Noname manuscript No. (will be inserted by the editor) Herschel SPIRE Fourier Transform Spectrometer: Calibration of its Bright-source Mode Nanyao Lu · Edward T. Polehampton · Bruce M. Swinyard · Dominique Benielli · Trevor Fulton · Rosalind Hopwood · Peter Imhof · Tanya Lim · Nicola Marchili · David A. Naylor · Bernhard Schulz · Sunil Sidher · Ivan Valtchanov 4 thedateofreceiptandacceptanceshouldbeinsertedlater 1 0 2 Herschel is an ESA space observatory with science instruments provided by European-led n PrincipalInvestigatorconsortiaandwithimportantparticipationfromNASA. a J N.Lu 9 NHSC/IPAC,100-22Caltech,Pasadena,CA91125,USA E-mail:[email protected] ] M E.T.Polehampton RALSpace,RutherfordAppletonLaboratory,Didcot,Oxfordshire,OX110QX,UK,and I Institute for Space Imaging Science, Department of Physics & Astronomy, University of . Lethbridge,Lethbridge,ABT1K3M4,Canada h p B.M.Swinyard - RALSpace,RutherfordAppletonLaboratory,Didcot,Oxfordshire,OX110QX,UK,and o Dept.ofPhysics&Astronomy,UniversityCollegeLondon,GowerSt,London,WC1E6BT, r UK t s D.Benielli a [ Aix Marseille Universit´e, CNRS, LAM (Laboratoire d’Astrophysique de Marseille) UMR 7326,13388,Marseille,France 1 T.Fulton v Institute for Space Imaging Science, Department of Physics & Astronomy, University of 5 Lethbridge,Lethbridge,ABT1K3M4,Canada 4 0 R.Hopwood 2 PhysicsDepartment,ImperialCollegeLondon,SouthKensingtonCampus,SW72AZ,UK . P.Imhof 1 BlueskySpectroscopyLethbridgeUniversity,Lethbridge,Canada 0 4 T.Lim 1 RALSpace,RutherfordAppletonLaboratory,DidcotOX110QX,UK : v N.Marchili i Universita´diPadova,I-35131Padova,Italy X D.A.Naylor r InstituteforSpaceImagingScience,DepartmentofPhysics,and a AstronomyDepartment,UniversityofLethbridge,Lethbridge,AB,Canada,T1K3M4 2 NanyaoLuetal. Abstract The Fourier Transform Spectrometer (FTS) of the Spectral and Photometric Imaging REceiver (SPIRE) on board the ESA Herschel Space Observatoryhastwodetectorsettingmodes:(a)anominalmode,whichisop- timizedforobservingmoderatelybrighttofaintastronomicaltargets,and(b)a bright-sourcemoderecommendedforsourcessignificantlybrighterthan500Jy, within the SPIRE FTS bandwidth of 446.7-1544GHz (or 194-671microns in wavelength), which employs a reduced detector responsivity and out-of-phase analog signal amplifier/demodulator. We address in detail the calibration is- sues unique to the bright-source mode, describe the integration of the bright- mode data processing into the existing pipeline for the nominal mode, and show that the flux calibration accuracy of the bright-source mode is generally within 2% of that of the nominal mode, and that the bright-source mode is 3 to 4 times less sensitive than the nominal mode. Keywords Instrumentation · Calibration · Herschel space observatory · Fourier transform spectrometer · Sub-millimeter astronomy 1 Introduction The Spectral and Photometric Imaging REceiver (SPIRE; Griffin et al. 2010) is one of three focal-plane instruments on board the ESA Herschel Space Ob- servatory (Herschel; Pilbratt et al. 2010). It contains an imaging photometric camera and an imaging Fourier Transform Spectrometer (FTS). The SPIRE FTS employs two detector arrays of spider-web neutron transmutation doped (NTD) bolometers (Bock et al. 1998), biased by a sinusoidal AC voltage with a 160Hz frequency: a short-wavelength array (SSW) of 37 bolometers cov- ering 959.3-1544 GHz in frequency (194-313 µm in wavelength) and a long- wavelength array (SLW) of 19 bolometers covering 446.7-989.4 GHz (303-671 µm). Two detector-setting modes are available: (a) a nominal mode with the detector arrays optimally biased (by voltages of amplitude of 36 and 31 mV for SSW and SLW, respectively) to achieve the highest detection sensitivity, and(b)abright-sourcemodewiththedetectorssubjecttoamuchhigherbias voltage(ofamplitudeof176.4mVforbothSSWandSLW)toyieldareduced detector responsivity. In the bright-source mode, the analog square-wave am- plifier/demodulator,assketchedinFig.11,isfurthertunedtobeabout70and 68 degrees, for SSW and SLW detectors, respectively, out of phase with the B.Schulz NHSC/IPAC,100-22Caltech,Pasadena,CA91125,USA S.Sidher RALSpace,RutherfordAppletonLaboratory,Didcot,Oxfordshire,OX110QX,UK I.Valtchanov HerschelScienceCentre,ESAC,P.O.Box78,28691VillanuevadelaCan˜ada,Madrid,Spain 1 ThisfigurewasadaptedfromSchulzetal.(2008),basedonthedocumentofTheSPIRE Analogue Signal Chain and Photometer Detector Data Processing Pipeline, available at http://herschel.esac.esa.int/twiki/pub/Public/SpireCalibrationWeb/Phot−Pipeline−Issue7.pdf SPIREFTSbright-sourcemode 3 Fig. 1 An illustration of the first part of the SPIRE FTS signal chain: The AC signal fromthebolometerpassesthroughtheJFETamplifier,abandpassfilterthatfiltersoutthe DCsignal,asquare-waveamplifier/demodulatorthatisnormallylockedinphasewiththe bolometerbiasvoltagetoturnthenegativepartofthesignalintopositive,andfinallyalow passfilterthatgeneratesaslowlyvaryingDCsignal.Inthebright-sourcemode,thesquare- wave demodulator is de-phased from the bolometer signal to further damp the output DC signal. detectorsignaltofurtherreducethechanceofsaturatingtheanalog-to-digital converter. As a result, the bright-source mode results in a much larger dy- namic range in flux, allowing for sources as bright as 25,000Jy to be observed withoutserioussaturation.Incomparison,thenominalmodeisrecommended for sources fainter than ∼500Jy within the SPIRE FTS bandwidth. The overall strategy for the in-flight calibration of the SPIRE FTS is out- lined in Swinyard et al. (2010). The actual calibration and pipeline imple- mentation of the FTS nominal mode are described in detail by Swinyard et al.(2013)andFultonetal.(2013;2010),respectively.Thebright-sourcemode calibration went through a major upgrade in March 2013 [coinciding with Version 11 of Herschel Interactive Data Processing Environment (HIPE); Ott (2010)]. Prior to HIPE 11, the data of the bright-source mode was processed in the temperature domain, with some residual bolometer nonlinearity passed through to the end of the pipeline. Starting in HIPE 11, we adopted a new fullnonlinearitycorrectionschemeandintegratedthebright-sourcemodedata processingdirectlyintotheexistingpipelineforthenominalmode.Asaresult, the agreement between the bright-source and nominal mode flux calibrations has been improved from within ∼10% in HIPE 10 to within ∼2% in HIPE 11. In this paper, we address those calibration issues unique to the bright- sourcemode(§2),describehowthedataobtainedinthebrightmodeisfolded intothenominal-modepipelinefordataprocessing(§3),demonstratethatthe bright-sourcemodefluxcalibrationaccuracyiswithinabout2%ofthatofthe nominal mode (§4), and finally summarize our results (§5). 4 NanyaoLuetal. 2 Calibration Scheme for the Bright-source Mode Thecalibrationstrategyforthebright-sourcemodeistoutilizethecalibration productsderivedforthenominalmodewhereverpossible,inordertominimize the overall FTS calibration effort and keep the pipeline simple and robust. In thisapproach,thebright-sourcemodeneedsonlythefollowinguniquecalibra- tionproductsorprocedures:(1)aphase-relatedgaincorrectionfactor,G , phase fortheout-of-phaseanalogamplifier/demodulator,(2)adetectornonlinearity correction calibration product, (3) a zero-point, DC-type gain correction fac- tor, G , which aligns the linearized signal scale of the bright-source mode to 0 thatofthenominalmode,and(4)apossiblefrequency-dependentgainfactor, G . The last one may result from effects such as a dependence of bolometer f response time constant on the bias voltage. We address each of these issues in more detail below. 2.1 Phase-related Gain Correction The square-wave analog amplifier/demodulator is locked in phase with the bolometer AC signal in the nominal mode, but is intentionally kept at φ diff outofphaseinthebrightsourcemode;whereφ ,measuredforeachdetector, diff variesslightlyfordifferentdetectorsofthesamebolometerarray,andisaround 70 and 68 degrees for SSW and SLW detectors, respectively. As discussed in Swinyard et al. (2013), the effective R-C circuit of the bolometer JFET amplifier and harness introduces a gain factor, G , and a signal phase shift, cab φ , as follows: t (cid:115) 1 G = , (1) cab 1+ω2 cr φ =atan(ω )+φ , (2) t cr off where φ is a constant phase offset (= 11.4 and 13.6 degrees for SSW and off SLW, respectively) and ω = 2πω R C , with ω being the detector cr bias tot H bias bias frequency, R is the total resistance of the bolometer readout circuit, tot including the bolometer itself and the load resistors, and C (=20pF) is the H cable capacitance of the FTS readout system. Since the bolometer resistance depends on the optical load, so does G . In practice, G and ω are de- cab cab cr termined in an iterative way. The gain factor related to the adjusted phase is given by: G =cos(φ −φ ). (3) phase diff t This G is divided into each voltage sample for the bright-source mode phase at the engineering data conversion stage in the pipeline. Typically, G is cab always close to unity and G is on the order of 0.5 (e.g., it is around 0.54 phase for SSWD4 and around 0.67 for SLWC3). SPIREFTSbright-sourcemode 5 2.2 Detector Nonlinearity Correction Since the nonlinear responsivity of a bolometer depends on its bias voltage, it is necessary to derive a separate nonlinearity correction calibration product for each of the two detector-setting modes. For the SPIRE bolometers, the signal linearization can be done in an analytic way as, following Swinyard et al. (2013) and Bendo et al. (2013), V −K V(cid:48) =K (V −V )+K ln( m 3), (4) 1 m 0 2 V −K 0 3 where V and V(cid:48) are the observed and linearized bolometer voltages, respec- m tively,V isanarbitraryreferencevoltage,andK ,K andK aretheparam- 0 1 2 3 eters characterizing the detector nonlinearity. For the nominal mode, these K parameters were derived from a physical bolometer model (e.g., Sudiwala et al. 2002; Woodcraft et al. 2002) using a bolometer analysis package developed at the NASA Herschel Science Center (Schulz et al. 2005) with both labora- tory and in-flight measured detector parameters (Nguyen et al. 2004). For the bright-source mode, which requires nonlinearity correction over a much larger flux range, these K parameters were determined directly from the calibration data taken on flashes of the SPIRE internal photometric calibrator (PCAL; Pisano et al. 2005). Each astronomical FTS observation contains 9 pairs of PCALflashesontopofthebackgroundemissionatthepositionofthetarget. Some dedicated PCAL calibration observations were also obtained in order to expand the background flux coverage. For each PCAL flash of power off and power on, we can write 1/δV =K +K /(Voff −K ), (5) m 1 2 m 3 where δV is the instantaneous bolometer voltage change when the PCAL m poweristurnedonandVoff isthevoltagereadingjustbeforethePCALpower m is turned on. We can write eq. (5) because the PCAL power and illumination pattern remain fixed over the entire mission (as well as between the nominal and bright-source modes) and because an arbitrary common scaling factor is allowed for K and K . (This scaling factor gets folded into the zero-point 1 2 gain correction factor in §2.3.) As examples, Fig. 2 shows two independent sets of PCAL flashes on the detectorSSWD4.Theleft-handsideplotrepresentsasetofPCALflashestypi- callyseeninanastronomicalobservation.Notethatthereisaslightdownward signal drift when the PCAL power is on, illustrating a possible heat input to the detectors from the PCAL power. One of our dedicated PCAL calibration observations towards the Galactic center is shown in the right-hand plot to illustrate a typical PCAL observation when the telescope was pointed at a bright discretesource. Thestrong baselinedrift overthe on-offcyclewas are- sult of the jitter in the telescope pointing. Our PCAL data reduction pipeline module fit a linear function independently to each on and off signal plateau (after excluding a certain percentage of the data points at the beginning and 6 NanyaoLuetal. Fig. 2 Two examples of PCAL flashes on the detector SSWD4. The actual data samples areshowninblackcurves.Thefitstothepower-offandpower-onsignalplateausareshown asredorbluelines,respectively. end of each plateau; see the SPIRE pipeline description document2 for more details)anddeterminesfromthesefitsVoff andVon (=Voff+δV )perPCAL m m m m flashpair.Finally,foreachPCALobservation,themedianVoff andδV values m m overallofitsPCALflashpairswerederivedforuseinourdetectornonlinearity characterization. Fig. 3 illustrates our fitting of eq. (5) to the bright-source mode PCAL data pairs (i.e., Voff and δV ) we accumulated over the entire mission for the m m two central detectors, SSWD4 and SLWC3. The voltage coverage along the X-axis ranges from dark sky observations (at the high voltage end) to those fromtwoSaturnobservations.Itisevidentthatthedatapointsarestillsparse atthelowvoltageend,leadingtopossiblyalowerfluxcalibrationaccuracyfor brighttargetssuchasSaturn.Foreachdetector,wedefinedavoltagerangeof Voff(min)toVoff(max),withinwhichthenonlinearitycorrectionbasedonthe m m fitisdeemedtobevalid.ThevalueofVoff(min)issetto5%belowthesmallest m voltagesampleobservedandthatofVoff(max)to3%abovethelargestvoltage m sample we have. If we compare the PCAL δV values on the same background source between the bright-source and nominal modes (for the nominal mode counterparttoFig.3here,seetheirFig.5inSwinyardetal.2013),ingeneral, the detector responsivity in the bright-source mode is about a quarter of that in the nominal mode. 2.3 Zero-point Gain Correction As an example, Fig. 4 shows the linearized PCAL voltages [via eq. (4)] for the detector SSWD4 in the nominal mode (on the left-hand side) and bright- sourcemode(ontheright).Theseplotsalsoillustratethat,forthemajorityof 2 The SPIRE pipeline description document, available at http://herschel.esac.esa.int/twiki/bin/view/Public/SpireCalibrationWeb, will be updated toreflectthePCALdatareductionalgorithmdescribedhere. SPIREFTSbright-sourcemode 7 Fig.3 Bright-sourcemodebolometernonlinearresponsivityfitsofeq.(5)tothePCALdata for the two central detectors, SSWD4 (on the left-hand side) and SLWC3 (on the right). The median results from individual sets of PCAL flashes are shown in black crosses. The two data points with the lowest observed base voltages are from PCAL flashes on Saturn. Thebestfitisshownasaredcurve. Fig. 4 LinearizedPCALsignalsforthecentraldetectorSSWD4,fromthenominalmodel ontheleft-handsideandfromthebright-sourcemodeontheright. the detectors, the typical sample standard deviation for the linearized PCAL signalsisoftheorderof2%forthebright-sourcemodeandislessthan1%for the nominal mode. Since the linearized voltage is proportional to the optical load on the detector, and the PCAL power and illumination pattern was kept the same for both detector modes, the linearized voltage ratio of the nominal mode to the bright-source mode gives a zero-point gain scaling factor, G , 0 from the bright-source to nominal mode. The resulting G varies between 4.1 0 and 5.4, depending on specific detector. 2.4 Frequency-dependent Gain Correction In addition to G , which was derived from low-frequency PCAL signal time 0 lines,wealsoexpectanadditionalfrequency-dependentscalingfactorbetween the two detector setting modes to account for the higher frequency signal 8 NanyaoLuetal. modulations in interferograms. This can arise from the fact that bolometer time constant depends on the bias voltage used. While there is a correction forthefinitebolometertimeconstantimplementedinthepipeline,anyresidual effect from imperfect correction could lead to some spectral shape distortion. Inthenominalmode,thispotentialresidualspectralshapedistortionissimply corrected for at the flux calibration step in the pipeline. To make use of the same flux calibration product for the bright-source mode, we introduced a frequency-dependent gain factor, G , which is to be applied to bright-source f mode spectra. Fig. 5 shows a number of pair-wise ratios of the nominal to bright-source modeforthetwocentraldetectors,SSWD4andSLWC3,usingdarkskyspec- tratakeninthelowspectralresolutionconfiguration.Onlythezero-pointgain correction, G , has been applied to the bright-source data here. Each pair of 0 observations were carried out close in time so that the telescope emission, which dominates the signal observed, remained unchanged over the observa- tional pair. Apart from an increased uncertainty at the low frequency end, where the removal of the instrument emission, which is significant only near that end of SLW, introduces additional flux uncertainties, the ratios show ap- proximatelyalineardependenceonfrequencyforeachdetectorarray.Alinear fit was applied to the data of SSWD4 or SLWC3, resulting G as a function f of frequency. Note that the correction associated with G is less than ∼5%. f Similar fits were obtained for all the other detectors. 3 Pipeline Implementation Fig. 6 is a flow chart illustration on how the bright-source mode data pro- cessing is folded into the standard nominal-mode pipeline. The bright-source mode pipeline processing is the same for the nominal mode except for the following three stages in the pipeline: (a) The phase-related gain correction option is turned on for the bright-source mode at the step of the engineering data conversion. (b) While both detector modes share the sample nonlinear- itycorrectionmodule,theyuseseparatenonlinearitycalibrationproducts.(c) AftertheFouriertransform,thereisanextrastepforthebright-sourcemode, i.e., each spectrum is multiplied by a combined gain correction factor, G G , 0 f which is a linear function of frequency for each detector. 4 Calibration Results The validity and consistency of the calibration scheme described above can be studied by comparing the bright-source pipeline results with those from the nominal mode for some bright sources that are observable in both observ- ing modes or with independent flux models of very bright celestial standards. Fig. 7 checks the pipeline results of all the high-resolution dark sky observa- tions taken in the bright-source mode over the entire mission. The two cen- tral detectors, SSWD4 and SLWC3, are shown here. Individual spectra have SPIREFTSbright-sourcemode 9 Fig.5 Frequencydependencyofpairwisedarkspectrumratiosofthenominalmodetothe bright-sourcemodeforthetwocentraldetectors,SSWD4andSLWC3.Onlythezero-point gain correction, G0, was applied to the data of the bright mode here. Independent data pairsarecodedindifferentcolors.Thespectrahavebeenslightlysmoothedinfrequencyto reducenoise.ThetwothickbluelinesarethebestlinearfitstothedataoverSSWD4and SLWC3,respectively. Fig.6 Aflowchartillustrationontheintegrationofthebright-sourcemodedataprocessing intotheexistingnominalmodepipeline.Thethreeprocessingstageswherethetwodetector modesaretreateddifferentlyaremarkedby(a),(b)and(c),respectively,andaredescribed inthetext. 10 NanyaoLuetal. Fig.7 Pipelinespectraofallthehigh-resolutiondarkskyobservationsmadeinthebright- sourcemodeovertheentiremission.Shownherearethetwocentraldetectors,SSWD4and SLWC3.Eachspectrumhasbeensmoothedtoreducenoiseandfringes. been smoothed to reduce effects of noise and fringes. A dark observation is dominated by the warm telescope emission, with an in-beam flux density of ∼200−800JyoverthewholeFTSbandwidth.Aperfectfluxcalibrationwould yieldaflatspectrumat0Jyfortheseobservationsasthetelescopeemissionis removedinthepipeline.Itisevidentthatthemeanfromthesedarkskyspectra iscloseto0Jy.Thesamplestandarddeviationisoftheorderof0.5Jy,except for the low frequency end of SLWC3 where the scatter is somewhat elevated mainly due to the fact that the removal of the instrument emission, which is significantonlynearthatendofSLW,introducesadditionalfluxuncertainties. Fig. 8 shows the (smoothed) spectral ratios of the nominal mode to the bright-sourcemodeforafewindependentobservationsofNeptuneandUranus in the central detectors, SSWD4 and SLWC3. Neptune and Uranus are the main photometric flux standards for SPIRE and span a flux-density range from a few tens of Jy to 220 and 500Jy, respectively, within the SPIRE FTS bandwidth. It is evident that these spectral ratios are all within 2%. Fig. 9 shows the (smoothed) spectral ratios of the nominal mode to the brightmodefortwomassivestars,EtaCarandAFGL2591.WithintheSPIRE FTS beams and bandwidth, these two sources span a flux density range from a few tens of Jy to about 600 and 1,000Jy, respectively. The flux differences betweenthebright-sourceandnominalmodesareagainwithinabout2%here. For sources even brighter than those in Fig. 9, there is no longer any nom- inal mode data for comparison as severe saturation would have occurred. For a few bright planets, there are reasonably accurate model spectra available. We compared our bright-source mode observations with the model spectra