The AzTEC mm-Wavelength Camera

0.67 MB·

by  G. W. Wilson

Mon.Not.R.Astron.Soc.000,1–??(2007) Printed2February2008 (MNLATEXstylefilev2.2)
The AzTEC mm-Wavelength Camera
G.W. Wilson,1 J.E. Austermann,1 T.A. Perera,1 K.S. Scott,1 P.A.R. Ade,2
J.J. Bock,3 J. Glenn,4 S.R. Golwala,5 S. Kim,6 Y. Kang,6 D. Lydon,1
8 P.D. Mauskopf,2 C.R. Predmore,7 C.M. Roberts,1 K. Souccar1 and M.S. Yun1
0
1Department of Astronomy, Universityof Massachusetts, Amherst, MA 01003.
0
2Physics and Astronomy, Cardiff University,5, The Parade, P.O. Box 913, Cardiff CF24 3YB, Wales, UK.
2
3Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109.

n 4Centerfor Astrophysics and Space Astronomy, 389-UCB, University of Colorado, Boulder, CO, 80309.
a 5California Institute of Technology, 1200 East California Boulevard, MC 59-33, Pasadena, CA 91125.
J 6Astronomy & Space Science Department, Sejong University,98 Kwangjin-gu, Kunja-dong, 143-747, Seoul, South Korea.

7 7Predmore Associates, South Deerfield, MA 01373.
1

]
h
p 2February2008
-
o
r ABSTRACT
t
s AzTECisamm-wavelengthbolometriccamerautilizing144siliconnitridemicromesh
a detectors. Herein we describe the AzTEC instrument architecture and its use as an
[
astronomicalinstrument.Wereportonseveralperformancemetricsmeasuredduringa

1 threemonthobservingcampaignattheJamesClerkMaxwellTelescope,andconclude
v with ourplans forAzTEC asa facility instrumentonthe LargeMillimeter Telescope.
3
8 Key words: instrumentation:detectors, submillimetre, galaxies:starburst, galax-
7 ies:high redshift
2
.
1
0
8 1 INTRODUCTION is well-suited for extragalactic surveys of dusty, opti-
0
cally obscured starburst (or AGN-host) galaxies detected
: In the era of ALMA,the scientific niche for large mm-wave
v by their sub-mm/mm emission (so-called Sub-Millimetre
telescopes comes in the ability to use array receivers to
i Galaxies, or SMGs). As the far-infrared (FIR) peak of
X make large surveys of the sky. The last decade has seen
the cold dust emission is increasingly redshifted to mm-
r a series of major advances in mm-wavelength bolometric wavelengths with increasing distance, AzTEC is equally
a instruments and it is now commonplace for ground-based
sensitive to dusty galaxies with redshift z > 1. The
telescopes to have mm/submm wavelength cameras with
AzTEC/JCMT05B surveys of the Lockman Hole and the
tens to hundreds of detectors: for example, SCUBA on the
SubaruDeepField(Austermannetal.inpreparation),COS-
JCMT (Holland et al. 1999), MAMBO on the IRAM 30 m
MOS (Scott et al. 2008), and GOODS-North (Perera et al.
telescope (Kreysaet al.1998),and Bolocam on theCaltech
inpreparation)representthelargestSMGsurveyswithuni-
Submillimeter Observatory (Haig et al. 2004; Glenn et al.
formhighsensitivity(1σ 1mJy)everconducted,success-
2003). AzTEC is a millimetre-wavelength bolometer cam- ∼
fullydemonstratingthesuperbmappingspeedandstability
era designed for theLarge Millimeter Telescope (LMT). Its
of the instrument and providing a new insight on the clus-
144siliconnitridemicromeshbolometersoperatewithasin-
tering and cosmic variance of the SMG population. In a
glebandpasscenteredateither1.1,1.4,or2.1mm,withone
similar vein, AzTEC’s high mapping speed and high sensi-
bandpassavailable perobserving run.AzTEC was commis-
tivity make it an interesting instrument for studies of cold
sioned at 1.1 mm during an engineering run at the James
dust in the Milky Way and in nearby galaxies.
Clerk Maxwell Telescope (JCMT) in June 2005 and com-
pletedasuccessfulobservingrunattheJCMTfromNovem- Configured for 1.4 mm and 2.1 mm observations,
ber 2005-February 2006 during the JCMT05B semester. In AzTECwillbetunedtomakehighresolutionimagesofthe
2007 the instrument was installed on the 10 m Atacama Sunyaev-Zel’dovicheffect in clustersof galaxies. AzTEC on
′′
Submillimeter Telescope Experiment (ASTE) where it will the LMT will have a per-pixel resolution of 10 at 2.1 mm
resideasafacility instrumentthrough2008.AzTECwillbe (5′′ at 1.1 mm) and will therefore be an unprecedented in-
installed on the LMT in early 2009. strumentfor thestudyof theenergetics of thefree electron
gas in clusters.
Inits1.1mmwavelength configuration,AzTECissen-
sitive to the Rayleigh-Jeans tail of the thermal contin- In thispaper we report on thedesign and performance
uum emission from cold dust grains. Consequently, AzTEC of the AzTEC instrument. In Section 2, we describe the
2 G.W. Wilson et al.
port is along the optical axis and has a resonant frequency
>400Hz.TheUCandICtemperaturesshowednocorrela-
tion with cryostat tilt oreffectiveatmosphereoptical depth
duringtheJCMT observations.
ThecustomdesignedAzTECcryostatisa’wet’system
with liquid helium (23 liters) and liquid nitrogen (26 liters)
tanks.Thedesignandpreparationofthecryostatresultedin
operational (full loading, including daily fridge cycles) hold
times of 3 days for liquid helium and over 7 days for liq-
uid nitrogen. The cryogen tanks are concentric annuli with
nearoptimumlength/diameterratiostomaximizetheirhold
times. The innerdiameter of thehelium tank is 15.25 cm –
afeaturethataddsvaluablevolumeandaccessibility tothe
shielded 4 K workspace without compromising cryogen ca-
pacity.
One challenge of working with semiconductor thermis-
torsistheneedtoimpedancetransformusingwarmJFETs.
Figure 1.A cut-away view of the AzTEC cryostat. Mechanical AzTEC’slow-noiseU401JFETsamplifieraresuspendedin-
supportshavebeenremovedforclarity.
sidetheinnerdiameteroftheliquidheliumtankbyasetof
reentrant G-10 tubes with tuned conductivity to allow the
componentsoftheinstrumentanditsconfiguration.Theob- stage to self-heat to 130 K where the JFETs have empiri-
serving software, mapping strategies, and practical observ- cally been found to exhibit low voltage and current noise.
ing overheads are addressed in Section 3. We describe the Anintermediatestageofthesuspensionissunktothe77K
calibration ofAzTECdatainSection 4andinSection5we bathandinterceptstheheatdissipatedbytheJFETs.These
list the sensitivity and mapping speeds of AzTEC as mea- 130 K and 77 K components are radiatively insulated with
suredattheduringtheJCMT05B run.Weconcludewitha severallayersofmulti-layerinsulation(MLI).Theinnerwall
brief discussion of the futureof AzTEC in Section 6. ofthe4Khelium tankispaintedwith anIR-blackpaintto
absorb stray thermal radiation.
Thecryostatisdesigned tominimizeEMI/RFIsuscep-
tibility. Two nested high conductivity faraday shields are
2 SYSTEM DESIGN
formed through which all electrical connections pass via
Acut-awayviewoftheAzTECinstrumentisshowninFig- imbedded pi-filters. The “cleanest” volume is bounded by
ure 1. Each subsystem labeled in the figure is described in theinnershieldwhichsurroundsthe4Kworkvolume.Cus-
detail below. tomcompactin-linecryogenicpi-filterswereconstructedfor
AzTEC in coordination with the manufacturer. The outer
faraday cage is formed by the cryostat vacuum jacket and
2.1 Cryostat and Cryogenics
electronics enclosure.
The AzTEC detector array is cooled using a three-stage,
closed-cycle 3He refrigerator (Bhatia et al. 2000) mounted
2.2 Detectors
to a liquid cooled 4 K cryostat. Operation of the AzTEC
refrigeratorisfullyautomatedandremotelycontrolled.The The AzTEC array is a 76 mm diameter monolithic sili-
refrigerator takes approximately 140 minutes to cycle. Op- conwafercontaining151silicon nitridemicro-mesh(spider-
timization of the fridge cycle has led to stable hold times web)bolometers with neutrontransmutation doped(NTD)
well over 24 hours at sea level with a full optical load. The Ge thermistors (Bock et al. 1996; Mauskopf et al. 1997;
operationalholdtimeoftherefrigeratorwas36hoursatthe Turneret al. 2001). The wafer is organized into six pie
JCMT(4092m)and42hoursatthehigherelevationASTE shapedregions(hextants)thattogethercontainthe144op-
telescope (4860 m). ticallyactivebolometersandtheirwiring.Thedetectorsare
Therefrigerator’s ultra-cold(UC)stage,operatingata arranged in a close-packed hexagonal configuration with a
temperatureofapproximately250mK(256.5 1.9 10mK spacing set by the detector feed-horn aperture of 5 mm.
± ±
where the 10 mK uncertainty comes from uncertainty in Thebias for each detector is provided viaa symmetric pair
the calibration of the GRT), provides the thermal sink to of 10 MΩ resistors contained in a separate module located
the bolometer array. An intercooler (IC) stage, operating closetothedetectorarray.The7“blind”bolometersof the
at approximately 360 mK, heat sinks an intermediate tem- array are not biased or read out.
perature stage of the detector array support structure to Thebolometerbehaviorisdictatedbythreesetsofpa-
thermally intercept heat flowing along the mechanical and rameters:thethermistorproperties,theamountofabsorbed
electricalconnectionstotheUCstage.BoththeICstageand optical and electrical power, and the thermal link between
the detector array assembly are thermally isolated and me- the bolometer and the cold heat sink of the cryostat. The
chanically supported using two sets of short, hollow Vespel thermal link is defined by a Au film deposited on one of
standoffs in a radial configuration. The resulting structure themesh support legs connected to thewafer substrate.Its
is extremely compact and stiff and the symmetry of the heat conductance has the form g0Tβ with β in the range
system prevents net motion of the optical axis under ther- 1.2-1.6 (Haig et al. 2004; Glenn et al. 2003) where T is the
mal contraction. The weakest vibrational mode of the sup- bolometeroperatingtemperature.Thethermistorresistance
The AzTEC mm-Wavelength Camera 3
has the form R0exp ∆/T with targeted R0 and ∆ values ellipsoidfocallength 645.2mm
of100Ωand42Krepspectively.Inprinciple,g0aswellasthe ellipsoidreflectionangle 37◦
electrical bias power are chosen to minimize detector noise
UHMWPElensfocallength 163.6mm
given a known optical loading. For the AzTEC array we f/#cass 12.0
choseaconservativelyhighg0 thatresultsinaheatconduc- f/#Lyot 11.8
tanceof167pW/Kat300mKtomitigatetheeffectsofany f/#horns 3.2
unanticipated excess optical loading and as a compromise Lyotstopdiameter 50mm
for the different amounts of loading expected in the three edgetaperatLyotstop -5.2dB
possible optical passbands. For thischoice, thenoise equiv- imageofprimarymirrordia. 52mm
planewavecouplingefficiencyatLyotstop 0.62
alent power (NEP)predicted from ourbolometer model for
theopticallyloadeddetectorsis5 7 10−17 W/√Hz,which detector spacing 1.4fλ
− ×
is comparable but sub-dominant to the photon-noise back-
Table 1. AzTEC design optical parameters for coupling to the
groundlimit(BLIP)inthe1.1mmpassbandforanticipated
JCMT. The calculation of the plane wave coupling efficiency at
operating conditions. The expected detector time constant,
theLyotstopismadeassuminggaussianopticsandfollowingthe
τ, of 3-4 ms also applies over the range of reasonable oper- perscriptionof(Goldsmith1998).
atingconditions.Thethermistorresistanceandbiasresistor
values are chosen such that the detector and photon noise
are dominant over noise sources from circuit elements fur- 52mmandsotheLyotstop,whichhasadiameterof50mm,
therdown thereadout chain. provides a cold guard ring to minimize the spillover at the
Once AzTEC is installed on a telescope, the only free primary mirror edge while allowing an aggressive edge ta-
parameter for optimizing sensitivity is the amount of elec- per.Thefoldingflatandellipsoidalcouplingmirroraresized
trical (bias) power dissipated at the bolometer and hence suchthattheedgetaperoftheoutermostbeamoneachsur-
the bolometer operating temperature T. Following Mather face is less than -30dB. Both mirrors and the cryostat are
(1984), thebalance between phonon noise, which rises with supported rigidly to minimize optical microphonic pickup.
T, and Johnson noise, which falls steeply with rising T, re- Inside the cryostat, the detector array assembly is sit-
sultsinanoptimumvalueofthebiasvoltageforagivenset uated behind a 4 K biconvex aspheric lens fashioned from
of detector properties and a given optical loading. In prac- ultra-highmolecular weightpolyethylene(UHMWPE)with
ticesincethereisasmallspreadindetectorpropertiesacross index of refraction, n = 1.52. The lens surface is grooved
the array and a slowly varying, but unpredictable, optical with a series of concentric trenches of depth λ/4√n which
loadingduetotheatmosphere,wefixedthethermistorbias results in an anti-reflection surface tuned for the system
amplitude for all detectors in all hextants at 62.5 mV over passband.Consideringtheopticalpathinreverse,chiefrays
theentireJCMT05B observingrun.Foranatmosphericop- fromeachofthethedetectors’feedhornsarefocusedbythe
ticaldepthat225GHz,τ225,of0.1,thisconservativelyhigh lens onto thecenter of the blackened 4 K Lyot stop so that
bias results in a sensitivity 10% worse than expected for all detectors illuminate the 4 K edge of the Lyot stop with
a bolometer that has the design parameters. As shown in an edge taper of 5.2 dB in power. A 4 K folding mirror
−
Section 5, nightly load curves, noise estimates, and beam betweenthelensandtheLyotstopallows foralongoptical
map observations during the JCMT05B run indicate that pathinacompactconfigurationtoaidinkeepingtheoverall
the detector sensitivity was near optimum over the entire structuremechanically rigid and isothermal.
observing run for this choice of bias. The250mKbolometerarrayismechanically andther-
mally supported between monolithic gold-plated aluminum
arrays of tuned backshorts and feedhorns in a manner
2.3 Optics identical to that of the Bolocam instrument (Glenn et al.
2003). The array of backshorts and integrating cavities sits
A schematic diagram showing the relative positions of the
λ/4 behind the detectors in order to optimize detector
opticalelementswithAzTECintheJCMTreceivercabinis
absorption and minimize optical cross-talk between detec-
showninFigure2.Ourdescriptionofthecouplingopticswill
tors (Glenn et al. 2002). The array of single-moded conical
start from the JCMT’s tertiary mirror unit (TMU) located
feedhornsliesinfrontofthedetectorsandproducesasetof
insidethereceivercabin.AcompletelistingofJCMToptical
quasi-gaussian beams, each with gaussian beam radius, w ,
specificationsmaybefoundontheJCMTwebsite.Alistof a
of 1.84 mm at the horn aperture for the1.1mm wavelength
optical constants for the AzTEC instrument and its JCMT
configuration. This corresponds to an f/# for each horn of
coupling optics is given in Table 1.
f/3.2.
At the JCMT, photons from the secondary mirror re-
flectoffoftheTMUwhichislocatedbehindthevertexofthe
primarymirror.FortheAzTECsystemtheTMUisoriented
2.3.1 Optical Filtering
to direct the incoming beam upwards at an angle of 53.2◦
withrespecttothevertical.Aflatfoldingmirrorthendirects An exploded view of the internal optical layout including
thebeamdownwardstoanellipsoidalmirrorwhichconverts thefiltersisshown inFigure3.Ashortstubofsingle-mode
thef/12beamtoanf/11.8beamanddirectsthebeaminto waveguide (3.4 mm length for the 1.1 mm passband con-
the cryostat. The JCMT’s cassegrain focus resides between figuration) at the detector end of each horn provides the
the folding flat and the ellipsoidal mirror. The AzTEC de- high-pass filter to the system passband. All of the low-
warislocatedsuchthatablackenedLyotstop,ataphysical pass filters in the system are quasi-optical metal mesh fil-
temperature of 4 K, is coincident with an image of the pri- ters consisting of a series of resonant metallic meshes sepa-
mary mirror. The image of the primary has a diameter of ratedbytransmissionlinesections.Theresonantmeshfilters
4 G.W. Wilson et al.
Figure 2. Schematic of AzTEC mounted in the JCMT receiver cabin with the associated coupling optics. All filters and mechanical
supports have been omitted for clarity. Zemax produced rays are shown for the central and outermost detectors in the plane of the
drawingtoillustratetheopticalpathofthevariousbeams.TheCassegrainfocusoftheJCMTislocatedattheconvergenceofthebeams
justbelowthefoldingflat.
are fabricated from thin copper films deposited on plastic
(polypropylene or mylar) substrates and patterned into in-
ductive or capacitive grids (Tucker& Ade 2006; Adeet al.
2006). In the 1.1 mm configuration, a set of low-pass fil-
ters, one rolling off at 360 GHz and the other rolling off
at 310 GHz, are mounted to the feedhorn array and define
the upper portion of the passband. For configurations with
otherbandcenters,thebackshortarray,thefeedhornarray,
and these two low-pass filters are replaced.
Further optical filtering is required to suppress har-
monicresponsefromthe250mKlow-passfilters.Alow-pass
capacitivemeshfilter(390GHzedge)andananti-reflection
coated 1050 GHz low-pass filter are mounted on the tele-
scopesideoftheLyotstopat4K.A77K,540GHzlow-pass
filter is mounted to the liquid nitrogen shield 2 cm further
downtheopticalaxis.Finally,aroomtemperatureZotefoam
PPA30(ZotefoamsPLC)windowprovidesIRscatteringand
Figure 3. An exploded layout of the internal AzTEC optics.
a vacuum seal that is transparent at 1.1 mm wavelengths
Components are: (a) integrated detector array, backshorts, and
tobetterthan 1% according toin-lab measurements with a conicalfeedhorns(250mK);(b)310GHzlow-passfilter(250mK);
FourierTransformSpectrometer.Theresultingsystempass- (c) 360 GHz low-pass filter (250 mK); (d) bi-convex ultra-high
band is shown in Figure 4. molecular weight polyethylene lens (4 K); (e) folding flat (4 K);
Measured optical characteristics are described further (f)50mmdiameterLyotstop(4K);(g-h)390GHzand1050GHz
in Section 5.1. low-passfilters(4K).Notshownareanadditional77K540GHz
low-passfilterandaZotefoam PPA-30cryostatwindow.
2.4 Signal Chain
A schematic for the detector signal chain is shown in Fig- Thebiasfrequencyof200 Hzischosen tobefast compared
ure 5. Each of the 144 bolometers is pseudo-current biased tothedetectortimeconstant,τ >3ms,butslowenoughto
by a differential 200 Hz sine wave produced by a digital to avoid excessive attenuation from the effective low-pass RC
analogconverter(DAC)–oneDACperhextant–andapair filter arising from the bolometer/load resistor combination
of10MΩloadresistors.Noiseinthebiasiscorrelatedacross ( 3 MΩ) and the net parasitic capacitance of the signal
∼
alldetectorsofahextantandisremovedinthedataanalysis. lines and JFET gates (Cparasitic 60 110 pF).
∼ −
The AzTEC mm-Wavelength Camera 5
Figure 5.Schematic ofthedetector readoutchain.Allanalogelectronicsarereplicated144timesinthesysem.
precision 24-bit delta-sigma A/D converters (Texas Instru-
ments ADS1252) which produce an output data stream at
a sampling rate of 16 kHz.
Once digitized, a Field Programmable Gate Array
(FPGA) serializes each hextant’s data into a single data
stream.Allprocessinguptothispointhappensinacustom
electronics enclosure attached to the cryostat (the “front-
end” electronics). This data is sent via fiber optics to the
off-cryostat back-end electronics where a single FPGA col-
lects all six hextants’ data streams. The back-end FPGA
presents this data in parallel to a single Digital Signal Pro-
cessor(DSP)fordemodulationofeachchannel(twice)using
thereferencebiassignalanditsquadraturephase.TheDSP
cachesthedemodulateddatastreamsinmemoryandapplies
asharpblockFiniteImpulseResponsefilter,whosecutoffis
justbelowtheNyquistfrequencyofthe64Hzfinalsampling
rate.
Bias generation and housekeeping electronics are also
Figure 4. The AzTEC system bandpass for a flat spectrum in-
partofthefront-endelectronicslocatedatthecryostat.The
putsource,normalizedtothepeakresponse.(Thelowfrequency
bolometer bias signal is generated from a digital sine wave
portionofthespectrumisslightlynegativeduetoasmallphase
stored in memory in the back-end FPGA and continuously
offset inthe fouriertransformspectrometer usedtomeasure the
fed via the fiber-optic connection to the DACs that gener-
response.)
ate the analog sine wave for each hextant. The bias ampli-
tude and phase are controlled digitally and in a real-time
fashion. The bias frequency for the JCMT05B observations
Following each bolometer is a matched pair of low-
was200Hz.TheHousekeepingcardreadsoutinternalther-
noiseU401JFETs,inthesource-followerconfiguration,that
mometervoltages, providespower, andcontrols the 3Here-
transformsthedetectoroutputimpedancefromseveralMΩ
frigerator.
to 350Ω.SignalsfromtheJFETamplifiersexitthecryo-
∼
stat via pi-filters at 4 K and 300 K and are read out by The back-end electronics are controlled by a dedicated
room-temperature, low-noise (1.7 nV Hz−1/2) instrumenta- MotorolaPower-PCrunningthereal-timeVxWorksoperat-
tion amplifiers (Texas Instruments, type INA103) with a ingsystem.While areal-time system isnot required bythe
fixed gain of 500. Each detector signal is then high-pass fil- data acquisition architecture, a real-time system simplifies
tered at 100Hz and low-pass filtered at 300Hz before a pro- synchronizationofthenativeAzTECsignalswithtelescope
grammable attenuation is applied by a digitally controlled pointing and other environmental signals. From the user’s
resistive divider (see Figure 5 for part numbers.) The out- perspective,AzTECiscontrolled anddataisstoredusinga
put of the divider is then digitized at 6.144 MHz by high- standardLinuxPCrunningtheLargeMillimeter Telescope
6 G.W. Wilson et al.
Monitor and Control (LMTMC) system which was devel-
oped at UMass-Amherst for general use byall LMT instru-
ments (Souccar et al. 2004). All communication and data
transfer between the Linux machine and the Power-PC is
viadedicated Ethernet.Includingpointingsignals andtele-
scopehousekeepingsignals,thetotaldatarateofAzTECis
80 kB/s.
∼
3 OBSERVING
TheAzTECdatasetislogicallygroupedinto“observations”
which include engineering tests (e.g., calibration, pointing,
or focus observations) as well as scientific observations. All
dataisstoredinamachineindependentbinaryformat,Net-
workCommonDataForm(NetCDF).AzTECNetCDFfiles
are self-describing in that they include all information re-
quiredtoproduceanoptimalimage,includingheaderinfor-
mationdefiningthedataaswellasallobservingparameters.
Figure 6.Jiggle-mapof QSO J1048+7143. A gaussianfitted to
Real-timemonitoringofAzTECisconductedusingvar-
the peak flux gives the telescope boresight offset from the refer-
ious tools in the LMTMC package and the KST plotting encebolometer(Section3.2.4).
suite. Quick-look software, custom written in the IDL pro-
gramminglanguage,isusedtoproduceimagesimmediately
afterthecompletionofeachobservation.Thesesimplemaps
allowtheusertoquicklyevaluatetelescopepointingandfo-
cus,andprovidesanassessmentoftheoverallqualityofthe
data.
3.1 Observing Modes
AzTEC is a passive instrument with respect to the tele-
scopeandoperatescompletelyindependentlyoftheobserv-
ing mode. Since themapping efficiency is a strong function
of the observing mode and observing parameters for maps
on-orderthefieldofviewoftheinstrument,wehaveadopted
two primary modes in the existing data analysis suite.
3.1.1 Jiggle-Mapping
Inthejigglemappingmode,thesecondarymirrorischopped
between on-source and off-source sky positions while step- Figure 7.Thecorrespondingweight map(inmJy/beam−2)for
ping through a series of small offsets (jiggles) to fill in gaps
thesignalmapinFigure6.Thelackofcoverageduetoaclustering
′′
in coverage left by theapproximately 20 spacing of thede-
of inactive detectors is indicated by the low weight region near
tectors. Differencing chopped data at a single jiggle posi- thecenter ofthemap.
tion removes the effects of low-frequency atmospheric and
instrumentaldrifts.Chop frequenciesduringtheJCMT05B
run ranged from 2 to 4 Hz with a secondary mirror tran- AzTEC jiggle-maps taken during the JCMT05B observing
sit time of less than 30 ms, and a jiggle frequency of 1 Hz. run, as shown in the Figures 6 and 7, where a contiguous
Once the full jiggle pattern is completed, the primary mir- region of detectors were inoperable. This resulted in a por-
ror “nods” to put the off-source beams on source. Averag- tionofthefieldbeingseverelyundersampled.Forjiggle-map
ing data from opposite nods further suppresses differential observationsofsourceswithaknownlocation,weadjustfor
pickup from spillover at the primary mirror, temperature the non-uniform coverage by offsetting the boresight point-
gradientsontheprimary,andscatteringfromthesecondary ing of the telescope so that the target falls in a region fully
support structure. populated by working detectors.
For fields on the same scale as the array field of view
(FOV)orsmaller,jiggle-mappingisthehighestefficiencyob-
3.1.2 Raster-Scanning
serving mode for a given target sensitivity. However, when
imagingfieldsthataremuchlargerthanthearrayFOV,the For raster-scan observations the telescope boresight sweeps
overheadsofjiggle mappingand thehighvariability in cov- across theskyalong onedirection, takesasmallstep in the
eragemakescanmappingessential.Thecoveragemapfrom orthogonal direction, and then sweeps back along the next
a jiggle-map is highly sensitive to the distribution of detec- row of the scan. This pattern is repeated until the entire
tor sensitivities on thearray. This is particularly evidentin fieldhasbeenimaged.Becauseofthelow-frequencystability
The AzTEC mm-Wavelength Camera 7
is thesecondary position where thebeam full-width at half
maximum(FWHM)isminimizedandthepeakamplitudeof
the signal is maximized. For the JCMT05B run we focused
the telescope at the beginning and mid-way through each
night of observing.
3.2.2 Relative Bolometer Offsets
Sincethearrayorientationisfixedinazimuthandelevation,
the relative offset on the sky between any two detectors is
constant. We determine these offsets by mapping a bright
point source each evening prior to science observations and
after focusing thetelescope. Ahigh-resolution map is made
such that the point spread function (PSF) of each detector
′′
inthearrayissampledwithatleast4 resolutioninorderto
determinetherelativebolometerpositionstoanaccuracyof
5% of thePSF FWHM. These “beam map” observations
≈
alsoprovidetheabsolutecalibrationofdetectorsinthearray
Figure 8. The integration time in each 2′′pixel for a 25′ 25′ (see Section 4).
×
raster-scan map with scanning done in the elevation direction
and with 10′′step sizes in azimuth, demonstrating the uniform
coverageachievedwiththisobservingmode.
3.2.3 Loadcurves
Loadcurve measurements – sweeping the detector bias
of the detectors, we do not chop the secondary mirror. In throughitsfullrangeofcommandablevalueswhileviewinga
fields where no bright sources are expected, removal of the blankpatchofsky–aremadeeacheveningfollowingthede-
atmospheric contamination is currently accomplished via a termination oftherelativebolometer offsets. Fromtheload
principalcomponentanalysis(PCA)ofthefulldatastream curveswedeterminethetotalopticalpower,Q,absorbedby
on a scan by scan basis (Laurent et al. 2005). The PCA each detectoraswell as theresponsivity,S, (theconversion
technique is used in the analyses leading to the instrument from Volts read out of the detector to Watts absorbed) of
performance and characteristics detailed in Section 5. eachdetector.Bymeasuringtheresponsivityinthismanner
Themajorbenefittoraster-scanningisintheabilityto undera wide range of atmospheric opacities weconstruct a
map a large area of sky in a single observation with very correction to the non-linearity of the detector response due
uniform coverage, asshown in Figure8.Thedistribution of to the overall variation in the atmospheric optical loading.
inoperable detectors on the array only affects the ultimate This process is described furtherin Section 4.1.
sensitivityofthemap,andnottheuniformityofthecoverage Figure9showsthecontributionsoftheatmosphereand
inthemap.Scanspeedsarelimitedbythedetectortimecon- other sources to the total optical power absorbed by the
stantandstability ofthetelescopebutaregenerally chosen AzTEC detectors. The telescope, coupling optics, and any
basedonthesciencetargetandtheatmosphericopacity.For parasiticopticalloadingsonthedetectorsdeliveracombined
point source observations at the JCMT05B run, high scan 9 pW of power – corresponding to an effective blackbody
speedswerepreferredinordertomovethesignalbandwidth temperature of 39 K. The optical loading due to the atmo-
above the knee frequency of residual atmospheric contami- sphereislinearwiththeopacityasmeasuredbytheCaltech
nation(seeFigure14),howeveratacostinoverallobserving Submillimeter Observatory (CSO) tau meter at 225 GHz
efficiency due to the fixed length turnaround time (5 s) of overtherange of opacities suitable for observing.
the telescope. Coadding tens of raster scanned maps taken
with different skyposition angles reduces scan-synchronous
effects in the maps and offers excellent cross-linking of the 3.2.4 Pointing Observations
data.
Pointing measurements are performed approximately every
hour in order to measure the absolute pointing offset be-
3.2 Observing Overheads tween the telescope boresight and a reference bolometer.
An optimal pointing source is bright (> 1 Jy), unresolved,
WhileAzTECwasontheJCMT,thefollowingancillaryob-
and located near the science target. Pointing observations
servationsweremade,averaging24.7%ofthetotalavailable
typically bracketed a series of science observations so that
observing time.
slow drifts in the residuals to the telescope pointing model
couldbefitout.Sinceonlyafewbolometersmustpassover
the source, pointing observations are usually carried out in
3.2.1 Focus Observations
jiggle-map mode. We fit a 2-dimensional Gaussian to the
Focus measurements consist of a series of jiggle-map obser- point source image, and the best-fit location of the peak
vationsonabright( fewJy)pointsourceasthesecondary signal gives the boresight offset as shown in Figure 6. To
∼
mirror steps through different focus settings. For each fo- correct the pointing signals for a given science observation,
cussetting,wefita2-dimensionalGaussiantotheregion of a pointing model is generated by interpolating between the
theimagecontainingthesource.Theoptimalfocuslocation pointing measurements taken overa night.
8 G.W. Wilson et al.
Figure 10. Left: Responsivity versus bolometer dc-level for a
typical detector as determined from all of the loadcurves taken
during the JCMT05B run. A best-fit line is over-plotted. Right:
Opacityversusbolometerdc-levelforthesamedetector.
(see Section 3.2.3). For the range of total power loading
on the detectors observed during the JCMT05B run, the
responsivityislinearlyproportionaltothedemodulateddc-
levelofeachbolometer’stimestreamsignal.Theresponsivity
Figure 9.The effective atmosphere temperature inthe 1.1 mm ofatypicalbolometerversusthedc-levelmeasured fromall
AzTEC bandpass as a function of the atmospheric opacity at of the loadcurve observations taken during the JCMT05B
225 GHz as measured by the CSO. The axis on the right gives run is shown in the left panel of Figure 10. The solid curve
the corresponding amount of optical power absorbed by the showsthelinearfittothemeasurements.Wederivethebest-
AzTEC detectors. The horizontal line shows the effective tem- fit offset and slope for each bolometer separately since the
perature/power ofthetelescopeandcouplingoptics.
spread in these parameters is large compared to the formal
errors on thefits.
4 CALIBRATION Theatmosphericextinctione−τeff iscorrectedinasimi-
larway.Alinearcorrelationexistsbetweentheatmospheric
The output of bolometer i at sky position α, bi,α (in units opacity, τeff and the bolometer dc-signals (see right panel
of V), is given by of Figure 10). For the JCMT data we use the atmospheric
∞ opacity at 225 GHz as determined from the CSO tau mon-
bi,α =Si(Q)AeffηZ dνf(ν)Z dΩPi(Ωα−Ω)e−τeffIν(Ω),(1) itor, which records τ225GHz (zenith) every 10 minutes, to
0 sky calibrate therelation with thebolometer dc-levels.
where S (Q) is the responsivity (in V/W), Q is the opti-
i
cal loading (in W) dominated by the telescope and atmo-
sphere,A istheeffectivetelescopeaperture,η istheopti- 4.2 Measured FCF from Beam Map Observations
eff
calefficiency,f(ν)isthepeak-normalizedAzTECbandpass,
Todetermine thefluxconversion factor for each bolometer,
P (Ω Ω)isthepeak-normalizedAzTECbeampatternfor
boilomαe−teriatskypositionΩ ,τ istheopacity,andI (Ω) FCFi,webeam mapa primaryorsecondary fluxcalibrator
is the source intensity on theαskeyff(in Jy beam−1). Asνdis- each night.Thetimestream bolometer signals are corrected
for extinction and the responsivity is factored out. An iter-
cussedbelow,S andτ ,bothofwhichdependonobserva-
i eff ativecleaning techniqueis used to minimize errors in fitted
tional conditions (e.g. weather) and change significantly on
parameters due to atmospheric contamination. In the final
thetimescaleofhours,aremodeledasfunctionsofthe“dc”
iteration,mapsaremadeforeachbolometerseparately and
level of the bolometer signal.
we fit a 2-d Gaussian to each map. The best-fit amplitude
The flux conversion factor for bolometer i, FCF (in
unitsof Jybeam−1W−1),is an expression involvingallifac- combined with the known flux of the source gives the FCF
for each detector.
torsthatare,inprinciple,constant(i.e.sourceandweather
To correct the flux for the angular size of Uranus, we
independent) over the entire observing run, and is defined
as assume that Uranus is a disk with angular radius ΘU and
brightness temperature T so that I(Ω) = T Φ(Ω), where
b b
FCFi= Aeffη 0∞dν f(ν) 1skydΩPi(Ωα−Ω). (2) 1Φ.(7ΩK) =(G1riffifonr &θ 6OrΘtoUn a1n99d3)0.oTthheerawviesreagaendfluTxbo=f U9r2a.6nu±s
R R for bolometer i is then given by
Forapointsource located at (θ0,φ0) withaverage fluxden-
sity I¯over theoptical bandpass, thebolometer output is ∞dν f(ν)2kTb,oν2 dΩP (Ω Ω)Φ(Ω)
bi(θ0,φ0)= Si(QF)CeF−iτeffI¯. (3) I¯= R0 R0∞dν fc(2ν)RsRksykydΩPi(iΩαα−−Ω) , (4)
whichcanthenbeusedinEquation3.Uranusissmallcom-
4.1 Responsivity and Extinction Corrections pared to the AzTEC detector PSFs on the JCMT (2ΘU ≪
θ , where θ is the true beam FWHM for bolometer i)
bi bi
The optical loading on the detectors and the bolometer re- andsowillappearapproximatelyGaussianwithameasured
sponsivities are determined from loadcurve measurements FWHMθ , where
mi
The AzTEC mm-Wavelength Camera 9
ln2
θb2i =θm2i − 2 (2ΘU)2 (5)
(Baars1973).Wemeasureθ fromUranusbeammaps,and
mi
useEquation5todetermineθ andsubsequentlyP (Ω).We
bi i
then calculate the integral in Equation 4 to determine the
correction factor for theflux of Uranus.
ForbeammapsofUranus,thereisastatisticallysignifi-
cantincreaseinthemeasuredFCF formeasurementstaken
i
within one hour after sunset at the JCMT. Measurements
takenafterthistimehaveconstantFCF .Thisisconsistent
i
with rough estimates of the telescope’s thermal time con-
stant. For this reason, we determine the average flux con-
versionfactorforeachbolometer<FCF >byaveragingover
i
all FCF measured from Uranusbeam mapstaken >1 hour
i
aftersunset,andweusethesevaluestocalibrateall science
observations taken after the telescope has settled. A linear
correction factor derivedfrom beam mapstakenwithin one
houraftersunsetisappliedtosciencedatatakenduringthis
period.Wemodel thiscorrection factor f FCF /<FCF >
i i
≡
as a linear function of thetime after sunset, such that
f = 1 if HAS>1
Figure 11. AzTEC’s “footprint” on the sky at the JCMT05B
= 1+m (HAS 1) if HAS<1 (6)
· − run.Thesixhextantsarelabeled“Hex”1-6.Thealternateshad-
whereHASisthetimeaftersunsetmeasuredinhours,mis ingindicates which bolometers arelocated ineach hextant. The
size of the ellipse corresponds to the bolometer’s FWHM, mea-
the same for all bolometers and continuity at HAS= 1 has
suredintheazimuthandelevationdirections.
been applied. We fit the measured FCF /<FCF > for all
i i
bolometersandallbeammapswithHAS<1simultaneously
to Equation 6 to find m= 0.115 0.002 hr−1.
− ±
4.3 Calibration Error
ForagivenscienceobservationwithresponsivityS andex-
i
tinction e−τeff (measured from the bolometer dc-levels) the
calibrated timestream bolometer signals I¯(t) are given by
i
b (t) <FCF > f
I¯i(t)= i ·Si e−τefif · (7)
·
where f is determined from the HAS that the observation
took place and the empirical formula derived above (Equa-
tion 6). The error on the calibrated bolometer signals is
therefore equal to the quadrature sum of the errors on all
four factors in Equation 7 and is typically 6-13% for the
JCMT05B data.
Figure 12. A cut in the elevation direction through the beam
5 PERFORMANCE
responsepatternofatypicalAzTECpixel.
5.1 Array Layout and System Efficiency
The array “footprint” on the sky for the JCMT05B run is all beam maps of Uranus taken at the JCMT. The size of
shownisFigure11,centeredonthereferencebolometerand theellipse is equalto thebeam FWHM, which is measured
withthesix hextantslabeled. Weexcludealldetectorsthat in the azimuth and elevation directions and is on average
werenotfullyoperationalduringthisrun,mostofwhichare 17′′ 1′′ inazimuthand18′′ 1′′ inelevationattheJCMT.
± ±
locatedinhextants1and2.Thisleftuswith107operational Anoff-axisellipsoidalmirrorintheopticschainleadstothe
pixels for theJCMT05B season. The majority of failures in slightelongationofthebeamintheelevationdirection.The
′
thesignalchainshaverecentlybeentracedtobrokenJFETs array FOVisroughly circularwith adiameter of5.Acon-
and these are expected to be repaired by2009. stant azimuth cut through a typicalbolometer’s beam map
Eachbolometer’slocation onthearrayanditsPSFare is shown in Figure 12. The beams are nicely gaussian down
determinedfrombeammapobservationsasdescribedinSec- to thefirst sidelobe response at 20dB.
∼−
tion3.2.2.ThepositionsandbeamsizesdisplayedinFigure Table 2 lists the measured optical characteristics of
11 were determined by averaging the measurements from AzTEC in the 1.1 mm configuration for the JCMT05B ob-
10 G.W. Wilson et al.
Bandcenter frequency 270.5GHz
Effectivebandwidth 49.0GHz
Effectivethroughput(AΩη) 0.2 0.014mm2 sr.
BeamFWHM(azimuth) ±17′′ 1′′
BeamFWHM(elevation) 18′′±1′′
±
Table 2. AzTEC 1.1 mm optical parameters for the JCMT05B
observations. The effective bandwidth is calculated assuming a
flat-spectrum source.
servations. The effective throughput(AΩη) is measured for
eachdetectorfromeachbeammapofUranus.Itiscalculated
using Equation 2 along with AzTEC’s measured bandpass,
f(ν) and themeasured fluxconversion factor of that detec-
tor,FCF .Thevaluelisted in Table2is anaverage overall
i
working detectors.
The optical efficiency η of the system may be written Figure 13.Thenoiseequivalentfluxdensity(NEFD)foratyp-
icalbolometer.Inorderof decreasingdarkness,thethree shades
astheproductofthetelescope’sefficiency,η ,andthecou-
tel
plingefficiencyoftheinstrumentηinst.Weestimatethetele- tchoircrkeesprocnudrvteos reeffpercetsievnetτr2a2w5 dvaaltuaewshoifle0.t1h1e,t0h.i1n6nearndcur0v.2e1s.shTohwe
scope emissivity, ǫ, as the ratio between the effective tele-
datathathavebeencleanedusingaprincipalcomponentanalysis
scopetemperatureofFigure9(dashedline)andthetrueav-
(PCA).Thecleanednoisespectraarehigheratsomefrequencies
erage temperature of the telescope surface. This emissivity because they have been scaled to account for point-source at-
estimategivesηtel =1 ǫ=0.85.Removingthisfactorfrom tenuation due tocleaning. The lower-opacitydata benefits more
−
the total throughput, we find that the instrument through- from the common-mode removal effected by cleaning as well as
put,AΩηinst,is0.23mm2-sr.Comparingthistotheidealized by the reduced optical loading. The dashed lines represent the
throughputofasinglemodedsystemat1.1mmwavelength corresponding NEFDs for a detector with targeted detector pa-
with the same telescope efficiency (AΩη =λ2η =0.99) rametervaluesatourJCMTbiaslevelalongwithanestimateof
tel tel
optical loading (details in the text). The dotted curve indicates
we find the optical efficiency for the AzTEC radiometer to
theapproximateopticalbandwidthofapointsource(inarbitrary
be 0.19. Given that the cold Lyot stop leads to a 38% re- units)forascanvelocityof180′′/s.
duction in throughput, this leaves an overall efficiency of
31%. This is very similar to the optical efficiency measured
fortheBolocam instrument(Glenn et al.2003)whichhasa
pact on sensitivity. Table 3 lists the expected detector op-
very similar optical design.
erating parameters and the breakdown of thermodynamic
noise contributions for conditions at the JCMT assuming
5.2 Noise and Sensitivity τ225 =0.11.Theactualpost-cleaningnoiseequivalentpower
measured for a typical detector is dominated by residual
Blank field observations done in raster-scan mode are used low frequency noise due to the atmosphere and, based on
to estimate detector noise and sensitivity. Figure 13 shows our point-source sensitivity, is equivalent to having a white
the noise equivalent flux density per beam (NEFD) for a noise level of 1846 aW/√Hz.
typical detector in three weather conditions at the JCMT. The increase in noise at low frequency highlights the
The thicker curves show the raw NEFDs while the thin- importance of scanning at high speed. The dotted curve of
ner curvesof thesame shade show the improvement dueto Figure 13 indicates the approximate optical bandpass of a
a time stream cleaning method based on a principal com- detectoratascanvelocityof180′′/s,whichisalsothedetec-
ponent analysis (PCA). The low-frequency features, domi- torresponsetoapointsourceat thatscan speed.Through-
nated by atmospheric fluctuations, are not completely pro- out the course of the observing run, we used scan speeds
jectedoutbythePCAcleaning.TheflatterNEFDathigher from 30′′/s to 270′′/s based on balancing the opposing ef-
frequencies that does not benefit from cleaning can be at- fects of higher sensitivity and larger turn-aroundtime frac-
tributed to the irreducible noise floor due to the photon tion at higher scan speeds given a map size, as explained
background limit (BLIP) and detector noise. in section 3.1.2. With scans along the elevation direction,
The three dashed lines indicate the thermodynamic we do not see vibration-induced noise increases within this
noiselimitsforanidealAzTECdetectorgiventhebiasvolt- range of velocities during the JCMT05B observations. We
age, expected atmospheric optical loading for three opaci- didmeasureexcessnoisenear 2Hzonbad-weathernights
∼
ties,andopticalloadingfromthetelescope(assuminga15% attheJCMT,possiblyduetowind-inducedsmallmotionsof
effective emissivity). In each case theactual high-frequency opticalloadsand/oroptics,liketheJCMT’sGore-texcover.
noiselevelisconsistentwiththedetectorandloadingmodel The abrupt noise cut-off near 32 Hz is due to a digital
to within 10-20%, well within our uncertainty of the total filterthatconditionsthedemodulatedbolometersignalsfor
optical loading on the detectors. The nearly constant ratio 64-Hzdecimation. Theattenuation with frequencybetween
betweentheachievedandexpectednoiselevelsovervarying 20and30Hzisduetothebolometertimeconstant.Theline
weather conditions indicates that the operationally conve- at 25Hzislikelyasidebandcausedbythethirdharmonic
∼
nient choice of using a constant bias voltage for the entire of AC power (or “60 Hz”) mixed with the 200 Hz demod-
observingrun(seesection 2.2)hadlittleifanynegativeim- ulation waveform. When analyzingraster-scan observations