It just gets better

As I wrote last weekend, Saturday night I had some of the best seeing I have ever had.  I had written a post a couple of weeks ago about the image quality we have been getting, based on a frame that Daniel took.  I thought it might be interesting to go back my focus frames to investigate the image quality on this exceptional night.  I took the raw frame that was closest to best focus and measured a half dozen or so stars in each OTA.  Here is the average on each OTA.

 

Wheras the previous measurements, which averaged about 0.6 arcseconds, showed little difference from detector to detector, these show substantially more.  Not surprising - as we get to better image quality, the focus or telescope-induced image quality differences become easier to see.  I have yet to  see whether I can detect focus differences from OTA to OTA, but I did notice that the bottom two OTAs (OTA 00 and OTA 61) had elongated images.  This suggests that we are seeing optical aberrations (either telescope or instrument) such as coma or astigmatism.   The telescope has some adjustments that we can explore, primarily with the active support system on the primary, that may improve these outlying fields.  However, we are seeing pretty good image quality throughout the central "science field", barely over 0.40 arcseconds on average.  

We are collaborating with the LSST wavefront group to determine the wavefront errors and how to minimize them.  This process involves calculating the sensitivity of the wavefront errors (in terms of Zernikes) to each adjustable parameter of the telescope or instrument. Then, we will look at out-of-focus images on each OTA and compare them with those predicted by the ideal optical system.  This will tell us what adjustments to make.  We'll iterate on this until we are satisfied. Of course, we'll have to wait for another night of exceptional seeing to really test the outcome, but the bottom line is we already are getting spectacular image quality, and we think we can do better.

Todd

First baby steps with the ADC

After all these nice images of the last few days I decided that we are having way too much fun up here, and this night we return to the technical evaluation of pODI.

One problem one encounters when observing closer to the horizon is so called atmospheric dispersion. While it is a tiny effect, the atmosphere acts on light coming from outer space like a prism, and disperses it into its colors. The effect is about a factor of 200 too small to be seen by the naked eye and goes unnoticed in everyday life. For a high-resolution imager like pODI it can be a significant effect, though. To counter for this atmospheric dispersion, ODI has two large prisms in its optical train that can be tuned to compensate for the distortion of the atmospheric dispersion (hence ADC: Atmospheric Dispersion Compensator). So far we operated the ADC in its neutral configuration, but tonight we observed some first test data, albeit with a poor seeing of about 1.4 arc seconds. At an airmass of about two we took some in the g' band (which is the most affected band) with the ADC in both neutral and in active state. We still need to fine-tune the strength of the dispersion (during better seeing conditions!), but the image improvement by the ADC is apparent in the example below.

For the tuning of the ADC we will use a "UG 5" filter with transmission peaks at 3500 A and 7200 A, but we have to wait until the filter adaptors are ready. At an high air mass, this filter will create double peaks of a single star (corresponding to the separation of the red and blue part of the star's spectrum), and the goal will be to tweak the ADC such that the two stars will merge into one single one. Although one shouldn't cross the beams.

Daniel

Image and contour plot of the same star observed in the g' band at an airmass of ~ 2. Left: ADC is neutral, right: ADC is engaged. Note the improvement of the roundness. 

Grasping StarGrasp

Engineering verification is split between daytime and nighttime activities, which is a good thing, considering the weather.  This week, we have two of the StarGrasp team, Peter Onaka and Greg Ching, here to help us with the tuning of the system and to do some technical training of mountain support personnel.  StarGrasp is the controller system that runs the detectors - voltages and signals - and receives and organizes the data coming back.  We had a very successful training session this morning in which Peter and Greg went through the entire system architecture and design, discussed how one diagnoses problems and what to do if the system is not working properly. The support model for StarGrasp is that we have a set of spare modules that can replace failed ones, and that the failed ones are then returned to Hawaii for repair.  The StarGrasp group has a second set of spares, and they send us their spare upon receipt of the failed one from us.  They then repair the failed one and it becomes the spare for next time.  For this to be effective, we will depend on the KPNO Electronics Maintenance staff to help diagnose which module has failed.  Daniel is still the local authority on StarGrasp, and he will coordinate this effort.  We will be putting this system into place over the next few months as we move into scientific commissioning and then shared risk observing.

Todd

Shooting the moon

It's raining again - not much chance of getting anything useful on the sky tonight.  I wanted to do a bit more with the frame of the moon that I took Saturday evening (In memory of Neil Armstrong).  When the first mosaic CCD cameras were built in the mid-90s, it became common to show their field of view by imaging the moon. To astronomers who had traded in their large photographic plates for tiny 800 X 800 CCDs, this was a welcome advance.  However, the surface brightness of the moon is higher than the daytime sky, and so one used to have to play tricks like partially closing the mirror cover, or occulting part of the telescope beam with the dome.  This was in addition to putting in a filter that gave you the part of the spectrum where your CCDs were least sensitive so you could take a 1 second exposure, which was about the limit of reliability and uniformity for the shutters we used.

In an earlier post, I noted that because of the design of the Bonn shutter we have on pODI, we can take exposures that have a very short effective exposure time but are still relatively uniformly exposed.  This is because the shutter consists of two sliding blades that are controlled independently by stepper motors. A short exposure can be obtained by effectively moving a narrow slit across the focal plane, with the two edges of the blades close together.  We first used this to take twilight sky flats even before the sun had set.

It was still an hour before sunset when I decided to take an exposure of the moon.  I put in the z' filter (where the sky is pretty dark and the CCDs are the least sensitive).  I set the telescope focus to the value I had used the previous night.  I tried a 5 millisecond exposure and I got about 3000 counts per pixel.  Good enough.  Finding the moon was not trivial.  John Thorstensen was my savior.  His Skycalc program gave me RA and Dec for the moon right now, right here.  It was very close.  Center it up, and - snap - 5 milliseconds - perfect exposure. About 30,000 counts per pixel in the moon.  The image that I posted was almost raw. It had the overscan level subtracted, and the display script tries to adjust the stretch to match up the individual OTAs, but it doesn't do so well for something so extended with such large dynamic range.

So tonight I did a very simple reduction.  I subtracted a bias and divided by a dome flat.  Here is the result.  This is just the central 3 X 3 OTA part of the focal plane.  You can see the three bad cells. You can see that the image display doesn't make the detector to detector gaps quite big enough.  But overall, pretty amazing.  The moon in the daytime with a 3.5m telescope.

Todd

You gotta love this telescope

Last night was my first night at WIYN where I knew it was going to be good.  A few puffy clouds floating around in the afternoon, but the weather forecast was "clear".  This was the first night predicted to be clear since June.  The wind was blowing just hard enough to feel its motion through the open dome.  I always enjoy standing on the WIYN dome floor right after sunset.  You feel like you are outside and have a clear view in every direction.

I thought the seeing might be good, so I put in the i filter before I started a focus sequence.  I've been looking at a lot of out of focus images (see my vignette on vignetting) and as I stepped by 20's through the focus values, I noticed that even as the image got smaller and smaller, it kept its donut-like shape. I could see the hole in the middle, and some consistent bright spots and irregularities in the brightness. It got very small before the images turned into blobs with a central peak.  I cut my steps down to 10, and I passed through focus and came out the other side.  Then I went back and measured the FWHM values of the images.  0.42 arcseconds.  That was the best.  That was among the best seeing I've ever seen.  It stayed that good - or almost that good - all night.  I worked mostly in g (where the detectors are very sensitive), and I took frame after frame with images 0.5 to 0.6 arcseconds FWHM; a few below 0.5.  These were guided 5 minute exposures.  Guided with the telescope - not with OT shifting on the detectors. I took a bunch of frames of a field in Stripe 82. In each one, I could see little galaxies, but they were not just smudges; they had spiral arms and nuclei and, when they were interacting, I could see knots in the streamers that were flying off them.  It was almost like looking at HST images.  I've inserted images of a couple of cells, but these don't really do justice to the data.  Remember that each of these cells is about 1 arcminute across and pODI has 13 X 64 = 832 of them.

I've got more stories from this night - including one that taught me to ALWAYS check my shoes for scorpions in the morning before I put them on - but that's for another entry.

Todd

In memory of Neil Armstrong (1930-2012)

vignetting

One of the things I have been curious about in ODI is the vignetting.  While the optical design was based on a one degree field of view, the square format of ODI actually puts the corner pixels as far as 44 arcminutes from the field center.  I have played a bit with trying to compare the flat field illumination with the photon statistics in order to separate the gain from the vignetting, but I wasn't happy with that approach.  Last night, I got a good sequence of frames going through the best focus, and so I took the most out of focus frame and looked at the donuts - the pupil images.  I took the best image from each OTA and made up a little picture that shows them placed on top of the position of that OTA.  As you can see the illumination is remarkably uniform until you get to the far corners of the focal plane.  The two OTAs that are in one square from the corner have centers about 28 arcminutes from the field center, and you can see that there is really minimal vignetting at that point.  In the one OTA in the real corner, centered about 39 arcminutes from the field center, the vignetting increases rapidly and is about 60-70% at the outermost pixel. 

Todd

On ODI's throughput

ODI at WIYN's sensitivity has been predicted based on modeling all optically relevant components. The throughput model includes: The atmospheric extinction and telluric absorption lines, the three telescope mirrors, absorption losses in the ODI optics (two lenses and two ADC prisms), anti-reflection coating performance, and the quantum efficiency of the OTA detectors. The predicted throughput has been verified on sky in the g', r', i', and z' bands within reasonable error margins, and the differences we found are now included in the throughput model as a fudge factor. The resulting as-build throughput of ODI is shown the figure below. We will keep tweaking this model as we get more on-sky data; I am in particular looking forward using a u'-band filter at some point in the future and verify the blue performance

The blue cut-off in ODI's sensitivity  is governed by special glass (O'Hara PBL6Y) in the ADC prisms. The fall-off in the red is driven by the vanishing quantum efficiency of the ODI detectors. The peak throughput of order of 55\% is the sum of all losses in the system; the most significant limit on the peak throughput is set by the losses in the three WIYN mirrors, though. 

 

Daniel

Total system throughput of the WIYN telescope and ODI, including atmospheric absorption.

Watching the grass grow

Kitt Peak is very green right now.  The rain and overcast continues, and so I decided to delay the start of my two days of sitting in the control room until Friday.  The weather report for Friday night is partly cloudy, and for Saturday night it is clear, so hopefully.....

In the meantime, Daniel has been up working on adjusting OTA voltages, and Andrey and Erik Timmermann, from the NOAO Science Data Management group, have been installing and testing out the observing GUI that Erik has developed.  Downtown, I have been working on getting the filter inserts made  that will hold NOAO Mosaic filters (5.75 inches square) - no, I'm not cutting metal, but negotiating the schedule with the machine shop. Also, thanks to Di and Charles Harmer, I located a Schott UG 5 filter that we can use for adjusting the positioning of the ADC.

Next Monday we have a visit from the Stargrasp group - Peter Onaka and Greg Ching - who will help us with final tuning and hold a series of training sessions for the mountain engineering group.

I hope to have something more exciting to report in a couple of days.

Todd

First reduced pODI image

Today - a guest post from Frank Valdes

One of the early first-light priorities was to take a dataset for producing the initial world coordinate system (WCS) to be added to the raw data.  A set of 18 forty-five second unguided exposures, in pairs within a 9 point dither pattern, was taken Aug 7th of IC 4756 in the SDSS g filter.  Note the seeing was only around 1" while pODI will, hopefully, often produce even better image quality.

A WCS using an optical model of the radial distortion across the field and a general polynomial fit to the remaining residuals was determined.  A check of this coordinate system is to use it, after tweaking it for each exposure, for remapping and stacking the dither followed by examining the stars to see if they are well registered across the field of view.

The images below show the large and small scale results of this first stack.  The processing was done using the IRAF ODI and MSCTOOLS packages.  The pipeline is current being worked on to automate this same processing.  Note there are a many things that still need to be addressed -- use of bad pixel masks, cross-talk subtraction, bleed trail masking, photometric registration, etc.  Because these instrumental features are not yet handled, the quick and simple way to produce a clean looking image is by simple median stacking, in this case without any photometric scaling of the exposures.

Frank

Figure 1: The full field of the central 9 OTAs trimmed to the common area of the dither.  The size is 13.6K by 12.3K at 0.11"/pixel.

Figure 2: A close up of an approximately 4' region indicated in the full field.



Still here, in the clouds

The new week did not bring the weather improvement we were hoping for. The afternoon gave us rain and rainbows, and the clouds keep loitering around throughout the night. The small holes in the clouds have not permitted us to open the dome, yet.

What is to be done, then? I went though some older images and worked on the software module that will one day autonomously find guide stars from a short snapshot image. The same tool also allows to find stars in a longer exposure, and based on the star's size to then automatically to judge the image quality.  I came across  one shorter (60 seconds) r' band image that seemed to have a delivered image quality more around 0.55'' throughout the image. 

As the clouds might finally break up, it would be nice to demonstrate what ODI can do with long exposures. Let's give it another hour.

Daniel

View from the WIYN parking lot.

View from under the clouds

The weather has taken a turn for the worse, and it looks like we won't get any more data on the sky this week.  We are ready to test a new, improved version of Daniel's guiding software, and Andrey has been working on collecting all the instrument and telescope data with each exposure to put into the headers. This is not real exciting stuff, and I, for one, am eager to see the end of the monsoon.

One new item for prospective pODI proposers is the ODI web site (still under construction, but you can see the layout and some of the content at http://www.wiyn.org/ODI/wiynodi.html) and, in particular, the page we have put together to tell you everything you need to know to write a proposal (http://www.wiyn.org/ODI/Observe/wiynoditools.html).  The call for proposals will be out around September 1, and we will update this page as we get more information.  One graphic I put together for this page is the layout of pODI with the formats, sizes, and gaps indicated.

Todd

First look at image quality

On Monday night, Daniel took a 200 second guided exposure in the r' band that he noted had stellar images with FWHM about 0.6 arcseconds.  This is the best we have seen so far, and so I took some time to examine the images in a bit more detail.  I measured the FWHM of a dozen or so stars in each OTA, and I found that the image quality was good and quite uniform over the entire field.  The picture shows the measured average (in arcsecs) in each OTA.  I did not notice any significant departure from roundness in any of the OTAs. 

The worst FWHM, 0.615 arcseconds corresponds to an additional contribution of 0.22 arcseconds (in quadrature) over the best FWHM.  Also, the OTA with the worst FWHM, OTA 55, failed its metrology test - that is, it is not within the specs for final height of the detector plus package.  Thus, part of the poorer image quality might be a focus effect, though it is not much worse than the OTAs around it.  Finally, I note that we have not yet begun to adjust the telescope optics to optimize image quality over the wide field; that is an activity planned for early September.

All in all, it looks like pODI (and later, ODI) will have the ability to deliver quite good images over its entire field of view.

Todd

Learning more about guiding & photometric zeropoints

The hailstorm (with about an inch diameter sized hailstones) yesterday afternoon curbed our hope for a clear night, so we used the afternoon for a new look at the noise characteristics, and we can confirm a similar read noise level as we did in the lab in Tucson. So far so good.

Todd took a first shot at the photometric zeropoints in the four ODI filters, and they are within 0.1 mag of the prediction we made a few years ago. As time permits, we will update the ODI exposure time calculator.

Surprisingly, the sky cleared up before 10pm, giving us the opportunity to spend more time to tune telescope guiding. A few hours of debugging and optimizing later, we ended up with a workable configuration of guide parameters that allows more stable guiding. Getting the guider fully tuned to a critically damped state will take some time, though.

In one instance we recorded a single bright star at a rate of >30Hz. While the telescope guider filters that signal to a slower rate, it also calculates the power spectrum of the image motion (the panels with the green histogram in the picture; ignore the units on the Frequency axis - they are meaningless at this time). Most of the time that power spectrum is boring, but when some wind blew into the dome, a clear spike showed up in the power spectrum, indicating that the telscope mount might have gone into its 8 Hz resonance. With still some bandwidth left in the system, this measurements is encouraging for the prospects of the coherently corrected OT modes with pODI, where this resonant image motion could be compensated in the detectors.

Daniel

Guiding - a major milestone

One of the things that makes ODI unique is its detectors, orthogonal transfer arrays (OTAs), which are more complicated than the typical CCDs found in most astronomical imagers.  The OTAs in ODI are each divided into an 8 X 8 array of "cells", each of which contain 480 X 494 pixels.  The reason for the separation into cells is so that the guide function can be accomplished by reading out a part of a single cell frequently to measure the position of a guide star without affecting the other 63 cells on the detector.  Ultimately, this will allow us to assign 4 cells on each OTA to the guide function and independently guide each quadrant to remove the local motion from atmospheric turbulence and telescope shake.  At this point (the first week on the telescope), the challenge has been to read out a cell at video rate, extract the guide star information, and send it to the telescope every second or so. 

We accomplished this last night for the first time.  The first picture shown is the OTA Listener, which displays the entire pODI focal plane on the left and one of the OTAs on the right.  For this exposure, we had pointed at M57 the ring nebula in order to take some long exposures and measure the crosstalk between cells.  The OTA shown is OTA 33, the center one in the 3 X 3 science field.  If you look closely at the lower left corner cell, you can see that it is truly black.  This is because during the exposure (about a minute long), we read out a piece of that cell and used it to guide the telescope.  This is the first step towards OT operation, and it is prerequisite to taking exposures longer than a minute or two.  The other thing you will notice are the amplifier glows. In order to read out an OTA, we have to turn the amplifiers on (throughout that entire OTA), and they glow in this batch of detectors.  The amps are off until readout on the OTAs not used for guiding - one of these, OTA 44, is shown in the lower picture.  Our planned mode of operation for pODI is to use one of the outer OTAs for the guide function, leaving the 3 X 3 "science field" free of amplifier glow.  For the full-up ODI, we'll manufacture new detectors that do not have the amplifier glow.  The images shown are raw - no flat fielding applied yet, so small variations in gain among cells are apparent.

Todd and Daniel

A 60 second long exposure of M57, with guiding, shown in the OTA Listener

OTA 44 from the above exposure - note that the amplifiers are not glowing

A good night to write documentation

We made some progress on the video mode today, but it is extremely unlikely that we will ever open the dome tonight. But we need to write documentation and procedures anyway...

Daniel & Todd

A haboob in the valleys around Kitt Peak.

Hell is breaking loose over the WIYN 0.9m telescope.

So this is why no one else is up here observing

This night was particularly frustrating.  Clear all day, clear at sunset.  An hour later thin clouds forming above us and getting thicker.  We tried to shoot between the clouds for a couple of hours until the overcast was solid.

One OTA with a very bright star, showing the crosstalk ghosts in cells in the same row.

Even so, we accomplished a few things and learned a bit about ODI.  We discovered that the shutter is so good that we can do twilight flats (8 millisecond exposures) while the sun is still up.  This means that you can get good flats for all filters without having to race against the darkening sky.  We worked on video readout for guiding, and, while we didn't get it going, we found and fixed some of the bugs.  This is our highest priority for the rest of the week.  We measured the crosstalk ghosts that result when you observe a bright star, and we found that they are limited to the row of cells in which the bright star appears (see picture).  The ghosts are about 1/10,000 of the original image, so they only show up where the original is saturated.  Finally, we obtained a complete dither sequence so that the pipeline developers can have some realistic data to begin playing with.



Todd and Daniel

More stars, more rain, and many lessons learned

The control room was filled with people tonight to witness the official first light tonight (as opposed to the dress rehearsal last Friday). So far everything has gone well other than the weather. Around 9:30 pm Krissy opened the dome (as she predicted after dinner), allowing us to take even more images of star clusters (mostly chosen to make focussing easy). Unfortunately, only half an hour later we had to close the dome again due to imminent thunderstorms and rain. 

Nevertheless, we made progress on several fronts today: 

1. The ring we saw on Friday's image was most likely caused by condensation on the dewar window (Thanks for the suggestion, George J). This was resolved by increasing the dry air flow. After some time spent wondering, this has turned out to be no issue at all. 
2. The baffle to block a stray light path (from the tertiary mirror via the primary mirror into the instrument) was installed today. We tested the effect of the baffle, and on first look it seems to remove the predicted scattered light component.  
3. We saw a << 1% pupil ghost in r' band flat fields, but apparently in no other filter. We expect that Frank Valdes' & Rob Swater's pipeline can readily handle the ghost. 
4. Examining an 20 second exposure of a 2nd magnitude star, we have not seen any indication of strong crosstalk yet. However, before making a bold claim here, we will obtain more suitable observations to quantify any crosstalk. 
5. A first experiment indicates that we have no obvious light leaks in the instrument – detailed testing will follow soon. 

Today marks the transition from instrument installation into the engineering commissioning phase. This week we will concentrate on fundamental detector operations, harvesting header information from the telescope, and enabling telescope guiding with ODI guide star videos. During this phase we might not produce as many spectacular images, but we will continue to keep you up to date on our progress.  

Daniel & Todd

More stars, nicely packaged into an open cluster (do you recognize it, Bob?). 30 seconds in r', bias and flat field corrected. Seeing was about 1"

pODI first light - a sneak preview

While we plan to have official pODI first light tonight, weather permitting, we were a bit fearful of discovering some difficult problem in front of an audience.  So - on Friday, we opened the dome for an hour to make sure that we could focus, point, etc.  Our first light preview went remarkably well, though we have done no quantitative analysis of the details yet.  We only obtained a couple of images and inspected them.  Here, for your viewing pleasure, is an image of M13, 30 seconds through the r' filter, no guiding.

Todd and Daniel

This is the central OTA (about 8 arcmin square) of the 3 X 3 central array.  All of the images here have been overscan-subtracted only.

This is the central 3 X 3 array of OTAs (about 24 arc min square).  There are 3 non-working cells.  The interesting artifact is the ring, which is centered just above the field center.  This appears in dome flats as well, and we are trying to figure out what this is and how we can remove it. 

This is the full field of pODI.

  

Loading the filters into ODI, detectors are alive

Over the course of several hours we loaded the four ODI filter into the instrument today. Each filter is about 40 cm x 40 cm large, and each costs the equivalent of a my dream sports car (I am thinking Audi TT here, if you would be so kind...). SA lot of time was spent practicing the procedure with aluminium blanks instead of glass. By the time we started the installation, we were very confident in the procedure.  The pictures of the installation are posted below. 

As of now we have four filters for pODI: SDSS g', r', i', and z'. This is the entire filter set for now, but yesterday Gary Muller released the drawings for filter adaptors that will allow Mosaic (about 6 inch size) filters to be used with pODI. We expect to have them by November or so. 

Andrey spend most of the day adjusting the data acquisition software to the new mountain environment.  At the end of the day, he was able to take the first bias and dark exposures to proof that the detectors are alive and well – a very good conclusion for the second week of installation. 

Daniel

Gary Poczulp installs a filter into its module.

Lifting the filter module to the instrument. 

Charles verifies that the filter is properly seated.  

The z' filter is moving into the instrument. 

Almost there

ODI is now cabled up and cold.  We spent the day watching the vacuum and temperatures carefully as the instrument cooled down.  This is the first time we will leave it cold while we are not around, so Daniel is briefing Krissy, the Observing Assistant who will be on duty tonight, how to do an emergency shutdown if necessary.  The rest of us will depart and return tomorrow for a day of biases and dome flats.  If all goes well, we will be opening the dome next week. 

Daniel and I worked out a schedule today for the engineering verification activities.  Since we are still in the monsoon, we will take advantage of the clear nights when they occur, and otherwise work on the stuff that can be done in the daytime.  All this is subject to change as we proceed, but the priorities are:

Aug 6-10: focus sequence; manual guide star acquisition; focus w/ guide star; telescope guiding; telescope telemetry

Aug 13-17: focus sensors; world coordinate system; on-axis wavefront adjustment; filter offsets

Aug 20-24: Image characterization; science GUI; data for PPA

Aug 27-31: ADC; begin photometric characterization; OT preliminary testing

Sept 3-7: photometric characterization

Sept 10-17: wide-field wavefront adjustment; image quality analysis; update exposure time calculator

It will be a busy period, but this should put us in good shape to begin scientific commissioning in early October.

Todd