Twenty Rows

As part of our next steps in detector characterization, we need to address the '20-row' issue we observed in our IRTF data.  A basic statement of the problem (please chime in if this isn't accurate!):

-  A quasi-fixed pattern exists in nod-pair results, with bright/dark horizontal banding on ~20-row centers.
-  This pattern does not subtract out, and swamps all but the brightest signals for astronomical sources.
-  Because it does not subtract out it must vary over time, or possibly with illumination (as the odd/even effect seems to).
-  This effect has (so far) only been noticed with the science array (serial 115).

First things first....here's a picture demonstrating the effect:

Difference of two scans looking at the dark flag

of the cal wheel @IRTF, 1s integration time.

In the cut it appears that the '20-row' effect in fact has a spatial period of about 16 rows, and an amplitude of about 30DN p-p (~6DN rms).  With a 1s scan time, the 16 row period is about 16ms, which has vary much the fragrance of 60Hz about it.  1DN is roughly 126µV, or 30 electrons, so this amplitude is about 900 electrons (180 rms).  In a single scan, the odd/even effect is similar in amplitude (at least in non-illuminated regions).

Since the period is so close to 60Hz, it worth wondering whether the '20-row' effect is just a pickup artifact...if so it should be asynchronous and should roll through the frame over time.  The file collected just after the image above shows the same amplitude and period, but is shifted in phase.  Here is an image showing the average these two difference frames:

Difference of the image above and the subsequent frame. dark flag

of the cal wheel @IRTF, 1s integration time

Here the phase difference is about pi, so the pickup about cancels and we can even see the faint cross-dispersed spectrum in the average frame.  In this averaged result the odd-even remains, but would probably disappear if the frames were differenced.

Not airtight, but I'm suspecting our '20-row' effect is just 60Hz pickup....

Slit scan in the lab

On Tuesday, Mark ran a slit scan of the BB with a 25 micron aperture across our slit 4, which is 229 microns (design). If everything is perfect, the spot at the slit should be diffraction limited or about 50 microns (f * lambda). Along the slit, this should be preserved. We stepped the slit by 12 motor steps per image, which should be around 60 microns. The slit should be just under 4 positions wide.

The data were an alternating series of viewing the BB and viewing it blocked off by a mirror (behind the aperture). I subtracted the pairs and extracted a single column of the difference (to avoid rotation). Combining the various slit positions gives me a map of the spot profile across our slit.

The top image shows the spot with the maximum of the color scale set to half the maximum. For the axes, I took the size in microns and converted to arcseconds to allow comparison to the IRTF spots in the earlier post. The lab scan is much better than the IRTF. The bottom image, I put the top of the color scale to 5% of the maximum to show the noise. There is a hint of a ring, but the spacing is wrong for the Airy ring. There is also some light scattered along the slit. And it seems to me there may be a change in detector response after the bright spot.

I've put cuts across and along the slit in the next figure. Along the slit (vertical in the images above), the spots are more Gaussian, although they have some wings at the bottom. The FWHM of a Gaussian fit is 79 microns wide. Across the slit, the spot is very flat topped.

Effects of foreoptics clocking errors

Is spot size sensitive to foreoptics clocking errors?




With a clocking error of 0.05° (equivalent to 44μm along the circumference of the 4" diameter ellipsoid) the spot is still diffraction limited.
A clocking error of 0.25° (equivalent to 220μm along the circumference of the 4" diameter ellipsoid) generates a spot between 300 and 400μm in diameter.
A clocking error of 0.70° (equivalent to 616μm along the circumference of the 4" diameter ellipsoid) generates a spot that is about 900μm in diameter.
The spot size is about 10 times less sensitive to clocking errors of the hyperboloid. That is, a clocking error of 0.50° of the hyperboloid creates about the same spot size as a clocking error of 0.05° of the ellipsoid ... and so on for the larger angles. However, the hyperboloid is much smaller, such that the 0.50° clocking error is equivalent to 111μm along the circumference of the hyperboloid. Similarly, a 2.50° clocking error is equivalent to 554μm along the circumference of the hyperboloid, and a 7.00° clocking error is equivalent to 1,552μm along the circumference.


What if both the ellipse and the hyperbola are mis-clocked?





The second set of spot diagrams shows what can happen when the ellipse has a clocking error of +0.40° and the hyperbola has one of a range of clocking errors. In this case, the hyperbola can actually mitigate the effects of the ellipse clocking error somewhat. Or it can make things even worse.

What happens at the detector as the spot size at the foreoptics focus changes?

In the high-res mode, spot sizes at the detector are comparable to those at the foreoptics focus - about 10% smaller.

focus scans from IRTF

Here are some images from when we scanned a star across the slit at different telescope focus positions.

The slit runs vertically, as it does in the lab. The scan direction is from left to right. At each point in the scan, we did two integrations each with non-destructive, spin, desructive (ND,S,D) at 1 second framerates. When we finished the second destructive sample, the telescope was commanded to move by 0.25" perpendicular to the slit. The scan started on blank sky and went across the slit.

In the data reduction, I took all the save frames from a scan, computed ND-D (gives positive numbers for more light) for each pair, and then looked at just the second integration from each position. I subtracted the first scan position from each position to get rid of background emission. I rotated the spectrum in software so it ran along rows and compressed all the spectral information from each spectrum so that for each scan position, I had an intensity map along the slit. I subtracted the median of these slit-intensity maps to account for variations in sky brightness during the scan. By putting the scan positions together, I get a full map.

There are two presentations of the same four map postage stamps that come from these scans. The telescope focus is at the top of each stamp. In the first representation, I set the saturation level at the FWHM of each postage stamp. I also have calculated the size of each pixel along the slit in arcseconds and use those units for the map. At far focus positions, the spots are clearly awful. You can see that at 0.96 it is narrower along the slit at each position than at 1.15, but it is spread out more in the scan direction.

The second presentation uses the peak pixel value of all the postage stamps, from focus = 1.15, to define the color scale. Also, the axis are in pixels.

How precisely can we place the IRTF focus?

Yet another factor to keep in mind in assessing EXES performance is the precision of placement of the IRTF focus (by changing telescope primary-secondary spacing) and the effect of this placement on spot size at the EXES detector.

The attached image shows the progression of detector spot sizes (on-axis, in high-res mode, at 7μm wavelength) as you change the IRTF telescope spacing over a 2mm range.

At the 'design' spacing, spot size is close to diffraction limited.

Plus and minus 1mm from design spacing of the IRTF mirror, spot size is ten times the diffraction limit.

Scanning Across the Slits

Having found what looks like decent focus at the slit based on the knife-edge tests in pupil mode, and choosing for the moment to attribute the squirrelly camera-mode image quality to our fifty-cent alignment mirror, we decided to move to lo-med mode and try scanning the slitwheel features across the IR spot to check how well we could cut off the image with narrow slits.  This is something we had trouble with on point targets at the IRTF.

Here's our first try, scanning the fattest (flux) slit across the IR spot at our knife-edge-test-indicated best focus:

Photometry-box plots of image strength as a function

of slit position, taken near knife-edge best focus.

The three curves are generated in the FORCAST gui for boxes placed at various points on the image.   This is a strip-chart plot of the signal as the slit is slowly scanned across the spot, the timing is such that each data point in the plot corresponds to abut two steps of the slitwheel motor.  I think two steps is worth roughly 10 µm of slit movement, and by eyeball I count about eight data points for a 20%/80% 'risetime' in the plot.  If this is right, we cut off the spot in about 80µm of slit travel.  The red line is from a small box placed on one of the narrow mode lines created by the neutral-density filter....we can ignore the green and blue curves. 

For comparison, here is the equivalent slit-scan for the nominal best focus position...it's pretty clear this is not the right place.

Photometry-box plots of image strength as a function

of slit position, taken the nominal focus position.

Finally, here is an equivalent plot using the twin 'V-Slit' feature, showing that the spot is small enough to be almost completely blocked by the narrow separator between the V-Slits:

Photometry-box plots of image strength as a function

of slit position, taken near knife-edge best focus.

More Knife Edge

After trying a few more focus positions, we've been using this point, which puts the blackbody aperture about 40mm closer to the ellipsoid from the 'design' best focus position.  Here the image snaps on fairly quickly, and it's hard to say from which side (click for the magic):

After seeing this, Dana used his Zemax model to evaluate the focus for different wavelengths and positions.  Here is what he came up with (NFP = Nominal Focus Position, units are µm):

	Telescope separation	IRTF focus position delta	Entrance focus position delta	Slit focus position delta	7micron Beam size at slit	7micron Airy disk at slit	7micron Beam size at detector	7micron Airy disk at detector	12micron Beam size at slit	12micron Airy disk at slit
	7021.096	-20000	-8535	-3936	368	174	430	174	300	298
	7021.048	-10000	-6545	-3025	292	174	336	174	222	298
NFP	7021.012	-1764	-5116	-2475	236	174	266	174	165	298
	7020.960	10000	-3000	-1700	150	174	164	174	82	298
	7020.916	20000	-1290	-872	86	174	91	174	16	298
	7020.872	30000	534	-122	28	174	74	174	60	298
	7020.820	40000	2856	732	65	174	142	174	142	298
	7020.784	50000	3852	1340	122	174	200	174	200	298
	7020.736	60000	5795	2131	197	174	284	174	277	298

We're running at 5µm, and Dana designed the system for 12µm light.  His 7µm data point shows best focus for 7µm is 32mm away from the design focus, so our finding of a ~40µm difference at 5µm isn't absurd.

Having found this out, we decided to run a focus check in camera mode (knowing the dubious character of the alignment mirror) to see whether we might be fooling ourselves.  Here is what the spot looks like as we run the blackbody source over a90mm range along the Z-axis:

In this movie the BB is moved along Z toward the entrance window at 10mm/frame.  The two spots may be coalescing (or simply passing one another) as we get closer to the entrance window, but it's difficult to be quantitative.  This data is stored on the mac mini as 'Camera Mode BB Focus Run' if anyone wants to look at it more closely.