Neural networks are increasing in scale and sophistication, catalyzing the need for efficient hardware. An inevitability when transferring neural networks to hardware is that non-idealities impact performance. Hardware-aware training, where non-idealities are accounted for during training is one way to recover performance, but at the cost of generality. In this work, we demonstrate a binary neural network consisting of an array of 20,000 magnetic tunnel junctions (MTJ) integrated on complementary metal-oxide-semiconductor (CMOS) chips. With 36 dies, we show that even a few defects can degrade the performance of neural networks. We demonstrate hardware-aware training and show that performance recovers close to ideal networks. We then introduce a robust method – statistics-aware training – that compensates for defects regardless of their specific configuration. When evaluated on the MNIST dataset, statistics-aware solutions differ from software-baselines by only 2 %. We quantify the sensitivity of networks trained with statistics-aware and conventional methods and demonstrate that the statistics-aware solution shows less sensitivity to defects when sampling the network loss function.
BICEP Array is a degree-scale Cosmic Microwave Background (CMB) experiment that will search for primordial B-mode polarization while constraining Galactic foregrounds. BICEP Array will be comprised of four receivers to cover a broad frequency range with channels at 30/40, 95, 150 and 220/270 GHz. The first low-frequency receiver will map synchrotron emission at 30 and 40 GHz and will deploy to the South Pole at the end of 2019. In this paper, we give an overview of the BICEP Array science and instrument, with a focus on the detector module. We designed corrugations in the metal frame of the module to suppress unwanted interactions with the antenna-coupled detectors that would otherwise deform the beams of edge pixels. This design reduces the residual beam systematics and temperature-to-polarization leakage due to beam steering and shape mismatch between polarized beam pairs. We report on the simulated performance of single- and wide-band corrugations designed to minimize these effects. Our optimized design alleviates beam differential ellipticity caused by the metal frame to about 7% over 57% bandwidth (25 to 45 GHz), which is close to the level due the bare antenna itself without a metal frame. Initial laboratory measurements are also presented.
BICEP3 is a 520mm aperture on-axis refracting telescope observing the polarization of the cosmic microwave background (CMB) at 95GHz in search of the B-mode signal originating from in ationary gravitational waves. BICEP3's focal plane is populated with modularized tiles of antenna-coupled transition edge sensor (TES) bolometers. BICEP3 was deployed to the South Pole during 2014-15 austral summer and has been operational since. During the 2016-17 austral summer, we implemented changes to optical elements that lead to better noise performance. We discuss this upgrade and show the performance of BICEP3 at its full mapping speed from the 2017 and 2018 observing seasons. BICEP3 achieves an order-of-magnitude improvement in mapping speed compared to a Keck 95GHz receiver. We demonstrate 6.6μK√s noise performance of the BICEP3 receiver.
Bicep Array is a cosmic microwave background (CMB) polarization experiment that will begin observing at the South Pole in early 2019. This experiment replaces the five Bicep2 style receivers that compose the Keck Array with four larger Bicep3 style receivers observing at six frequencies from 30 to 270GHz. The 95GHz and 150GHz receivers will continue to push the already deep Bicep/Keck CMB maps while the 30/40GHz and 220/270GHz receivers will constrain the synchrotron and galactic dust foregrounds respectively. Here we report on the design and performance of the Bicep Array instruments focusing on the mount and cryostat systems.
Targeting faint polarization patterns arising from Primordial Gravitational Waves in the Cosmic Microwave Background requires excellent observational sensitivity. Optical elements in small aperture experiments such as Bicep3 and Keck Array are designed to optimize throughput and minimize losses from transmission, reflection and scattering at millimeter wavelengths. As aperture size increases, cryostat vacuum windows must withstand larger forces from atmospheric pressure and the solution has often led to a thicker window at the expense of larger transmission loss. We have identified a new candidate material for the fabrication of vacuum windows: with a tensile strength two orders of magnitude larger than previously used materials, woven high-modulus polyethylene could allow for dramatically thinner windows, and therefore significantly reduced losses and higher sensitivity. In these proceedings we investigate the suitability of high-modulus polyethylene windows for ground-based CMB experiments, such as current and future receivers in the Bicep/Keck Array program. This includes characterizing their optical transmission as well as their mechanical behavior under atmospheric pressure. We find that such ultra-thin materials are promising candidates to improve the performance of large-aperture instruments at millimeter wavelengths, and outline a plan for further tests ahead of a possible upcoming field deployment of such a science-grade window.
Bicep Array is the newest multi-frequency instrument in the Bicep/Keck Array program. It is comprised of four 550mm aperture refractive telescopes observing the polarization of the cosmic microwave background (CMB) at 30/40, 95, 150 and 220/270 GHz with over 30,000 detectors. We present an overview of the receiver, detailing the optics, thermal, mechanical, and magnetic shielding design. Bicep Array follows Bicep3's modular focal plane concept, and upgrades to 6" wafer to reduce fabrication with higher detector count per module. The first receiver at 30/40GHz is expected to start observing at the South Pole during the 2019-20 season. By the end of the planned Bicep Array program, we project 0.002 ⪅ σ(r) ⪅ 0.006, assuming current modeling of polarized Galactic foreground and depending on the level of delensing that can be achieved with higher resolution maps from the South Pole Telescope.
H. Hui, P. Ade, Z. Ahmed, K. Alexander, M. Amiri, D. Barkats, S. Benton, C. Bischoff, J. Bock, H. Boenish, R. Bowens-Rubin, I. Buder, E. Bullock, V. Buza, J. Connors, J. Filippini, S. Fliescher, J. Grayson, M. Halpern, S. Harrison, G. Hilton, V. Hristov, K. Irwin, J. Kang, K. Karkare, E. Karpel, S. Kefeli, S. Kernasovskiy, J. Kovac, C. L. Kuo, E. Leitch, M. Lueker, K. Megerian, V. Monticue, T. Namikawa, C. Netterfield, H. Nguyen, R. O'Brient, R. Ogburn, C. Pryke, C. Reintsema, S. Richter, R. Schwarz, C. Sorensen, C. Sheehy, Z. Staniszewski, B. Steinbach, G. Teply, K. Thompson, J. Tolan, C. Tucker, A. Turner, A. Vieregg, A. Wandui, A. Weber, D. Wiebe, J. Willmert, W. L. Wu, K. W. Yoon
BICEP3, the latest telescope in the BICEP/Keck program, started science observations in March 2016. It is a 550mm aperture refractive telescope observing the polarization of the cosmic microwave background at 95 GHz. We show the focal plane design and detector performance, including spectral response, optical efficiency and preliminary sensitivity of the upgraded BICEP3. We demonstrate 9.72 μKCMB√s noise performance of the BICEP3 receiver.
K. Karkare, P. A. Ade, Z. Ahmed, K. Alexander, M. Amiri, D. Barkats, S. Benton, C. Bischoff, J. Bock, H. Boenish, R. Bowens-Rubin, I. Buder, E. Bullock, V. Buza, J. Connors, J. Filippini, S. Fliescher, J. Grayson, M. Halpern, S. Harrison, G. Hilton, V. Hristov, H. Hui, K. Irwin, J. Kang, E. Karpel, S. Kefeli, S. Kernasovskiy, J. Kovac, C. L. Kuo, E. Leitch, M. Lueker, K. Megerian, V. Monticue, T. Namikawa, C. Netterfield, H. T. Nguyen, R. O'Brient, R. Ogburn, C. Pryke, C. Reintsema, S. Richter, M. St. Germaine, R. Schwarz, C. Sheehy, Z. Staniszewski, B. Steinbach, G. Teply, K. Thompson, J. Tolan, C. Tucker, A. Turner, A. Vieregg, A. Wandui, A. Weber, J. Willmert, C. L. Wong, W. L. Wu, K. W. Yoon
BICEP3 is a small-aperture refracting cosmic microwave background (CMB) telescope designed to make sensitive polarization maps in pursuit of a potential B-mode signal from inflationary gravitational waves. It is the latest in the Bicep/Keck Array series of CMB experiments located at the South Pole, which has provided the most stringent constraints on inflation to date. For the 2016 observing season, BICEP3 was outfitted with a full suite of 2400 optically coupled detectors operating at 95 GHz. In these proceedings we report on the far field beam performance using calibration data taken during the 2015-2016 summer deployment season in situ with a thermal chopped source. We generate high-fidelity per-detector beam maps, show the array-averaged beam profile, and characterize the differential beam response between co-located, orthogonally polarized detectors which contributes to the leading instrumental systematic in pair differencing experiments. We find that the levels of differential pointing, beamwidth, and ellipticity are similar to or lower than those measured for Bicep2 and Keck Array. The magnitude and distribution of Bicep3’s differential beam mismatch – and the level to which temperature-to-polarization leakage may be marginalized over or subtracted in analysis - will inform the design of next-generation CMB experiments with many thousands of detectors.
J. Grayson, P. A. Ade, Z. Ahmed, K. Alexander, M. Amiri, D. Barkats, S. Benton, C. Bischoff, J. Bock, H. Boenish, R. Bowens-Rubin, I. Buder, E. Bullock, V. Buza, J. Connors, J. Filippini, S. Fliescher, M. Halpern, S. Harrison, G. Hilton, V. Hristov, H. Hui, K. Irwin, J. Kang, K. Karkare, E. Karpel, S. Kefeli, S. Kernasovskiy, J. Kovac, C. L. Kuo, E. Leitch, M. Lueker, K. Megerian, V. Monticue, T. Namikawa, C. Netterfield, H. Nguyen, R. O'Brient, R. Ogburn, C. Pryke, C. Reintsema, S. Richter, R. Schwarz, C. Sorenson, C. Sheehy, Z. Staniszewski, B. Steinbach, G. Teply, K. Thompson, J. Tolan, C. Tucker, A. Turner, A. Vieregg, A. Wandui, A. Weber, D. Wiebe, J. Willmert, W. L. Wu, K. W. Yoon
Bicep3 is a 520mm aperture, compact two-lens refractor designed to observe the polarization of the cosmic microwave background (CMB) at 95 GHz. Its focal plane consists of modularized tiles of antenna-coupled transition edge sensors (TESs), similar to those used in Bicep2 and the Keck Array. The increased per-receiver optical throughput compared to Bicep2/Keck Array, due to both its faster f=1:7 optics and the larger aperture, more than doubles the combined mapping speed of the Bicep/Keck program. The Bicep3 receiver was recently upgraded to a full complement of 20 tiles of detectors (2560 TESs) and is now beginning its second year of observation (and first science season) at the South Pole. We report on its current performance and observing plans. Given its high per-receiver throughput while maintaining the advantages of a compact design, Bicep3- class receivers are ideally suited as building blocks for a 3rd-generation CMB experiment, consisting of multiple receivers spanning 35 GHz to 270 GHz with total detector count in the tens of thousands. We present plans for such an array, the new "BICEP Array" that will replace the Keck Array at the South Pole, including design optimization, frequency coverage, and deployment/observing strategies.
The inflationary paradigm of the early universe predicts a stochastic background of gravitational waves which would generate a B-mode polarization pattern in the cosmic microwave background (CMB) at degree angular scales. Precise measurement of B-modes is one of the most compelling observational goals in modern cosmology. Since 2011, the Keck Array has deployed over 2500 transition edge sensor (TES) bolometer detectors at 100 and 150 GHz to the South Pole in pursuit of degree-scale B-modes, and Bicep3 will follow in 2015 with 2500 more at 100 GHz. Characterizing the spectral response of these detectors is important for controlling systematic effects that could lead to leakage from the temperature to polarization signal, and for understanding potential coupling to atmospheric and astrophysical emission lines. We present complete spectral characterization of the Keck Array detectors, made with a Martin-Puplett Fourier Transform Spectrometer at the South Pole, and preliminary spectra of Bicep3 detectors taken in lab. We show band centers and effective bandwidths for both Keck Array bands, and use models of the atmosphere at the South Pole to cross check our absolute calibration. Our procedure for obtaining interferograms in the field with automated 4-axis coupling to the focal plane represents an important step towards efficient and complete spectral characterization of next-generation instruments more than 10000 detectors.
Searching for evidence of inflation by measuring B-modes in the cosmic microwave background (CMB) polarization at degree angular scales remains one of the most compelling experimental challenges in cosmology. BICEP2 and the Keck Array are part of a program of experiments at the South Pole whose main goal is to achieve the sensitivity and systematic control necessary for measurements of the tensor-to-scalar ratio at σ(r) ~0:01. Beam imperfections that are not sufficiently accounted for are a potential source of spurious polarization that could interfere with that goal. The strategy of BICEP2 and the Keck Array is to completely characterize their telescopes' polarized beam response with a combination of in-lab, pre-deployment, and on-site calibrations. We Sereport the status of these experiments, focusing on continued improved understanding of their beams. Far-field measurements of the BICEP2 beam with a chopped thermal source, combined with analysis improvements, show that the level of residual beam-induced systematic errors is acceptable for the goal of σ(r) ~ 0:01 measurements. Beam measurements of the Keck Array side lobes helped identify a way to reduce optical loading with interior cold baffles, which we installed in late 2013. These baffles reduced total optical loading, leading to a ~ 10% increase in mapping speed for the 2014 observing season. The sensitivity of the Keck Array continues to improve: for the 2013 season it was 9:5 μK _/s noise equivalent temperature (NET). In 2014 we converted two of the 150-GHz cameras to 100 GHz for foreground separation capability. We have shown that the BICEP2 and the Keck Array telescope technology is sufficient for the goal of σ(r) ~ 0:01 measurements. Furthermore, the program is continuing with BICEP3, a 100-GHz telescope with 2560 detectors.
The SPTpol camera is a two-color, polarization-sensitive bolometer receiver, and was installed on the 10 meter South Pole Telescope in January 2012. SPTpol is designed to study the faint polarization signals in the Cosmic Microwave Background, with two primary scientific goals. One is to constrain the tensor-to-scalar ratio of perturbations in the primordial plasma, and thus constrain the space of permissible in inflationary models. The other is to measure the weak lensing effect of large-scale structure on CMB polarization, which can be used to constrain the sum of neutrino masses as well as other growth-related parameters. The SPTpol focal plane consists of seven 84-element monolithic arrays of 150 GHz pixels (588 total) and 180 individual 90 GHz single- pixel modules. In this paper we present the design and characterization of the 90 GHz modules.
The SPTpol camera is a dichroic polarimetric receiver at 90 and 150 GHz. Deployed in January 2012 on the South Pole Telescope (SPT), SPTpol is looking for faint polarization signals in the Cosmic Microwave Background (CMB). The camera consists of 180 individual Transition Edge Sensor (TES) polarimeters at 90 GHz and seven 84-polarimeter camera modules (a total of 588 polarimeters) at 150 GHz. We present the design, dark characterization, and in-lab optical properties of the 150 GHz camera modules. The modules consist of photolithographed arrays of TES polarimeters coupled to silicon platelet arrays of corrugated feedhorns, both of which are fabricated at NIST-Boulder. In addition to mounting hardware and RF shielding, each module also contains a set of passive readout electronics for digital frequency-domain multiplexing. A single module, therefore, is fully functional as a miniature focal plane and can be tested independently. Across the modules tested before deployment, the detectors average a critical temperature of 478 mK, normal resistance RN of 1.2Ω , unloaded saturation power of 22.5 pW, (detector-only) optical efficiency of ~ 90%, and have electrothermal time constants < 1 ms in transition.
SPTpol is a dual-frequency polarization-sensitive camera that was deployed on the 10-meter South Pole Telescope in January 2012. SPTpol will measure the polarization anisotropy of the cosmic microwave background (CMB) on angular scales spanning an arcminute to several degrees. The polarization sensitivity of SPTpol will enable a detection of the CMB “B-mode” polarization from the detection of the gravitational lensing of the CMB by large scale structure, and a detection or improved upper limit on a primordial signal due to inationary gravity waves. The two measurements can be used to constrain the sum of the neutrino masses and the energy scale of ination. These science goals can be achieved through the polarization sensitivity of the SPTpol camera and careful control of systematics. The SPTpol camera consists of 768 pixels, each containing two transition-edge sensor (TES) bolometers coupled to orthogonal polarizations, and a total of 1536 bolometers. The pixels are sensitive to light in one of two frequency bands centered at 90 and 150 GHz, with 180 pixels at 90 GHz and 588 pixels at 150 GHz. The SPTpol design has several features designed to control polarization systematics, including: singlemoded feedhorns with low cross-polarization, bolometer pairs well-matched to dfference atmospheric signals, an improved ground shield design based on far-sidelobe measurements of the SPT, and a small beam to reduce temperature to polarization leakage. We present an overview of the SPTpol instrument design, project status, and science projections.
The Bicep2 and Keck Array experiments are designed to measure the polarization of the cosmic microwave background (CMB) on angular scales of 2-4 degrees (ℓ = 50–100). This is the region in which the B-mode signal, a signature prediction of cosmic inflation, is expected to peak. Bicep2 was deployed to the South Pole at the end of 2009 and is in the middle of its third year of observing with 500 polarization-sensitive detectors at 150 GHz. The Keck Array was deployed to the South Pole at the end of 2010, initially with three receivers—each similar to Bicep2. An additional two receivers have been added during the 2011-12 summer. We give an overview of the two experiments, report on substantial gains in the sensitivity of the two experiments after post-deployment optimization, and show preliminary maps of CMB polarization from Bicep2.
The Keck Array (SPUD) began observing the cosmic microwave background's polarization in the winter of 2011 at the South Pole. The Keck Array follows the success of the predecessor experiments BICEP and BICEP2, 1 using five on-axis refracting telescopes. These have a combined imaging array of 2500 antenna-coupled TES bolometers read with a SQUID- based time domain multiplexing system. We will discuss the detector noise and the optimization of the readout. The achieved sensitivity of the Keck Array is 11.5 μKCMB√s in the 2012 configuration.
Between the BICEP2 and Keck Array experiments, we have deployed over 1500 dual polarized antenna coupled bolometers
to map the Cosmic Microwave Background’s polarization. We have been able to rapidly deploy these detectors because
they are completely planar with an integrated phased-array antenna. Through our experience in these experiments, we
have learned of several challenges with this technology- specifically the beam synthesis in the antenna- and in this paper
we report on how we have modified our designs to mitigate these challenges. In particular, we discus differential steering
errors between the polarization pairs’ beam centroids due to microstrip cross talk and gradients of penetration depth in the
niobium thin films of our millimeter wave circuits. We also discuss how we have suppressed side lobe response with a
Gaussian taper of our antenna illumination pattern. These improvements will be used in Spider, Polar-1, and this season’s
retrofit of Keck Array.
In January 2012, the 10m South Pole Telescope (SPT) was equipped with a polarization-sensitive camera, SPTpol, in order to measure the polarization anisotropy of the cosmic microwave background (CMB). Measurements of the polarization of the CMB at small angular scales (~several arcminutes) can detect the gravitational lensing of the CMB by large scale structure and constrain the sum of the neutrino masses. At large angular scales (~few degrees) CMB measurements can constrain the energy scale of Inflation. SPTpol is a two-color mm-wave camera that consists of 180 polarimeters at 90 GHz and 588 polarimeters at 150 GHz, with each polarimeter consisting of a dual transition edge sensor (TES) bolometers. The full complement of 150 GHz detectors consists of 7 arrays of 84 ortho-mode transducers (OMTs) that are stripline coupled to two TES detectors per OMT, developed by the TRUCE collaboration and fabricated at NIST. Each 90 GHz pixel consists of two antenna-coupled absorbers coupled to two TES detectors, developed with Argonne National Labs. The 1536 total detectors are read out with digital frequency-domain multiplexing (DfMUX). The SPTpol deployment represents the first on-sky tests of both of these detector technologies, and is one of the first deployed instruments using DfMUX readout technology. We present the details of the design, commissioning, deployment, on-sky optical characterization and detector performance of the complete SPTpol focal plane.
KEYWORDS: Control systems, Sensors, Telescopes, Data archive systems, Antennas, Human-machine interfaces, Bolometers, Data acquisition, Detection and tracking algorithms, Data storage
We present the software system used to control and operate the South Pole Telescope. The South Pole Telescope is
a 10-meter millimeter-wavelength telescope designed to measure anisotropies in the cosmic microwave background
(CMB) at arcminute angular resolution. In the austral summer of 2011/12, the SPT was equipped with a new
polarization-sensitive camera, which consists of 1536 transition-edge sensor bolometers. The bolometers are read
out using 36 independent digital frequency multiplexing (DfMux) readout boards, each with its own embedded
processors. These autonomous boards control and read out data from the focal plane with on-board software
and firmware. An overall control software system running on a separate control computer controls the DfMux
boards, the cryostat and all other aspects of telescope operation. This control software collects and monitors
data in real-time, and stores the data to disk for transfer to the United States for analysis.
The Keck Array (SPUD) is a set of microwave polarimeters that observes from the South Pole at degree angular scales in search of a signature of Inflation imprinted as B-mode polarization in the Cosmic Microwave Background (CMB). The first three Keck Array receivers were deployed during the 2010-2011 Austral summer, followed by two new receivers in the 2011-2012 summer season, completing the full five-receiver array. All five receivers are currently observing at 150 GHz. The Keck Array employs the field-proven BICEP/ BICEP2 strategy of using small, cold, on-axis refractive optics, providing excellent control of systematics while maintaining a large field of view. This design allows for full characterization of far-field optical performance using microwave sources on the ground. We describe our efforts to characterize the main beam shape and beam shape mismatch between co-located orthogonally-polarized detector pairs, and discuss the implications of measured differential beam parameters on temperature to polarization leakage in CMB analysis.
C. Sheehy, P. Ade, R. Aikin, M. Amiri, S. Benton, C. Bischoff, J. Bock, J. Bonetti, J. Brevik, B. Burger, C. Dowell, L. Duband, J. Filippini, S. Golwala, M. Halpern, M. Hasselfield, G. Hilton, V. Hristov, K. Irwin, J. Kaufman, B. Keating, J. Kovac, C. L. Kuo, A. Lange, E. Leitch, M. Lueker, C. Netterfield, H. T. Nguyen, R. Ogburn, A. Orlando, C. L. Pryke, C. Reintsema, S. Richter, J. Ruhl, M. Runyan, Z. Staniszewski, S. Stokes, R. Sudiwala, G. Teply, K. Thompson, J. E. Tolan, A. Turner, P. Wilson, C. L. Wong
The Keck Array is a cosmic microwave background (CMB) polarimeter that will begin observing from the South
Pole in late 2010. The initial deployment will consist of three telescopes similar to BICEP2 housed in ultracompact,
pulse tube cooled cryostats. Two more receivers will be added the following year. In these proceedings
we report on the design and performance of the Keck cryostat. We also report some initial results on the
performance of antenna-coupled TES detectors operating in the presence of a pulse tube. We find that the
performance of the detectors is not seriously impacted by the replacement of BICEP2's liquid helium cryostat
with a pulse tube cooled cryostat.
A new 10 meter diameter telescope is being constructed for deployment
at the NSF South Pole research station. The telescope is designed for
conducting large-area millimeter and sub-millimeter wave surveys
of faint, low contrast emission, as required to map primary and secondary anisotropies in the cosmic microwave background. To achieve the required sensitivity and resolution, the telescope design employs an off-axis primary with a 10 meter diameter clear aperture. The full aperture and the associated optics will have a combined surface accuracy of better than 20 microns rms to allow precision operation in the submillimeter atmospheric windows. The telescope will be surrounded with a large reflecting ground screen to reduce sensitivity to thermal emission from the ground and local interference. The optics of the telescope will support a degree field of view at 2mm wavelength and will feed a new 1000-element micro-lithographed planar bolometric array with superconducting transition-edge sensors and frequency-multiplexed readouts. The first key project will be to conduct a survey over &dbigwig;4000 degrees for galaxy clusters using the Sunyaev-Zel'dovich Effect. This survey should find many thousands of clusters with a mass selection criteria that is remarkably uniform with redshift. Armed with redshifts obtained from optical and infrared follow-up observations, it is expected that the survey will enable significant constraints to be placed on the equation of state of the dark energy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.