Memo 144. Final Report of Eleven Feed Project: Development of Broadband Cryogenic Frontend Prototype for the SKA

Transcription

1 Memo 144 Final Report of Eleven Feed Project: Development of Broadband Cryogenic Frontend Prototype for the SKA M. Pantaleev J. Yin M. Ivashina J. Conway May

2 Final report Eleven Feed project: development of broadband cryogenic frontend prototype for the SKA By Miroslav Pantaleev, Jungang Yin, Marianna Ivashina and John Conway (with inputs from Per Simon Kildal, Jian Yang, Leif Helldner, Hasan Raza; Oleg Iupikov, Wan Chun Liao, Benjamin Klein and Arnold van Ardenne) Table of Contents 1 CONTRACTUAL INFORMATION OVERVIEW Progress and Achievements Work Packages DETAILED REPORT Antenna performance Folded dipole feed Circular Eleven Feed Circular Eleven feed optimised for the 1.2 4GHz band version without dielectric support Circular Eleven feed optimised for the 1.2 4GHz band version on PCB Receiver integration in Cryostat and Tsys measurements Noise modelling Expected performance in reflector systems Comments on frequency variations in Eleven feed response CONCLUSIONS REFERENCES Page 1 of 24

3 1 Contractual information The work reported in this document has been carried out as part of EC FP 7 funded project: A Preparatory Phase project for the Square Kilometre Array. Grant agreement no.: The goal of the project was to develop a broad band receiver system for the SKA as described in the project proposal submitted to SPDO in December The work was carried out as a collaboration between Onsala Space Observatory and the Antenna Group at the Department of Signals and Systems at Chalmers University. The people who have worked on the project are: Dr. Miroslav Pantaleev (project leadership and system tests), Jungang Yin (electrical design and test measurements), Leif Helldner (mechanical design and assembly), and Dr. Marianna Ivashina (modelling of the antenna performance in reflector systems and system nose temperature modelling), all from the Onsala Space Observatory (OSO), Department of Earth and Space Sciences, Chalmers. OSO also provided the fabrication of mechanical components (workshop at OSO) and a cryogenic technician (Christer Hermansson). Prof. Per Simon Kildal, Dr. Jian Yang (electrical design of the feed and measurements of the feed patterns and S parameters), and Hasan Raza (design of a passive balun network), Wan Chun Liao, a Master student (modelling of the antenna performance in reflector systems) all from the Antenna group, Department of Signals and Systems, Chalmers. Benjamin Klein, a PhD student at Hartebeesthoek Radio Astronomy Observatory (HartRAO), South Africa (the system noise temperature modelling). 2 Overview In section 2.1 we summarise the project achievements. In section 2.2 we describe the completion of the work packages. Technical details are given in section 3 and conclusions in Section Progress and Achievements During the project we have: 1) Measured beam pattern of the previously developed Folded dipole Eleven feed model (see Figure 1a) 2) Designed, fabricated and measured a Circular Eleven feed covering for the GHz band (see Figure 1b). 3) Optimised the design of the above Circular feed for the 1.2 4GHz band. The optimisation includes also a novel feeding structure and the study of all metal petals (see Figure 1c). 4) Designed, fabricated and measured the reflection coefficient and beam parameters of a novel feeding structure based on passive balun. 5) Measured the Tsys performance of a cryogenically cooled version of an older geodetic version of the Eleven feed covering 2 14GHz (version not shown on Figure 1) integrated in a cryostat with LNAs from different suppliers. The measured receiver noise temperature was about 20K over the 3 10GHz band (see Figure 14). 6) A previously developed noise modelling tool was further improved and showed very good agreement with measured receiver noise temperature. 7) A study was done to compare the performance of the Eleven feed when used in different reflector systems Page 2 of 24

4 (a) (b) (c) diam. 280mm, height 68mm diam. 210mm, height 68mm diam. 290mm, height 100 mm GHz GHz 1.2 4GHz Figure 1: Three versions of the Eleven feed developed during this project. (a) A version with straight dipoles with the outer dipoles folded down. (b) A circularised version covering GHz and (c) a circularised design optimised for the 1.2GHz 4GHz range. The first two versions have been constructed and have had their EM properties measured experimentally while the last version has had its performance simulated. In addition to these three versions an older geodetic VLBI2010 version (not shown) with straight unbent dipoles covering 2GHz 14GHz has been used for making cryogenic Tsys measurements. 2.2 Work Packages ID Uppgiftsnamn dec jan feb mar apr maj jun jul aug sep okt nov dec jan 2012 feb mar 1 WP1 Solve the stability problem of the present measurement setup 2 WP2 A Integrate and perform tests on VLBI2010 Eleven feed model with different LNA alternatives. 3 WP2 B Integrate and perform tests on SKA eleven feed model with different LNA alternatives. 4 WP3 Complete noise models 5 WP4 Set a desired goal for further system optimization 6 WP5 Feed optimization 7 WP6 Manufacture and measure the optimized feed 8 WP7 Report results of the study Table1: Gantt chart of modified project plan Table 1 gives an overview of the activities on the work packages according to the proposal submitted to SPDO with one correction, reported in the report for the second quarter, where WP2 was divided into two sub tasks: WP2A measurements of the VLBI2010 Eleven feed model and WP2B measurements of the PrepSKA (1.2 10GHz) Eleven feed model. WP1 This WP included improvements to the cryostat to increase the holding time and repair leaky O rings and also the repair of the harness of the LNA bias and temperature sensors. All the work was done during January and February 2011 and after that the cryostat was successfully used for tests of the Eleven feed. Later in August 2011 there were problems with the cold head which Page 3 of 24

5 were successfully resolved. The work on this WP1 is accomplished and details can be found in the report for the first Quarter of WP2 The work on this WP included Tsys tests of the 2 12GHz feed with three types of LNAs: 4 8GHz LNAs from GARD, differential 1 12GHZ LNAs from Caltech and 50ohm 1 12GHz LNAs from Caltech. As reported previously, two of the differential LNAs were not operational and they were sent for repair in March 2011, however these are still not repaired. Details on this work package are presented in the reports for Quarters 2 and 4. In the final report we give an overview in section 3.2. Due to the delay of the repair of the differential LNAs work package 2B has not been fully accomplished. Details on the 2 12GHz VLBI2010 Eleven feed can be found in [1]. Details on the design of the first SKA folded dipole version (see Fig 1a) covering GHz can be found in [2] and [3]. WP3 The noise model gives an important tool to predict the overall noise performance and the expected Aeff/Tsys taking in to account all system parameters. The work on this WP was successfully accomplished and used to estimate the measured receiver noise in WP2. Details are given in the reports for Quarter 2 and 3. Section 3.2 in the current report gives a summary on the achievement. The work on this WP resulted in the paper[4]. WP4 The goals for further optimisation were defined in the report for Quarter 2 as improvement of the reflection coefficient, aperture efficiency, including BOR 1 and efficiency for the 1.2 4GHz band. WP5 For further optimisation we developed a new Eleven feed geometry in which the feed contains circularly curved folded dipoles on a flat microwave substrate. Following the goals as defined in WP4, the work on this work package was accomplished in two steps: the first step was to optimise the performance over the GHz band with the purpose of keeping a good reference to the conventional design which used straight folded dipoles. The design of this circularized feed is described in the reports for Quarters 3 and 4. The second step was to improve the performance over the 1.2 4GHz band. This work was done during January March 2012 and details are presented in section and of this report. WP6 Test measurements of the EM properties of Circular eleven feed (1.2 14GHz version) were made during the first quarter of Details are given in section WP7 The work on WP7 is accomplished with the current report. 3 Detailed Report Here we present detailed technical information related to the development done during the project with relation to the corresponding work packages Page 4 of 24

6 3.1 Antenna performance Here we present a summary of Eleven Feed electrical design and testing, including measurements of reflection coefficient, antenna beam and efficiencies. This section is related to WP4, WP5 and WP Folded dipole feed The design of an GHz Eleven feed (see Figure 1a) was started in 2010 as a collaboration project between Caltech, the Chalmers Antenna group and OSO where the Antenna group was responsible for designing the feed, OSO was involved with the mechanical design and prototype manufacturing and in exchange, Caltech were going to deliver four Ultra Wide Band cryogenic differential LNAs. The design of the GHz Eleven feed is based on folding down the outermost elements of the array thus creating a new compact Eleven antenna with only 40% of the volume of the original standard Eleven antenna for the same frequency band [1]. a) (b) Figure 2: (a) Measured and simulated reflection coefficient and (b) efficiencies from measured far field functions for folded dipole Eleven feed. As seen in Figure 2a the measured reflection coefficient is below 10dB over the 1.2 2GHz and in the 3 4GHz is even below 14dB. The aperture efficiency is about 1.5dB (see Figure 2 right) but drops to 3dB at 1.2GHz. The directivity is 11dBi across the band, but drops to 9dBi at 1.2GHz (for details on measured beam patterns and directivity, refer to the report for the second quarter) Circular Eleven Feed In this section we report the measurement results for the circularised feed covering the wide band GHz (see Figure 1b). In this version the petals are manufactured via standard PCB technology on TMM3 microwave substrate. As described in the previous report (and as can be seen from the dimensions in Figure 1) the circularised Eleven Feed is a very compact structure. Compared to the configuration with folded dipoles for the same band, the feed diameter is 70mm smaller. Figure 3 below shows the simulated and measured reflection coefficient for this feed design. The measured S11 degrades for the lowest part of the band with respect to the simulations. So far we do not have explanation for discrepancy Page 5 of 24

7 0 Reflection Coefficient (db) Simulated Measured Frequency (GHz) Figure 3: Simulated and measured reflection coefficient for the circularised Eleven feed covering GHz (i.e before optimisation for the 1.2GHz 4 GHz band). The measured beam patterns are given in Figure 4. As it can be seen the beam width is almost constant across the band. 0 1GHz 1.3GHz 2GHz GHz 4GHz 5GHz -30 Amplitude (db) GHz 9GHz 7GHz 10GHz 8 11GHz GHz 13GHz 14GHz Figure 4: Measured BOR1 patterns for unoptimised circular Eleven feed (b) Figure 5 shows the aperture and spill over efficiencies calculated from measured beam patterns assuming a reflector with F/D=0.42 corresponding to half subtended angle of 60 degrees. The aperture efficiency is about 1.5dB (60 70%) and it is nearly constant across the band. In Page 6 of 24

8 comparison with the Folded dipole model the Circular eleven feed has much better efficiencies (compare Figure 2b and Figure 5a) Efficiencies (db) e sp e ap Frequency (GHz) (a) (b) Figure 5: (a) Aperture efficiency and its individual contributions [%] calculated from the measured beam patterns of the circularised Eleven feed (see Fig 1b) when placed in a reflector with F/D=0.42 corresponding to half subtended angle of 60 degrees (b) Aperture and spillover efficiencies [db] Circular Eleven feed optimised for the 1.2 4GHz band version without dielectric support This section presents details on the optimisation of the circularised feed for the 1.2 4GHz band. This work was done during February March 2012 and was based on the experience gained from the design of the circularised Eleven feed for the GHz band. With reference to WP4 in which we define the optimisation goals we considered that the efficiency achieved by the circularised Eleven feed was already very good, so what was needed was improvement of the reflection coefficient. We used as inputs for the optimisation process the measurement data for the Circular Eleven feed. We decided to investigate also the possibility to decrease the number of ports and study different descrambling alternatives. We were also interested to make thick metal dipoles and omit the dielectric substrate aiming a cheaper manufacturing cost with the view of possible future mass production for the SKA. In the beginning, a circular Eleven feed made by pure metal material was planned to be built. This would be a different construction from the previous 1.2GHz 14GHz circularised Eleven feed whose petals were constructed via printed circuit boards. This all metal solution was meant to be lower loss and eliminate potential concerns related to different expansion coefficients of metal and dielectric materials when the feed is cooled down. An all metal dual polarized circular Eleven feed was first designed and optimised. The final design consisted of 12 folded dipoles on each petal, and the width of the innermost dipole strips was about 0.6 mm. However, the thickness of the dipole strips on antenna petals was found to be critical in this model for operation over GHz, as verified by a study of two designs differing only in the dipole thickness. As can be seen from Figure 6, the matching is very good when the dipole thickness is 0.10 mm; however, the increase of thickness from 0.10 only to 0.25mm severely degrades the matching at high and low frequencies. As estimated via analysis of the mechanical stiffness of such feed petals construction, the metal thickness should be at least around 0.5 mm to achieve the necessary structural strength in room temperature as well as to lower the risk of structural deformation after cryogenic cooling. We concluded it was far Page 7 of 24

9 too difficult to realise the all metal solution over GHz, and so we decided to abandon the all metal solution for this frequency band. Reflection coefficient [db] petal thickness 0.10 mm petal thickness 0.25 mm Frequency [GHz] Figure 6: CST simulated reflection coefficient of a GHz all metal circular Eleven feed with two different dipole strips thickness. Nevertheless, we consider the all metal solution as feasible for lower frequency bands, e.g GHz. If the GHz all metal model (0.1 mm dipole thickness) is scaled by a factor of 4, then the dipole thickness becomes 0.4 mm and can be manufactured; in addition, it is possible to build somewhat thicker far out dipoles to achieve even better strength of structure. Future numerical studies would to investigate the feed performance at these lower frequencies would be interesting Circular Eleven feed optimised for the 1.2 4GHz band version on PCB We turned to next consider a PCB solution since we choose to abandon the all metal design. This solution is based on previous designs of circular Eleven feed [1] [4] made by using Rogers TMM3 printed circuit boards (0.07 mm copper layer on mm TMM3 substrate), in particular, the manufactured GHz circular Eleven feed consisting of 17 folded dipoles on each petal. This prototype is log periodically extended from a small Eleven model intended for Genetic Algorithm (GA) optimization, which consists of 8 folded dipoles on each petal, making it doable to perform GA optimization linked with CST on our computer server. It took about one hour to finish one single simulation run in CST, and the good chromosomes were derived after hundreds of simulation runs when 5 generations were used to achieve the convergence [8] [9]. The GHz prototype had pretty good reflection coefficient between 1.7 and 12.2 GHz according to the measurement results. In order to optimise to GHz, we tried in the first place to remove 4 innermost folded dipoles while adding 2 extra far out folded dipoles by log periodic scaling. The scale factor k is the same as before, about 1.2 on the same mm TMM3 boards, ending up with 15 folded dipoles on each petal (the diameter of the wall cylinder becomes about 290 mm, see Figure 1). The corresponding frequency range can be estimated as follows: f lo = 1.7/(1.2) 2 = 1.18 GHz f hi = 12.2/(1.2) 4 = 5.88 GHz The high end of the range, e.g GHz is greater than 4.0 GHz, because we want to have some margin for adding extra matching lines, coupled lines and coaxial cables, which will affect the performance particularly at high frequencies Page 8 of 24

10 Different from previous designs using a symmetrical 8 port center puck, we decide to apply different feeding schemes for X and Y polarizations, respectively using approach similar to previous design described in [5]. By doing so, the numbers of ports and LNAs will be decreased so as to lower the cost of the whole feed system. Polarization-Y Case: Figure 7: The feeding scheme for polarization Y of the circular Eleven feed model; commercial 50 Ohm mini coaxial cable pairs are used. First, we simulate the model with only the antenna petals fed by 2 discrete ports (ref: 400 Ohm) for one polarization (pol Y); we call it Ant only model. Then, as shown in Figure 7, we build exponentially tapered matching lines on the tilted circuit boards, connected by microstrip coupled lines (Teflon PTFE support); 2 commercial 50 Ohm mini coaxial cables (0.86 mm outer diameter) are chosen as feed in the center of the microstrip coupled lines. We call it Ant + feeding (1) model. Third, in order to improve the matching at low frequencies, we add one more extended outermost folded dipole while removing one innermost folded dipole on each petal (still 15 folded dipoles on each petal, but the diameter of wall cylinder increases to about 350 mm); in the meanwhile, we make TMM3 boards for the longest 3 folded dipoles as thick as mm so as to enhance the structural strength. We call it Ant + feeding (2) model. CAD 3D model of the design is shown in Figure 1c. Presented in Figure 8 are the simulated S11 results (at ports for pol Y) of the above mentioned three models. As can be seen, the S11 of the Ant + feeding (2) model is largely better than 10 db over GHz, but for a couple of narrow spikes below 2 GHz. Adding one more extended folded dipole is expected to improve the matching at low frequencies, but at the cost of an unsatisfactorily large size. An alternative to make the feed compact is to bend the other most dipoles the same way as it was done for the GHz feed (see Figure 1, a). Due to time limitation we did not investigate this alternative, but we consider that the feed diameter can be made considerably smaller using this approach Page 9 of 24

11 S11 Magnitude [db] Ant+feeding (2) Ant+feeding (1) Ant only Frequency [GHz] Figure 8: CST simulated reflection coefficients (at ports for pol Y) of circular Eleven feed design optimised for GHz operation (see Fig 1c). The radiation performance is shown in Figure 9 in terms of aperture efficiency calculated from simulated beam patterns when the feed is placed at the prime focus of an axisymmetric parabolic dish (120 o subtended angle) also shown are 5 factorized sub efficiencies, which are related as: e AP = e BOR1 + e SP + e ILL + e PHASE + e POL (in db) It is observed that the BOR1 as well as the spillover efficiency is better than 0.5 db (90%) over GHz, while the aperture efficiency is better than 2.0 db (63%) over most of the band. Efficiencies [db] e BOR1 e AP e ILL e PHS e POL e SP Frequency [GHz] Figure 9: Aperture efficiency and sub efficiencies (for pol Y) of an axisymmetric paraboloidal reflector (120 o subtended angle) fed by Ant + feeding (2) model for circularised Eleven feed optimised for 1.2 4GHz the calculated (based on simulation of the feed model as shown on Figure 1c) Page 10 of 24

12 Polarization-X Case: (a) (b) Figure 10: Two feeding schemes (power divider (left), descrambling (right)) for polarization X of the circular Eleven feed model; commercial 50 Ohm coaxial cable pairs are used. As shown in Figure 10a, a PCB (RT 5880 with mm thickness; copper with mm thickness) 2 port feeding puck is mounted on the backside of the ground plane, where the Y shaped (by linear tapering) power dividers are printed on top followed by commercial 5.7 mm coaxial cables. The antenna petals and the feeding puck are connected by vertical twin lines that go through holes on the ground plane. A second feeding scheme is illustrated in Figure 10b, where a 4 port descrambling board (same PCB board as the former) is used, similar to previous design of 8 port feeding center puck. The 300 Ohm ports of coupled twin lines can be transformed to the 50 Ohm ports of the uncoupled coaxial cables with the help of linearly tapered microstrips power divider descrambling S11 Magnitude [db] Frequency [GHz] Figure 11: CST simulated reflection coefficients (at ports for pol X) of circularised eleven feed optimised for 1.2 4GHz (see Fig 1c) using different feeding schemes. The simulated reflection coefficients are shown in Figure 11. The S11 magnitude is better than 8 db over GHz in both cases, which is quite similar to that of previous circular Eleven models fed by an 8 port center puck. The radiation performance is presented in Figure 12 in terms of efficiencies with the same assumptions as above. Although different feeding schemes are applied for two polarisations, this only makes a slight difference to the radiation patterns of these two polarisations. It can be found, similar to the Pol Y case, that when mounted in an axisymmetric dish the BOR1 and spillover efficiency (except for a drop at 1.2 GHz, this needs to be Page 11 of 24

13 confirmed by further investigation) are better than 0.5 db (90%) over GHz while the aperture efficiency is better than 2.0 db (63%) over most of the band. Efficiencies [db] e BOR1 e AP e ILL e PHS e POL e SP Frequency [GHz] Figure 12 simulated expected aperture efficiency and sub efficiencies (for pol X) for circularised 11 feed optimised for 1.2 4GHz (see Figure 1c) when placed in a paraboloidal reflector (120 o subtended angle) fed by a feeding puck on the backside of ground plane. 3.2 Receiver integration in Cryostat and Tsys measurements. The work presented here is related to WP2A and WP2B. The main goal of these work packages involved the integration of an older 2 14 GHz geodetic Eleven feed version with straight unbent dipoles into a cryostat in order to verify experimentally the receiver noise temperature, the reflection coefficient and also the reliability of the feed for recycling to cryogenic temperatures. Figure 13 shows the geodetic VLBI2010 2GHz 12GHz Eleven feed mounted on the cold plate of the cryostat together with three different LNAs from different suppliers used for testing. Figure 13: The geodetic VLBI GHz Eleven feed (top left) in the test cryostat showing integration with different LNAs. 4 8GHz LNAs from GARD/OSO/Chalmers, differential LNAs from Caltech (bottom left) and 50ohm LNAs from Caltech (bottom right) Page 12 of 24

14 Testing was done using hot/cold load measurements in the open air pointing at the zenith (with a pyramidal shield to reduce pickup effects from the ground, see Figure 15b). The measured noise temperatures using different LNAa are compared in Figure 14. The main difficulty with the noise measurement is the strong interferences from the mobile phone bands at frequencies below 2.5GHz. This explains the sharp rise in the receiver noise at that frequency. The best results are achieved with the 50ohm LNAs from Caltech: the noise temperature is below 20K over the 3 8GHz band. The receiver noise with the 4 8GHz LNAs is higher. This is due to the fact that these LNAs are designed for optimal noise match (S11 is poor) and are supposed to be used with isolators, but we did not have such available for the tests. Three of the four available differential LNAs failed and the result in Figure 14 (blue dashed line) is for one petal only. We consider that the noise temperature with differential LNAs will be somewhat lower than the 50ohm LNAs. This is mainly because the 50ohm LNAs are connected via a 200ohm to 50 ohm transition, while the differential LNAs do not use an impedance transformer and the lower reflections will result in lower noise temperature. Figure 14: Comparison of the receiver noise temperature of the geodetic VLBI GHz eleven feed with different LNAs as measured in cryostat. Note that the result for differential LNAs is for one petal only. The noise performance for the Circular eleven feed optimised to 1.2 4GHz is expected be very analogous to that experimentally measured with the 2 14GHz feed since the impedance matching between the LNAs and the feed is quite similar. 3.3 Noise modelling The work presented here is related to WP3. The work on the noise model is presented in detail in the reports for Quarters 2 and 3; additionally the final numerical results and complete description of the overall system model can be found in the conference paper of the EuCAP2012 [4]. This paper presents a mathematically general method for exporting the far field patterns, radiation efficiency and impedance data, for multi port antennas, from EM solvers to the microwave circuit simulator of the CAESAR software in order to perform an analysis of a multiport antenna receiver system. This hybrid modeling approach offers a few important advantages over earlier developed modeling methods and standard simulation tools; including the capability to predict the correlated noise contributions due to both internal and external noise sources, while exploiting the translation/rotation symmetry of the antenna structure to reduce computation time. In this WP the noise model is verified by comparing its predictions with measurements made on the geodetic VLBI2010 Eleven feed covering 2GHz to 12GHz Page 13 of 24

15 The resulting modelling analysis and comparison to measurements are shown in Figure 15a e. The red curve in Figure 15a gives the noise temperature of a single LNA which was measured with a 50 Ohm resistor (that would be the case for the ideal feed); the black curve shows the predicted noise temperature due to 8 LNAs connected to the ports of the Eleven feed (4 LNAs per polarization), including the feeding/combining network and 2 nd stage amplifiers. The black curve gives the contribution of the LNAs including three components (1) the LNA minimum noise temperature Tmin when LNAs are optimally noise matched (this contribution is independent from the source impedance); (2) the noise mismatch contribution due to the difference between the LNA optimum reflection coefficient opt and the active reflection coefficient of the Eleven feed (thus is the contribution due to the reflection and mutual coupling between the 8 antenna ports); and (3) losses of the feeding/combining network. As one can see, the 50 Ohm resistor contribution is of the order of 5 Kelvin from 2 to 11 GHz and increases up to 12 Kelvin at higher frequencies, as the 50 Ohm source does not provide the optimal noise match condition for this type of the LNA. The noise contribution of the integrated Eleven feed receiver is equal to that of the 50 Ohm resistor with a single LNA at multiple frequency points where the black curve crosses the red one (these are the points where the antenna impedance is close to 50 Ohm). At frequencies in between these points, the noise values vary with the average value of 7.5±2.5 Kelvin over the plotted frequency range. Figure 15b represents the modelled ground noise pick up when the feed points towards the sky and the thermal ground noise enters into the systems through the far side and back lobes. The ground noise pick up is shown both for the setup when antenna is placed inside a pyramidal shielding horn and without it. Figure 15c and 15d show, respectively, a photo of the measurement setup with the antenna placed inside a pyramidal shielding horn that was used to remove some fraction of the noise contributed from ground and trees in the environment, and the smoothed noise temperatures when measured with and without the horn. Note that the measured data are smoothed in order to reduce random measurement noise in the evaluation of the noise increase effect. The relative difference between the measured values in these setups is about 4 5K over most of the bandwidth, but increases up to 10K at frequencies higher than 11 GHz and up to 20K at frequencies lower than 3 GHz. This increase is the result of the broadening of the antenna pattern at low frequencies and stronger back radiation at high frequencies. It is clear that only a fraction of the far side lobe noise contribution is removed using the shielding pyramid. Similar observations can be made based on the simulation results shown on Figure 15b. The predicted and measured system noise temperatures are plotted on Figure 15e. All the measurements were done with the feed placed looking at the sky. This setup of course significantly increases the measured noise, as compared to when such a feed mounted in a parabolic reflecting antenna, due to the additional ground noise pick up component. The results in Figure 15 are consistent with a receiver noise contribution for our measured hardware, which after excluding ground pick up effects is <10K. In conclusion, we can see that; (1) The developed system model works well for predicting the system noise temperature of the integrated antenna receiver system. To the authors best knowledge, it provides the best agreement with measured temperatures thus far presented for the UWB feed system within the SKA community. (2) The data collected are consistent with very low receiver noise for an LNA coupled to an Eleven feed Page 14 of 24

16 (a) (b) (c) (d) (e) Figure 15: Comparison of measured and modelled noise temperature of Eleven feed receiver, as shown on Figure.1a, with 50ohm Caltech LNAs. (a) Measurement results for a single LNA connected to the 50 Ohm resistor and simulation results (based on measured S parameters of the antenna) giving predicted noise contributions due to 8 LNAs when connected to the Eleven feed (4 LNAs are connected to one polarization and 4 other LNAs act as active loads), including the effect of the feeding and combining network; (b) predicted ground noise pick up when antenna is pointed at the zenith (with and without shielding horn). (c) Photo of the measurement setup with the shielding horn. (d) Comparison of the measured temperatures with and without the shielding horn (smoothed data points) and (e) Comparison of measurements and predictions of the noise model Page 15 of 24

17 3.4 Expected performance in reflector systems The study of the Eleven feed performance in combination with reflector systems considered for the SKA is not part of the current project. Nevertheless this is a vital question to consider and therefore we provide a brief overview of the current progress of work within the Chalmers Antenna Group and OSO in this direction. More details of this work can be found in [10] [12] Axi symmetric reflector antenna systems Determination of the optimal F/D and position of struts (0.5 4 GHz) The geometry of the reflector antenna (F/D) and configuration of its aperture blocking structures represent important design parameters of radio telescopes which affect the antenna efficiency and overall system sensitivity. Figure 16 presents the simulation results for the key performance parameters (including the antenna aperture efficiency, system noise temperature Tsys and receiving sensitivity (Aeff/Tsys) obtained for the reflector with the half subtended angle ranging from about 40 to 70. The measured patterns of the compact Eleven feed (as shown on Figure 1a) were used in these simulations. The results are shown in Figure 16. Note that herein the total system noise temperature is defined as Tsys = Tsp +Trec+Tsky, where Tsp is the spillover noise contribution due to the thermal ground noise pick up, Tsky=5K and Trec is the receiver noise temperature contribution due to the combined effects of the antenna and LNA impedance noise mismatch, ohmic losses in the antenna structure, the feeding network and the second stage amplifiers. The Trec is assumed to be constant and is set to values of 5 K, 15 K and 25 K the range which is relevant to cooled and uncooled systems. The Tsp component was evaluated for different antenna pointing evaluation angles, while the azimuth angle was set to a constant value of 0. The elevation angle (El) is defined as the pointing angle up from the ground (horizon). El = 90 is the zenith pointing angle. Often in published results on reflector antenna and feed designs, the elevation angle is assumed to be constant and equal to 90. However, it is critical to take into account the total range of angles El when optimising the system parameters. The analysis shows that the optimal value of the reflector angle for the maximum aperture efficiency is close to 50, this optimum is the result of the interplay between the antenna spillover and illumination efficiency contributions. For the maximum sensitivity, however, the optimum shifts towards the wider angle optics and depends on the value of the receiver noise contribution Trec. For receivers with the noise temperatures in the order of Trec = 15 K the optimal value of the reflector angle is close to Page 16 of 24

18 Figure 16: Determination of the optimal angle of the reflector optics for the axi symmetric system: The upper figures show the aperture efficiency and its dominant contributions, and the bottom figures show the system noise temperature and receiving sensitivity, as computed at 5.6 GHz. The solid lines represent the results obtained for elevation angles 90, and the dotted lines are the averaged values of Tsys and sensitivity over the range of elevation angles El = For the sensitivities given in the last panel antenna diameter of 15m is assumed. These results demonstrate that when considering a noise temperature of the receiver that is close to 15K, the performance of the Eleven feed is virtually independent on the subtended angle of the reflector (see the red curves on the figure in the bottom right corner). The same conclusion can be likely drawn for Off set Gregorian systems with the same range of subtended angles sub =55 70 (see the next section). Furthermore, we have also in a separate study analysed how the telescopes performance changes due to the aperture blockage provided by the feed cabin and its supporting struts. This study has been carried out for two examples: Example 1 a set of two compact wideband Eleven antenna feeds of about 4:1 bandwidth each (like the one shown on Figure 1 (c)) and Example 2 a set if multiple octave horn feeds that cover the same total frequency range. Our results show that in the optimal operational range of the 15 m reflector system (from about 1 GHz and higher), the relative increase of the spillover noise temperature Tsp and the aperture efficiency loss are less than 5 Kelvin and 5% (corresponding to the aperture blockage efficiency of 95%), respectively, in comparison with the unblocked aperture antenna with the two compact Eleven feeds. Note that when the size of the feed cabin and blockage area of struts are larger (that would be the case for an equivalent set of octave horn feeds), the efficiency loss and Tsp increase will be higher. For example, if the cross section of the struts (such as used for the analysis for the case of the Eleven feed) is doubled, the aperture blockage efficiency lowers to 92% and Tsp increase becomes 8K with respect to the unblocked aperture antenna fed with the horn feeds. It is worth mentioning that the degradation of the system performance due to the blockage effects in prime focus axi symmetric systems is a combined effect of the struts and the shape of the feed pattern. We have observed that in this respect, a wideband feed like the Eleven antenna has an advantage over the conventional horn feeds, as its radiation is mainly Page 17 of 24

19 concentrated in the region of the reflector s subtended angle with a sharper cut off towards its rim than that of the horn. For comparison, see the far field patterns of the horn and Eleven feed on Figure 17. (a) The measured pattern of the Eleven Feed (see Figure 1b) at 5.6 GHz (b) Simulated pattern of the octave horn feed (designed according to the procedure in [13]) at 5.6 GHz. (c) The scattered far field from the four struts due to the Eleven Feed (d) The scattered far field from the four struts due to the octave horn feed Figure 17: Comparison of the far field patterns of (a) the Eleven antenna feed and (b) octave horn feed, as measured and simulated respectively, at 5.6 GHz. Comparison of the scattering effects of four struts (no reflector) when the struts are illuminated by (c) Eleven feed and (d) octave horn feed. Note the high levels of the scattered field around the strut locations Page 18 of 24

20 3.4.2 Off set Gregorian reflector system (US SKA design) with the Circular Eleven Feed We show in Figure 20 below our analysis of a scaled version of the circularised (1.2 14GHz version) Eleven feed (see Figure 1b) using the measured parameters of this feed combined with the US design offset Gregorian antenna. The scaled feed was assumed to have a size 30% larger to improve the low frequency performance. An analysis has yet to be done using the optimised (1.2 4GHz) Eleven feed (see Figure 1c), which should have even better performance, especially over the 1.2GHz 2GHz range (which we don t show in the plots below). (a) (b) results from [14] (c) (d) Figure 20: The predicted performance parameters of the 30% scaled circularised Eleven feed (Figure 1b) combined with the US off set Gregorian reflector optics (a) Overall geometry, (b) Antenna efficiency of (b) Antenna aperture efficiency of octave band feeds and ideal feed optimized at each frequency (see [14]); (c) Antenna efficiency of the 30% scaled circularised Eleven feed (based on the measured patterns when the feed), compared to the ideal feed in [14]: (d) Antenna noise temperature due to the ground noise pick up Page 19 of 24

21 The results in Figure 20b show that our modelled feed/gregorian reflector system has peak aperture efficiency over 83% approaching that of the ideal horn feed that is optimized at each frequency. Over the range from 2 to 6 GHz the aperture efficiency stays above 78% and from 1 to 2 GHz it is above 68%. Over the plotted bandwidth of factor 6, the antenna efficiency of the circular Eleven feed is only slightly lower than the aperture efficiency of the ideal feed and on average higher than that of narrow angle octave horns shown in panel b. Figure 20c shows the antenna temperature contribution for our Eleven feed and an ideal feed, both for an elevation angle of 90. The predicted performance of the ideal feed seems to be based on the simulations of the feed patterns. In contrast to this ideal feed model, the antenna noise temperature of the Eleven feed was computed from the measured patterns of the feed. The antenna noise temperature Tsp of our Eleven feed is smaller than 10K from 2 to 6 GHz and has the minimum contribution of 6K at around 5GHz. The Tsp increases at < 2 GHz due to widening of the pattern of the electrically smaller feed structure and higher side lobe levels of the feed pattern at high frequencies due to the presence of surface waves in the dielectric material at frequencies above 7 GHz. It is worth mentioning that the Eleven feed patterns were measured for the feed alone (thus in the absence of the cryostat). When the feed is integrated inside the cryostat, the measured patterns become slightly narrower at low frequencies due to the effects of the cryostat window that starts acting as a part of the feed design at these frequencies. As the result, the spillover efficiency increases with about dB (while the total antenna efficiency remains high due to the interplay between other efficiency components) as can be seen by comparing the values obtained from the measurement data with and without the cryostat on Figure 21 (a) and Figure 21 (b), respectively. These results demonstrate that the antenna noise temperature of the feed integrated inside the cryogenic system will be lower than the values shown on Figure 20d, and close to 10 Kelvin. (a) (b) Figure 21: The predicted efficiency and its spillover contribution based on the measured patterns of the Eleven feed (see the model on Figure 1a) when tested (a) without and (b) with the cryostat supporting structure. Note that when the feed is integrated inside the cryostat, the measured patterns become slightly narrower at low frequencies with the result that the spillover efficiency (see red curves) increases with about dB, while the aperture efficiency (see black curves) remains approximately the same. Figure 22a shows the final predicted A eff /T sys versus frequency, the peak performance of which matches that which is theoretically possible. Figure 22b shows a comparison with the estimates made by [14] for various horn feeds including the ideal feed optimized at each frequency, and octave narrow and wide opening angle horn feeds Page 20 of 24

22 (a) Figure 22: (a) The predicted receiving sensitivity (Aeff/Tsys) of 30% scaled circularised Eleven feed (Figure 1b) combined with the US off set Gregorian reflector optics compared to the ideal feed (b) comparison with octave horn feeds with the wide and narrow opening angles in [14]: The system noise temperature Tsys is defined as a sum of the antenna noise temperature as shown in panel d of Figure 20, the sky contribution of 5K, and the receiver noise contribution that was assumed to be constant and equal to Trec=10K for both the Eleven feed and horn feed cases. Note that the sensitivity of the Eleven feed was computed from the measured patterns of the feed (without the effect of the cryostat window), while the sensitivity of the horn feeds is based on the simulated patterns. Note that the sensitivity predicted for the 30% scaled circularised Eleven feed does not take into account the expected reduction of the antenna ground noise pick up as the result of the higher spillover efficiency of the feed when integrated inside the cryostat. After including this effect in the analysis, the antenna noise temperature should remain virtually constant and close to 10 Kelvin in the range of GHz. Since, our expectation is based on the actual measurements with and without the cryostat supporting ring, we believe that the resultant sensitivity at the low frequencies will be rather constant and in the range of m 2 /K for this feed design. Furthermore, it is worth mentioning that the 1.2 4GHz optimised version of the circularised Eleven feed (see Figure 1c) should have even better performance than that plotted in panel d. This design has a rather flat predicted behaviour of the spill over efficiency from 1.3 to 4 GHz, so it is expected that the ground noise pickup can be even lower than 10 K. (b) 3.6 Comments on frequency variations in Eleven feed response. One concern that has been raised about the Eleven feed are the variations in performance with frequency seen in earlier designs and whether these would be a problem for calibration. A very significant improvement has been achieved in this area by using the new circularised Eleven feed geometry; in this new design the A eff fluctuations versus frequency (see Figure 5) have been much reduced compared to the old VLBI2010 Eleven feed model with straight dipoles. These variations in A eff,, which is the critical quantity for determining the spectral bandpass shape, are now less than 6% over 1GHz. It is expected that the performance of the 1.2GHz 4GHz circularised Eleven feed will be even better in this regard. It is true that the circularised feeds still have significant variations with frequency in reflection coefficient (see Figures 3, 8 and 11), however the peak values of these reflection coefficients are now relatively small (better than 10dB except for occasional peaks at 8dB) and furthermore they only give rise to relatively small variations in Tsys versus frequency. The reason for this is that at cryogenic temperatures Tsys is dominated by antenna noise pickup effects (see Figures 15 and 20). For an interferometer, as opposed to a single dish, these resulting small Tsys variations with frequency Page 21 of 24

23 do not affect the spectral baseline, only the amplitude of the Gaussian noise contribution versus frequency. Two more general points about frequency variable effects and spectral baselines can be made which are relevant to all feeds, not just the Eleven feed (1) The critical quantity is the residual spectral baseline after calibration on astronomical sources. If spectral variations in performance are stable over a calibration switching cycle they can be completely removed. Additionally, given the extremely high sensitivity of the SKA, there will be plenty of in beam continuum calibrator sources, these allow for the continuous calibration and removal of spectral effects. (2) For spectral line work, where spectral baseline issues potentially have the largest effect, the observed lines are virtually always less than 300km/s wide and hence narrower than 5MHz at 5GHz, hence even if UWB feed designs have modest gain variations over 500MHz bandwidths the narrow frequency ranges covered by spectral lines means the response will be well modelled by low order polynomials. 4 Conclusions Within the current project we have designed, manufactured and tested Eleven feed versions optimised for the 1.2 4GHz band which show improved beam pattern symmetry, spillover efficiency and reflection coefficients compared to previous designs. The highlights of our project work include. (A) Measurement at cryogenic temperatures using the old VLBI2010 feed of on sky system temperatures of 20K over wide bands. A large fraction of this figure is likely to come from pickup from the immediate environment, a contribution which is still present even after installing pyramidal shielding. When mounted in a suitable reflecting antenna to minimise ground pick up system temperatures would likely be <15K. (B) Design, manufacture and measurement of a new circularised Eleven feed geometry (see Figure 1b). Combining beam measurements with simulations of performance of this feed in axi symmetric and offset Gregorian dishes predicts aperture efficiencies of 65% and 83% respectively. (C) The new circularised designs also show smaller variations of aperture efficiency with frequency than the previous straight dipole designs, which eases concerns about poorer spectral baselines compared to other UWB feed designs. (D) During the last months of the project we have developed a novel feeding network decreasing the number of ports from eight to four. (E) We have designed but not yet manufactured or measured a version of the circularised Eleven feed optimised over the band 1.2 4GHz (see Figure 1c). This version has been designed to improve the reflection coefficient and efficiency over this frequency range in particular over the GHz region where performance for the original circularised Eleven feed was poor. Computer modelling suggests that this design has good reflection coefficient and efficiency performance over the band from 1.2 5GHz. We hope as next step to construct and measure the EM properties of this feed and then test its noise performance cryogenically Page 22 of 24

24 Improvements in Eleven feed performance over the last two years has been significant and we believe it should be possible in the near future to achieve comparable Aeff/Tsys performance to octave feeds over a frequency range of 3.5:1 (plus reasonable performance simultaneously over higher frequency ranges). As well as good performance such a feed has some practical advantages (small size and weight plus lower construction and maintenance costs) compared to a suite of octave feeds). We also note the significant potential scientific advantages of UWB feeds for continuum observations and astrometric observations. Octave feeds only have <1GHz bandwidth at frequency <2GHz, Eleven feeds allow larger bandwidths (i.e 1 4GHz) which, if sampled and processed, give higher continuum sensitivity (and better mapping fidelity using Multi frequency Synthesis). Additionally by making simultaneous wide frequency range observations UWB feeds allow for the very accurate estimation of atmospheric and ionospheric delays (as achieved by geodetic VLBI). Combining large separation angle UWB observations to estimate such delays with rapidly switched narrow angle target to calibrator observations provides the best possible astrometric accuracy. This technique can potentially greatly improve the accuracy of pulsar astrometry measurements (part of one of the key SKA phase 1 science goals at >1.4GHz). We hope to investigate the scientific advantages of SKA UWB operation in a future SKA memo. Sweden is presently in the process of signing the papers to join the SKA organisation. Our ambition is to continue our work on UWB single pixel feeds as part of our contribution to the pre construction phase of SKA. A modest amount of funding to continue this work in the immediate future will be provided funding via the DIVA JRA of Radionet 3. We have also recently submitted a large proposal to the Swedish Research council for Project Planning Grant, aiming to continue the work on SKA Ultra Wide Band Feeds. This includes also optimising other designs (i.e quad ridge) and testing their performance against the Eleven feed. The main work tasks in the submitted proposal are: (1) Further optimisation of the Eleven Feed including feeding structure and interface to LNAs; (2) design optimisation, prototype fabrication and testing of quad ridge feed, (3) Integration of both UWB feeds in cryogenic receiver with wideband LNAs (including locally produced, now commercially available, Chalmers/MC2/Low Noise Factory MMIC devices) and (4) evaluating combined receiver performance/optics design of reflector feed combinations. In addition to the above work specifically for SKA, OSO/Chalmers/Omnisys plan to continue development of UWB feeds for geodetic VLBI. OSO has recently received funding (3M ) from the Wallenberg foundation for a twin 12m antenna system at Onsala, including money for feeds and receivers. These antennas will provide a useful test bed for evaluating the performance of UWB feeds using astronomical observations Page 23 of 24