NICMOS Focus Field Variations (FFV) and Focus Centering– · NICMOS Focus Field Variations (FFV) and Focus ... A.Suchkov & G.Galas March 16, 1998 A ... we present empirical results

1

NICMOS Focus Field Variations (FFV)and Focus Centering±

A.Suchkov & G.GalasMarch 16, 1998

ABSTRACT

NICMOS foci are known to vary across detector’s field of view. These variations must betaken into account when determining ‘‘best’’ focus from a particular stellar image. Focusfield variations (FFV) may introduce substantial positional dependence in aperture cor-rections, thus they may also be important for high precision stellar photometry. In thisreport, we present empirical results of a study of time dependence as well as PAM settingdependence of FFV for the NICMOS cameras. Although the results are not easy to inter-prete because of substantial errors, they still suggest that some amount of both time andPAM position dependence of FFV is consistent with the cameras 2 and 3 data. There is apreliminary understanding of why FFV should depend on PAM setting and may changewith time. Howevera detailedinterpretationof theactualFFV timeandPAM dependenceis beyond the scope of this report and is left for future analysis. As a practical result, wepresentthecoefficientsof thesecondordersurfacefit to theFFV dataaveragedover timeand PAM positions. They are recommended for use in the NICMOS focus monitoring pro-gram to reduce focus to the detector center (focus “centering”).

1. Intr oduction

Along with temporalfocusinstability, NICMOSfocuswasfoundto varyacrossdetec-tor’s field of view (Krist, 1997, Suchkov, Bergeron, & Galas 1997). For camera 2 andcamera 3, the magnitude of the effect is as large as ~1.5 mm in PAM (Pupil Alignment

Mechanism) space, comparable to the nominal tolerance of . This is a fairly big

amount, which makes it necessary to take focus field variations into account as camera’sfocusis derivedfrom PSFmeasurementsataparticularplacein thedetectorfield of view.FFV mean that there is essentially no single focus for a camera. One has to refer to a par-ticular position on the detector, e.g., the detector center, when characterizing focus. Thusthe knowledge of the geometry of focus variations across the detector became an impor-tant component of the NICMOS focus monitoring program. Until now, focus from phase

1 mm±

Instrument Science Report NICMOS-98-005

2

retrieval analysis was reduced to the detector center using the FFV parameters derivedfrom the data which were obtained on August 12, 1997. We have undertaken the presentstudy to obtain more robust centering parameters using the data from many observations(visits of the focus monitoring program 7608). Also we wanted to see if those parametersarechangingwith timeand/orPAM positionandquantifythesechanges.Onemayexpectto see some PAM dependence because at different PAM positions the detector’s field ofview is imagedondifferentpartsof thecurvedfocal surface.Also a timedependencecan-not be ruled out because of the continuing mechanical processes within the cameras. Sothe idea was that a subsequent analysis of any FFV dependence on time and PAM settingmight help to better understand the changing NICMOS optics and/or mechanics of thedetector motion in NICMOS cameras.

2. Data

To quantify FFV dependence on time and PAM position, we have used the data fromthe first 8 visits of the NICMOS focus monitoring program 7608 (Suchkov, Bergeron, &Galas 1998). The data cover the period of about 4 months, from June 9, 1997, throughSeptember 24, 1997. The program was designed so that in each visit a stellar field of thesame target, the open cluster NGC 3603, is observed at 17 different PAM positions rang-ing from ~-8 mm to ~+8 mm in PAM space. This provides 17 image frames per cameraper visit. We have selected frames taken at three different PAM positions, which made 72framesto studyandabout1600starsto dophaseretrieval on.Threeframesin avisit werethought to be enough to check the magnitude and the sense of the FFV dependence onPAM position while leaving the amount of efforts to do all the data reduction within rea-sonable limits. Two of the PAM settings were chosen to be adjacent and not too close tocameras’1 and2 bestfocus(which is basicallyin themiddleof thePAM rangein thecaseof cameras 1 and 2, and beyond the PAM range in the case of camera 3). The third PAMposition was taken at the opposite side of the PAM range.

For camera2 andcamera3, mostof theframeshaveabout~25starswhichsamplerea-sonablywell thedetector’sfield of view andaresuitablefor focusmeasurementsbasedonphase retrieval (phase retrieval technique is described in Krist & Burrows 1995, its appli-cation to measurements of NICMOS foci is given in Burrows & Krist 1997 and Suchkov,Bergeron,& Galas1998).Thestellarfield in camera1 framesis moresparse,andsomeoftheframeshave too little starsto providegoodsamplingof thefield of view, whichmakesthe results more noisy than for the other two cameras (because of that, camera 1 is dis-cussed below much less then cameras 2 and 3).

3

Figure1: Timedependenceof NICMOSfocus(mmof PAM space)at thedetectorcorner, asobtainedfrom thefirst ordersurfacefit to thefocusvaluesacrossthe

detectorfield of view at threedifferentPAM positions.R is thecorrelationcoefficient,andthethick straightline is thelinearfit. A gapseenin theupperpanelsis dueto visit 5 omit-ted in NIC 1 results.

(x 0 y, 0)= =

4

Figure 2:Time dependence of the FFV first order surface fit coefficients for NIC 1 atthree different PAM positions.

The results of phase retrieval were visually examined in the focus–XY-coordinate dia-gram to identify and remove outliers as well as to get an idea on the magnitude andpossible geometry of FFV. Only a few frames were found to have obvious one or two out-liers which turned out to be either too faint stars or stars with overlapping PSFs. Thosestars were removed from subsequent analysis. In the case of camera 1, we dropped allthreeframesfrom visits1 and5 asthey hadeitherinsufficientnumberof starsor producedtoo noisy focus results.

5

Figure 3:Time dependence of the FFV first order surface fit coefficients for NIC 2 atthree different PAM positions

6

Figure 4:Time dependence of the FFV first order surface fit coefficients for NIC 3 atthree different PAM positions

7

Figure 5:Time dependence of the FFV second order surface fit coefficients for camera 1at three different PAM positions (separated by thick horizontal bars)..

8

Figure 6:Time dependence of the FFV second order surface fit coefficients for camera 2at three different PAM positions (separated by thick horizontal bars).

9

Figure 7:Time dependence of the FFV second order surface fit coefficients for camera 3at three different PAM positions (separated by thick horizontal bars).

10

3. Results

The actual results of phase retrieval for chosen visits and PAM positions are given inAppendix(Figures 11 through19)asthebestfocusat thepositionof aselectedstar(mmin PAM space) plotted againstx- andy-coordinates of the star (coordinates are given inpixels).Overplottedis asecondorderpolynomialfit to theobservedfocusvalues.Onecansee that even a 1–D fit displays a sustained pattern across most of the frames, at least forcameras 2 and 3. The pattern is obviously suggestive of a nonlinear focus trend acrossdetector’s field of view.

We have fitted the data points in Figures11 to 19 with both a first order surface,

and a second order surface,

Thefirst orderfit providesaneasyway to seeif thereis any timeand/orPAM positiondependence in detector’s tilt with respect to the focal plane. The fitting results are dis-played in Figures 1 to 4 (see Table 1 for correspondence between visit number and the

date of observation). Figure 1 gives the focus at the detector corner for

eachcameraat threedifferentPAM positions.Onemaynoticeaslightly smallercamera2focus values at the negative side of PAM settings and an upward trend of camera 3 focuspersistent across PAM positions. With regard to FFV coefficients given in the next threeFigures, the most conspicuous feature is probably a substantialy-tilt in camera 2. The tiltseemsto beabit largerat thenegativePAM settings.Onemayalsonoticea trendin thex-tilt, seemingly dependent on PAM position. Real or not, this trend does not present a bigproblem for focus centering since thex-tilt is small. But, if real, it may be of interest forinterpreting camera’s detector–optics connections.

Thereis nodoubtthattheactualNICMOSfocusfield variationshaveasubstantialsec-ondordercomponentin x andy, andwewantto take it into accountin focuscentering.Toderivecenteringcoefficients,wehavefitted thephaseretrieval resultswith asecondordersurface given by equation (2). Time dependence of the coefficients of equation (2) isshown in Figures 5 through7.

focus a0 axx ayy , (1)+ +=

focus a0 axx ayy axxx2

ayyy2

axyxy. (2)+ + + + +=

(x 0 y, 0)= =

11

Table 1.Date of observation for the visits of the focus monitoring calibrationprogram 7806 used in this study.

Inspection of Figures 5 to 7 reveals that, of all the cameras, only NIC3 shows signifi-

cant correlation of with the visit numbers at . However,

most of the correlation is due to the first two visits. If those are disregarded, one can statethat there is no or a very weak evidence that NICMOS FFV systematically changed overthe period under consideration. The first two visits are also the major contributors to the

large scatter in the fitting coefficients at .

As for the PAM dependence, there are indications that, for camera 2, for which focus

variation occurs most significantly alongy-axis, they-tilt, i.e., , is a bit larger at

, just as in the case of the first order surface fit. A similar pattern can be

seen for camera 3 if the first two visits are ignored.

Weleaveadetailedinterpretationof theresultsfor futureanalysis.For thepurposesoffocus centering, we prefer, for the time being, to average the fitting coefficients over allvisitsandPAM positions.Table2 presentstheresultsof thisaveraging.In thecaseof cam-era 3, the first two visits were rejected for the reasons discussed above.

visit numberdate of

observationday sinceJanuary 1

1 9-Jun-97 160

2 30-Jun-97 181

3 14-Jul-97 195

4 28-Jul-97 209

5 12-Aug-97 225

6 25-Aug-97 237

7 8-Sep-97 251

8 24-Sep-97 267

ax and axy PAM 6 mm–=

PAM 6 mm=

ay

PAM 6 mm–=

12

Table2. Coefficientsof equation(2) obtainedby averagingtheresultsof individualfits for18 frames for camera 1 and camera 3, and 24 frames for camera 2.

Figure 8 displays the surface describing camera’s FFV in terms of equation (2) withcoefficients from Table 2. It differs little from Figure 7 in Suchkov, Bergeron, & Galas(1998) which displays the FFV fitting results obtained from a single visit (these resultswere used in focus centering till the end of the focus monitoring program 7608 in Febru-aryof 1998).Becauseof that,focuscenteringbasedonequation(2) andTable2 shouldbeexpected to produce the results not very much different from those in Suchkov et al(1998). Comparison of Figure 9 and Figure 10 shows that this is indeed the case.

Despite the little difference between the FFV fit used before and the present “aver-aged” fit, we recommend to use for focus centering the results based on equation (2) withcoefficients from Table 2 as they are statistically more accurate and have well defineduncertainties.

camera

NIC1 -2.869 +/- 1.529 1.383 +/- 0.542 1.199 +/- 0.416 0.229 +/- 0.145 0.068 +/- 0.227

NIC2 -1.304 +/- 0.211 2.819 +/- 0.168 0.397 +/- 0.056 0.605 +/- 0.061 0.614 +/- 0.079

NIC3 -3.682 +/- 0.159 -6.578 +/- 0.120 1.580 +/- 0.066 1.710 +/- 0.039 -0.055 +/- 0.062

ax 103× ay 10

3× axx 105× ayy 10

5× axy 105×

13

Figure8: NICMOSFFV asgivenby equation(2) with coefficientsfrom Table2 andfocusset arbitrarily to 0 at x=y=0.

4. References

Burrows, C.J. & Krist, J.E. 1997, Memorandum, March 19, 1997, ‘‘NICMOS Focus’’

Krist, J. 1997, in: 1997 HST Calibration Workshop, STScI.

Krist, J. E. & Burrows, C.J. 1995, Applied Optics, 34, 4951

Suchkov, A., Bergeron, L., & Galas, G. 1997, in: 1997 HST Calibration Workshop.

Suchkov, A., Bergeron, L., & Galas, G. 1998, ISR NICMOS-98-000

5. Appendix

The following Figures include the plots of NICMOS focus history (Figures 9 and 10)and actual focus measurements used in this report to study FFV (Figures 11 through 19).

14

Figure9: NICMOSfocushistoryfrom Suchkov etal (1998).Focusfrom phaseretrieval iscentered with FFV coefficients from a single visit used through February 1, 1998.

15

Figure 10:Same as in previous Figure but for focus centering based on equation (2) withcoefficients from Table 2.

16

Figure11: Focusin camera1 measuredatdifferentx andy positions.IndicatedareimageID and PAM position. Focus is given in mm of PAM space.

17

Figure 12:Same as in previous Figure but for a different PAM setting.

18

Figure 13:Same as in previous Figure but for a different PAM setting.

19

Figure 14:Focus in camera 2 measured at differentx andy positions. Indicated areimage ID and PAM position. Focus is given in mm of PAM space.

20

Figure 15:Same as in previous Figure but for a different PAM position.

21

Figure 16:Same as in previous Figure but for a different PAM position.

22

Figure 17:Focus in camera 3 measured at differentx andy positions. Indicated areimage ID and PAM position. Focus is given in mm of PAM space.

23

Figure 18:Same as in previous Figure but for a different PAM position.

24

Figure 19:Same as in previous Figure but for a different PAM position.