Photo courtesy of Chia-Yu Shih circa ~2008 shortly after the telescope was upgraded from 7 to 13 elements. The 13 antennas and associated receivers operate from 86 to 102 GHz and are mounted to a rigid carbon fiber 6-m platform. The platform position is controlled by a unique hexapod drive system.
I began my career with ASIAA on this project back in 2002 when it was named AMiBA (Array for Microwave Background Anisotropy) and have stayed involved until today. I have many fond memories of working closely with several of the staff members who are still at ASIAA.
2014-Oct - As part of an effort to repurpose the AMiBA telescope from a wide-band analog correlator to a digital correlator with higher spectral resolution for CO intensity mapping, I was tasked to develop a microwave design architecture that can digitize any 2 GHz portion of the 2-18 GHz intermediate frequency (IF) band. I was initially struggling with the concept of using switched RF filter banks because of its high cost and reliability issues. Then it dawned on me to use an I/Q down converter similar to my first job at Boeing 3 decades earlier. We hashed out the pros and cons over the next several months and here we are today digitizing any 4 GHz (LSB + USB) portion of the IF band. First light of the full dual polarization system was obtained in April 2018. Major kudos to Ranjani for working on the software and commissioning work (D. Kubo, C.T. Li, H. Jiang).
System block diagram:
2015 Gantt Schedule:
2016 overall hardware status report:
2018 SPIE conference article:
2016 - Designed, assembled, and tested by John K., this photo is of the first production I/Q Down Converter bracket developed for YTLA's CO intensity mapping project. A total of 18 (+ 2 spare) assemblies were produced to support 8 antennas with dual polarization receivers.
This bracket down converts any 4 GHz portion of the 2-18 GHz IF by tuning the LO (4-16 GHz) and maintains decent amplitude balance and phase quadrature. Residual amplitude and phase correction is performed in the digital domain to maintain >/= 20 dB sideband rejection. The YTLA has completed its commissioning phase and has started early science as of April 2018. Much appreciation to John K. for the technical craftsmanship he put into the final production design which resulted in a beautiful DC-2.24 GHz output spectra (design - J. Kuroda, D. Kubo, assembly/test - S. Ho, R. Srinivasan, J.C. Cheng, C.T. Li).
Description of this production bracket on the ASIAA website:
2016 test report:
2015 - This was one of two prototype I/Q Down Converter plates that were designed and assembled by John K. in Hilo. Each plate processes a single IF input signal and provides in-phase (I-channel) and quadrature (Q-channel) down conversion to baseband. The I and Q baseband outputs are fed to the ROACH-2 for digitization and processing.
We sent these two prototypes to Taipei for initial sideband separation characterization in Taipei where the digital backend hardware was being developed. After the tests and characterizations were completed the prototype plates were returned to Hilo and installed on the telescope for single baseline astronomical tests. A decision was made shortly afterwards to proceed with the production design of the I/Q down converters (J. Kuroda, D. Kubo, C.T. Li, H.M. Jiang).
Description of this prototype plate on the ASIAA website:
2014 - This was one of two prototype conventional Down Converter plates that were designed and assembled by John K. in Hilo. Each plate accepts a pair of IF input signals and provides down conversion to baseband using a fixed LO of 5.85 GHz. The plates were installed onto the telescope to perform 4-element test observations with the ROACH-2 digital correlator which was in development (J. Kuroda, D. Kubo).
Description of this prototype plate on the ASIAA website:
2016 - This noise + tone calibration system covers 2-18 GHz and was developed in Hilo using left over components from the decommissioned analog correlator system. Channel to channel isolation is > 120 dB and was achieved using cascaded SPDT RF switches. As with most things engineering, there was a lot more that went into this design than meets the eye of the casual observer. (D. Kubo)
Description of this noise/tone plate assembly on the ASIAA website:
2015 - This Clock Distribution unit was developed in Hilo and assembled in Taipei. The design is based around a Valon 5008 programmable synthesizer module controlled by USB interface. The clock is phase locked to the GPS 10 MHz system reference and is distributed to 16 outputs to the ROACH-2 F-engines (8 per polarization). The initial clock rate was set to 1.6 GHz during the ROACH-2 bit code development and was increased to it's final rate of 2.24 GHz (D. Kubo, J.C. Cheng, C.C. Han).
Description of the unit on the ASIAA website:
2003/2009 - The IF distribution system supports 2-18 GHz and consists of 3 cascaded sections of amplification and 4-way Wilkinson power dividers. The 1st and 2nd sections were developed in Hilo by Peter and myself using connectorized components and the 3rd section was designed by Chao-Te in Taipei using a custom PCB design.
The original AMiBA telescope began with 14 of these 1st section modules and was expanded to 26 modules in 2009 to support 13-element dual polarization receivers. This photo is of the 1st Section module designed to fit into a standard Eurocard chassis. The black heatsink shown in the photo was later increased in size to facilitate cooling of the Celeritek power amplifier (D. Kubo, P. Oshiro).
Report on the 13-element IF distribution system:
2004/2009 - These 2nd Section plates were designed and assembled in Hilo and support 2-18 GHz IF distribution. There are two dash versions, -1 allows for swapping X with Y polarization for cross polarization observations (which we never did) and the -2 without the switch. The original 7-element system utilized 2 plates (-1 and -2) and the 13-element upgrade in 2009 required 2 more for a total of 4 plates (D. Kubo, P. Oshiro).
2004/2009 - This 4-lag analog Correlator Module was designed in collaboration with Ference Marki of Marki Microwave. It accepts 2-18 GHz IF inputs from a pair of antennas via left and right connectors and a final set of 4-way power division to 4 analog mixers. The 4 mixers are staggered to provide 90 degree phase delay at 10 GHz between each mixer. The low frequency signal product is output via pins on the rear of this module and fed to a DC amplifier board. The original 7 element system required 49 modules and in 2009 we upgraded to 13-elements which required an additional 120 for a total of 169 modules (design F. Marki, D. Kubo, J. Peterson - CMU). Photo courtesy of F. Marki.
Description of this module on the ASIAA website:
2010 Astrophysical Journal article:
2004/2009 - Another view of the Analog Correlator module with solid absorber installed over the mixers, note the Duroid substrate. Many thanks to Ferenc for putting in countless hours assembling these modules for our AMiBA project.
One of the more challenging aspects of the analog correlator design was how to physically interconnect 78 correlator modules so that each processed a unique pair of antenna baselines. We accomplished this with a rectangular matrix consisting of horizontal 4-way power dividers on the front and vertical power dividers on the rear.
Drawing of physical configuration:
2004/2009 - A DC amplifier with gain of 1000 V/V was designed to interface directly to the Analog Correlator Module. A DC block was introduced to remove the undesired DC component from the mixers (varied as a function of input power and temperature) phase switching allowed the desired signal to pass through and be amplified. This design utilized low noise Burr Brown instrumentation op-amps to minimize the Johnson noise effects (D. Kubo, W. Wilson - ATNF).
2006 - I added this early photo of the AMiBA platform to fill this space. This hexapod drive system turned out to be quite complex as it requires the the platform to be very stiff. We ended up having to design and add a steel ring structure (P. Raffin) between the hexapod legs and CFRP platform to reduce deformation. Even with the steel ring we still see ~1 mm of platform deflection at the edges which is non-trivial for 3 mm observation wavelength. The science team has largely modeled the deformation and removes most of it's effect in post processing.