Observing*Checklist:*A3ernoon*

Transcription

1 Ay#122a:# Intro#to#Observing/Image#Processing# (Many&slides&today& c/o&m.&bolte)& Observing*Checklist:*A3ernoon* Set*up*detector*(format,*gain,*binning)* Bias*frames* Dark*frames*(if*necessary)* Arc*Lamps*(for*spectroscopy)* Upload*coordinate*lists* procedures*in*place.* Check*weather.*

2 Detector*Decisions*L*Gain* gain ** units*of*electrons*(e L )*per*pixel*/*analogltoldigital*unit**(adu)*per*pixel.* Note*that*by*conven@on*this*is*inversely*propor@on*to*the*voltage*gain*of*the* system.* usually*in*range*of*1 10*e L /*ADU.**** some@mes*can*be*set*by*observer.* ADU*are*also*known*as*DN*=* data*numbers *or* counts.* The*largest*number*of*counts,*corresponding*to* satura@on *is*set*by*the* A/D*converter,*at*(2 #*bits* *1).**Want*to*choose*the*gain*to*equate*this* maximum*roughly*to*the*fulllwell*e L *depth*on*the*pixel.** for*n=16*bits,*maximum*counts*is*??*****2 16 1*=*65535** for*e.g.*300,000*fulllwell*capacity*of*deep*deple@on*ccds,*gain*should*be*??**** The*smallest*number*of*counts*is*set*by*the*RN.*Want*to*set*the*gain*to* sample*the*read*noise,*but*not*overlsample*it.* * Detector*Decisions*L*Binning* * OnLchip*binning*produces:* Smaller*files*L*not&so&important& Faster*readout*@me*L*can&be&important& Less*readout*noise*per*area*L*can&be&important* For*direct*imaging*in*broadband*filters,*readnoise*is*very* rarely*an*issue.*you*want*the*fwhm*of*point*sources*to* be*at*least*2.5*pixels*in*order*to*properly*sample*the* point*spread*func@on.*if*you*are* oversampled,*that* doesn t*usually*have*any*dire*consequences.** Undersampled *data,*by*contrast,*can*be*a*problem*for* good*photometry.**the*readout*@me*can*be*a* determining*factor*in*deciding*whether*to*bin.* For*spectroscopy,*readnoise*some@mes*is*a*significant* component*of*the*noise.*having*fewer*pixels*under*the* spectrum*reduces*the*noise*in*a*resolu@on*element.**if* spa@al*informa@on*is*unimportant,*binning*in*the*spa:al& direc@on*has*merit*though*it*can*compromise*the*fidng* and*subtrac@on*of*sky*lines.**binning*in*the*spectral& direc@on*can*reduce*spectral*resolu@on,*so*use*cau@on.* spectral* direc@on* spa@al* direc@on*

3 Noise Sources in a Digital Image Readout Noise: Caused by electronic noise in CCD output transistor (uncertainty in A/D conversion) and in the external circuitry (spurious electrons); typically σ RN ~ 1-3 e - for CCDs and σ RN ~10 e - for IR arrays. Can be reduced by binning. Dark Current: Caused by thermally generated electrons (since operating detector at T > 0 K), and mitigated by cooling. Generally negligible for CCDs and now <1 e - /second/pixel for IR arrays. Can be measured with Dark Frames. Pixel Response Nonuniformity: Also called Pattern Noise. QE variations due to defects in the materials and manufacturing. Removed by Flatfielding. Photon Noise: Also called Shot Noise. Photons arrive in a random fashion described by Poisson statistics. Present in all of your data and your calibration frames. CCDs Are Not Cosmetically Perfect Bright Column (charge traps) Hot Spots (high dark current, but sometimes LEDs!) Dark Columns (charge traps) QE variations Cosmic rays

4 Bad*Pixels,*Cosmic*Rays,* other* image*defects * [R.*Chromey]* Preliminary*Processing* There*are*two*types*of*instrumental*signature* to*remove:* Addi@ve:** Bias*Level*or*Offset* Bias*Structure* Dark*Counts* Fringing*effects* * Mul@plica@ve:** Q.E.*varia@ons*on*all*scales* Fringing*effects* Constant*#*of*counts*added* independent*of*the*detector* illumina@on.* Constant*frac:onal*effect,* depending*on*illumina@on.*

5 Reducing a CCD or IR Array Image Raw Data Science Frame Flat = image of a uniformly illuminated surface Bias = a zero level image Calibration exposures Dark or Bias Flat Field Bias Image Sci. -Dark Flat -Bias Sci-Dk Flt-Bias Output Image that you fluxcalibrate with images of standard stars and then analyze. (Non)linearity* Want*detector*response*to*be*directly* propor@onal*to*incident*signal*and*thus*t.* CCDs*are*quite*linear*right*up*to*satura@on.* But*IR*arrays*are*inherently*nonLlinear!*** For*a*source*that*would*produce*an* intrinsic * count*rate*r*(adu/sec),*we*may*measure*n* counts*a3er*t*seconds.**define*linearity* correc@on*λ(n):* Λ =**N**/*(R&t)* This may be calibrated using a series of exposures with different integration times taken of a constant source (e.g. a stable flat field lamp). Λ(N) is usually different for each pixel, and correction should be applied in a two-dimensional fashion. But many people are lazy and don t do this.

6 Flat*Fielding* such*as*dust*on*filters,*the*dewar*window*or*detector.* For*spectroscopy*refer*to*the* Ideally,*illuminate*the*detector*with*a*source*that*is*as* flat*on*the*sky*as*the*background*and*collect*at*least*a* million*photolelectrons*per*pixel.*then*could*flatlfield*to********* *****************************************************=0.1%.* 10 6 = Twilight*Flats* Sunlight*scapering*at*twilight*illuminates*the*telescope* just*like*the*dark*night*sky.* Take*a*sequence*of*spa@ally*dithered*frames*with* telescope*tracking*on*(otherwise*can*have*star* streaks).* Sky*brightness*changes*rapidly.**Can*be*challenging*to* Work*very*well*for*imaging.*** Usually*difficult*to*get*enough*counts*for*spectroscopic* observa@ons.*also,*have*to*watch*out*for*effects*of* lines*(e.g.*oh)*in*the*background.**

7 DarkLsky*Flats* Can*median*combine*all*of*your*onLsky*data* frames*together*to*make*a* super*sky*flat. * Generally*want*to*smooth*this*a*liple.* These*tend*to*work*very*well*for*imaging* when*feasible*since*they*match*the*sky* illumina@on*and*any*gradients*perfectly.* But*they*have*much*lower*signalLtoLnoise* ra@o*than*dome*or*twilight*flats.* And*they*some@mes*contain*fringes.* Example*Flats*

8 ILband* Imaging*flat*field*examples* Dust*on*filter* VLband* Note*differences*with*color.*This* means*that*objects*with*different* spectra*will*be*flatlfielded*slightly* incorrectly.* Rings*due*to*non*uniform*thinning* ULband* Two*amplifer*readout*

9 Flatfields: sky vs. dome screen 10% variation! Fringing

10 (From T. Dahlen)

11 PointLLike*and*Small*Sources* ** Dither the telescope through some non-repetitive pattern in order to place objects on different portions of the f.o.v. with each dither position Extended*Sources* offset field 1 object field offset field 2 target object faint (hopefully) objects in background fields Combine images of offset fields (w/ medians or masking objects) and subtract from object image

12 Gaps:*nonLuniformity,*nonL rows*(dither*strategy)* Gain,*QE,*uniformity,* detector!** NonLphotometric*data,*effects* of*airmass *now*need*some* thought* Lflat*fields*not*necessarily* flat L* why?* want*1*adu*(count)*to*mean* the*same*thing*everywhere* produce*stacks*