More Stray Light

I looked at files that were darks (detector blanked off) along with high and low flux images from  5 different filters (18a,  bp14100, N10692-9, Ocli O, BBP5030_5725).  I corrected for number of coadds and took the ITIME parameter as the integration time for each frame.  There were a couple filters where the data were taken with SUTR (4 frames).  I took the last frame except for one case where the signal was going non-linear, so I took the 2nd of the 4 frames.

Comparing the 4 frames in SUTRs, there is a clear difference between the "dark" regions when there is strong light and when there is not.  For the case where very little light is coming through, getting about 2 ADU/sec at 5 microns.  When there is a fair bit of light, the "dark" regions have more signal when the orders themselves have more signal.  Values are something like 5 ADU per second (peak is around 20 ADU per second)  This suggests there is light from echelon orders spilling into inter-order regions.  Consistent with echelle scattering 20% of the light at 5 microns.  


Looking at all frames, I also see clear correlation of "dark" region signal and peak signal.  For 18a, BP14100, N10692-2, and OCLI O, the extra light is at about the same level, ~10 ADU per second.  The amount of extra light at BBP5030 is more.   This suggests the echelle scattering is not as bad at long wavelengths.  


These results are not definitive, but they make some sense given what I expect for the grating performance (more scattered light as going to shorter wavelengths).

A better test for CD14 will be to include a dark when doing filter scans and to take the data with SUTR and shortening the exposure time as needed.

Deflection estimates

To evaluate the use of the triangles, Jeff is going to do some modeling.

3 configurations: No triangles, triangles attached only to plate, triangles attached to plate and EXES

2 orientation: Triangles vertical on plate (telescope at 40 degree elevation) and SI flange rotated 20 degrees (telescope at 20 or 60 degree elevation)

2 gravity: 1g (static) and 0.9g (assuming 0.1g gets through to SI flange)

Looking at deflection in z at 300 mm forward of SI flange (nominal SOFIA focus) and at 762 mm forward of SI flange (nominal cold stop location).

Believe can evaluate sag based on motion in EXES focal plane scaled to SOFIA focal plane and motion of cold stop as fraction of secondary image.

EXES focal plane platescale is roughly 8"/mm and SOFIA focal plane platescale is ~4"/mm.  Want deflection to be <1" with roughly 4" images.   This gives 0.250 mm deflection at SOFIA focal plane.

Secondary mirror is 350 mm diameter and, from zemax file, we found it to be 4.895m from EXES entrance lens.  The entrance lens images the secondary onto the cold stop.  The cold stop is 38mm diameter and, from Autocad file, is 711 mm from entrance lens.  If the sag should be <2.5% of the cold stop diameter, then we get that a deflection of 0.95mm is acceptable.

The deflections of interest are not the static ones, but deflection between the vertical 0.9g case and the 20 degree/1g case.

Best focus position of the pupil lens

We have the laser fiber in the bumper and the stopped-down pupil mask with a bearing placed in the 1/4" hole to provide a target on which to test the focus.

The target image appeared to be very sharp on the grating and at the collimator (seen below on a card just forward of the collimator mirror).

As a starting point for the 1" KRS5 pupil lens, Joe and I placed 14 mm into its lens wheel barrel, (see post "Current Lens Wheel Configuration")

For this test, I took images with the bare-MUX and red laser with the KRS5 lens placed at different positions in the barrel.

From left to right, these distances are: 13.1 mm, 14.3 mm (closest to calculated value), 15.2 mm, 16.2 mm, 17.3 mm, 18.3 mm.
Visually, 17.3 mm seems to be the best focus. Using the 1" KRS5 lens in this position we will try to align the ellipsoid while EXES is on the TAS.

Current lens wheel configuration

When Joe was here on 4/30/12, we installed and adjusted the lenses in the lens wheel. For future reference, here's some measurements and a record of the current configuration of the lens wheel.

distances:
detector to O-box wall (CD013):   10.9 mm
detector to O-box wall (proposed for CD014) :  7.7 mm
O-box wall to lens front of wheel: 14.4 mm 
focal length of 1" KRS5 lens:  38 mm (measured at optical)
0.63um focal length of 1.5" KRS5 lens: 38 mm
0.63um focal length of 1.5" ZnSe lens: 38 mm

To calculate how deep to place the lenses in the lens barrels we calculated d_i from the lens equation:
d_i= (1/f - 1/d_o),  where we used 1000mm for the value of d_o.

For the 38mm focal lengths, this was d_i=39.5 mm

n(0.63um) =2.57
n(11um)=2.37

f_mir= 38mm * (2.37/2.57) = 35 mm

d_i=36.27 mm


For the ZnSe lens, we chose the closest possible distance to the detector,  ~ 25.3 mm in order to see the effect of an extreme case on the focus.

For the 1" KRS5, we chose the best 0.63 um focus distance for the current config (with the D-box standoffs the same as CD013),   total distance is 10.9+14.4+7.9mm = 39.5 mm

Before the next cool down (CD014) we will move the lens 3.3 mm away from the detector, to compensate for the 1/8" inward change of position of the detector.

BBP5030_5725

Same basic process as for 18a.  Images saved as groups of 4 SUTR, so photon response is easy using image differences.

First off, here is wavelength as a function of echelle angle.

Shows that when we're in main part of filter window, things are ok.  Have a light leak at near 45 degrees near center of an echelle order and near 55 degrees near edges of echelle orders.

Next is echelon order spacing as a function of wavelength and angle.  This is near the short wavelength extreme for EXES.  Note that we can get almost 50 echelon orders in a single image, which would be about 30 cm-1.

I got data numbers from the difference between the first and last images in the SUTR.  I took the 75%ile of pixels which were over 150 DN in the difference.  Plotting row 550 of all the images in the series is in the figure below.  The value I take for each image is the red dot.

Finally, assuming 9 seconds of integration between first and last images and a resolution of 4 km/s, I get throughput numbers as below.  Again, they are too high.  I wonder if they are even higher than the factor of 4 too high seen for 18a.

18a filter info

I've finally gone through the 18a filter.  From the logbook, we used ctdelay-110 (3 seconds) and the 3.5 km/s slit (which gives something more like 3.8 km/s at 18 microns).

Joe looked at the images and decided if the signal for each echelle angle was on, off, or weak.  He also recorded if the data frame was clean (only one echelle order) or not.  This info was saved in a csv file.

I read this into IDL, used quantities from the blog regarding echelle angle and wavelength for given voltage, and worked out the wavelength for each setting.  To do this I started with the assumption that the wavelength getting through was in echelle order = 3 and then adjusted the lower angle measurements to order = 2 so that the wavelengths that came through, the number of orders, and the filter warm curve made sense.

I saw that the array had some weird data behavior on the first row and last row (see blog post before this one), so I reset those, as well as pixels with extreme values, with an estimate of the median reset value.

I used a function I call ORDERSPACING to determine a single value for the number of pixels between echelon orders for each clean file (see figure below).  These can be used to check models for the instrument.

I then got an estimate of a "typical" value for each good image by looking at values where there was clearly light, requiring there to be >5000 such pixels, and selecting the 25%ile (more light means lower numbers) as "typical".  I subtracted the no-light values I found for each image to get the number of data numbers corresponding to the incident light.  The plot below shows a cut along row 510 of each good image (there are 25 good images) using the small dots.  These values go across orders near the peak of the blaze.  My typical values are shown in red with each frame having a single value.  For most of the frames, these values are not at the top, but do reflect larger than median signal.  For frames with low response, the "typical" values seem too high.  The green dots are my "no-light" values.  Note that I believe there is some light on the detector beyond what is landing in the orders (see comment to post on scattered light).

I finally calculated throughput using logbook/blog values for the integration time (ctdelay = 110 or 3 seconds), gain (35.9 e- per DN), slit resolution (3.8 km/s), and assuming emissivity =1 and temperature = 295 K.

I calculated the area*omega product in two ways.  First, I used the half angle that we measured during CD13 where we mapped a soldering iron tip across the beam and divided by an estimate for the distance to the focus (note that we had a KRS-5 lens not designed for EXES, so it just has a stock focal length that is close).  This half-angle = 0.0251 radians.  I calculate the corresponding solid angle as 2*!pi*(1-cos(0.0251)) = 0.001979 sr.  I next took EXES design numbers for performance on SOFIA.  Since area*omega should be invariant, if we had the proper SOFIA lens in place, this should be equivalent.  We didn't have that lens in place, so I view this more as a sanity check if we're close.  Using the SOFIA telescope = 2.375m diameter (accounts for undersized cold stop), EXES as an f/10 system, and 25 micron pixels, I get a platescale of 0.217 arcsecond per pixel or 1.053e-6 radians per pixel.  The net result ends up that area*omega from CD13 is about 4 times larger than using SOFIA design parameters.

Anyway, this entire digression of area*omega is possibly relevant because when I calculate the number of photons that should be incident per pixel assuming 100% throughput compared with the record number of electrons, I get unrealistically high numbers for the CD13 experimental area*omega and realistic numbers for SOFIA design parameters.  I'm hoping someone will find an error in my calculations.

Below is my plot for throughput.  White is with CD13 area*omega and red is the SOFIA design.  I wouldn't be surprised if the true throughput is even a factor of two below the red points for this wavelength.  I plot values as a function of wavelength and as a function of echelle angle.  It is apparent the echelle is more efficient at angles <55 degrees.  You can also see that we need a blocker for the filter to get the 16-17.5 wavelength region.

Some strange data values

Looking at 18a filter scan data, I noticed ALL the data frames (single destructive reads) have some funny behavior.  In the plot, the white dots are the first row of 1024 pixels, the green are row 510 (but are generally true for rows 1 to 1022), and the red dots are the last row of 1024 pixels.  

You can't really see it from this plot, but the first row clearly shows the 4 readout chains while the others don't show that imprint clearly.  

The first row, after 4 reference pixels near 15,000 ADU, jump to 16,000 ADU and slowly climb to 16,200 before a short jump down to the last 4 reference pixels.  

The middle rows look like like good data with a sharp jump on the end 4 reference pixels the only big weirdness.  It is a little weird to me that reference pixels at start have more charge than data.

The last row almost seems pinned at a saturation value (3200 ADU).  First four reference pixels are all around 14,500 ADU and last four reference pixels are 10,500; 10,500; 10,500; 2500 ADU.

Don't know if this is partially coming from ebox interacting with array, but something looks screwy.

Finding a Common Best Focus Position for High & Medium Resolution Modes


Last week I finished running scans of both Medium and High Resolution Zemax models for different settings of collimator distances. I used my Medium Res results to pick a detector distance of 95mm from the turning mirror and 35 turns on the Echelle Collimator.  The blue straight line on the graph is shows Medium Res best focus position for different positions of the Echelle Collimator.  The brown curve shows the Medium Res best focus geometric spot radius for those different Echelle Collimator positions.

  Then I went to my High Res results, found where the detector distance of 95mm and Echelle Collimator setting of 35 turns coincided (roughly) with an Echelon Collimator setting of 10 turns. The individual straight lines represent best focus detector positions for different Echelle Collimator positions at one Echelon Collimator position. The extrapolation on the High Res graph indicates about 13 turns needed on the Echelon Collimator to coincide with the Medium Res best focus position.

 So... going to my Zemax models and setting these conditions I get an RMS spot size (radius) of ~18microns in Medium Res (vs Airy radius 167microns) and an RMS spot size (radius) of ~101microns in High Res (vs Airy radius 143microns.)  The best setting for the Echelon Collimator was about 11.4 turns instead of 10 or 13 ... not wildly different considering how I got all of these numbers.

The logarithmic PSFs for both modes are essentially the same with both having FWHM of 200 microns or 8 pixels at 13.7 microns wavelength:

 

In other words, this method works pretty well - in the models.  Also, my value of 95mm is, I think, indistinguishable with the current measured value of 93mm.

Now - how do we know when we've set the real collimators at some 'n' number of turns with respect to the nominal Lacy model positions?

Distance from turning mirror to detector plane

We took advantage of having the flip mirror side-wall exposed today and measured the distance from the center of the turning mirror to detector plane.

Using calipers, the center of the turning mirror to the outer edge of the O-box wall is 3.24"~ 82mm

The detector plane is 10.9 mm from the outer edge of the O-box wall.
This value was obtained using the SolidWorks drawing of the D-box, and adjusting/assuming the following values:
        1) The G-10 standoffs are 4.445mm in the drawing, but for CD013 are actually 3.82mm.
        2) Assumed that the detector plane is 6.35mm behind the front cover of the detector socket.

Our calculation puts the detector plane 10.9 mm away from the outer edge of the O-box wall or
93 mm (3.66") from the center of the turning mirror (for CD013).