known technical issues
This page gives an account of past and current technical problems experienced
with ESPaDOnS, along with the fixes that were applied or are planned in the
During engineering and commissioning, we observed that the instrument is less efficient
than planned bluewards 700nm (by typically 20% below 600nm, see plot on the right and
information on instrument response).
The origin of this chromatic
loss is not identified yet. One possibility is that the detector has a lower than expected
quantum efficiency in the blue. No CCD upgrade is planned.
An additional 20% loss over the whole spectral domain
is also observed. The origin of this
loss is not identified yet. One possibility is that light losses within the fibre are larger
than expected. New fibre bundles are being built with a newer (and hopefully better) design
and should be implemented in early 2006.
This achromatic loss reached 1.5 mag (a factor of 4) from Feb. to Jun. 2005. This enhanced
loss was due to a severely damaged fibre bundle. Changing the fibre to a spare bundle (early
July 2005) brought the efficiency back to a level comparable with the previous estimate (about
30% below normal).
New fibres bundles, featuring improved connectors and AR coating on both sides, were
implemented in 2006 May; apparently, the situation improved slightly, but the issue is
not fixed completely.
During engineering and commissioning, we observed that the instrument behaves as expected
when measuring circular polarisation and that the detected Stokes V signatures from
well-known magnetic Ap stars are perfectly compatible with values published in the
literature. Zeeman signatures in linear polarisation contain a small but significant
fraction (about 20%) of circular polarisation, witnessing
a crosstalk from V to Q and U
(and vice versa).
The origin of this crosstalk was successfully identified and attributed to the first
triplet lens within the polarimeter (located above the polarisation module) whose barrel
contracts in the cold and produces birefringence stesses in the glass. This triplet lens
was changed once; the crosstalk was reduced (to about 10%) but not completely suppressed.
A new lens was constructed, allowing enough play between the glass and barrel to avoid
stresses at cold temperatures. This new triplet was implemented within ESPaDOnS in 2006
May and tested on the sky in early June. The crosstalk is considerably reduced, down to
a level of only a few percent (typically 2 to 3%). An example of Zeeman signatures
(in both linear and circular polarisation) measured from a hot chemically peculiar star
(the Ap star 78 Vir) is show on the right. It is pretty clear from this plot (click on the
graph to get a larger image) that circular polarisation (V, expanded 5 times and shown in
red) does not leak by more than a few percent into linear polarisation (Q and U, expanded 10
times and shown in green and blue repectively): the line feature at 502nm shows for instance
a large circular polarisation signature but no visible signature in linear polarisation,
demonstrating that the polarisation crosstalk is very small.
Other similar results are visible on the page presenting
examples of collected spectra.
As expected from the instrument conception and as confirmed by tests, ESPaDOnS is poorly
competitive at measuring
polarisation; this reflects essentially the fact that
fibre-fed instruments are not able to achieve the photometric accuracy that continuum
polarisation measurements require.
During enginneering and commissioning, the CCD detector assigned by CFHT to ESPaDOnS was
an enginnering grade detector locally known as EEV1e. Since December 2004, ESPaDOnS is
equipped with a science grade CCD detector known at CFHT as EEV1. The readout noise level
measured during enginneering, commissioning and most of the first runs (up to 2005
Septemeber included) was significantly higher than normal (by typically 50% to 100%) and
strongly variable from one run to the next (and sometimes even from one night to the
This problem was fixed in 2005 December by installing additional grounding of all detector
electronics, of the detector dewar and even of the optical table itself. The problem is
thought to be due to electrical glitches induced by the newport table driving the dioptric
camera of ESPaDOnS (and used to achieve spectrograph focussing). Noise levels are now
back to the normal values (ie to what they were during acceptance
tests at OMP, see information on CCD readout modes).
The thermal control of the detector was slightly inaccurate, the minimum temperature it
reached after each refill changing by several degrees (up to 5 degrees from night to
night). By softening the copper link ensuring the thermal transfer between the dewar
and the detector, this problem was apparently fixed (on 2005 December).
The thermal regulation of the detector was also showing a slight defficiency, with signs
of interruption during readouts. This problem was recently identified and attributed to a
defect in the controler utility board (2005 December). This problem is now fixed and
testing is under way.
In spectral calibration frames (Thorium lamps), a clear charge leak
is observed from strong
lines in the readout direction (ie towards top of CCD, see image on the right). Although
very obvious in strong lines, it is probably also present everywhere at a much weaker level,
and may reflect a charge transfer efficiency (CTE) problem. This problem is not fixed yet,
and may also explain the resolution issue mentioned below.
The resolution achieved with ESPaDOnS with the engineering grade EEV detector (EEV1e) is
about 68,000 in
both 'polarimetric' and 'star plus sky' modes, and 81,000 in the 'star only' mode (see
information on spectral resolution). This is already slightly smaller
than what the instrument was theoretically supposed to reach (70-75 K and 90-100 K for both
modes respectively). We suspected that CTE problems within the engineering detector was
causing this slight loss of resolution power.
Changing from the engineering grade detector (EEV1e) to the science grade detector (EEV1)
made the spectral resolution drop even further, down to about
63,000 and 75,000 for both modes respectively. Given the apparent CTE problem that EEV1
is experiencing (witnessed from the charge leak noticed in Th calibration exposures, see
above), we suspect that our interpretation is correct and that the actual resolution of
the instrument is mostly limited by the CCD itself. No CCD upgrade is planned at the moment.
While the thermal stability of the instrument compares well with expectations (with
temperature changes within the spectrograph
over one night of order 0.1 degree, see plot on the right, dashed
curve), the radial velocity variations induced in the spectra (full line)
is about twice larger than
what is expected from previous laboratory experiments
(of order 3 km/s per degree change).
We suspect that the slow nitrogen evaporation within the dewar, that can produce
a very slight time-variable drift of the dewar head throughout the night (of order of a few
microns), could be responsible
for this effect. Testing along these lines should take place in 2006.
At the moment, this effect is measured (and corrected for) by looking at drifts of telluric
lines within stellar spectra. The relative radial velocity accuracy of ESPaDOnS in this
setup is of order 5 m/s (rms) over one night, and about
15 m/s over a week.
© Jean-François Donati, last update 2006 Jun 09