Seeing Measurements at UKIRT

Seeing Measurements at UKIRT

Need for objective seeing measurements

The quality of the delivered images is the true measure of how much the improvements to UKIRT are benefitting the telescope and its usefulness. Unfortunately the need to guide at sampling rates of tens of Hz on the faintest possible guide stars means that the fast guider CCD operates in a quadrant sensing mode and can’t tell us about image quality, only about image movement. Our trusty but obsolescent imager IRCAM, in its standard mode, has a pixel FOV of 0.”280, far too large to sample properly any but the images we would rather not know about, while its add-on magnifiers (X2, giving 0.”140/pixel and X5, giving 0.”056/pixel) degrade sensitivity and are therefore used quite infrequently.

Consequently, until recently quantitative image quality information at UKIRT has really been obtained only on engineering nights by occasional special efforts, and we have had only a hazy idea of the image quality distribution over many nights. This is clearly unsatisfactory, especially as we enter the phase of the telescope development programme where we are trying to learn to exploit the Dome Ventilation System and the soon-to-be-completed mirror cooling system. Anecdotal information simply will not let us learn how these systems are best used. 

A measurement protocol

As a result we have devised a standardised protocol for measuring the seeing FWHM on K-band (2.2 µm) images on a regular basis, twice a night on most or all nights on which IRCAM is used. This has been in regular operation since 13 February 1998 and at the time of writing has produced 143 measurements.

The observers are asked to take appropriate images, and to process them in a defined manner to determine the FWHM, on all nights on which IRCAM (imaging) data is taken, using whatever magnifier the observer happens to be using. The observation involves securing a set of three to five images of stars of suitable K-band brightness, just after focussing the telescope, roughly midway through each of the first and second halves of the night. 

Correction for undersampling

A key to the campaign has been the development of a way to use the mostly-undersampled images secured with IRCAM with no magnifier (the “X1” mode) as it is used 2/3 or 3/4 of the time. The need for this became obvious when we realised that observers – understandably – tend not to use the x2 magnifier in poor seeing, so that the better-sampled dataset taken with the X2 is significantly biassed.

To determine the undersampling corrections we have used engineering time on a number of nights to secure consecutive sets of images using no magnifier (X1), the X2 magnifier and on just one glorious night so far (see below) the X5 magnifier, in each case just after focus checks. The comparison of image FWHMs derived using 0.”28/pixel and 0.”14/pixel showed dramatic effects of undersampling, which becomes perceptible at about 3.7 pixels/FWHM and reaches a factor >2 at 2 pixels/FWHM for both pixel scales. The X1/X2 measurement sets, taken on different nights and in differing seeing, nevertheless fall on a well-defined curve, indicating that a single correction function can be used to convert FWHM results from variously sampled images into “true” results. The construction of the function assumed that undersampling of large images at 0.”28/pixel has similar effects to a similar degree of undersampling of smaller images at 0.”14/pixel. This assumption appears to have been justified: see below. 

The data set: seeing distibution at UKIRT

As X1/X2 data pairs have become available the correction function has been progressively refined. Using it we convert FWHM estimates from undersampled data (taken in both X1 and X2 modes) to fully-sampled values. The data histograms below show first the raw FWHM estimates made using IRCAM in X1 (0.”280/pixel) and X2 (0.”140/pixel) modes, and then the distribution after correction for undersampling except for three measurements in x5 mode which are used uncorrected. (Note that the labels refer to the upper (higher-value, right-hand) edge of the bins of the histograms, not the centres of the bins.) 

The raw K-band image quality data set (uncorrected) up to 29 September 1998, showing the marked difference in the distributions of the measured FWHM from the x1 and x2 magnifiers 
The histogram of the “corrected” K-band FWHM data set. Note the apparent excess of measures in the smallest bin, at the diffraction limit. 
The cumulative distribution. The curve indicates that one can expect to get K-band images with FWHM <0.”9 more than 90% of the time, <0.”45 arcsec around 50% of the time and <0.”25 around 20% of the time.

We currently have 139 image quality measurements taken between 13 February 1998 and 29 September 1998. For the corrected data set the median FWHM is 0.”433. For comparison with data from other sites, more commonly measured in the optical, we note that this is nominally equivalent to 0.”57 at the optical V band (0.55 µm, scaling as lambda -0.2).

As the data set increases we plan to investigate dependences of seeing on such variables as wind direction (likely to be linked with boundary layer effects), dome and mirror air temperatures, wind velocity and Dome Ventilation System settings. The programme of seeing monitoring will be continued with the UKIRT Fast Track Imager (UFTI) after its delivery. Its 0.”09 pixels will remove most (but evidently not quite all! – see below) of the need for undersampling corrections. 

Overcorrection, or do the images REALLY get that good?

The distribution of the corrected FWHM values extends right down to the diffraction limit (0.”108 at K) where there is actually an apparent excess of images, suggesting that images of this sort of quality really do occur on nearly 10% of nights and “pile up” in the histogram. (There are 13 images in our sample of 139 which are nominally less than 0.”15 FWHM!)

Since the majority of these (apparently) very good images come from the X1 sample we have been anxious to validate the correction function. The bump near the diffraction limit, of images with corrected FWHM less than 0.”15, derives from much larger measured values of <0.”51 with X1 and <0.”28 with X2, so direct confirmation was clearly essential before we would be justified in believing that we really do sometimes see images this close to the diffraction limit at K.

We were therefore extremely pleased that recent data taken with the X5 magnifier (PFOV = 0.”056) as well as X2 and X1 has validated the correction function to better than 10%, not far from the small-FWHM end where we were most concerned. 

Image of a star through the 5X magnifier (0.”056 /pixel) obtained on 8 September 1998.  Note the scale of the image – the box is only 2 arcseconds square! The graphs below give cross sections of the image in RA and Dec directions (formal FWHM = 0.”171; Strehl ratio ~0.25).

The image above is the best of three consecutive 20-s K-band exposures which produced images with near-diffraction-limited cores (the FWHM of the others were 0.”183 and 0.”180) secured by Antonio Chrysostomou and Thor Wold on the night of 8 September 1998 UT. We believe they are the best images ever obtained from the ground without the use of higher-order adaptive optics (at any rate, they are the best we have heard of!). We note too, with surprise, that Thor, the TSS, did not consider the night to be one of exceptionally good seeing! The correction function then in use, applied to the X2 images taken just before these, gave FWHM = 0.”17 and the corresponding X1 images yielded 0.”16: both slightly – but pretty insignificantly – optimistic relative to the 0.”178 of the (slightly undersampled!) X5 data. We have adjusted the correction function to match the X5 measures and the data set (displayed in the histogram above) has been re-analysed. 

The end of an era?

The 30th of September 1998 was the first engineering night of the new UKIRT Fast Track Imager (UFTI). We will therefore probably not be collecting much more seeing data with IRCAM, and future measurements will benefit from the smaller pixel scale offered by UFTI.

The table below shows a month-by-month analysis of the corrected image quality data collected using IRCAM since the campaign began in February 1998. 

The number of measurements in a given month is mainly determined by the availability of IRCAM on the telescope.

IRCAM has certainly left us with excited expectations of the results from UFTI. September was a spectacular month, with 20% of the images measured being nominally diffraction limited. Furthermore, the results suggest that this has been achieved at some time in nearly every other month! It now remains to be seen if UFTI confirms this expectation. Watch this space…