 260 MHz LPF 22 db Z IN Z OUT Figure 3.5. Block Diagram of the L-band to Baseband Downconverter 13

25 Figure 3.6 shows a picture of the L-band to baseband downconverter. The L-band part is on the left side and the baseband, on the right side. The two connectors in the middle are test points to monitor the power of the L-band signals. Figure 3.6. Picture of the L-band to Baseband Downconverter LO Filters After splitting the LO signal in order to send them to the mixers, two band-pass filters are used to improve the channel-to-channel isolation. The LO filters are 3-section ceramic filters manufactured by Lorch Microwave with part number 3DF4-1300/12-M. Multiple filters were built, the best and worst measured parameters of the filters are listed, next to the design specifications, in Table 3.3. Table 3.3. Measured Parameters of the LO Filter Parameter Design Value (best) Value (worst) Center Frequency 1300 MHz MHz MHz Lower Cutoff Frequency 1294 MHz MHz MHz Upper Cutoff Frequency 1306 MHz MHz MHz Insertion Loss (1300 MHz) 5.5 db 3.01 db 4.09 db 14

26 L-band Filters The L-band filters are 5-section ceramic filters manufactured by Lorch Microwave with part number 5DF4-1160/230-M. These band-pass filters are used to reject the image band 3 (see Figure 3.1) and to filter noise. The parameters of the filters are listed in Table 3.4, like in the previous section. Table 3.4. Measured Parameters of the L-band Filter Parameter Design Value (best) Value (worst) Center Frequency 1160 MHz MHz MHz Lower Cutoff Frequency 1045 MHz MHz MHz Upper Cutoff Frequency 1275 MHz MHz MHz Insertion Loss (1160 MHz) 1 db 0.29 db 0.53 db Rejection (930 MHz) 25 db db db Baseband Filters After the mixers, there are two low-pass filters to reject the intermodulation products and to filter noise. The baseband filters are 260 MHz bandwidth filters manufactured by Lark Engineering with part number XLMS260-4CC. Figure 3.7 shows the measured S-parameters by the manufacturer. Return Loss (S11) [db] Bandwidth Transmission (S21) [db] Frequency [MHz] Figure 3.7. Measured S-Parameters of the Baseband Filter 15

27 3.3 Transmitter The transmitter consists of the dual upconverter (i.e. two stages of frequency translation). Both stages are is in a single PCB. An external waveguide amplifier after the upconversion may be necessary to achieve the required transmitted power Ka-band Dual Upconverter The previous design of the Ka-band dual upconverter [15] with 100 MHz bandwidth and an output frequency of GHz is modified to increase the bandwidth to 200 MHz and the output frequency to GHz. The upconverter is a four-layered PCB used for RF, ground, power and telemetry signals. The top dielectric layer is RT/duroid 6002 for the RF signals and the other dielectric layers are standard FR-4. Figure 3.8 shows the block diagram of the board. BB q in MHz L-band LO 1.3 GHz L-band LO to DDC S Balun BPF 1300/12 MHz LSB: MHz USB: MHz LSB: MHz S BPF 2Way /230 MHz 19 db LSB: GHz USB: GHz USB: GHz BPF GHz 19 db BPF 19 db GHz To Antenna BB i in MHz Balun Ka-band LO GHz S Ka-Band LO to DDC Figure 3.8. Block Diagram of the Ka-band Dual Upconverter Figure 3.9 shows a picture of the Ka-band dual upconverter. The Ka-band part is on the right side and the L-band LO part, on the left side. The two baseband inputs are the connectors in the middle of the board, one up and another down. The other three connectors are unused, they were used with the original design with another baseband mixer. It is possible to revert the changes placing 0 W resistors as jumpers. 16

28 Figure 3.9. Picture of the Ka-band Dual Upconverter Baseband Mixer The image band 3 (see Figure 3.1), produced by the first upconversion, is nearly impossible to reject completely with only a filter because is too close to the desired signal. A single-sideband mixer, shown in Figure 3.10, is used instead. The only problem is the non-availability of 90 hybrids with this bandwidth at baseband. Figure Single-Sideband Mixer HMC497LP4 17

29 The HMC497LP4 from Hittite is a low noise high linearity Direct Quadrature Modulator RFIC that can be used as single-sideband mixer. And to solve the previous problem, instead of dividing the input signal internally, two signals (q in and i in ) have to be generated externally by a signal generator. The mixer will be configured to generate the lower sideband and to reject the upper sideband L-band LO Power Splitter The L-band LO power splitter is a microstrip ring hybrid with a 180 phase shift between the two output ports [6]. It is designed and simulated using the software from Agilent, Advanced Design System (ADS). Figure 3.11 shows the layout of the splitter. 3λ/4 P4 w r P2 λ/4 λ/4 λ/4 P3 P1 Figure Layout of the LO Power Splitter The next equations give a good starting point for the radius (r) in the simulations, to find the best value. υ p = c λ = υ p /f 1 µr ɛ r C = 2π r = 6 4 λ (3.1a) (3.1b) (3.1c) 18

30 The simulations are made using Agilent s Momentum which is integrated with ADS. It employs frequency-domain Method of Moments (MoM) technology to simulate passive circuits with multiple layers. This method is the appropriate for planar structures like microstrips. The parameters of the splitter, shown in Figure 3.11, are listed in Table 3.5. The substrate is RT/duroid 6002 dielectric, the same used in the PCB. Multiple simulations with different radii were made before finding the final radius. Table 3.5. LO Power Splitter Design Parameters Parameter Description Value r Radius 54 mil w Width 12 mil h Substrate thickness 10 mil ɛ r Substrate rel. permittivity 2.94 Figure 3.11 shows the simulation made with the parameters listed in Table 3.5. The marker B indicates the S-parameters at GHz, the new LO frequency. The previous LO frequency (34.59 GHz) is also indicated with the marker A. db(hybrid_33_7_b_54_0_a..s(4,1)) db(hybrid_33_7_b_54_0_a..s(3,1)) db(hybrid_33_7_b_54_0_a..s(2,1)) db(hybrid_33_7_b_54_0_a..s(1,1)) AB Frequency [GHz] S-Parameter 33.7 GHz A freq= 33.68GHz S 11 A db db(hybrid_33_7_b_54_0_a..s(1,1))= S 21 A 3.30 db(hybrid_33_7_b_54_0_a..s(2,1))= db(hybrid_33_7_b_54_0_a..s(3,1))= S 31 A 3.08 db(hybrid_33_7_b_54_0_a..s(4,1))= S 41 A B freq= 34.60GHz db(hybrid_33_7_b_54_0_a..s(1,1))= S-Parameter GHz db(hybrid_33_7_b_54_0_a..s(2,1))= S 11 B db(hybrid_33_7_b_54_0_a..s(3,1))= db(hybrid_33_7_b_54_0_a..s(4,1))= S 21 B 3.24 S 31 B 3.13 db db S 41 B Figure Simulated S-Parameters of the LO Power Splitter 19

31 Ka-band Filters The Ka-band filters are 4-section coupled line bandpass filters made with microstrip transmission lines [6]. The design and simulations are made again with ADS, following the same process used in the previous section. The function of these filters is to increase the rejection of image bands 1 and 2, and to avoid to transmit power out of the signal bandwidth. The image band 3 is too close to the desired signal to reject it, which is done at L-band. Figure 3.13 shows the layout of the filter. P1 s1 1 1 w1 2 w2 l 1 s2 2 2 l P2 Figure Layout of the Ka-band Filter The first simulation, shown in Figure 3.14, is the original filter with GHz output. The parameters are l 1 = l 2 = 52.5 mil and the others are listed in Table A B 200 db(bpf_52_5..s(2,1)) db(bpf_52_5..s(1,1)) Frequency [GHz] S-Parameter S 11 A S 21 A GHz db 2.57 db A B C freq= 34.76GHz Figure Simulated freq= S-Parameters 34.96GHz of the Ka-band Filter (52.5 freq= mil) 34.76GHz db(bpf_52_5..s(1,1))= db(bpf_52_5..s(1,1))= phase(bpf_52_5..s(1,1))=104.8 db(bpf_52_5..s(2,1))= db(bpf_52_5..s(2,1))= phase(bpf_52_5..s(2,1))= S-Parameter S 11 B S 21 B phase(bpf_52_5..s(2,1)) phase(bpf_52_5..s(1,1)) GHz db 2.77 db Fr 0 m3 m1 freq=33.74ghz Eqn Bandwidth=ba

32 New filters were designed for the new output frequency of the upconverter, GHz. But there was a mistake and the upconverter was built using the original filters. To temporally fix the problem, the filters were cut manually to adjust the frequency. These modifications were tested and the results can be seen in Section The new design of the filter is included for a future upconverter. The parameters, shown in Figure 3.13, are listed in Table 3.6. Table 3.6. Ka-band Filter Design Parameters Parameter Description Value l 1 Length 1 51 mil l 2 Length 2 51 mil w 1 Width 1 16 mil w 2 Width 2 17 mil s 1 Separation 1 6 mil s 2 Separation 2 26 mil h Substrate thickness 10 mil ɛ r Substrate rel. permittivity 2.94 The center frequency is directly related to the length of the microstrip lines, f 1 l [14]. Knowing this, multiple simulations were made changing the lengths before finding the final l 1 and l 2. Figure 3.15 shows the simulated S-parameters of the signal bandwidth and the image band 2. db(bpf_51_0..s(2,1)) db(bpf_51_0..s(1,1)) C D A B S-Parameter S 11 A S 21 A S-Parameter S 11 B S 21 B S-Parameter S 11 D S 21 D GHz db 2.62 db dB Frequency [GHz] F A B freq= 35.66GHz Figure Simulated freq= S-Parameters 35.86GHz of the Ka-band Filter (51E mil) db(bpf_51_0..s(1,1))= db(bpf_51_0..s(1,1))= freq= 35.66GHz db(bpf_51_0..s(2,1))= db(bpf_51_0..s(2,1))= phase(bpf_51_0..s(1,1))=104. phase(bpf_51_0..s(2,1))=-31.5 C D freq= 33.34GHz freq= 33.54GHz 21 db(bpf_51_0..s(1,1))= db(bpf_51_0..s(1,1))= db(bpf_51_0..s(2,1))= db(bpf_51_0..s(2,1))= phase(bpf_51_0..s(2,1)) phase(bpf_51_0..s(1,1)) GHz 2.68 db GHz 3.08 db db

33 CHAPTER 4 RADAR HARDWARE EVALUATION 4.1 Ka-band to L-band Downconverter The Ka-band to L-band downconverter was originally designed to work with an input frequency of GHz and a LO of GHz. With these frequencies, the output was MHz. To work with the new designed Ka-band dual upconverter and L-band to baseband downconverter, the LO frequency should be GHz to obtain an output of MHz, a 140 MHz change in the LO frequency. The gain performance of the downconverter with the new frequencies is compared with the original design to check that is possible to change the LO frequency without problems. The phase performance is evaluated in the next chapter Gain Measurement The gain (G) is defined as the ratio of the signal output power to the signal input power in the same channel. It is measured following the next equation, G [db] = P out [dbm] P in [dbm]. (4.1) The measurements are made injecting different tones with the Agilent E8257D Signal Generator at a known power (P in ) and reading the output power (P out ) with the Advantest U3772 Spectrum Analyzer, for each channel. The first measurement compares the gain with different LO frequencies, maintaining the same input, to see the bandwidth of all of the filters affected by the LO. Figure

34 shows the results. The gain at 1300 MHz (f LO = GHz) and at 1160 MHz (f LO = GHz), with an input frequency of GHz, is the same at each channel. However, the gain of channel ( ) is 8 db lower than in channel (+). This difference was likely caused by an unanticipated cut on the transmission line between the LO splitter and the mixer of channel ( ). This problem was fixed wire-bonding over it. Before fixing it, the difference was bigger than 20 db, but maybe there is still some loss and the mixer is not receiving enough LO power Gain [db] flo = GHz 20 Channel (+) Channel ( ) Frequency [MHz] flo = GHz Figure 4.1. Ka-band to L-band Downconverter Gain with different f LO After checking that a LO frequency of GHz was functional, the next step is to change the input frequency of the signals to be downconverted. Figure 4.2 shows the gain over the bandwidth of the downconverter. The gain variation of each channel is approximately 5 db. This variation was also observed with the design LO frequency (34.45 GHz) and it may be caused by some of the filters. Comparing these results with Figure 4.1 suggests that the problem is not in the low-pass filters or high-pass filters at L-band. It looks like that there is some error in the bandwidth of the band-pass filters at Ka-band. 23

35 70 65 Channel (+) Channel ( ) 60 Gain [db] Bandwidth Frequency [MHz] Figure 4.2. Ka-band to L-band Downconverter Gain (f LO = GHz) 4.2 L-band to Baseband Downconverter The new design of the L-band to baseband downconverter uses a LO frequency of 1300 MHz to produce an output of MHz with the input, from the Ka-band to L-band downconverter, of MHz Gain Measurement The gain of this downconverter is measured using the same method explained in the previous section, but with the new frequencies. Figure 4.3 shows the gain of each channel of the downconverter. The gain variation of channel (+) is < 2 db, but in channel (+) is < 5 db. This difference produces that in some parts of the signal bandwidth, the gain of channel ( ) is 9 db lower than in the other channel. A temporary way to fix this problem and the same problem in the other downconverter is connecting the channel (+) of one downconverter to the channel ( ) of the other and vice versa, to compensate both errors. 24

36 50 45 Channel (+) Channel ( ) Gain [db] Bandwidth Frequency [MHz] Figure 4.3. L-band to Baseband Downconverter Gain Image Rejection Ratio The image rejection ratio (IMRR) is defined as the ratio of the signal output power produced by the desired input frequency to that produced by the image frequency. The input power used in both measures should be exactly the same. It is measured following the next equation, IMRR [db] = P signal [dbm] P image [dbm]. (4.2) The image is attenuated before the mixer with a band-pass filter centered at 1160 MHz with 200 MHz bandwidth. But this image at MHz is too close to the desired signal, so it is nearly impossible to obtain a good rejection of the lower frequencies. Figure 4.4 shows the IMRR of each channel following the shape of the filter. At 40 MHz there is only a rejection of 6 9 db, but at higher frequencies the filter is more effective and it improves to db. The difference between the channels can be explained by the gain difference between channels seen in Figure

37 Image Rejection Ratio [db] Channel (+) Channel ( ) Bandwidth Frequency [MHz] Figure 4.4. L-band to Baseband Downconverter Image Rejection Channel-to-Channel Isolation The channel-to-channel isolation is defined as the ratio of the signal output power produced by the signal input power in the same channel to that produced by the signal input power in the opposite channel. The input power used in both measures should be exactly the same. It is measured following the next equation, Isolation 1 2 [db] = P 2 2 [dbm] P 1 2 [dbm]. (4.3) This measurement is important for an interferometer because the interference between channels can produce errors during the phase estimation. Figure 4.5 shows a isolation between channel (+) and channel ( ) from 60 db to 74 db. This experiment was made without the enclosure. The difference between the two measurements can be explained by the gain difference between channels. The main path between the channels is the LO network, because connects directly with both channels. To reduce the inter-channel crosstalk, there are two narrow band-pass filters centered at the LO frequency (1300 MHz) that reject all the other 26

38 Channel-to-Channel Isolation [db] Ch( )-to-ch(+) Ch(+)-to-Ch( ) Bandwidth Frequency [MHz] Figure 4.5. L-band to Baseband Downconverter Channel Isolation frequencies. One way to increase the isolation is to put the PCB inside an enclosure that physically separates the different sections of the downconverter. Another way is to use metallic shields around the mixers and the DC power components to reduce the spurious radiation Phase Performance The phase difference between the two channels and its stability over time is the most important measure of an interferometer. It is measured in a similar way as described in Section 5.2 of the next chapter. Figure 4.6 shows the phase difference between the two channels with an input frequency of 1060 MHz and an output frequency of 240 MHz, for more than 60 minutes. It also shows the residual power, the error of the phase estimation algorithm (see Section 5.1). Around minutes after switching on the L-band to baseband downconverter, the residual power and the temperature of the board is stable, and then the phase is stable too. 27

39 Phase difference [deg] Residual power Time [minutes] Figure 4.6. L-band to Baseband Downconverter Phase Difference (f = 240 MHz) The horizontal green line in Figure 4.6 represents the measured mean phase difference and the red lines represent the measured standard deviation (σ φ ), from the minute 25 when the phase is stable. In this case (f out = 240 MHz), the standard deviation is 33 mdeg. The same measurement was made for an output frequency of 140 MHz (f in = 1160 MHz), the standard deviation is 38 mdeg, and 40 MHz (f in = 1260 MHz), the standard deviation is 43 mdeg. 4.3 Ka-band Dual Upconverter The new design of the Ka-band dual upconverter transmits a signal with a frequency of GHz, produced with two input signals of MHz and a separation of 90. It needs two LO signals, 1.3 GHz and GHz. 28

40 4.3.1 Output Power The objective of the first measurement is to see the bandwidth of the two band-pass filters at Ka-band. It is made changing the LO frequency and measuring the output power with the Advantest U3772 Spectrum Analyzer, maintaining the same input. At this frequency, there are some losses in the cables and the spectrum analyzer. To correct them, these losses were measured before with a power meter. 5 0 Output Power [dbm] Image 1 Image 2 Signal Image Frequency [GHz] Figure 4.7. Ka-band Dual Upconverter Output Power with different f LO Figure 4.7 shows the output power over the signal bandwidth and the images bandwidth. But more important, it also shows the shape of the band-pass filters. The images 2 and 3 are attenuated more than 15 db compared with the signal. The power of the signal is 3 dbm, but the design was meant to be db higher and the bandwidth of the filters to be narrower. As explained in Section , there was an error and the microstrip filters were longer, changing the center frequency. These filters were cut manually to adjust the frequency, but this increased the attenuation in the pass band and increased the bandwidth too. 29

41 After seeing that the signal was within the bandwidth of the filters, another measurement was made to check the output power of the signal and the three images produced by the dual upconversion. These images can be seen in Figure Output Power [dbm] Image 1 Image 2 Signal Image Frequency [GHz] Figure 4.8. Ka-band Dual Upconverter Output Power (f LO = GHz) Figure 4.8 shows the output power of the signal and the images that are transmitted too. The variation over the bandwidth is approximately 2 db and the output power of the images is 35 db below the power of the signal. All the measurements of the upconverter are made using the Tektronix AFG3252 Function Generator to produce the i in and q in separated by 90. They have to be generated externally due to the non-availability of 90 hybrids at baseband with 200 MHz bandwidth. With these two signals, the single-sideband mixer at baseband attenuates the image (upper sideband, image 3), that it is too close from the signal to be completely filtered later. In some cases it can be difficult to produce these signals and it would be easier to use the same signal in both inputs. Figure 4.9 compares the output power of the signal and the image 3 in both cases with a baseband frequency of 40 MHz, the worst case. It shows that the output power is 4 db lower. 30

42 5 0 5 i in (0 ) and q in ( 90 ) i in (0 ) and q in (0 ) Output Power [dbm] Signal Image 3 Signal Image Frequency [GHz] Frequency [GHz] Figure 4.9. Ka-band Dual Upconverter Output Power (f in = 40 MHz) When the same input signal is used, the only way to reject the image 3 is with the band-pass filter, centered at 1160 MHz and 200 MHz bandwidth, after the first upconversion. Note the band-pass performance for this image reject filter is shown in Figure 4.4. As it can be seen in Figure 4.9, the 40 MHz image is only attenuated by 10 db and there are some harmonics within the desired bandwidth. 31

43 CHAPTER 5 PHASE CHARACTERIZATION The phase performance of the Ka-band to L-band downconverter has to be measured very accurately. To achieve the required accuracy a maximum-likelihood algorithm was previously developed and implemented to measure the phase [2, 9]. This algorithm is used to measure how the temperature affects the phase difference between the two channels. 5.1 Phase Estimation Algorithm The phase estimation algorithm is based on a time-domain maximum-likelihood estimator. It works with a single-tone signal that has been sampled using a digital oscilloscope. The sampled waveform is compared to the estimated frequency and phase. The accuracy of these values depends on the number of samples of the input signal and its SNR. Increasing either of them will improve the accuracy. The vector (x) of N samples is the measured voltage of the input signal and should cover multiple cycles. The steps are explained below. (A) Determine the mean of the signal The DC offset is determined by the arithmetic mean ( x) of the sampled signal x following the next equation, x = 1 N N x i. (5.1) i=1 32

44 (B) Estimate the signal amplitude The amplitude (A) of the sampled signal x is estimated following the next equation, A = 2 1 N N (x i x) 2. (5.2) i=1 The amplitude-normalized signal (xn) is calculated to estimate the phase and to determine the residuals. xn = x x A. (5.3) (C) Estimate the signal phase The signal phase estimation is the main part of the algorithm. As explained before, is based on a time-domain maximum-likelihood estimator that minimizes the sum of squared residuals. The residuals are the difference between the estimated values and the sampled values. It is explained in detail in [9]. (D) Correct the signal frequency The error in the phase estimation is linearly correlated to the frequency error. To reduce this error, the sampled vector is divided into blocks of equal length and the phase of each block is estimated with the previous algorithm. A new corrected frequency is calculated using the estimated phase of the blocks. This process is explained in [2]. (E) Repeat steps C and D Step C is repeated with the corrected frequency obtained in D. To improve the accuracy of the estimated phase, this iteration can be repeated several times. Once obtained the phase and the frequency, the estimated waveform (x est ) can be written as, x est = sin(2π f est t + φ est ). (5.4) 33

45 (F) Determine the residuals The residuals (r) are the difference between the estimated waveform and the sampled signal, r = x est xn. (5.5) The residual power can be considered the error of the estimation and it is determined by the next equation, Residual Power = 1 N N (r i r) 2 (5.6) i=1 where r is the arithmetic mean of the residuals r. 5.2 Experimental Setup The experimental setup for measuring the phase difference between the two channels of the Ka-band to L-band downconverter is shown in Figure GHz Agilent E8257D 20 GHz GHz Agilent E8257D 40 GHz HP 83554A Freq. x2 Narda DB R serie Mini-Circuits ZFM-2000 S Ka-DDC S Spin Stamp Microcontroller Mini-Circuits SLP GHz HP 8648B serie ethernet Agilent MSO6104A Oscilloscope Mini-Circuits ZFM-2000 Mini-Circuits SLP-70 Figure 5.1. Experimental Setup for measuring the Phase Stability The Ka-band input signal (35.75 GHz) is split by a waveguide magic-t in the two channels needed by the downconverter. The Ka-band LO signal (34.45 GHz) has 34

46 to pass through a frequency doubler and a power amplifier because the maximum frequency of the signal generator is 20 GHz. The L-band output signals (1.3 GHz) should be sampled by a digital oscilloscope, but its maximum frequency is 1 GHz. Then the output signals are downconverted to 40 MHz using a symmetric circuit (a mixer and a low-pass filter) for both channels, before sampling them. The sampled values are sent to a computer using an ethernet connection. The temperatures of the different sensors of the downconverter are read by a microcontroller and sent to the same computer through a serial cable Microcontroller The Spin Stamp microcontroller uses a Propeller processor that can be programmed in Spin, a high-level language. It is connected to the downconverter and the computer with serial connections. The main function is to read the temperature sensors of the downconverter and convert the digital values to millivolts. These values are sent to the computer to analyze them. The microcontroller can also switch on and off the four resistors next to the Ka-band lines to change the temperature. The complete code is listed in Appendix B MATLAB Program The computer runs a MATLAB program that has several functions. First, it receives the sampled data over 100 µs from the digital oscilloscope with a 4 GHz sampling rate, and the temperatures in millivolts from the microcontroller. The next step is convert the temperatures to degrees Celsius ( C) and use the algorithm, explained in Section 5.1 and implemented in MATLAB code, to estimate the phase of each channel. Finally, it calculates the phase difference and presents the results in some plots in real time. The speed of the program depends mostly in the phase estimation algorithm and the computer, here there is an iteration every seconds. 35

47 To switch on and off the two resistors, the program can send different commands to the microcontroller that will send them to the downconverter. The code of the program called phase temperature.m is listed in Appendix A.1. All data is stored in the computer and can be analyzed later with the code in Appendix A Phase Stability Phase stability is a very important characteristic in a downconverter that is going to be used in an interferometer. The phase should change only when there is some change in the phase of the two input signals. However, there are other factors affecting the phase. This section evaluates how the temperature of the downconverter affects the phase. All tests are done in two different versions of the downconverter, one from October 2011 and another from May Both boards were built using the same design files, so the results should be very similar. The two temperature sensors chosen for most plots are the sensor 8, LNA (+), and the sensor 12, LNA ( ). They were chosen because are the closer sensors to the Ka-band lines and to the controlled resistors in each channel October 2011 Version During the first measurement, the downconverter was left open to the ambient temperature of the laboratory. Figure 5.2 shows the phase difference between the two channels as a function of time and the temperature of one sensor of each channel. A periodic variation can be seen in the phase difference with a cycle of about 20 minutes, but not in the temperature sensors inside the board next to the signal paths. These variations are quite similar to those seen in [15] with the previous downconverter with 100 MHz bandwidth. They were produced by the cyclical air conditioning system 36

48 in the laboratory. To prove this hypothesis, the downconverter was placed inside a thermally isolated environment in the next test. Phase diff. [deg] Temperature [ C] Channel (+) Channel ( ) Time [minutes] Figure 5.2. Phase Difference in the Ambient Temperature of the Laboratory As seen in Figure 5.1, the downconverter and the splitter (a waveguide magic-t) are placed in a expanded polystyrene box to isolate them from the temperature changes of the laboratory. Figure 5.3 shows the phase difference without the previous sinusoidal variations once the temperature is stable. However, the final temperature was higher due to the dissipated heat by the components of the downconverter. The standard deviation of the phase difference (σ φ ) is 23 mdeg, represented by the horizontal red lines. The green line represents the measured mean. This suggests that if there is a temperature control, the phase difference can be very stable. The next step is to characterize the phase difference between the two channels versus the temperature. The easiest way to increase the temperature inside the box was to use the dissipated heat by the downconverter. The measured temperature used 37

49 Phase diff. [deg] Temperature [ C] x 10 Residual power Time [minutes] Figure 5.3. Phase Difference in a Thermally Isolated Box to make the plot is the arithmetic mean of all sensors connected to the ADC2 (see Table 3.2 and Figure 5.4) because they are the eight sensors in the RF board of the downconverter along the paths of both channels. Figure 5.4. Temperature Sensors (ADC 2) and Temperature Control Resistors 38