Initial evaluation of
Fiber guiding/focussing
vs.
Pellicle guiding/focussing

Niall Gaffney and John Booth

June 7, 1999

The purpose of this investigation was to characterize the performance of guiding and maintaining focus of the HET with a 10 arc-second coherent fiber bundle, and to compare fiber bundle performance with the performance of pick-off pellicle.

Background:

This has long been the design baseline method of guiding and focussing for the telescope. A probe consisting of a pick-off prism optically coupled to a coherent fiber bundle is inserted in the output beam of the Spherical Aberration Corrector just in front of focus. The output end of the coherent bundle is imaged onto the PXL guide camera at approximately the prime focus plate scale of 0.12 arcsec/pixel. Two to three stars were thought to be necessary for tracking, rotation, and focus.

In the late spring of 1998, hardware fabrication and the relative complexity of implementing this method began to jeopardize scheduled initial PFIP installation. The fiber guiding was shelved in favor of getting on the telescope more quickly with a pellicle. The pellicle is inserted into the SAC output beam well in front of focus, and reflects 8% of the light over the entire 6 arc minute field to acquisition focus. The remaining 92% of the light is transmitted to the FIF focus.

The advantage of fiber guiding was always thought to be the ability to use 100% of the telescope aperture at the science focus, while using a fully-illuminated guide star for offset guiding (with a 90% filling factor of the round fibers in the square bundle).

Clear, demonstrated advantages of the pellicle are the ability to guide on the science object, and the minimal setup time required acquiring a science target and beginning an exposure.

Unforeseen limiting factors:

There are two limiting factors that hinder the use of fiber guiding as originally conceived. The first of these is that there are significant variances in the fiber-to-fiber transmission (see image to on the next page of the fiber bundle transmission). As designed, each fiber covered four pixels on the CCD array. This distorted images through the fibers. In order to smooth over variations 4x4 co-addition is required, making each co-added pixel subtend 0.48 arcseconds on the sky; larger than acquisition camera images.

Second there were significant problems with the polishing of the fiber bundles. This created the gradient across the image. This lead to an apparent shifting of the image across the array as focus was changed, as the centroid of the image shifted because of this ~20% variation across the array.

Third, the small fiber bundle size combined with large stellar images make it difficult to estimate sky brightness. This in turn may have also lead to problems in determining the image sizes in the fiber bundles, as the zero point for the image fitting routines depend on this value.

Finally, as installed the fiber bundle only transmitted ~45% of the light that fell on each fiber. This was measured by imaging a star simultaneously through both systems with the same integration time. This estimate was made based on the assumption that 8% of the total light is transmitted by the pellicle and that the round fibers have a filling factor of 90% of the square field. For faint stars (e.g. the background-limited case) the S/N of the pellicle image is only lower by a factor of 2. Thus, to achieve a similar image through the pellicle, one need only integrate 4 times longer to achieve the same image quality. Experiments with this system demonstrated that even with the current tracking accuracy of the HET,

  1. 10 second exposure times were able to keep an object centroided in the fiber bundle to within ±0.2 arc seconds (and 20 second exposure times seemed only 0.1 arc seconds worse!)
  2. The pellicle was able to guide on an approximately 17th magnitude star (while the fiber image was not) with a 20 second exposure time 2.5 arc minutes from the center of the field during bright time. Note also this was done with the current acquisition camera images, in which significant coma is present in the outer regions of the field.

Guiding sensitivity compared:

We attempted to guide with the fiber with all three forms of co-addition. The ability of one system to guide relative to the other was estimated by the scatter in the centroids found by one system when the other system was guiding. For example, the fibers guiding ability was evaluated by turning on autoguiding in the fiber image and looking at the centroids in the acquisition image. Guiding was evaluated on a 14th magnitude star with 5-second integration times on both cameras. The results were as follows:

·  1x1 co-added fiber images guiding had a 0.3 arc-second larger scatter than did the acquisition image. This was due to the fact that the acquisition image guiding held the star better in place on the fibers, leading to more consistent image shapes (and hence more consistent centroids).

·  2x2 Some improvement found in the fibers ability to guide, though it was not ever any better than the acquisition camera.

·  4x4 binning got rid of “fiberization” of image, effectively changed plate scale to ~0.5 arcsec/pixel. While this lead to a stronger peak flux in the image (~3 times as many counts), it did not allow us to guide any better on this 14th magnitude star. Because of the small field and flat field structure, we could not guide on a 17th magnitude star even with 4x4 binning, while we could do so with the pellicle images.

Thus, there was no advantage to fiber guiding on this moderately bright guide star. When guiding on fainter stars, we found other complications. Identifying a 17th magnitude star on the fiber bundle was not an insignificant task (especially during bright time where there is a significant amount of sky background in the fiber image), though one assumes it would be easier with a better polished bundle. Identification of a similar star in the acquisition camera was not difficult.

Focus sensitivity compared:

Focusing on anything but the 4x4 coadded fiber image was impossible due to the fiber-to-fiber variations. With four fibers per ccd pixel, stellar images began to look like stellar images. However, focus curves derived from ee50 measurements in the acquisition and fiber guider images demonstrated that even with the 4x4 coaddition, the fiberization added significant noise and hindered our ability to measure focus with the fiber images. Focusing with the fibers at 4x4 was limited to +/- 0.2 mm while that of the acquisition camera was good to +/- 0.05 mm.

Focus offset values derived from the acquisition camera images over a trajectory held constant until the primary became significantly unstacked (e.g. required single ring stacking). However the absolute focus offset number varied after each stacking by 0.5 mm. The offset changed for observations at the same azimuth. This may suggest that our ability to focus the telescope absolutely is limited by some aspect of stacking. With the telescope in its current state, our ability to focus during a trajectory is not a significant limitation to getting light down a fiber/slit.

Miscellany:

Guiding exposure limits: seem to be about 1 second minimum in order to sample the guide star well enough to provide an accurate centroid, to about 20 seconds maximum. For focusing, 2 second or longer exposures are needed. To create focus curves during periods of variable seeing, using the smallest measured value of ee50 from three exposures in a row reduced the scatter by removing periods of spuriously high seeing. Upper limit is where we started to see the autoguider produce significant guide corrections (>0.3 arcsec). Note that this upper limit will only increase as tracking and pointing become better with improvements to the mount model.

Setup times for fiber guiding will always be longer than acquisition image guiding. At the least, we will be limited by the ~30 to 60 second travel time of the fiber probes moving about the field. Given the current enthusiasm towards preselecting guide stars (e.g. none from the users), we will not be able to preposition the fiber bundle. This will change somewhat once we begin to fully queue schedule the telescope, as we will have time to select potential guide stars prior to moving to the next science field. Fiber guiding will require the following setup steps for faint object guiding:

  1. Identifying the field in the acquisition frame with the mirror in, selecting guide stars and centering the object on the science fiber/slit
  2. Selecting guide stars
  3. Positioning the fiber
  4. Putting in the pellicle
  5. Verifying the fiber has a guide star
  6. Re-center the object on the science fiber/slit
  7. Pull out the pellicle and start guiding (once the guide star reappears)

Pellicle guiding requires only steps 1, 2, 4, and 6. Thus ~30% more setup time will be needed when using fiber guiding. This may be reduced by 50%, but not all of this setup time will be removed with propositioning the guide probe! Further, there are many regions of the field in which fiber guiders will not be able to acquire stars, as their presence will vignette the science beam. Unless we find a way for fiber guiders to guide on significantly fainter stars, there will always be more guide stars available to the acquisition image than to the fiber guider.

Finding guide stars near the NGP was tested. We went to four random fields within 30 minutes of time from the NGP and found that in all four fields, we had a guide star that could be seen with the acquisition camera through the pellicle. In only one of the four fields did we have a star fainter than 15 (this was our 17th magnitude star). While this sampling is far from complete, it does suggest that the need for more photons in the guide images is not as critical as was initally expected. Note that in Larry Ramsey’s initial document about guiding he assumed we needed 2 stars to accurately guide. We have found that 1 star is good enough as field rotation is not a problem for guiding. Using his numbers I find that we should have 3.5 stars of V > 17 per field at the NGP. Thus, we can expect to find at least one guide star in a NGP field 97% of the time. Bu integrating a bit longer in the other cases, we will be able to acquire a fainter star in the field. Note this is 2.25 times larger than the number for guide probes if the probes are limited to the 2 to 3 arc minute radius annulus proposed for guiding. Thus, in the fiber guiding case, we can expect to find a single guide star of V> 17 only 78% of the time.

Telescope tracking seemed quite good in the regions we tested, and has not been specifically addressed as an area to be improved. We are simply tracking based on the theoretical trajectories, modified by the various pointing and mount models now incorporated in the tracker software. Tracking may be expected to improve further once laser alignment of the SAC over the tracking range is completed.

Offset guiding at the edge of the field is subject to the problem of differential centroid drift. This problem arises from the fact that images at the field edge are formed by a different part of the pupil than images at the field center. As the pupil migrates during a track, the centroids of these images shift slightly relative to each other. We measured a differential drift during one track of about 0.3 arcsec, about what Zemax predicts. The effect can probably be modeled out, but adds to the complexity of fiber guiding (or pellicle guiding) near the field edge.

Using the pellicle, the operator usually has the option of guiding either directly on the science target, or on an object much nearer center, where the effect of differential centroid drift is negligible. This cannot be done with guide probes, as the probe holder must be far enough out of the science beam to not vignette it.

Conclusions:

Many of the complications associated with fiber guiding have been revealed by our recent tests. We conclude that, given the current state of the HET, fiber guiding is not a productive solution. The 8% gain in photons at the focus of the telescope will be overwhelmed by the lost exposure time during setup of the fibers, the limited ability to centroid, and the limited number guide stars in the smaller guiding region. Further, fiber guiding does not help in focusing the telescope. Because of these limitations and the time it would require to overcome them, we recommend shelving fiber guiding until the project reaches a point where our achievements are limited by the throughput of the telescope.