The NRC Herzberg
News E-Cass Report
These reports appear
in each issue of E-Cass with the goal of informing the Canadian astronomical
community on the activities at NRC Herzberg.
Feedback is welcome
from community members on how NRC-Herzberg is doing in fulfilling our mandate
to “operate and administer any astronomical observatories established or
maintained by the Government of Canada” (NRC
Act).
General News
May 13, 2014 was the 100th anniversary of
the public announcement that a 72-inch telescope, now named the Plaskett telescope, would be built in Victoria. As the
front-page article of the Daily Colonist proclaimed “Huge Telescope for Victoria”. Indeed, the Plaskett would be the world’s largest operating telescope for a few months before the
100-inch on Mt. Wilson was successfully put into operation.
The May 13, 1914 edition of the Daily Colonist is online for anyone
interested in reading the coverage of this monumental event in the history of
Canadian astronomy.
Space
Astronomy
The James Webb Space Telescope
(JWST), a joint project of NASA, ESA and CSA, continues to make excellent
progress. Most of the flight hardware has been manufactured and the focus is
now firmly on integration and testing with a scheduled launch in 2018. All four
science instruments have been delivered and integrated into the Integrated
Science Instrument Module (ISIM). The second cryo-vacuum
test of the ISIM begins in June 2014. Over a three month period all the
instruments will get extensively tested to verify their performance. The
enormous Chamber A at Johnson Space Center in Houston, once used to test the
Apollo spacecraft, has been extensively renovated in preparation for testing of
the telescope in 2017.
Figure
1
The integrated ISIM with all four science instruments labelled, shortly before the
second cryo-vacuum test at Goddard Space Flight
Center. Photo credit: NASA.
On the Canadian side, there is considerable activity supporting the
second cryo-vacuum test this summer and completing
the fabrication and testing of new detectors, dual wheel motors, a grism and an electronics board which will be swapped into
the FGS/NIRISS instrument later this year. The science team is active
simulating the main NIRISS observing modes and making plans for commissioning,
calibration and operations.
Optical
Astronomy
Plaskett Telescope
For more than a decade it has been known that the quality
and durability of the aluminum mirror coating on the Dominion Astrophysical
Observatory’s (DAO) 1.8-m Plaskett telescope’s
primary mirror has been less than optimal because of an aging vacuum coating
facility that produced coatings with poor adhesion properties and, as a result,
relatively poor reflectivities. Unfortunately, because of commitments of NRC
Herzberg staff to other higher priority international instrumentation and
telescope projects, scheduling time for the DAO Plaskett and 1.2-m telescopes is not always easy! However, this spring a lengthy project to upgrade the telescope’s
coating chamber was finally completed and in mid-May, just after its 96th
anniversary of first light, the Plaskett telescope
was finally treated to a new high-quality aluminum coating for its primary
mirror. Shortly thereafter the Cassegrain secondary was also swapped for a spare with a
fresh enhanced aluminum coating with a reflectivity of approximately 95%. The accompanying photograph shows the primary
mirror shortly after it was recoated and before
it was returned to service.
Observations
obtained shortly after the mirror installations suggest that the throughput of
the telescope has improved by a factor of 3 to 3.5 with the new mirror
coatings. No, that is not a typo - the
mirrors were not in good shape at all! A likely incomplete list of NRC Herzberg staff involved in the coating
project includes Jim Jennings, Gordon Hnylycia, Colin Ganton, Felipe Miranda, Les Saddlemyer and Dmitry Monin. I was lucky enough to be the first user of the ‘new’ telescope on the
May long weekend and can confirm that the telescope is acting like a 2-m class
facility again. NRC Herzberg’s long-term
‘business plan’ now also includes an annual commitment of staff for work on the
two DAO telescope primary mirrors and so we expect each telescope to receive
fresh aluminum in alternating years in the future.
The focus of our work on the Plaskett telescope will now be concentrated on completing efforts to enable robotic
operation of the facility using lessons learned from automation of the 1.2-m
telescope and high-resolution McKellar spectrograph.CADC
After 23 years of operation, the HST archive is still
alive and well at CADC. During those 23 years, technological and operation
improvements have made the HST archive more scientifically relevant than ever.
Astronomy
Technology
GRACES:
A successful first light for the longest astronomical fibre system
The
GRACES (Gemini Remote Access to CFHT ESPaDOnS ) experiment completed its very successful first light run
( May 6th to May 19th) with the performance of the system exceeding the
predictions. GRACES is an innovative partnership between Gemini
Observatory, the Canada-France-Hawaii Telescope (CFHT) and National Research
Council (NRC) Herzberg in Victoria B.C. with the development being done across
the three organizations.
GRACES is a groundbreaking experiment in fibre fed spectroscopy and intra-observatory facility
cooperation. The experiment links the Gemini North telescope to the ESPaDOnS ( Echelle SpectroPolarimetric Device for the Observation of
Stars) spectrograph at CFHT with a 270m fibre link.
Light is injected into the fibre at the Gemini
telescope via an injection module installed in the GMOS spectrograph ( replacing the GMOS intergral field unit). Two 270m fibers carry the light (
star & sky ) from the Gemini telescope to the CFH dome then to the ESPaDOnS spectrograph in the CFH Coude room. A receiver unit feeds the light into the spectrograph providing two
operation modes: a star/sky mode with a resolution of 37k and a star only mode
with a resolution of 55k. The receiver unit is designed to allow a remote
operation change between ESPaDOnS being feed by the
CFHT telescope and ESPaDOnS being feed from Gemini
telescope.
NRC Herzberg was responsible for the design and
construction of the opto-mechanical system, with the fibre cable being developed in partnership with FiberTech Optica in Kitchener
Ontario. NRC Herzberg delivered the system to the telescopes at the end
of april with the fibre being installed between the two telescopes on April
24th, and the integration of the opto-mechnical systems at the telescope starting on April 27th. First light was achieved on
May 6th, on schedule! ( a testament of the superb
teamwork between Gemini, CFHT and NRC Herzberg staff). On sky performance test
have confirmed the spectrograph resolutions of 37k/55k and transmission has
exceed expectations with a SNR=1 sensitivity as high as 22 mag per 1 hour
exposure. The combination of the silver coatings on Gemini,
the superb red efficiencies on ESPaDOnS ( with the prism cross dispersers and the silver coated relay mirrors) make
GRACES a competitive 8m class spectrograph over the designed wavelength
range ( 450nm to 1000nm).
Gemini
High-Resolution Optical Spectrograph (Ghost)
Sometimes
GHOSTs can be real!
Merriam
Webster dictionary defines a ghost as: a disembodied soul; especially : the
soul of a dead person believed to be an inhabitant of the unseen world or to appear
to the living in bodily likeness.
It
would be fair to say that the notion of ghosts has endured from generation to
generation and has managed to capture the imagination of people from many
cultures across the planet. Perhaps more famously ghosts are associated with
movies; where they appear as typically malicious spirits bent on haunting an
innocent and unsuspecting family which moves into an otherwise perfectly
seeming house in some innocuous suburban neighbourhood.
It’s
this notion of “haunting” that brings us to the point. For the past few years
the Canadian and broader astronomical community have been haunted by the
promise of adding a workhorse spectrograph to the Gemini suite of tools. A few
years back just such an instrument seemed poised for design, fabrication and
deployment but like all good stories there was a twist. The Gemini
High-resolution Optical Spectrograph (GHOS) seemed set to launch. Led by a
coalition of Australian partners (Australian Astronomical Observatory (AAO),
Australian National University (ANU) and KiwiStar Optics Ltd.) the project seemed set to roll forward.
Unforeseen
in this mix was that KiwiStar would be bought out by
a parent company which would lose appetite to continue with the GHOS project.
For two years GHOS appeared dead in its track but like all good spirits, its
memory haunted the community until a new partner emerged to replace KiwiStar and provide the spectrograph portion of the
instrument. This partner is Canada’s very own NRC Herzberg. In an effort led by
the AAO, NRC Herzberg has entered the mix as a sub-contractor tasked with
designing and building the spectrograph portion of the instrument. ANU will
continue-on in their role of leading the software design and AAO will provide
the Cassegrain interface and the fiber feed
connection to the spectrograph routed through their slit assembly.
GHOS
was originally envisioned as a 4-arm white pupil Echelle spectrograph - GHOST, as it has been officially renamed, launched mid-April
2014. This workhorse is being designed to provide high-resolution spectra with
large simultaneous wavelength coverage and has the ambitious aim of being
on-sky mid-2017.
The
NRC team has recently returned from upbeat kickoff meetings with our
counterparts in Sydney as well as the Gemini project leadership and is excited
to be moving forward with this long awaited project! The NRC team will feature
John Pazder as Optical design lead and Project Engineer; Project Scientist:
Alan McConnachie; Mechanical Engineering: led by
Andre Anthony with support from Ivan Wevers;
Detectors: led by Greg Burley; Software: led by Jennifer Dunn with support from
Bob Wooff. The project will be managed by Eric
Chisholm. As the GHOST project comes back to life we’ll update you frequently
on both NRC’s and the overall progress of the project.
Facility
Adaptive Optics System – NFIRAOS
In 2013, two major trade studies took place to re-examine
the NFIRAOS baseline configuration since its Preliminary Design Update (PDU) in
2011.
1. TRADE-OFF STUDIES FOR A COMMON-SIZE OF
DEFORMABLE MIRROR
During 2013, we undertook an extensive trade study on
changing the size of one or both deformable mirrors (DMs), with the aim to
reduce risk and cost by have two common size DMs that are interchangeable with
one conjugates at ground level, i.e. DM0, and one conjugates at 11.2 km, i.e.
DM11.2. By having common DMs, the DM0 position will always be filled should one
of the two DMs were to break.
On the face of it, if DM0 were to break and with no
replacement, then NFIRAOS would be inoperable and resulting severe setback
during integration and commissioning. Full commissioning and even first light
on TMT (defined as capturing of the first diffraction-limited images) would be
delayed for several years until a replacement DM can be obtained. Budgets and
manufacturing capacity make it unlikely that we actually would have a spare DM
on-hand during this critical phase. In principle, if DM11 broke, a flat mirror
could replace it and NFIRAOS will operate in LTAO mode. Although LTAO mode falls far short of the
scientific aspirations of MCAO mode, meaningful astronomical observations are
still possible.
For the above reason, we studied the possibility of
redesigning NFIRAOS with two DMs of the same size. In that way, DM11.2 could
serve as a spare for DM0. We considered having both DMs equal in size to DM0,
or both equal to the DM11.2 size, or to some intermediate size. We examined a
series of variants that changed the size of the pupil and metapupil (Table 1). The main challenge was to find an optical prescription that meets
all the existing requirements such as distortion, exit pupil location, and
instrument back focal distance etc. while maintaining the current instrument
space envelope of NFIRAOS.
Table 1 Options of the common size DM trade study –
actuator counts indicate the physical DM diameters and the beam size required
to maintain “clear aperture” with Option A representing the baseline PDU
design. (Note: DM0 is mounting on the Tip Tilt Stage which provides tip/tilt
offload from DM0.)
Initially we believed if the ground-conjugated DM
diameter were larger than the baseline DM0 of 63x63 actuators, then we would
need to replace the existing Tip Tilt Stage with a larger unit, because DM0
nests inside. Since we already have acquired the Tip Tilt Stage and plan to
upgrade its electronics before installing it in NFIRAOS, buying a new stage
would be expensive and impractical. However, during the trade study under
Option C, we realized that DM11 could be mounted on the Tip Tilt Stage with an
adaptor. The adaptor would move the optical surface forward, but to compensate,
the stage itself could be moved back onto a second mounting location in
NFIRAOS. The stage would need additional counterweights to balance the larger
DM11 whose centre of gravity would then lie ahead of the stage tilt axis. We also verified that the additional moment
of inertia would be within the capability of the existing voice-coil actuators
to provide sufficient correction bandwidth and yield acceptable heat
dissipation.
As we tried to package the optical variants, we
recognized that increasing either DM diameter was expensive, and that the opto-mechanical effort to shrink DM11 and/or expand DM0 was
running into space constraints (Figure 1). The follow-on effort and risk to
rework the structure, subassemblies and thermal optics enclosure for NFIRAOS
posed a large risk to our schedule and resources. Taken together with the
realization that DM11 can act as a temporary spare for DM0, we decided to retain the PDU baseline
design with DM0 having 63x63 actuators and DM11 having 76x76 actuators as
outlined in Option C.
Figure
2
Comparison of footprint of the opto-mechanical structure: Option E exceeds the NFIRAOS
space envelope as defined by Option A
2. REAL-TIME COMPUTING ARCHITECTURE
TRADE-OFF STUDY
As well, during 2013, we revisited computing architecture
for the NFIRAOS Real Time Controller which processes the WFS data and creates command
vectors to control the DM shape for AO corrections. At the last 2011 review,
the baseline RTC used custom boards with Field Programmable Gate Arrays (FPGA)
running an iterative algorithm. Because the NFRIAOS RTC arithmetic operations
required are nearly one hundred times larger than any existing AO systems, the
iterative algorithm technique drastically reduced the number of computations
compared with the standard Matrix-Vector Multiplication (MVM) method. However,
the custom nature of these FPGA boards and the labour-intensive
specialist programming required meant that this approach was costly and
specialized acknowledge are difficult to develop and maintain.
Meanwhile advances in commercial computation hardware
offered hope that a RTC architecture using conventional CPU boards and techniques
is within reach. We benchmarked with MVM on clusters of CPU servers, with and
without accelerators such as Graphical Processing Units (GPUs), and Intel Xeon
Phi. Half a dozen dual-GPU boards can do the required computation in ~ 900 µs
with jitter of a few 10s of µs. This architecture provides sufficient
performance margin when compared with the required frame period of 1.25 ms. While we found a server
cluster with 12 Xeon Phis computes most frames quickly enough but occasionally
their internal operating system suspends the RTC work and resulting in
unacceptable jitter of up to 10 ms in the worst case. Since real-time control
is not the marketing niche for the Xeon Phi processor and its operating system
is proprietary with no real-time upgrade expected, we have ruled it out for our
RTC.
Using CPUs alone with a real-time patch applied to the
Linux operating system, our benchmark work showed that six servers, each with
two dual-CPU motherboards can process the WFS data and create DM command
vectors in typically 700 µs, with low jitter resulting in a worst case of
slightly more than 800 µs as shown in Figure 2. Although GPU accelerators can
reduce the number of server chassis and total power consumption, we concluded
that for ease of development and minimizing the variety of hardware and
software components in the RTC made an all-CPU solution attractive, especially
when considering maintenance over the lifetime of NFIRAOS. Our baseline is now an all CPU architecture.
Figure
3
2 Benchmark timing of CPU-based RTC -
For this benchmark, simulated pixels representing half of the LGS WFS are
streamed in a 500 µs burst over 10 Gb/s Ethernet to one server which processed
pixels (computing slopes and statistics) and then applied MVM
Contributions from David Bohlender,
Eric Chisholm, John Pazder, Kei Szeto and Chris Willott
