CFHT Information Bulletin, number 38, First Semester 1998

GECKO Spectrograph

Slavek Rucinski, Resident Astronomer, CFHT. (

Semester 97II observations

GECKO was used over 3 weeks in October 1997. According to the long-term CFHT employees, this run was one of the most quiet and uneventful since GECKO was commissioned. In fact, there were almost no technical problems, in spite of the use of a new, thin, blue-sensitive CCD device Loral-5 and of several setup changes during the run.

An interesting (but somewhat dubious) "discovery" has been made during the run: It was found that small changes in rotation of the grating produced relatively large changes of the effective linear dispersion and of the spectral coverage within a spectral window, as defined by the detector. In fact, this strong dependence of the dispersion and effective resolution on the positioning of the grating is an inherent property of GECKO. This property simply became less known as the time progressed. The background is in fact very simple: The spectrograph uses moderately high diffraction orders between m=6 and m=18. Within each order, the dispersion is expected to be non-linear (can be represented by a cosine curve). In contrast, the dispersion can be taken as approximately constant, even within short spectral windows, only in "traditional" coudé spectrographs which use the low dispersion orders of m=1 or m=2. Thus, the conventionally cited resolution of GECKO of R=120,000 should be taken as a rough approximation only. In fact, within each order, the resolution changes between rather wide limits, typically R=90,000 to R=150,000. This effect should be taken into account in planning of the observations. Some useful graphs are shown in a new version of the GECKO Manual. The Manual is currently modified and updated, and is still not complete because of the changes of the windowing system that will be included in its HTML version. At present, this incomplete Version 2 of the Manual (but with the dispersion graphs) is available as a PostScript version in:

The future GECKO observers should note that selection of the wavelength regions should be done carefully and prudently. Requests to prepare GEKO for observations for many different regions are troublesome for the CFHT staff. Since positioning of the various components of GECKO requires setting several parameters, preparation of the setup files may take many hours of hard work. Thus, if you really want to observe many spectral regions, that may be fine (although moving between regions requires re-focussing and thus leads to a drop in efficiency), but, please, do not request many regions "just in case".


The new CCD detector has been used in GECKO during the semester 97II. The detector has some very nice properties: it is very sensitive, especially in the blue part of the spectrum, and shows excellent linearity (typical deviations < ±0.05%, < ±0.1% at full well of 80,000 electrons). According to the manufacturer specifications, Loral-5 (L5) should be more sensitive than the popular thick chip Loral-3 (L3) as in the following table:
wavelength (nm)
















Some checks confirm the high sensitivity of Loral-5. Dan McKenna made a spot check in the CFH laboratory at the wavelength of 625 nm confirming QE = 81% ± 1%. We also noted high sensitivity in the blue, about 4-times higher relative Loral-3. However, this gain in the sensitivity may be only partly visible to the observer because a larger gain is used for Loral-5 (4.3 electrons/ADU, compared with 2.3 electrons/ADU for Loral-3). The larger gain is needed because of the higher readout noise of about 8 electrons.

The chip is not without some problems: Large areas close to the center and in the upper part of the detector suffer from relatively low full-well levels (75% nominal, i.e. about 60,000 electrons). However, wide bands along the edges of the chip are fully useful. Currently a 200 pixel-wide horizontal band at the edge (serial direction) is used. Loral-5 must be read slowly to avoid the Charge Transfer Efficiency (CTE) and vertical-smear problems affecting the upper up to 1/2 of the chip. The currently used clamp-to-sample time is 8 micro-seconds. The chip also shows some Modulation Transfer Function degradation in comparison with Loral-3; this can be explained by the charge spreading effect in the thin device. Thus, one should use a slightly larger value for the number of pixels per resolution element. Instead of the canonical FWHM = 2.5 pixels, the numbers that we have seen during the 97II semester are closer to FWHM of about 2.8 - 3.0 pixels. As a thin device, Loral-5 also shows fringing in the red part of the spectrum. The amplitude (peak-to-valley) grows approximately as:
wavelength (nm)






% amplitude






A major area of concern emerged in the summer, before the October run, when Loral-5 developed a large spot of decreased sensitivity which became more pronounced with time. Dan McKenna was able to stop and reverse this process by chemical and thermal treatment. The cause of the sensitivity loss was most likely contamination by water or hydrocarbons. It appears that the problem has been cured as no localized areas or single-pixel losses have appeared in two months since October 1997 (such phenomena are most characteristic for general sensitivity losses), but we will pay a special attention to this matter.

From now on, Loral-5 will be the designated detector for GECKO.

Plans for GECKO

The plans for GECKO include building of a fiber-feed system and simplification of the observing procedures.

The fiber feed will be built by Dr. Paul Felenbok at Observatoire Paris in Meudon. The system will permit by-passing the mirror train and will permit feeding the light directly from the Cassegrain Bonnette to the coudé slit room. The length of the fiber must be about 29-30 meters, which means that the transmission losses, especially in the red part of the spectrum, should be reasonably small (about 5-10% in the red part of the spectrum). Since setting up the coudé mirror train is a complex and lengthy procedure, the feed will ease scheduling of the coudé observations. The coudé observations would not have to be scheduled before or after the prime focus observations, the only time when the coudé secondary mirror can be used. The coudé secondary will no longer be useable when the Megacam is built, so the fiber feed will prolong the useful life of the GECKO spectrograph.

We are now planning to move the GECKO observations to the 4th floor, that is to the same place where other observations are currently done. This will permit direct communication between the observer and the Observing Assistant and will add to the overall safety. Currently, contacts are solely via an intercom and do not permit any continuous monitoring of the situation at both ends. GECKO observers were encouraged to come with an assistant, but this cannot be assured in all cases. Single observers, sitting in isolation in the 3rd floor room, can faint or fall sick and the OA would notice that something went wrong no sooner than in typical intervals of the telescope moves. This may be as long 20 to 40 minutes, or even several hours, which may be too long for some health-risk situations. The other reason of moving the observing environment to the 4th floor is that, with most operations now controlled remotely, the observers very seldom have any real needs of entering the slit and coudé rooms.

The coudé observations could be moved to the 4th floor any moment now. The only obstacle is the bulky exposure meter. We intend to move it upstairs or, preferably, to install some simple repeaters of its functions close to the 4th floor terminals.

GECKO not scheduled in the 98I semester

The spectrograph will not be used during the first semester of 1998. The reason is the small number of GECKO proposals of which only two were awarded grades by the time allocation committees above the acceptance cutoff level. Since scheduling of GECKO is difficult and is constrained by the upper-end changes, it was impossible to allocate any time to the GECKO proposals. Thus, this fine spectrograph, one of the best in the world in its class, will not be used this time. Although the low number of proposal may be a statistical fluctuation, it may also indicate a deeper trend which should be contemplated by all who practice high-resolution spectroscopy in France and Canada.

Editor: Dr. T. M. C. Abbott,
Copyright © 1998, CFHT Corporation. All rights reserved.