Med mode TAAS spectrum at 6.6 um

I quickly looked through the different wavelengths at which we did medium mode TAAS FCLS spectra.  The 6.6 micron filter was first and turned out to be very interesting because of atmospheric H2O lines.

A quick look at the other wavelengths shows that the long wavelengths (18a, Soifer) are totally not worth doing unless we cut down on the light or really speed up the frame time (partial array readout).
The BP14100 wasn't that good, either.  The N07904 with slit 1 and 0.5mm FCLS was brilliant, but no interesting atmospheric lines.

The rotation and shifting used at 6.6 worked well at 7.9, which is probably because the grating angle is the same for both.  The continuum in the 6.6 is spatially offset from in the 7.9.  The FCLS parameters are:  6.6) X,Z = 43,26 and 29,14.5 with 2mm aperture, slit 6.  7.9) X,Z = 43,26 and 29,14.75 with 0.5mm aperture and slit 1.  The offset could be due to the different aperture and EXES slit.   The separation between the continua is different between the two filters -- and in the wrong sense from what the FCLS coordinates suggest -- but I'm doubtful the data are good enough to make conclusions.

There was very strange fringing in the raw files (upper left, below).  Saw enhanced DN in relatively thin bands that look like wavey lines across the detector.  They are not evenly spaced in frequency, which basically means they can not be caused by an optic before dispersion (filter, window).  Best theory is that they are from the detector.  See same basic pattern at 6.6 and 7.9, although fewer fringes at longer wavelength.

I reduced the data by taking difference of the "nod" pair with greatest extent along the slit (upper right).  I then rotated the frame using IDL new_frame = ROT(old_frame,12.0) where the rotation angle was not what made the slit vertical in imaging mode, but 1.4 degrees more than that.  This probably depends on grating angle.  With atmospheric absorption lines in the 6.6 difference, I judged the rotation by eye.  At this point the continua still ran diagonally across the detector and the spatial dimension was correct.  I then shifted the continua at each spatial position so the spectra ran along rows.   I then took a box around where all the data were good (see lower left), summed the spectral information to create a spatial weighting, and made a weighted extraction of the spectrum (lower right).

The spectrum, I later noticed, has some variation with integration time.  Not too surprising since I didn't correct for non-linearity.  We will need to look at how to do that best before continuum.  The displayed data comes from analyzing the second samples in each ramp.

The final extracted spectrum was compared with the HITRAN emission strengths for H2O at room temperature using the hitran_exes IDL procedure that lets you specify a wavenumber, a grating angle,  a mode, and a grating angle (many other things possible) and get a predicted output spectrum.  I ended up using 51.885 degrees as the echelle angle - Curtis later told me there is an offset between hi-med and med mode spectra of about 100 pixels - and overplotted the results in the lower right.  It's not perfect, but it is pretty darn close.  I don't know what causes the additional absorption near 6.51 microns.  I have since found that CH4 has a couple lines there, but don't know that we have enough CH4 to cause that.  Also, it looks more like a tight band to me at 6.51 microns, so it could be a hot band that I don't know about or just haven't stumbled across.

Below I show the impact of integrating longer.  The data had four samples in each ramp.  The black line is for the first sample, green - second, red - third, and blue - fourth.  There could be an impact due to non-linearity, but there could also be an impact in the absorbers changed.  I have trouble believing the air in the TAAS changed substantially, but the line near 1537 cm-1 really looks like a line and not a non-linearity.

CD16: FCLS quick notes

Looking at three files from CD16 where we were imaging the FCLS through the flux slit (Slit 7) with the N04679 filter.  These are simple D's and the applied bias was set to 0.5 V (to avoid saturation?).  The three images are below along with the EXES x,y that I found today and the FCLS coordinates.  The EXES x,y that I found was off by 1-3 pixels compared to what was in the lab book, but I don't think that is a big deal.

The FCLS coordinates are, according to Ted Dunham's email in Feb of 2013 and stated in a previous TAAS post, millimeters and the DAOF TAAS FCLS corresponds to a SOFIA platescale of 4.16" per mm.  From the difference between top and bottom image locations, I calculate an EXES platescale of

17.96mm / 376.2 pixels = 0.04774 mm/ pixel

or

0.04774 mm / pixel * 4.16 " / mm = 0.199" / pixel

Our design number comes assuming we are F/10 and have 25 micron pixels and is:

2.06e5 arcsec * 25 microns/pixel / 2.5e6 microns / 10 = 0.206" arcsec/pixel

Looks like we are <4% off from the design number.  Good enough.

For rotation, I first considered the angle between our slit long axis and our columns.  This comes out to be 10.6 degrees (determined by using IDL ROT(image,angle) function and seeing what angle gives best alignment of top and bottom of the slit).  Our slit is rotated -10.6 degrees clockwise from our columns

Comparing our slit to the FCLS x,z coordinates, I first plotted the FCLS coordinates for the spot and noted that the top of the slit ends up in the lower left corner of the plot.  This means our slit is rotated 

180 + atan(fcls_x / fcls_z) = 231.3 degrees

clockwise from the FCLS Z axis.  Assuming the FCLS axes correspond to +elevation (+x) and +cross elevation (+z), this means we are rotated 231.3 degrees clockwise from the cross elevation axis.  

Noise vs. Temperature

Noise vs. Temperature

MIRI detectors have shown peaks in the detector readout noise at specific temperatures, in the range of 7K where we tend to operate. To check for excess noise at our operting temperature, or to optimize the temperature at which we should operate, a series of noise runs (3 sets each of 5 ramps of 8) were taken at temperatures from 6k to 9K on the way up and then 9K to 7K on the way down. The noise results are plotted here.

There is a great deal of scatter, but generally speaking the noise seems to increase with increasing temperature. The lowest noise actually occurred at 6K rather than at 7K where we have been operating, so it may be desirable to operate the array at a lower temperature in the future.

Reference Pixel Subtraction

Reference Pixel Subtraction

In an attempt to improve linearity of the ramp and noise over the full set of scans (rather than having to ignore the first one) reference pixel subtraction was explored. Initially the simple step of averaging all reference pixels in a scan from both the leading and trailing stripe of the array gave no real improvement. The second attempt was to subtract row by row the average of the 8 reference pixels in that row. This improved the linearity of the ramps, as shown below, but worsened the noise. Scan average is plotted vs. scan number in the data cube.

First using a set of noise data, 5 ramps of 8 scans for Fowler-4 noise, that was taken during grounding scheme optimization at 2V applied bias, the initial linearity was poor (data 413 from CD016) and the noise was 1.336 DN (47 electrons) full frame, all 5 scans, or 1.168 DN full frame only the last 4 scans.

By subtracting the reference pixels row by row, the observed linearity in the following plot was improved, especially for scans 1 and 8 of each ramp, but the noise increased to 1.56 DN (55 electrons) full frame all 5 ramps.

Now trying a different set of data taken at zero applied bias to eliminate any photon generated charge noise, the initial data is a 7K run in a set of noise-temperature measurements (data 1794 from CD016). The initial ramps had larger (0.4 DN) offsets of the first and eighth scans in each ramp as shown below, and the noise was 1.336 DN.

After subtracting row-by-row reference pixels, the offsets of the first and last scans improved a lot but the noise increased to 1.544 DN (54 electrons).

So, although the scan averages fall on a better line, the fact that we are not averaging thousands of reference pixels before our subtraction makes the subtracted data noisy and contributes to the overall noise significantly.

Noise vs. Grounding Scheme

Noise vs. Grounding Scheme

In an effort to minimize the readout noise of the array, we attempted different grounding configurations external to the dewar. The internal cold fanout board configuration for CD016 was:

JP3 IN = AGND jumped to DGND

JP1 IN = AGND jumped to Clock Shields

JP6 IN = AGND jumped to Bias Shields

JP10-14 IN = AGND jumped to Preamp Shields

The external ebox configuration was:

JP1 IN = AGND DC jumped to Bias Shields

JP4 IN = AGND PRE jumped to Preamp Shields

JP9 IN = DGND LEVL jumped to Clock Shields

The ebox Power Supply (P/S) configuration to start the run was:

Jumper wires from AGND DC ro DGND LVL to AGND PRE

Data was taken in the dark, with the mirror selected in the lens wheel for lowest photon noise contribution. The following sets of variations were performed in the external circuit. In each variation, three sets of data were taken, each one 5 ramps of 8 scans (14 5 egos) in order to extract Fowler-4 noise data. The noise was lower in all cases when the first of the 5 ramps was ignored, so only the last 4 ramps are used in the Fowler average. The noise plotted is the average of all 3 data sets at each configuration.

1               Baseline configuration as above

2               Remove jumper wires between DC P/S modules in ebox P/S

3               Remove ebox jumbers JP4 and JP1 for signal and bias shields

4               Replace jumper wires between DC P/S modules in ebox P/S

5               Replace ebox jumbers JP4 and JP1 for signal and bias shields

The noise is worse with the shield jumpers removed, whether with or without P/S jumpers. The noise is much worse with shield and P/S jumpers out, and somewhat worse with just the P/S jumpers out. Noise is lowest (runs 1 and 5) with the baseline configuration as given above.

Linearity

To analyze the linearity of the detector, 20 repeat frames were taken with the following configuration: 25 1 egos (N2 S0 N3 S0 N3 S0 N1 D0), frame time = 0.5s, slit 3, center_arc_middle decker, filter BP14100, echelle angle = 37, cal source = 22 degrees, high-resolution mode.  The data can be found under the directory C:/data/CD016/cd016_2013oct/52, frames 1817 - 1836.  Frame 1823 had a dropped read and so was not used in this analysis, leaving a total of 19 repeat frames.

Each read was subtracted from the first read, which was taken as a baseline.  A 5x50 pixel array [499:503,441:490] near the center of the detector that was approximately uniformly illuminated was used for this analysis.  A 3 sigma iterative rejection was used when calculating the statistics of each read.

(Top Panel)  Flux versus integration time is plotted for each individual pixel in the 5x50 array.  A linear equation is fit to the first 6 data points for each individual pixel, while an exponential equation is fit to data points 6 through 10.  The last three data points are clearly saturated and not used for any fit.  It is obvious that there are two distinct fits for both types of equations due to the offsets between even and odd rows in the detector.  Two separate fits are listed on the right-hand side of the panel.

(Bottom Panel)  Flux versus integration time is plotted for the average of all of the pixels at each integration time.  The first frame is not included in this average, due to it typically being offset from the rest of the points.  This left a total of 18 frames for the average.  Again, a linear and exponential fit is shown.  The average fits lay somewhere in between the individual pixel fits, as would be expected.  The equations stated on the right-hand side can be used to show how the detector is approximately 1% non-linear at a flux value of ~5050 DN.  Looking at the first read for each frame, the counts value is ~ 10,850 DN in the same 5x50 pixel array, which would mean that the detector response becomes 1% non-linear at a count value of ~5800 DN.  

Thoughput Filter Curves

The throughput for the different filters was measured over several different EXES cool downs.  The specific cool down where each filter's data in the plot below were taken is listed here: 18a (CD16), BBP4700 (CD16), BBP5030 (CD14), BBP5500 (CD14), BBP8000 (CD15), BP4900 (CD15), BP6000 (CD15), BP6600 (CD15), BP14100 (CD16), N04582 (CD15), N04679 (CD16), N07053 (CD16), N07466 (CD16), N07904 (CD16), OCLI_Q (CD14), OCLI_R (CD14), Reading6.2 (CD16), Soifer (CD16).  For all of the measurements, a gain of 35.9e-/DN was assumed and the acceptance half angle was 0.028, as measured from CD16.

Dark Frames

The goal of this analysis was to examine the stability of dark frames and to compare using the slit mirror to the lens mirror.

Two different data sets were taken: One with the slit mirror in place and one with the lens mirror in place.  The data can be found in directory C:/data/CD016/cd016_2013oct/56.  The slit mirror frame numbers are 2165-2184 and the lens mirror frame numbers are 2145-2164.  The data was taken using a frame time of 0.5s and a 24 1 egos pattern (N6 S64 N5 D0) with 20 repeats.

The first read of an exposure was taken as a baseline and subtracted from all other reads.  A 3 sigma iterative rejection was used when calculating the statistics for each frame.  The results are shown below.

(Top Panel) All twenty frames are plotted with flux in DN versus integration time.  The blue triangles are the slit mirror data, while the black open circles are the lens mirror data.  The first frame has abnormally high flux compared to the rest of the images and so it is not included in the average polynomial fits shown in the bottom right.

(Bottom Panel) This plot shows the flux per time versus time.  These data points represent the average of all 19 frames (ignoring the first frame) for each integration time step.  The corresponding fits to the data are stated in the upper right-hand corner.  By the end of the frame time (~40s), the data points have flattened out at a value of ~3 DN/s.   The red circles were calculated from data that can be found in C:/data/CD016/cd016_2013oct/52, frames 1877-1896 (there were dropped reads in image 1889, so it was not used), frame time equal 0.5s with a 25 1 egos pattern (N2 S0 N3 S0 N3 S0 N1 D0) using the lens mirror.  This data is over plotted to show how well the fits work with data at integration times not present with a 24 1 egos pattern.  The error bars on all points represent the standard deviation from the mean.  The lens mirror and slit mirror data appear comparable with respect to each other.