Text Size:
Smaller Text Normal Text Larger Text


Follow us on Twitter

Commissioning of FIES

Around the change of year 2005/2006 the high-resolution FIbre-fed Echelle Spectrograph (FIES) was moved from the observing floor to a dedicated building next to the telescope. This new location provides a stable environment to enable high-accuracy measurements to be obtained with FIES.

A new CCD was installed in FIES in the autumn of 2016, providing more wavelength coverage, with almost no fringing. Over the years FIES was equiped with new fiber bundles; the latest, bundle D, was installed in July 2017.

This page is targeted at (potential) users of FIES, and provides the current status of the development and commissioning of FIES. To keep this document focussed, i have taken away the information that only relates to old bundles A,B,C and the old CCD #13. See the previous commissioning document for that.

Papers describing FIES
FIES light-path overview
Fiber units
Calibration units
Wavelengths, dispersion, etc.
Inter-order distances and PSF
Wavelength zero-point stability
Shutter limitations
Exposure meter
Simultaneous-sky mode

Technical papers describing FIES

The FIES spectrograph report to STC (Lindberg & Frandsen, Sept 1998)

FIES technical report (Lindberg, Lens-Tech, Skellefteaa, Sept 1998)

FIES: a high resolution FIber fed Echelle Spectrograph for NOT (Frandsen & Lindberg, in Astrophysics with the NOT, page 71, 1999)

Design of the the Fiber bundles B and C (Lindberg, 2011)

FIES: The high-resolution Fiber-fed Echelle Spectrograph at the Nordic Optical Telescope (Telting et al. 2014, AN335, 41)

FIES light-path overview

The light from the telescope is picked off in the telescope adapter by an approximately 45 degree mirror that can be moved into the telescope beam at any time. The light from the sky is imaged onto small adjacent mirrors on the fiber head. Each mirror has a central hole, behind which a fibre is located. The fibres are running from the telescope adapter to the spectrograph: one approximately 180 micron fibre (low-res), and three approximately 90 micron fibres (med-res sky, med-res and high-res).

For target acquisition the standby imager StanCam is used. The light reflected from the small mirrors of the fibre head is imaged onto StanCam, to allow the telescope to be tweaked until the target is centered onto one of the fibres.

As of 2010 the atmospheric dispersion corrector (ADC) is installed under the 'roof' of the telescope adapter, and can be swung into the telescope light path before the pickoff mirror.

FIES detector

FIES is equipped with an e2v fringe-suppressed 2k x 2k detector. See CCD15 for details.

Detector characteristics:
The read-out noise at 100 KPix/s speed is about 3.5 electrons. Full-frame read-out time with a single amplifier is about 45 seconds.

FIES spectrograph

The Echelle spectrograph itself is described in detail in documents that can be found on the NOT web pages.
The FIES spectrograph report to STC (Lindberg & Frandsen, Sept 1998)
FIES technical report (Lindberg, Lens-Tech, Skellefteaa, Sept 1998)
FIES: a high resolution FIber fed Echelle Spectrograph for NOT (Frandsen & Lindberg, in Astrophysics with the NOT, 1999)
FIES: The high-resolution Fiber-fed Echelle Spectrograph at the Nordic Optical Telescope (Telting et al. 2014, AN335, 41)

The Echelle orders run vertically on the CCD, and in the blue region they are sufficiently far apart to allow for an interlaced calibration spectrum (ThAr or sky) to be recorded simultaneously (see below for details).

The spectrograph allows a fixed wavelength setting only. Except for the focus drive and spectrograph shutter there are no motorised moving parts in the spectrograph.

The spectrograph can be focused with the computer-operated focus drive, which moves some of the optics in the spectrograph camera towards or away from the CCD. The spectrograph shutter is operated by the CCD controller, and sits at the entrance of the spectrograph just behind the fibres.

Focussing the spectrograph:
Using up/down Hartmann masks we found a top/bottom tilt of the detector of half a pixel Hartmann shift (CCD13). If optimally focussed, this tilt does not seriously compromise the spectral resolution.

Fibre unit

New bundle: bundle D
In the Summer of 2017 the new bundle D, with octagonal fibers, was installed. The resolutions offered, R=25000, R=45000, R=67000, are the same as for the previous bundles. The octagonal fibers are used to significantly improve radial-velocity stability, as these provide better scrambling of the near-field illumination at fiber exit.

Currently the med-res resolution is not functional.

Differences between bundle C and D are

  • for D the stars are imaged onto the fiber entrance; for C the telescope pupil is imaged onto the fiber entrance
  • for D the bundle consists of 2 stretches (approx 42m + 1.5m), with a fiber-to-fiber coupling station near the spectrograph
  • for D the fibers are octagonal
  • for D the fibers are aligned as a square on the sky, for C the fibers formed an arc on the sky
  • for D the slit plate in the spectrograph contains 2 rows of fiber exits, for C the fibers were all in one row
  • for D the second stretch of high-res fiber is a rectangular fiber, for C the round 100 micron high-resolution fiber was equipped with an 50-micron exit slit
  • for D the profile of the high-res fiber has FWHM of 9 pixels; for C a FWHM of 4 pixels.
  • the throughput of bundle D is approximately 0.7 times that of C

    Additionally, bundle D provides accommodation for possible upgrades:

  • a slicer and double scrambler for the high-res mode
  • a 5th fiber for simultaneous wavelength calibrations
  • a return fiber towards the telescope for an etalon-based wavelength calibration mode
  • connections in the spectrograph slit plate for future polarimetry fibers.

    As the layout of the fibers on the telescope/sky is different for this new bundle, a new calibration-light feed and new calibration and masking mechanisms have been installed.

    Design documents by Julian Stuermer describing the new bundle:
    Fiber types
    Telescope coupling
    Spectrograph coupling
    Slit plate

    Spare bundle: bundle C
    In the autumn of 2009 the bundle C was installed. At present, the previously best performing med-res fiber F3 is broken and cannot be repaired.

    Design of the the Fiber bundles B and C (Lindberg, 2011)

    The fiber unit C has a total of 4 fibres: a 200 micron (Fib1, low-res R~25000 ), a 100-micron (Fib3, med-res R~46000 ), a 100-micron with 50 micron exit slit (Fib4, high-res R~67000 ). The sky apertures are 2.5 arcsec diameter for the low-res fibre and 1.3 arcsec for the med- and high-res fibres. The telescope and spectrograph are coupled through the science fibres that are about 40 meters long, Polymicro FBP type. The redundant med-res fibres allowed for a simultaneous interlaced sky spectrum to be obtained.

    Additionally there is a separate high-res calibration fibre for the simultaneous ThAr mode. This fibre can be used in combination with target fibres Fib3 (med-res) or Fib4 (high-res).

    At the telescope end, the starlight is imaged onto the target acquisition mirrors. The telescope pupil is imaged onto the fibre entrance by means of a spherical microlense which sits directly behind the aperture. Fibres are selected using a sliding mask which opens access to either one fibre or two adjacent fibres.

    At the spectrograph the 4 fibres end in a line, with a fifth fibre that comes from a calibration unit located in the spectrograph room. This fifth fibre allows for simultaneous ThAr spectra to be recorded. The echellograms produced by each of the fibres are slightly shifted in the cross-dispersion direction.

    At the fibre exit, the high-res and the calibration fibre are equipped with a slit of 50 micron width, which is needed to reach a spectral resolution of R~67000. The slit losses are approximately 40%. The light of all 5 fibres passes a focal extender that narrows the beam into the spectrograph to limit the grating-overfill losses due to focal-ratio degradation in the fibres. The lens multiplies the Fratio by a factor 1.76, resulting in a beam of approximately F/7.5 going towards the collimator.

    Calibration units

    Bottom calibration unit
    The fifth fibre connects to a calibration unit housing 4 or 5 lamps, of which usually only a halogen and a ThoriumArgon are used. The latter is used for the 'simultaneous ThAr mode', in which an interlaced ThAr spectrum can be recorded while exposing on-sky. The halogen lamp is used to define and trace the orders of the Echellogram of the fifth fibre.

    Top calibration unit
    A second calibration unit is located in the telescope adapter, which calibrates the science fibres. An arm, that transmits light from either a halogen or a ThAr lamp, can be located in front of one of the science fibres at the time. When the arm is in front of a fibre, the light from the telescope is blocked.

    This unit also has a mask that blocks light to fibers that are not in use. Normally, only access to one fiber at a time is admitted by the mask.

    Spectral resolution and wavelength range

    The CCD is not big enough to sample the full echellogram. As a result there is no order overlap in the red ( > 8300 Angstrom), and only a limited number of Echelle orders can be sampled. We have currently chosen a fairly blue setting that samples the Balmer jump and runs redwards from it.

    Due to the way the fibres are mounted at the spectrograph entrance the Echellograms of each of the fibres are a bit shifted with respect to each other in the cross-dispersion direction (X direction of the CCD). At the red end the CCD cuts off the most orders for the low-res fibre and the least for the high-res fibre: see the wavelength range in the table below. The spectral resolution is taken from the FWHM of arc lines at the blaze, and can be slightly lower at the extremes of the orders. Note that the lamp light does not have a central obscuration as the telescope has; in case of incomplete radial scrambling in the 40 meter long fibres, the spectral resolution for starlight might be somewhat lower than for lamp light.

    Fibre bundle C (spare bundle) and CDD13
    fibreFWHMresolutionwavelength range
    #1 low-res 6.4 pix 25000 3630 - 7170 (orders 157 - 80)
    #2 med-res 3.4 pix 46000 3630 - 7260 (orders 157 - 79)
    #3 med-res 3.4 pix 46000 3630 - 7260 (orders 157 - 79)
    #5 high-res calib 2.4 pix 65000 3640 - 7360 (orders 157 - 78)
    #4 high-res 2.4 pix 67000 3640 - 7360 (orders 156 - 78)

    Fibre bundle D (default) with CCD15
    fibreFWHMresolutionwavelength range
    #1 low-res 5.9 pix 25000 3620 - 8700
    #4 high-res 2.2 pix 67000 3640 - 9110

    A list of order numbers and wavelength ranges per order and dispersions per pixel can be found here . Note that the CCD is too small to sample all the listed orders together. With CCD15 order overlap is achieved only below 8300 Angstrom.

    System efficiency

    Note that this section was written based on results with CCD13, while the new CCD15 has much enhanced throughput in the i and z bands. The Exposure-Time Calculator v2.8 is up-to-date for both bundle D and CCD15.

    The plot below shows two measurements made at different nights, for a very blue and a very red wavelength setting, using CCD13. These settings were chosen to check the system efficiency throughout the optical domain. The low-res fibre was used, as this fibre suffers the least from seeing losses and losses due to atmospheric dispersion.

    The system efficiency includes the full light path: atmosphere, telescope, fibres, spectrograph and the detector. For the low-res fibre the system efficiency at the blaze peaks at about 9% in the R band. In regions where there is order overlap, the combined efficiency peaks at about 9.5%.

    The high- and med-res fibres suffer from seeing and atmospheric dispersion losses. If not using the ADC, it is important to center the star on the fibres using a colour filter in the beam to the target acquisition camera (StanCam) to minimise the atmospheric dispersion losses in the wavelength region of interest.

    Fibre bundle C (spare)
    fibre peak system efficiency
    #1 low-res ~9%
    #2 med-res ~8%
    #3 med-res ~9%
    #4 high-res ~5%

    Fibre bundle D (default)
    fibre peak system efficiency
    #1 low-res ~6%
    #4 high-res ~4%

    Inter-order distance and PSF

    For all fibres the inter-order distance (measured between adjacent order centers) decreases from 29.9 15-micron pixels at 4000 Angstrom to 14.7 pixels at 7000 Angstrom. For the old bundle C the profile FWHM of the orders is 7.3 and 4.3 pixels for the 200 and 100-micron fibres respectively. For fibre bundle D the profile FWHM of the orders is 7 and 9 pixels for the low- and high-res fibres respectively.

    Spare bundle C: this bundle allowed simultaneous-ThAr mode for Fib3 (now broken) and Fib4, and simultaneous-sky mode for Fib3 (target) + Fib2 (sky). At approximately 6300Å the sky spectrum starts to get close to the target spectrum of the neighboring order.

    On-blaze detail of an image using the bundle A/B/C high-res fibre (halogen light) with the simultaneous ThAr calibration fibre.


    The current CCD15 detector fringes minimally. Fringing is visible from 8000 Å and onward. At 9000 Å the peak-to-peak fringe amplitude is about 5%.

    Wavelength zero-point accuracy

    Scientific analyses of stellar spectra obtained during a single night or during a few-night run, point at the following achievable wavelength/velocity zero-point accuracy, for bundles A,B,C:

  • stellar spectrum followed by a single separate ThAr exposure using the same fibre: accuracy better than 150 m/s
  • stellar spectrum with simultaneous ThAr light in calibration fibre: accuracy better than 15 m/s
  • short stellar spectrum with separate ThAr spectrum before and after in same fibre: accuracy better than 15 m/s .

    Results show that for a series of blue-sky spectra (i.e. the solar spectrum) with 10000 e- peak flux in simultaneous-ThAr mode, with exposure time 15 seconds, a wavelength zero-point accuracy of better than 4 m/s can be achieved within a single day. We expect that the same holds for well-illuminated stellar spectra.

    From a sequence of ThAr spectra taken during a continuous 28 hours (6/10/2008), we have been able to measure the Echellogram shift due to atmospheric air-pressure variations. The ThAr spectrum shifts by approximately 120 m/s per millibar (HPa). Note that the diurnal atmospheric pressure 'tide' amounts to about 1 Hpa peak-to-peak amplitude. When using simultaneous-ThAr mode this air-pressure effect is largely calibrated out.
    Note that normally the drifts due to temperature changes should be (much) lower than those due to pressure effects, as the temperature is kept constant (within approximately 0.02 degree) in the FIES room and spectrograph. It is expected that the effect of 1 millibar pressure change is similar to that of 0.1 degree temperature change.

    Another result from scientific analysis: the RMS/sqrt(#lines) of the wavelength solution decreases by almost 40% when going from a 5s to 20s ThAr (med-res fiber, bundle B). Apparently the penalty from saturation and diffraction of the strong lines is less than the what is gained from making the forest of weaker lines stronger.

    More recent results for bundle C

    Results reported from planet-confirmation programs point at a problem with the observations made with Bundle C as of 2010. From on-sky tests it has become clear that the radial-velocity accuracy of the medium-resolution fiber F3 in this bundle depends on the atmospheric conditions. In case of stable seeing, intra-night radial-velocity accuracy of better than 10m/s can be achieved. However, in case of variable seeing the radial-velocity accuracy can be three times worse than that: 25-30 m/s.

    For the high-res fiber F4 in Bundle C the RV accuracy does not differ that much from Bundle B. Intra-night radial-velocity accuracy of 5-10 m/s has been observed regularly, regardless of the atmospheric conditions.

    Tests with the fiber shaker show that in cases of unstable seeing the poor radial-velocity accuracy of the medium-resolution fiber F3 in bundle C can be minimised. When shaking the fiber, the radial-velocity accuracy achieved is close to that in stable seeing conditions, and better than 10m/s.
    The shaking has no positive effect in normal seeing conditions.

    Latest results: bundle D

    In July 2017, over a course of 12 days, an RV RMS of 3.3 m/s was achieved for 7 observations of a K-dwarf, with S/N=70 at 6000 Å, using the high-res mode with ThAr exposures before and after each stellar exposure.

    Shutter limitations

    The new main shutter (1-inch diameter, located at spectrograph entrance) has a minimum opening time of 0.03 seconds.

    The exposure meter

    Since autumn 2008, FIES is equiped with an exposure meter that can be used to monitor the count rates during an ongoing exposure. The exposure meter picks off light inside the spectrograph that otherwise would have fallen outside the bottom of the CCD. As the red orders are longer than the blue orders, and hence more red light is falling off the CCD, the exposure meter is more sensitive for red stars than for blue stars.

    The exposure meter is placed under a small pickoff prism that intercepts part of the beam at the intermediate focal plane inside the spectrograph, after the light has been dispersed by the echelle grating, but before cross-dispersion. The placement of the pickoff prism is such that no photons are lost from the part of the echellogram that is sampled by the 2kx2k CCD.

    Below the pickoff mirror is a shutter that can be closed to protect the photomultiplier. The shutter will open when doing a non-dark exposure on the exposure meter.

    The Hamamatsu H8259-01 photomultiplier is continuously providing counts to an external counter that is placed in the front room. The photomultiplier is sensitive over the full FIES range, peaking between 300-600nm. The external counter can be undestructively read out every 3-4 seconds. The accumulated counts are written to the FIES status database by the FIES computer.

    The status program FIESTA is providing a graphical history of the exposure meter counts based on the accumulated count in the FIES status database.

    The dark current of the exposure meter is on the order of 150 counts per second, and has not been accounted for in the below table.

    The exposure meter was last calibrated for bundle C with CCD15, see the EMCC

    Atmospheric Dispersion Corrector (ADC)

    In 2010 the first on-sky tests with the ADC were performed. Although the ADC is primarily developed for use with FIES, it can for some cases be valuable for other instrument as well. The ADC is located under the 'roof' of the adapter, and swings into the telescope beam above the FIES and STANCAM pickoff mirrors.

    ADC transmission as observed by ALFOSC and NOTCAM
    U B V R I J H K
    88% 88% 90% 91% 89% 84% 85% 67%
    Note: in the K band significant emission from the ADC prisms should be expected, leading to about 0.5 mag higher background level.

    The ADC consist of a pair of rotating prisms, that when positioned adequately can fully correct for up to approximately AM=3 . The ADC provides an unvignetted FOV of 3.1 arcmin diameter, and hence can be used for ALFOSC/NOTCAM on-axis spectroscopy as well.

    When the ADC is in the beam, the telescope focus needs an offset of about 560 units.

    When in the beam, the ADC support arm will obscure some of the FOV of the autoguider. This needs to be sorted out. It is expected that the ADC obscures the top-left part of the U-shaped guide field.

    When using the ADC, note that the guide star is not corrected by the ADC. This means that gradually the target will move away from its initial position, if the ADC is used in 'continuous mode'. This may result in object shifts of about 1 arcsec in half an hour of guiding. Hence caution is required when using the 'continuous mode', and for many applications the mode 'automatic-at-preset' should be considered instead.

    Below some first results of using FIES with the med-res fiber F3, with and without the ADC, for different airmasses. Red-drawn spectra are taken with the ADC. All spectra are normalised to the (black) AM=1.1 spectrum taken without ADC. The spectra were obtained with the B-filter in front of the fiber-viewing camera. For AM>1.5 (or so) factors of 3-4 can be won at the extremes of the FIES wavelength coverage, when using the ADC.

    It was also found that the ADC is beneficial for accurate RV measurements at high airmass. Without ADC the RV varies parabolically as a function of echelle order, whereas this effect is much smaller when using the ADC.

    Simultaneous sky mode

    The simultaneous sky mode is not available for the moment.

    In 2010 the possibility of making simultaneous observations of a target star and the sky background was investigated using fiber 2 and 3 (both medium resolution) in fiber bundle C. Tests were made using blue-sky spectra.

    A throughput comparison between fiber 2 and fiber 3 has been made and shows the science fiber (fiber 3) to be 41% more efficient than fiber 2, in mean. A polynomial fit to the difference is shown below.

    When using the instrument in simultaneous sky mode the inter-order distance between the sky spectrum and the science spectrum decreases towards the red, causing problems with scattered light subtraction when FIEStool is used for the reduction. Below is shown the difference between a spectrum taken with fiber 3 and a spectrum taken in simultaneous sky mode. At a wavelength of 6700 Å the difference between the two spectra is larger than 2.5% with increasing discrepancy towards the red.

    Calibration frames can be obtained individually for the two fibers (F2 and F3) so two sets of flats and wavelength-definition frames should be made (in the afternoon). Also a set of two wavelength-calibration frames (one for each fiber F2 and F3) should be made every time an update of the wavelength solution is required, to get accurate wavelengths for both the sky spectrum and the science spectrum. To calibrate the throughput difference of the two fibers a set of blue-sky exposures should be made before sunset in an ABBA sequence (in this case F2 --> F3 --> F3 --> F2) to average out the effect of the changing light level, when merging the frames. Exposure times of 20 seconds are appropriate for this. Make sure the telescope is pointing AWAY from the Sun before opening the dome. Note that the dome-lights used for dome flats doesn't provide sufficient illumination for this calibration.
    Also note that a full reduction of simultaneous sky observations is not yet implemented in FIEStool, so the reduction of simultaneous sky frames requires reducing the science frames twice, using the corresponding order-definitions and wavelength definitions for each of the fibers respectively.

    John Telting

  • Back to top Last modified: 14-Nov-2017